S

SATELLITE REMOTE SENSING / Aerosol Measurements 1941 SATELLITE REMOTE SENSING Contents Aerosol Measurements Cloud Properties GPS Meteorology Precipitation Surface Wind Temperature Soundings TOMS Ozone Water Vapor Wind, Middle Atmosphere Aerosol Measurements Y J Kaufman, NASA Goddard Space Flight Center, Greenbelt, MD, USA D Tanré, Université de Sciences et Techniques de Lille, Villeneuve d’Ascq, France Copyright 2003 Elsevier Science Ltd. All Rights Reserved. Introduction Aerosols are submicron particles suspended in the air, in the form of smoke (Figure 1), urban and/or industrial pollution, or micron-size dust particles blown from the deserts. Aerosols are an important part of the atmospheric physical and chemical processes (see Aerosols: Climatology of Tropospheric Aerosols; Physics and Chemistry of Aerosols). Aerosols impact cloud properties and also affect precipitation formation (see Aerosols: Role in Cloud Physics. Satellite Remote Sensing: Cloud Properties; Precipitation). Aerosols scatter and absorb solar radiation and hence influence the planetary energy balance (see Aerosols: Role in Radiative Transfer). Measurement techniques have been developed to assess aerosol properties and interaction with the environment. In situ techniques (see Aerosols: Observations and Measurements) measure aerosol composition and physical and chemical properties. Chemical composition is used to relate aerosols to sources of air pollution or to natural processes. Models are used to assess the role of aerosols in atmospheric chemistry and climate. Early on, the scientific community recognized that ground-based or airborne in situ measurements (see Aerosols: Observations and Measurements), though providing very detailed information, have drawbacks that are not shared by satellite and groundbased remote sensing measurements. Satellite measurements determine a smaller array of aerosol parameters, but they measure the properties of the ambient aerosol particles without sampling them on filters and altering the physical environment. Satellites measure aerosol properties globally with a single instrument, ensuring a unified technique of comparing aerosol concentration and properties in different parts of the world. Satellites also integrate aerosol measurements over the entire atmospheric column, which is better for some applications (e.g., the effect on the radiation budget but less desirable for others (e.g., the impacts on human health and on visibility near the surface). Ground-based remote sensing shares some of the characteristics of satellite measurements and some of the in situ measurements. Ground measurements do not have a global coverage like satellites, although the measurements are obtained many times a day in some 100 locations all over the world. Ground-based remote sensors measure the ambient aerosol and provide a wide array of aerosol physical parameters, rather as do measurements in situ. Figure 2 shows aerosol climatology derived from the Aerosol Robotic Network (AERONET) of Sun/sky radiometers that provide daily information on the aerosol properties. Compilations of thousands of measurements during 2–6 years reveal considerable differences in aerosol size distribution, absorption efficiency of solar radiation, and optical refractive index between the major aerosol types. Even for the same aerosol type, e.g., urban regional pollution, aerosol properties vary with the geographic locations 1942 SATELLITE REMOTE SENSING / Aerosol Measurements Figure 1 MODIS remote sensing of fires and smoke in the wild fires in the North West US on 23 August 2000. The red dots are fires detected in the 3.9 mm channel, the rest of the image a visual red–green–blue composite. The black area are burn scars from previous days’ fires (white areas are clouds). MODIS, the Moderate Resolution Imaging Spectroradiometer, has taken measurements on board the Terra and Aqua satellites, at 10:30 a.m. and 1:30 p.m. local times, respectively. owing to differences in aerosol source and atmospheric processes. For example, aerosol in a site near Paris is characterized by lower single-scattering albedo than that in a site near Washington DC, probably owing to a higher concentration of black carbon associated with a higher rate of diesel fuel use. Emissions from Mexico and the Maldives have an even lower single-scattering albedo and higher concentration of the coarse mode, owing to the lower efficiency of fossil fuel consumption by transportation and industry, open fires and lack of filtering from the emissions of the large (coarse) aerosol particles. Satellite techniques have evolved in the last decade from merely reporting an effective measure of the aerosol column concentration (Figures 3 and 4) to the quantitative measurement of the aerosol optical thickness over land and ocean; the assessment of the individual column concentration of submicron(smoke and urban/industrial aerosol) and micronsize (dust) particles using spectral and polarization measurements; the measurements of the aerosol impact on cloud properties and precipitation; and measurements of the aerosol impact on the Earth’s reflection of sunlight to space (review: Kaufman et al. 1997). New satellites have been launched recently with new aerosol measurement capability (e.g., the EOS/ Terra with the MODIS and MISR instruments; Figure 5). Satellite data and ground-based remote sensing are being combined to measure detailed aerosol properties, such as absorption efficiency (expressed by the single scattering albedo, o0 ). Coordinated field experiments that include satellites, AERONET, and in situ measurements are used to characterize the specific aerosol type – e.g., biomass burning, Atlantic aerosol, and East Asian pollution. Remote Sensing of Aerosol Over Land Satellites observe simultaneously the Earth’s surface and the semitransparent aerosol layer above it. The land reflectance of sunlight is highly variable, owing to the variability of land surface cover, making a challenge the sensing of the semitransparent aerosol layer above it. Figure 6 illustrates the land and aerosol SATELLITE REMOTE SENSING / Aerosol Measurements Urban/Industrial aerosol GSFC Creteil/Paris Mexico City Mixed aerosol Maldives Biomass burning Amazonian forest South American cerrado African savanna Boreal forest Single-scattering albedo 1.00 Desert dust Bahrain/Persian Gulf Solar Village/Saudi Arabia Cape Verde Oceanic aerosol Lanai/Hawaii 0.95 0.90 0.85 n =1.39 n =1.40 n =1.47 n =1.44 0.80 0.75 440 870 1020 440 0.20 670 870 1020 Wavelength (Pm) D =1.95 D =1.85 D =1.95 D =1.96
=1.90
=1.80
=1.80
=1.55 0.25 n =1.55 n =1.58 n=1.48 n=1.36 n =1.47 n =1.52 n =1.51 n =1.50 670 0.30 dV/dln R (Pm3 Pm2) 1943 870 1020 440=0.7 440=0.15 440=0.7 440= 0.7 670
=1.10
=0.41
=0.36
=1.40 0.15 0.10 440 0.05 0.00 0.1 1.0 10 0.1 1.0 Radius (Pm) 10 0.1 1.0 10 Figure 2 Averaged optical properties of different types of tropospheric aerosol retrieved from the worldwide AERONET network of ground-based radiometers (http://aeronet.gsfc.nasa.gov). Urban/industrial, biomass burning, and desert dust aerosols are shown for optical thickness of text ð440Þ ¼ 0:7, except oceanic aerosol shown for text ð440Þ ¼ 0:15. Ångström parameter a is estimated using optical thickness at two wavelengths 440 and 870 nm (Reproduced with permission from Dubovik O, Holben BN, Eck TF, et al. (2001) Variability of absorption and optical properties of key aerosol types observed in worldwide locations. Journal of the Atmospheric Sciences 59: 590–608.). effect on the image brightness in the visible and mid-IR parts of the solar spectrum. In Figure 6A we can visually distinguish between regions with and without smoke only if the smoke is thick enough. Some parts of the land regions in the image are as bright as the smoke itself. Only recently algorithms for remote sensing of aerosol over land were developed for regional to global scales. New satellite sensors were developed for this purpose, using different techniques to separate the land contribution from that of the aerosol. These techniques can be grouped into spectral, angular, and polarization methods, examples of which are reviewed below. Spectral Technique In Figure 6, the smoke particles that obscure the image are very small, about 0.4 mm in diameter. Since aerosol particles interact most efficiently with radiation of a wavelength similar to the particle size, they interact very well with the visible radiation (0.47–0.66 mm in Figure 6A) and are almost completely transparent in Figure 6B for wavelengths larger than 1 mm. This property is used to separate the aerosol from the land. The satellite observes the land at 2.1 mm, where there is no obstruction by the atmosphere. An empirical relationship, based on many measurements in different parts of the world, is used to derive a simple relationship between (1) the surface reflectance in the red (0.66 mm) and blue (0.49 mm) wavelengths and (2) that at 2.1 mm. Once the surface reflectance in the visible channels is determined using this relationship, the aerosol opacity and column concentration are derived from the difference between the atmosphereplus-surface reflectance measured from a satellite and the actual surface reflectance derived for these 1944 SATELLITE REMOTE SENSING / Aerosol Measurements Figure 3 Three-month average aerosol effective optical thickness derived from the AVHRR. Color scale is given in the left bottom corner. Heavy aerosol is observed off the coast of Africa next to the Sahara (dust) flowing towards Central America. Biomass burning smoke is also observed flowing from Southern Africa. The Arabian Sea shows heavy dust aerosol, and urban/industrial pollution is observed off the coast of North America and East Asia. (Reproduced with permission from Husar RB, Prospero JM, and Stowe LL (1997) Characterization of tropospheric aerosols over the oceans with the NOAA Advanced Very High Resolution Radiometer optical thickness operational product. Journal of Geophysical Research 102: 16889–16909 – http://orbit-net.nesdis.noaa.gov/crad/sat/atm/aerosol/avhrr/ index.html) Figure 4 TOMS aerosol index showing heavy smoke aerosol from large wild fires in Mexico, 15 May 1998, transported to North America. (Reproduced with permission from J. Herman, NASA-GSFC–http://toms.gsfc.nasa.gov/aerosols/aerosols.html) SATELLITE REMOTE SENSING / Aerosol Measurements 1945 Figure 5 The launch of EOS Terra, 16 December 1999 from the Vandenberg air force base in California (Photograph from the NASA-GSFC archive–http://terra.nasa.gov/About/SC/ about_spacecraft.html) wavelengths. It is easier to view aerosol over darker land with a smaller contribution to the radiation detected by the satellite than over brighter land. Therefore only satellite observations over the darkest part of the image are used to derive the aerosol concentration. This method is applied to data from the MODIS instrument on the EOS-Terra satellite; Figure 7 shows results of aerosol remote sensing over land and ocean from MODIS, and the validation of the results over the land against the ground-based AERONET measurements. Validation campaigns show that MODIS can derive the aerosol optical thickness, t, over the land with an error of Dt ¼
0:05
0:20t. A different spectral technique is used by the TOMS instrument. Here differential absorption in two UV channels is used to view the aerosol. In 0.34 and 0.38 mm, the land and ocean are very dark and the aerosol absorption is decreasing with wavelength. The difference in satellite signal between these two wavelengths is used to detect the aerosol (Figure 4). Figure 6 Large fire near Cuiaba on 25 August 1995, taken from the ER-2 AVIRIS instrument during the Smoke Cloud And Radiation – Brazil (SCAR-B) experiment (Kaufman YJ, Hobbs PV, Kirchhoff VWJH, et al. (1998) The smoke, clouds and radiation experiment in Brazil (SCAR-B). Journal of Geophysical Research 103: 31783–31808. et al., 1998). The image is 10  20 km and 20 m resolution. (A) Heavy smoke emitted from the fire and flowing over Cuiaba. It resembles human vision and is composed of 0.47 mm (blue), 0.55 mm (green) and 0.66 mm (red). (B) 2.1 mm (blue), 1.2 mm (green), and 1.65 mm (red). The smoke is almost transparent at these longer wavelengths, and the fire is clearly seen with its three main temperature zones (blue – glowing; purple – smoldering, emitting the heavy smoke; and white – the fire front. Note that it is much easier to observe the surface features at the long wavelengths that penetrate the smoke. The AVIRIS data were provided by Robert Green from the NASA – Jet Propulsion Laboratory in Pasadena. Angular Technique Three satellite sensors, ATSR on ERS-2, MISR on EOS Terra, and POLDER on ADEOS can view the same location with 2, 9 or 14 different view angles respectively in several spectral channels, within 1946 SATELLITE REMOTE SENSING / Aerosol Measurements Aug. Aug. Sep. Sep. 0.0 (A) 0.2 1.5 0.4 0.6 Aerosol optical depth 0.0 0.8 2.0 1.0 3.0 4.0 Ångström exponent At 0.47Pm 0.75 R = 0.91 1.0 At 0.66Pm R = 0.85 MODIS MODIS 0.50 0.5 0.25 y = 0.86 u0.02 y = 0.86 u0.06 0.0 0.0 0.5 1.0 1.5 0.00 0.00 Sun photometer (B) 0.25 0.50 0.75 Sun photometer Total points = 315 Figure 7 (A) September 2000 average of the MODIS analysis of aerosol over land and removed the results for August. (i) the aerosol optical thickness at 0.55 mm; (ii) the Ångstrom exponent. Black regions are where no aerosol data were retrieved, due to lack of sunlight, ice, and snow cover or bright desert land cover. Note that the pollution in the eastern US in August is associated with elevated optical thickness and a higher Ångstrom exponent (small particles). The same is true for biomass burning in South America in September. Aerosol around the Sahara (in black) is associated with a low Ångstrom exponent, indicating large dust particles. (B) 315-point validation of the MODIS analysis of aerosol over the land, using most of the AERONET stations, in the blue and red spectral wavelengths. The dashed lines are the error predictions when the algorithm was perceived 3 years before launch. (See: http://modis-atmos.gsfc.nasa.gov) a few minutes of observations. Since the atmosphere is more obstructive in slant view directions, owing to the longer optical path through the aerosol layer, the sensors use the difference between the vertical and slant observations to derive both the surface properties and those of the aerosol layer above it. Polarization Technique Solar radiation has more characteristics than just the spectral wavelength and brightness. Solar radiation is an electromagnetic wave that can be visualized as similar to ocean waves. Polarization of the electromagnetic wave can be associated with the direction of variation of the height of wave. While the height of ocean waves is always perpendicular to the surface, light arriving from the Sun has no preferred direction of variation (zero polarization). Once reflected by the surface or atmosphere, the wave may acquire a preferential direction or polarization. (Note that polarized sunglasses preferentially transmit only one direction of polarization, to avoid observing the highly polarized glare over a wet road, say, or ice.) The POLDER instrument on the ADEOS satellite measures polarization. Since the polarization of small aerosol particles is much larger than the polarization of large nonspherical dust particles and the polarization of the Earth’s surface, POLDER can be used to determine the concentration of smoke or pollution independently of the presence of dust or surface reflectance (Figure 8). SATELLITE REMOTE SENSING / Aerosol Measurements 1947 Figure 8 POLDER measurements of small aerosol particles, mainly from biomass burning and urban/industrial activity over land and ocean. The aerosol is detected using the aerosol polarization. The results are given as average for four separate months. (See http://wwwprojet.cnes.fr:8060/POLDER/SCIEPROD/ae9611.htm) Remote Sensing of Aerosol over the Ocean Previous satellite measurements over the ocean were limited to reflectance measurements in one channel (from a geostationary satellite like GOES or METEOSAT) or two channels (from AVHRR/NOAA), and algorithms could derive only the total aerosol content, assuming a given aerosol model. The aerosol model was taken from the literature as the one most representative of the local conditions, and this method has been successfully applied over water with a particular emphasis on Saharan dust studies. An operational global product is provided by NOAA from AVHRR data over oceans (see Figure 3). The new generation of satellite sensors provides well-selected multispectral data – e.g., MODIS on the Earth Observing System/TERRA launched in December 1999 – or multiangular data provided by POLDER on ADEOS launched in August 1996 and by also MISR on EOS/Terra. Polarized reflectance is provided by POLDER. From such additional information it is possible to characterize the aerosol properties better and to derive the aerosol content or optical thickness more accurately. While the polarized reflectance is mainly sensitive to the particle refractive index and the directional reflectances to the optical thickness, it has already been demonstrated that the spectral dependence of the optical thickness carries information on the aerosol’s size distribution. 1948 SATELLITE REMOTE SENSING / Aerosol Measurements Aerosol reflectance, L FoPo 0.01 Dry smoke, reff = 0.10 Pm Urban, reff = 0.20 Pm Wet, reff = 0.25 Pm 0.001 Salt, reff = 1 Pm Dust, reff = 1 Pm Dust, reff = 2.5 Pm 0.4 0.5 0.6 0.8 1 Wavelength (Pm) 2 Figure 9 Typical spectral aerosol reflectance (brightness that corresponds to a given surface reflectance), for several aerosol types: small accumulation mode particles that correspond to smoke or dry urban/industrial aerosol (‘‘sulfate’’); wet larger ‘‘sulfate’’ particles, salt, dust, and a mixture of sulfate and salt. The aerosol optical thickness is 0.2 at 0.55 mm, the solar zenith angle 361, the view angle 241, and the azimuth 861, resulting in a scattering angle of 1351. Converting to reflectance, L is the radiance, F0 the solar spectral flux, and m0 the cosine of the solar zenith angle. Note the reduction with wavelength of the apparent reflectance for all aerosol types. Spectral Technique The best example is the MODIS spectral radiance measured over the dark ocean surface. With the MODIS instrument, we retrieve aerosol size information from the spectral signature of the radiances between 0.550 and 2.2 mm, as shown in Figure 9. The radiance is the product of the spectral dependence of the optical thickness and that of the phase function. Its sensitivity to details of the aerosol size distribution has been found to be lower than those of the optical thickness alone. Numerical analysis shows that it can be used to derive simultaneously the total aerosol optical thickness, a measure of the column loading, and two independent parameters describing the size. Therefore, we derive: (1) the ratio of the contribution to the radiance of micron size versus the submicron particles and (2) the specific size of the dominant aerosol mode. Examples of the three products for September 2000 are given in Figure 10. Angular Technique The directional characteristics of the solar radiation are observed from space by MISR and POLDER instruments in addition to more limited spectral information. The additional multidirectional information can be used to improve retrieval of the aerosol model, since its sensitivity to the particle size is different from that of the spectral information. It can also be used to distinguish spherical from nonspherical particles. The aerosol phase function that can be derived from the wide range of scattering angles is very sensitive to the aerosol shape. In fact, nonsphericity effects have been shown to be important for backscattering directions. For instance, dust particles, which are likely to be nonspherical, have a flatter behavior in backscattering directions of over 1401 than have spherical particles. Polarized Technique The POLDER instrument adds a new dimension to the remote sensing capability, namely the polarized ratio of the reflected radiance. The polarized light is sensitive to the real part of the refractive index, though preliminary analysis of the POLDER data shows that the problem is in fact more complex. The polarized light is sensitive to the refractive index of the small SATELLITE REMOTE SENSING / Aerosol Measurements 1949 Figure 10 Monthly mean statistics of MODIS aerosol retrievals over land and ocean for the optical thickness (center) and for the effective radius (top) and the ratio between the modes (bottom) over ocean. (See http://modis-atmos.gsfc.nasa.gov) particles – i.e., particles that are in the accumulation mode – but no information regarding the coarse mode can be derived. So the final product and its accuracy depend on the respective contributions of both modes to the total aerosol size distribution. In addition, particles that are within the coarse mode are more likely nonspherical (like dustparticles), which makes the interpretation of the polarized signal even harder. Remote Sensing of Single-Scattering Albedo, Fires, CO Satellite measurements are sensitive also to aerosol absorption of the solar radiation. The sensitivity increases over bright reflective earth surfaces, e.g., deserts, since the aerosol can absorb both the downward solar radiation and that which is strongly 1950 SATELLITE REMOTE SENSING / Aerosol Measurements 0.5 Apparent reflectance of the Earth surface plus atmosphere =0 0.4  =0.4 o =0.96 0.3 =0.8 =0.4 0.2 =0.8 o =0.87 0.1 = 0 0 0 0.1 0.2 0.3 0.4 0.5 Surface reflectance Figure 11 Calculated reflectance of the earth surface 1 atmosphere as observed from space at nadir (l ¼ 0:66 mm, y0 ¼ 32 ). Solid blue line – no aerosol (t ¼ 0), only molecular scattering; broken lines – aerosol with low absorption, o0 ¼ 0:96 (green), and high absorption, o0 ¼ 0:87 (red), respectively. The aerosol optical thickness, t, of 0.4 and 0.8 is indicated. reflected. As illustrated in Figure 11, surface brightness determines whether the aerosol increases (low absorption) or decreases (high absorption) the surface-plusatmosphere reflectance. Over the bright surface, satellites measure the balance between aerosol absorption of the solar radiation and scattering of sunlight to space. Therefore the change in the Earth’s brightness due to the additional aerosol in a hazy day vs. that on a clear one, works like a precise laboratory scale to provide an exact measure of the ratio of absorption to scattering (expressed as the singlescattering albedo, o0 – the ratio of scattering to scattering-plus-absorption). Recently, combination of satellite and ground-based remote sensing over west Africa and Cape Verde were used to determine the spectral absorption properties of dust. Landsat images with different column concentration of the dust were used and showed that dust brightens even the very bright Senegal desert landscape (Figure 12). This dust brightening allows the precise detection of dust absorption, and shows that dust absorption in the solar spectrum is much smaller than was previously used by radiative models. Even more important than dust absorption is the effect of human activity on aerosol concentration and the ability to absorb sunlight. Field measurements of biomass burning in the Amazon and regional pollution in the Indian Ocean have shown enhanced absorption by heavy aerosol concentrations (o0 in the range of 0.8–0.9). This enhanced aerosol absorption was shown to affect atmospheric dynamics and the formation of clouds and precipitation. To understand the effect of human activity on climate, through the emission of aerosol, it is important to distinguish among the different natural and man-made aerosol sources. This understanding is aided by satellite measurements of the presence and strength of fires (Figure 1) and the simultaneous emission of carbon monoxide that is largely a manmade product. Both fires and CO are measured simultaneously with the aerosol measurements by instruments on the EOS/Terra satellite. Aerosol Direct Radiative Forcing of Climate The presence of aerosols modifies the solar radiation reflected at the top of the atmosphere (TOA) as well as the radiation transmitted at the bottom. Aerosols also redistribute the energy within the atmosphere. Estimates of the aerosol direct radiative effect are at present based largely upon models that simulate the aerosol cycle (sources, transport, and sinks) and estimate their radiative properties from an a priori knowledge of the global aerosol distribution. Because computations of the model-simulated albedos require a rough description of the aerosol properties, this approach introduces significant errors into the estimates of forcing. For better assessment of the aerosol direct radiative effect, satellite measurements are very convenient, since they measure directly the radiances reflected by the atmosphere. Even should some shortcomings be SATELLITE REMOTE SENSING / Aerosol Measurements 1951 Figure 12 Landsat TM images of dust over the coast of Senegal. (A) 3 May 1987, dust optical thickness of t ¼ 0:8. (B) 17 April 1987, heavy dust with optical thickness t ¼ 2:4. Both images were made with the same color enhancement. Color scale: 0.49 mm (blue), 0.55 mm (green), and 0.66 mm (red). present in the derivation of the aerosol properties, we can still reasonably assume that the radiative quantities can be well restituted and that the conversion of the reflected radiances into fluxes is quite accurate. So, the aerosol properties derived from the measurements are used as inputs of the radiative code. Figure 13 shows the short-wave radiative flux change over ocean due to the presence of aerosols as estimated from the POLDER data. The global mean is around Aerosol flux perturbation (W m2) 4.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5 97 19 7 n. 99 Ju 97 y1 Ma 19 19 97 r. Ap r. 19 97 Ma 19 97 b. Fe 19 96 n. Ja c. De No v. 19 96 8.0 Figure 13 Global and hemispheric means in aerosol radiative perturbation from POLDER/ADEOS1. (Reproduced with permission from Boucher O and Tanré D (2000) Estimation of the aerosol perturbation to the Earth’s radiative budget over oceans using POLDER satellite aerosol retrievals. Geophysical Research Letters 27: 1103–1106.) 5:5 W m2 and is fairly constant over the 8 months of POLDER data. Contrast can be observed for both hemispheres; in the Northern Hemisphere the cooling is larger than in the Southern Hemisphere by 1 to 3 W m2 and depends on the season. Flux can also be derived directly from the radiances. Flux is the twodimensional angular integral on the radiance, and aerosols’ properties are needed only to extrapolate the radiance from a particular direction to the integrated flux value. Using this principle, one of MODIS’s benefits has been the reconstruction of the aerosol fluxes from the measured radiance and the derived aerosol products. To derive the aerosol complete radiative effect, the satellite measurements have to be complemented by measurements at the surface, and the AERONET network of sky radiometers has been shown to be very suitable. The instrument measures the sky radiance in the principal plane in four spectral bands. The sky radiance can be converted easily into flux, in a similar way to that used for the satellite radiances, but with even higher accuracy, since the angular representation of the sky radiance observed by AERONET in the principle plane is much better. One advantage of the technique is that the results are not contaminated by the presence of clouds, since the corresponding cloudy data are excluded from the computations. The measured 24-hour average aerosol impact on the solar flux at the surface per unit optical thickness has been estimated to be around DF=Dt ¼ 80 W m2 in sites where smoke, dust, or urban/ industrial pollution is present. In cases of high amounts of broken clouds, the effect is reduced to 50 W m2. 1952 SATELLITE REMOTE SENSING / Aerosol Measurements Figure 14 (A) Aerosol loading derived from POLDER measurement during the spring of 1997. The aerosol loading is described as aerosol index, a qualitative index that in some conditions is equal to the optical thickness. (B) Cloud droplet radius derived from POLDER measurement for the same period. The unit is microns. (Reproduced with permission from Bréon FM, Tanré D and Generoso S (2002) Aerosol effect on cloud droplet size monitored from satellite. Science 295: 834–838.) SATELLITE REMOTE SENSING / Aerosol Measurements Aerosol Indirect Radiative Forcing of Climate Aerosol may also have an indirect effect by acting as cloud condensation nuclei (CCN). Elevated aerosol concentrations can enhance CCN concentrations and correspondingly the density of droplets within the clouds. Consequently the droplet size decreases (for a similar liquid water concentration). Clouds with 1953 denser droplet populations are brighter because the distribution of drops has a larger surface area to reflect solar light. The smaller droplet size also slows the process of precipitation formation and thus decreases the precipitation rate. For a cloud to precipitate, the droplets have to reach a given minimum size (e.g., a radius of 14 mm) before gravitational settling can start the process of droplet coalescing and precipitation. Direct observation of the aerosol effect on the cloud Figure 15 Example of an AVHRR image over the Amazon basin, showing the presence of fires (red) high clouds that may not interact with the aerosol (T o270 K – blue), low clouds (white or bumpy yellow depending on the droplet size) that do interact with aerosol. Vegetation with varying concentration of smoke is shown from green to yellow. The black lines are of latitude and longitude. Droplet size is derived from the reflectance at the 3.7 mm channel. (Image produced by Rich Kleidman.) 1954 SATELLITE REMOTE SENSING / Aerosol Measurements albedo was performed from ship tracks on stratocumulus clouds off the coast of California, but there is at present no global assessment of such a relationship. From AVHRR/NOAA, a contrast of about 2 mm between the cloud droplets over land and ocean has been obtained. In addition, a similar pattern was observed over ocean between the spatial distribution of the aerosol optical thickness and the distribution of the cloud droplet effective radius. Owing to the unique capability of POLDER to detect the presence of anthropogenic aerosol over land, the results can be extended to the whole globe. In Figure 14, the POLDER results are shown. The largest values of the aerosol loading correspond to regions where the cloud droplet radius is minimum, e.g., over the Indian subcontinent and East Asia. The largest cloud drop radii are observed over remote clean oceanic regions. In Figure 15 we zoom in to the cloud–aerosol interaction in the Amazon Basin during the dry season with heavy and heterogeneous biomass burning smoke aerosol. In the figure the AVHRR image was processed 4 10 3 5 T (qC) H (km) 1 2 3 2 1 20 (A) 5 1 25 0 3 5 6 0 15 to show the aerosol effect on the cloud microphysics. The basic image is a composite of the AVHRR 0.63 mm (red), 0.83 mm (green), and reflective part of 3.7 mm (blue) channels. Fires were imposed on the image as red dots (when the 3.7 mm channel saturates at 325 K) and high clouds, that may not interact with the aerosol, were separated by dark blue colour. Note that in the smoke-clear conditions (dark green, in the top left corner of the image) the clouds are yellow, indicating large droplets (with a small reflection at 3.7 mm), while in the smoke regions the clouds are white, indicating small droplets (with a high reflection at 3.7 mm). Analysis of dozen of images showed that an increase in the smoke optical thickness from a background of 0.2 to 0.8 decreases the droplet size by 4– 6 mm. The decrease larger than that found in POLDER data corresponds to the larger change in the aerosol optical thickness. The effect of pollution sources in the otherwise clean Australian air is shown in Figure 16. The image shows the effect of pollution plumes on clouds and 30 35 Reflectivity (dBZ) 40 0 45 5 10 15 25 20 reff (Pm) (B) 5 1 6 2 3 4 Figure 16 Satellite visualization of the TRMM data of cloud microphysics and precipitation. 30 SATELLITE REMOTE SENSING / Aerosol Measurements precipitation over south-eastern Australia at 04.44 UT on 21 October 1998. The TRMM VIRS instrument visible and 3.7 mm channels show clouds with a large droplet size in the clean area as red (e.g., boxes 1 and 3 in the image) and clouds contaminated by aerosol, with small droplet size, as yellow (e.g., box 2 in the image). The white patches in the clean regions 1 and 3 denote precipitation echoes observed by the TRMM precipitation radar. This is an example of the effect on cloud microphysics and suppression of precipitation. Above the image are two plots that show the detailed measurements. Profiles of the radar reflectivity (Figure 16A) show a strong return in regions 1 and 3 and no return in the polluted region 2. The profile of the cloud droplet radius (as a function of the cloud top temperature) is shown in Figure 16B. The polluted clouds of box 2 show a very small variation in the vertical distribution of droplet size, not passing the 14 mm threshold needed to start the precipitation process. Conclusions In the last decade we have seen the deployment of new and exciting satellite and ground-based remote sensing instruments. These instruments are changing remote sensing into a highly sophisticated global laboratory to measure aerosol, their optical properties and their effects on clouds and radiation. Here we have 1955 given a few examples that show how the traditional role of remote sensing of the aerosol optical thickness is extending to new applications of precise measurements of the contribution of fine and coarse particles to the aerosol optical thickness, particle size, and absorption, which more resemble laboratory research than remote sensing. One advantage of remote sensing over in situ or laboratory measurements is that the aerosol is measured in its natural ambient form. There is no ambiguity regarding the humidity of the air and the state of the volatile organics in the aerosol properties. Although remote sensing cannot measure the aerosol composition, at least the optical properties and forcing of climate are obtained for aerosol in natural conditions. In the next decade, remote sensing will be further enhanced with the launch of lidars (Figure 17). Instruments are being designed to combine the spectral information measured by MODIS and the angular and polarization information from POLDER into a sophisticated mission for aerosol sensing. Together with field experiments, chemical analysis of aerosol composition, chemical transport models, and climate models, remote sensing may be expected, in the next decade to resolve some of the outstanding questions regarding the roles of aerosol in climate and in atmospheric chemistry, and also its influence on human health and life on this planet. Figure 17 A 3D visualization of data collected by the Lidar in Space Technology experiment (LITE) in September 1994, showing a deep haze layer (yellow to red) over the Eastern US and Atlantic Ocean. The yellow lines are wind back trajectories over 5 days period. The LITE data extend from the surface level up to 20 km. (Image reproduced with permission from the web site: http://www-calipso.larc.nasa.gov/ calipso.html) 1956 SATELLITE REMOTE SENSING / Cloud Properties See also Aerosols: Climatology of Tropospheric Aerosols; Observations and Measurements; Physics and Chemistry of Aerosols; Role in Cloud Physics; Role in Radiative Transfer. Dust. Satellite Remote Sensing: Cloud Properties; Precipitation. Further Reading Holben BN, Tanré D, Smirnov A, et al. (2001) An emerging ground-based aerosol climatology: aerosol optical thickness from AERONET. Journal of Geophysical Research 106: 12067–12097. Kaufman YJ, Tanré D, Gordon HR, et al. (1997) Passive remote sensing of tropospheric aerosol and atmospheric correction. Journal of Geophysical Research 102: 16815–16830. Kaufman YJ, Tanré D, Boucher O (2002) A satellite view of aerosols in the climate system. Review for Nature 419: 215–223. King MD, Kaufman YJ, Tanré D and Nakajima T (1999) Remote sensing of tropospheric aerosols from space: past, present and future. Bulletin of the American Meteorological Society 80: 2229–2259. Raes F, Bates T, Mcgovern F and Van Liedekerke M (1999) ACE-2 general overview and main results. Tellus 52B: 111–125. Ramanathan V, Crutzen PJ, Lelieveld J, et al. (2001) Indian Ocean Experiment: an integrated analysis of the climate forcing and effects of the great IndoAsian haze. Journal of Geophysical Research 106: 28371–28398. Ramanathan V, Crutzen PJ, Kiehl JTand Rosenfeld D (2001) Aerosols, climate, and the hydrological cycle. Science 294: 2119–2124. Cloud Properties P Yang, Texas A&M University, TX, USA B A Baum, NASA Langley Research Center, Hampton, VA, USA Copyright 2003 Elsevier Science Ltd. All Rights Reserved. Introduction On any given day, clouds cover about 65% of the planet. In a fairly stable atmosphere, clouds may be cellular in appearance (i.e., cumuliform), or may appear in sheets (i.e. stratiform) that may extend over large horizontal distances. While these clouds may extend over wide areas, their typical geometric thickness is less than 1 km. In unstable atmospheres, clouds may extend from near the planet’s surface to the upper troposphere. As most of the tropospheric water vapor resides near the surface, where temperatures tend to be relatively warm, low-level clouds tend to be composed of water droplets and are typically opaque to a viewer. The opacity is denoted in terms of a quantity known as optical thickness, or optical depth, and is a dimensionless measure of light attenuation caused by the scattering and absorption of energy by atmospheric particles. Clouds forming near the tropopause reside at very cold temperatures and are composed typically of ice crystals. For clouds at intermediate heights between the planetary boundary layer (B1 km above the surface) and the troposphere, clouds may be composed of a mixture of water and ice particles. Water and ice clouds interact with solar radiation differently and have a large influence on the Earth’s radiative energy budget. The energy budget is composed of both solar and terrestrial radiation. Solar radiation spans from ultraviolet (UV; lo0:4 mm where l is wavelength) to infrared wavelengths (l >5 mm). A portion of the incoming solar radiation may be absorbed at the surface and within the atmosphere by clouds, aerosols, water vapor, and other trace gases such as carbon dioxide and methane. Subsequently, absorbed solar radiation is re-emitted at longer wavelength from 5 mm to 100 mm. In recent years, data from operational meteorological satellites have been analyzed for global cloud macrophysical properties such as cloud height, phase (water, ice, or some mixture of both), and microphysical and optical properties such as optical thickness and particle size. Global cloud observations based on satellite measurements serve many uses. In numerical weather models, where the time scale of interest is on the order of hours to days, satellite-derived cloud and clear-sky properties from the geostationary satellites serve as initial conditions for the models, that is, where the clouds are at some given time. Numerical weather models may be regional in extent, covering a specific area such as North America, or global, in which case global and near real-time cloud and clear-sky properties are required for initialization of the models. Satellite-derived cloud properties are also useful for developing and testing global climate models where the time scale of interest is months to years. For this type of use, cloud properties need to be collected, analyzed, and ultimately reduced to a global gridded and time-interpolated product. An example of such a product would be one where each of the cloud properties retrieved during the course of a month is SATELLITE REMOTE SENSING / Cloud Properties 1957 reduced to a monthly average with a time resolution of every 3 hours. To date, meteorological satellites have recorded information over the Earth at a limited number of wavelengths through the use of specially designed filter radiometers. The filters allow radiation only over a very narrow wavelength range to pass through to the detectors. Such narrowband wavelengths are typically chosen in atmospheric ‘windows’, where the atmospheric constituents such as water vapor and carbon dioxide least attenuate the energy along the path from the surface, through the atmosphere, and finally to the satellite. At a minimum, operational satellite data are recorded at a visible (VIS) wavelength (e.g., 0:65 mm), a near-infrared (NIR) wavelength (3:82 mm) and an infrared (IR) wavelength (11 mm). Radiances at visible and near-infrared wavelengths are often converted to reflectances whose values range from 0 to 1. IR radiances are often converted to brightness temperatures through application of the Planck function. Because of the huge volumes of data collected by satellites, the data reduction effort can become quite complex. This article discusses some of the available methods to infer cloud properties such as cloud top pressure, phase, optical thickness, and particle size. Cloud Top Pressure, Height, and Temperature Over the past 20 years, a number of approaches have been developed to infer cloud top heights from satellite data. The simplest method is to use a measurement at a single wavelength – specifically, the IR radiance or, equivalently, the brightness temperature at 11 mm. Data on the atmospheric state (vertical profiles of wind, temperature, and humidity) are obtained from a global, gridded meteorological product available, for example, from the National Oceanic and Atmospheric Administration (NOAA), or by rawinsondes. Rawinsondes are comprised of instrumentation packages carried aloft by balloons, and are launched from a variety of sites around the globe twice per day. The cloud top is derived as the height where the measured brightness temperature matches the profile temperature. While the approach is simple, the operating assumption in this technique is that the cloud is opaque in the radiative sense, meaning that no radiation is being transmitted through the cloud from the atmosphere or surface below the cloud. Low-level clouds tend to be more opaque than high-level ones such as cirrus, but globally only a small percentage of the clouds fall in this category. The International Satellite Cloud Climatology Project (ISCCP) infers cloud properties such as cloud height and optical thickness using both the VIS and IR channels for data collected during daytime hours, but only the IR band for nighttime data. The ISCCP approach involves analysis of geostationary satellite data, which are recorded typically every 3 hours over a specific geographical region, and polar-orbiting platforms, which obtain data globally but see any specific region on the planet twice daily, once during daytime hours and once during nighttime hours. Cloud properties such as cloud height and optical thickness are inferred through comparison of measured VIS and IR radiances with simulations provided by radiative transfer models. The models provide a set of predefined radiative transfer calculations for water and ice clouds as a function of cloud height, temperature, viewing angles (solar zenith, viewing zenith, and relative azimuth angles), cloud phase, optical thickness, and other variables. The ISCCP approach is most accurate for those clouds that are optically thick, such as low-level water clouds and clouds generated in convective situations such as in midlatitude frontal regions. This bispectral approach is not optimal for inferring the properties of optically thin clouds such as cirrus. On many satellite platforms, measurements are obtained at wavelengths located in the 15 mm wavelength region, a region in which atmospheric transmission is dominated by atmospheric CO2. As the wavelength increases from 13:3 mm to 15 mm, the atmosphere becomes more opaque owing to CO2 absorption, thereby causing each channel to be sensitive to a different portion of the atmosphere. This sensitivity is demonstrated in Figure 1, which shows weighting functions for wavelengths ranging from 12 to 14 mm. Each channel has a peak in its weighting function that occurs at a different pressure level from the other channels. The 12 mm channel is shown for comparison – note that its weighting function peaks at the surface. This is a ‘window’ channel that is insensitive to CO2. In the 1970s, a technique known as CO2 slicing was developed to infer cloud top pressure and effective cloud amount (the product of the cloud fraction and the cloud emittance) from radiances measured at wavelengths between 13.3 and 14:2 mm. The pressure at cloud level is converted to cloud height and cloud temperature through the use of gridded meteorological products that provide temperature profiles at some nominal vertical resolution every 6 hours. One benefit to this algorithm is that cloud properties are derived similarly for both daytime and nighttime data, as the IR method is independent of solar illumination conditions. This approach is very useful for the analysis of midlevel to high-level clouds, and even optically thin clouds such as cirrus. The drawback in the use of the 15 mm channels is that the 1958 SATELLITE REMOTE SENSING / Cloud Properties MODIS weighting functions 100 12 Pm 13.3 Pm 13.6 Pm 13.9 Pm 14.2 Pm Pressure (hPa) 200 300 400 500 600 700 800 900 1000 0.1 0.2 0.3 0.5 0.4 0.6 0.7 Weighting function: dt /d(ln p) 0.8 0.9 1 Figure 1 Weighting functions derived for MODIS wavelengths ranging from 12 to 14:2 mm. The weighting function is the derivative of the transmittance profile as a function of pressure. The peak in the weighting function provides an indication of what levels in the atmosphere provide most of the upwelling radiance that will be measured by a satellite. signal-to-noise ratio becomes small for clouds occurring in the lowest 3 km of the atmosphere, making a retrieval problematic for low-level clouds. When low clouds are present, the 11 mm channel is used to infer cloud height. Cloud Thermodynamic Phase While the cloud phase is extremely important in radiative transfer simulations of clouds and the retrieval of cloud properties, it is not always straightforward to determine a cloud’s phase. If the cloud is located in the upper troposphere where the temperatures are extremely cold, it is assumed to be composed of ice. Conversely, if the cloud is located in the boundary layer over warm surfaces, it is assumed to be water. The difficulty lies in the inference of phase when the cloud top temperature lies between perhaps 240 and 273 K. If the cloud temperature is below 233 K, the homogeneous nucleation temperature, it will be composed of ice. If the cloud temperature is above 273 K, it will be composed of water. If the cloud has a temperature between 233 and 273 K, it might be ice, water, or some mixture of both. In the high-latitude storm tracks in either hemisphere, large-scale stratiform cloud decks tend to form with cloud top temperatures in the 250 to 265 K range, and cloud phase is quite difficult to discern. At temperatures below 273 K, the supersaturation of ice is much higher than the supersaturation with respect to water. If water vapor is present in an atmospheric layer at a temperature in this range, say 260 K, and both water and ice particles are present in this layer, then the water vapor will preferentially condense on the ice particles rather than the water particles. As the ice crystals become larger, which may occur over the course of seconds to minutes, the growing ice crystals will begin to fall through the cloud layer. The result may be that the top of the cloud layer is populated primarily by water droplets, with ice crystals falling through the cloud base. In the middle of the layer, the cloud particles may take on aspects of both ice and water, consisting of an ice core with water droplets that have attached to the icy surface through a process called riming. The inference of cloud phase from satellite data under these conditions is quite challenging. Two methods are presented here to infer cloud phase. One method involves IR radiances measured at 8.5 and 11 mm. The radiances are converted to brightness temperatures through the Planck function, and the phase is inferred from the brightness temperature difference (BTD) between the 8.5 and 11 mm brightness temperatures (BTD[8.5–11]) as well as the 11 mm brightness temperature. Ice clouds exhibit positive BTD[8.5–11] values, whereas water clouds tend to exhibit highly negative values. There are three SATELLITE REMOTE SENSING / Cloud Properties 1959 contributing factors to the behavior of the BTD [8.5–11] for ice and water clouds. First, the imaginary component of the index of refraction (mi ) differs for ice and water at these two wavelengths. Second, while the atmosphere is relatively transparent to gaseous absorption, absorption by water vapor in the atmospheric column above the cloud can still exert a considerable effect on the BTD values. As most of the atmospheric water vapor resides in the lower layers of the atmosphere near the surface, the BTD[8.5–11] values will be most affected by the water vapor in a column above low-level clouds rather than high-level clouds that reside above most of the water vapor. Third, while a small effect, cloud particles scatter radiation even at the IR wavelengths, and clouds with smaller particles will tend to scatter more radiation than those with larger particles. Multiple scattering radiative transfer calculations show that for ice clouds, the BTD[8.5–11] values tend to be positive in sign, whereas for low-level water clouds the BTD [8.5–11] values tend to be very negative (o 2 K). The second method is based on reflectances obtained at a visible (VIS) wavelength and a near-infrared (NIR) wavelength (e.g., 0:65 mm and 1:64 mm, respectively). At wavelengths less than about 0:7 mm, clouds composed of either liquid or ice tend to absorb very little solar radiation. However, at 1:64 mm, the mi values for both water and ice increase in comparison with those at the visible wavelength and diverge, with mi for ice being greater than the value of mi for water. From this line of reasoning, one might expect that for two different clouds (one ice, one water) of similar particle size and habit (or particle shape) distributions, the cloud reflectance at 0:65 mm would not depend much on thermodynamic phase, while the cloud reflectance at 1:64 mm would. In theory, given two clouds of differing phases, where each has a fairly high optical thickness and a similar particle size, one might expect the 1:64 mm reflectances for the ice cloud to be less than those for the water phase cloud. Cloud Optical Thickness and Particle Size The fundamental optical properties of clouds are cloud optical thickness and the single scattering properties of cloud particles, which include the single-scattering albedo, the scattering phase function, the scattering/absorption/extinction efficiencies, and the asymmetry factor of the scattering phase function. These parameters essentially determine how much incident radiation is reflected or absorbed by clouds. The single-scattering albedo is defined as the ratio of the portion of energy scattered by a particle to the total extinction (scattering 1 absorption) of energy by the particle. The phase function specifies the percentage of radiative energy that is not absorbed but is instead redistributed by the action of scattering by cloud particles when radiation impinges on clouds. The asymmetry factor of the phase function describes the ratio of forward scattered to backscattered energy, and is a quantity often used in radiative flux calculations. In practice, the single-scattering albedo and the asymmetry factor are parameterized in terms of analytical functions (normally polynomials) of particle effective size for both water and ice clouds. In many radiative transfer models, the radiative properties of clouds are described in terms of particle effective size and either liquid or ice water content (LWC or IWC), depending on the cloud phase. Cloud optical thickness and particle effective size are crucial to the accurate determination of bulk radiative properties of clouds. For this reason, substantial efforts have been made to retrieve cloud optical thickness and effective particle size globally from satellite data. Various methods have been suggested to derive the optical thickness and particle effective size based on narrowband radiometer measurements by airborne or satellite-based imagers. Operational methods tend to rely on IR bands or a combination of VIS and NIR bands. The IR approach depends on the spectral information from thermal emission of clouds, whereas the VIS–NIR approach is based on the reflection of solar radiation. Nakajima and King were among the first to use reflected solar radiation to retrieve cloud optical thickness and effective particle size simultaneously for water clouds. The typical infrared technique employs the brightness temperature or BTD values based on window channels at 8.5, 11.0, and 12:0 mm. Regardless of the detailed spectral information involved in these two methods, they are similar in that both depend on comparison of measured radiance data with simulated radiances derived for similar viewing and atmospheric conditions. The first step in this process is to discuss the generation of reliable libraries of simulated cloud radiances. Single-scattering calculations must be carried out regarding how individual cloud particles interact with incident radiation. For water clouds, the liquid droplets can be well approximated as spheres for light scattering. The scattering properties of an individual liquid sphere can be calculated by using the well-known Lorenz–Mie theory that has been documented in many texts. Hansen and Travis have extensively discussed the effect of size distribution on single-scattering properties of spheres. Their work provides a theoretical framework for using the bulk radiative properties of liquid droplet distributions, which is briefly recaptured here. 1960 SATELLITE REMOTE SENSING / Cloud Properties Within a given water cloud, liquid water droplets span a range of sizes that may be represented mathematically in terms of the gamma distribution, given by N0 ðreff Veff ÞðVeff 1Þ=Veff ð13Veff Þ=Veff nðrÞ ¼ r G½ð1  2Veff Þ=Veff    r ½1  exp reff Veff where N0 is the total number of the droplets in a unit volume; reff and Veff are the effective radius and effective variance that are defined, respectively, as follows: R r2 3 r nðrÞ dr reff ¼ R rr12 2 ½2 r1 r nðrÞ dr Veff ¼ R r2 r1 ðr  reff Þ2 r2 nðrÞ dr Rr r2eff r21 r2 nðrÞ dr ½3 Were one to plot the gamma distribution, one would find that the peak location of the distribution is determined by reff, and that Veff affects the width of the distribution. Typical values of the effective variance for water clouds range from 0.05 to 0.1. For a given size distribution, the bulk scattering properties of cloud droplets may be calculated. For example, the phase function averaged over a size distribution is given by R r2 2 r ss ðrÞPðy; rÞr nðrÞ dr ½4 hPðyÞi ¼ 1 R r2 2 r1 ss ðrÞr nðrÞ dr where ss is the scattering cross-section of droplets and Pðy; rÞ the phase function for droplets with radii of r, which describes the angular distribution of scattered radiation versus scattering angle y. Figure 2 shows the phase functions averaged for size distributions for water clouds at wavelengths 0.65, 1.63, and 8:52 mm. For the 0:65 mm wavelength, the phase function corresponding for a large reff displays scattering maxima at 1401 and 1801. Physically, the two maxima correspond to the rainbow and the glory, both characteristic features of Mie scattering. The phase functions at the NIR wavelength are similar to those at 0:65 mm, but the rainbow and glory peaks are somewhat reduced by absorption within the particle. At the IR wavelength of 8:52 mm, the scattering maxima of the phase function are largely smoothed out due to absorption. Another measure of the relative amounts of scattering versus absorption is provided by the singlescattering albedo. At 0:65 mm, the scattering of incident radiation by cloud droplets is conservative, meaning that energy may be scattered, but not absorbed, by the particles. Thus, the single-scattering albedo is unity at 0:65 mm but less than unity at 1:63 mm. The particle size also affects the singlescattering albedo at 1:63 mm. For example, for effective sizes 4 mm and 32 mm, the particle single-scattering albedo is unity at 0:65 mm whereas the corresponding values at 1:63 mm are 0.9976 and 0.9824, respectively. Because of the difference in single-scattering albedo at the two wavelengths, cloud reflection at 0:65 mm is essentially a function of optical thickness. At 1:63 mm, however, cloud reflectance is sensitive to droplet effective size. This feature of cloud reflectance provides a mechanism to retrieve cloud optical thickness and particle sizes using two channels at visible and near-infrared wavelengths, as will be further explained later in this section. Cirrus clouds are composed almost exclusively of nonspherical ice crystals with various sizes and habits (i.e., shapes). Observations based on airborne twodimensional optical cloud probes (2D-C) and balloonborne replicator data show that typical cirrus habits include relatively simple shapes such as bullet rosettes, solid and hollow columns, and plates, as well as more complex shapes such as aggregates. Research is underway to determine how to calculate accurately the single-scattering properties of these nonspherical ice crystals. In practice, methods such as the discrete dipole approximation (DDA), finite-difference time domain (FDTD) technique, or the T-matrix method are used to calculate the scattering properties of small ice crystals. For ice crystals with sizes much larger than incident wavelength, scattering calculations are performed using the ray-tracing technique based on the principles of geometric optics. Figure 3 shows the phase functions at 0:65 mm wavelength for four types of ice crystals: plates, hollow columns, bullet rosettes, and aggregates. In Figure 3, the particle shapes analytically defined for calculations of light scattering are compared with those obtained from observations in situ. Scattering calculations for aggregates include the effect of surface roughness but the other habits have smooth surfaces. Plates, hollow columns, and bullet rosettes display a strong scattering peak at 221, and are produced by the hexagonal structure typical of ice crystals. In addition to the peak at 221, plates and bullet rosettes display a small peak corresponding to a 461 halo. Bullet rosettes display a strong peak at 101 that is caused by ray refraction through the pyramidal part of the bullet elements. Compared with the phase function for pristine crystal habits such as plates and columns, the phase function for aggregates is essentially featureless owing to the roughened surface texture. The rougher the particle, the more featureless is the phase function. SATELLITE REMOTE SENSING / Cloud Properties 1961 Water Scattering phase function (0.65 Pm) 10 3 re = 4 Pm re = 8 Pm re = 16 Pm 10 2 10 1 10 0 10 1 10 2 10 3 10 4 0 20 40 60 100 80 Scattering angle 120 140 Scattering phase function (1.63 Pm) 10 3 180 re = 4 Pm re = 8 Pm re = 16 Pm 10 2 10 1 10 0 10 1 10 2 10 3 10 4 0 20 40 60 80 120 100 Scattering angle 140 10 3 Scattering phase function (11 Pm) 160 160 180 re = 4 Pm re = 8 Pm re = 16 Pm 10 2 10 1 10 0 10 1 10 2 10 3 10 4 0 20 40 60 80 120 100 Scattering angle 140 160 180 Figure 2 Phase function of water droplets calculated at three wavelengths at 0.65, 1.63, and 8:52 mm for effective radii of 4, 8, and 16 mm. In reality, cirrus clouds are composed of many different crystal habits. To derive the bulk radiative properties of cirrus clouds, we need to consider not only a particle size distribution but also the percentages of the various particle habits that comprise the cloud. For this reason, the derivation of accurate radiative transfer simulations of ice clouds is considered more difficult than for water clouds. For a given size distribution, a number of definitions have been suggested for the effective size. However, it has been found that the bulk optical properties of cirrus clouds are insensitive to the detailed structure of the size distribution if effective size is defined as the ratio of total volume to total projected area, that is reff RP fi Vi ðDÞnðDÞ dD 3 i ¼ RP 4 fi Ai ðDÞnðDÞ dD i ½5 where D is the maximum dimension of an ice crystal, fi is the percentage for ith particle habit, V and A are the volume of and projected area for an individual particle, n is number concentration, and D is the maximum dimension. Unlike the case of water clouds, there is no single analytical expression to describe particle size distributions of cirrus clouds. Only a limited number of measurements in situ are available from cirrus clouds. Part of the difficulty is that aircraft must be able to reach high altitudes, making this a more difficult process than for sampling low-level water clouds. Table 1 lists the effective radii for 12 in-situ size distributions for cirrus clouds. For crystals smaller than 70 mm, we assume that 50% of crystals are bullet rosettes, 25% are hexagonal plates, and 25% are hollow columns. For crystals larger than 70 mm, the percentage is assumed to be 30% for aggregates, 30% for bullet rosettes, 20% for hexagonal plates, and 20% for hollow columns. This 1962 SATELLITE REMOTE SENSING / Cloud Properties 104 103 102 101 100 Phase function 10 10 _1 _2 0 60 120 180 0 60 120 18 80 0 60 120 180 0 60 120 180 104 103 102 101 100 10 10 _1 _2 Scattering angle (q) Figure 3 Four ice crystal geometries commonly observed in cirrus clouds: plate, hollow column, bullet rosette, and aggregate. Also shown are the phase functions for the four shapes. microphysical model for cirrus clouds in terms of particle habit has been used in many practical retrievals of cirrus clouds from satellite imager data. Observations in situ indicate that the effective radius of ice crystals in cirrus clouds range typically from 7 mm to 60 mm. Larger particle radii might be expected for ice clouds formed in convective situations where the updrafts are much faster, at meters per second, than those found under conditions where optically thin cirrus clouds tend to form (centimeters per second). SATELLITE REMOTE SENSING / Cloud Properties 1963 Table 1 Effective size calculated for various sizes distributions obtained from observations in situ Cloud type Effective radius (mm) Cloud type Effective (FIRE-II data) radius (mm) Ci (cold) Cs Ci (warm) Ci uncinus Ci (T ¼ 20 C) Ci (T ¼ 40 C) Ci (T ¼ 60 C) 6.7 14.5 19.7 58.9 25.0 28.0 9.55 Oct. 22 Oct. 25 Oct. 28 Nov. 1 Nov. 2 47.4 50.7 48.6 32.0 40.0 Given the single-scattering properties, radiative transfer computations can be carried out for various cloud optical thicknesses and effective particle sizes for a number of solar and view angle configurations. To calculate the bidirectional radiance of clouds, one can use well-established discrete ordinate or adding/ doubling methods. Figure 4 shows the correlation of 1:64 mm reflectance and 0:65 mm reflectance of cirrus clouds for a number of optical thickness and effective sizes for a given incident-view geometry. The ice crystal shape is assumed as hexagonal plate or column in the simulation. Evidently, the 0.65–1.64 mm correlation is sensitive to particle shape. At higher optical thicknesses (meaning the cloud is more opaque), there is a ‘quasi-orthogonality’ between the optical thick- 0.7 0.7 0.6 0.6  = 12  = 20  = 40 r = 10 Pm =8 0.4 =6 =5 0.3 r = 15 Pm =4 =3 0.2 r = 20 Pm 0.5 Reflectance (O = 1.64 Pm) 0.5 Reflectance (O = 1.64 Pm) ness and particle size curves. As we have mentioned previously, the cloud reflectance at 0:65 mm is sensitive primarily to cloud optical thickness whereas the reflectance at 1:64 mm is sensitive to the particle size. This orthogonality forms the underlying principle for application of the two-channel correlation technique for retrieving cloud optical thickness and effective size. For example, assume the symbol ‘X’ in the left panel of Figure 4 represents the (0:65 mm, 1:6 mm) reflectance values for a satellite imager pixel. One may infer that the cloud optical thickness and effective particle size for that pixel are approximately 18 and 17 mm, respectively. As an alternative or as a complement to the VIS/NIR bispectral retrieval algorithm, infrared channels in the window region (8–12 mm) may be used for retrieving cloud properties. The window region in an important part of the IR spectrum because terrestrial thermal emission peaks within this spectral region. IR-based methods are useful because a single approach may be used for both daytime and nighttime conditions, thereby simplifying the data reduction effort and also the comparison between daytime and nighttime cloud properties. IR methods are insensitive to sun glint over water that is often present in operational data. Interpretation of data over reflective surfaces is often performed more readily using IR methods rather than those that involve VIS/NIR wavelengths. r = 10 Pm 0.4 r = 15 Pm r = 20 Pm 0.3 r = 30 Pm 0.2 r = 40 Pm r = 50 Pm r = 60 Pm =2 0.1 0 0 0.1 r = 30 Pm r = 40 Pm r = 50 Pm r = 60 Pm 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 Reflectance (O = 0.65 Pm) 0.9 1 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Reflectance (O = 0.65 Pm) 1 Figure 4 Theoretical relationship between the reflection function at 0.65 and 1:64 mm for various values of cloud optical thickness and effective particle radius. Hexagonal plate and column are assumed for the results shown in the left and right panels. The solar zenith angle is assumed to be 30 degrees with a nadir view geometry. 1964 SATELLITE REMOTE SENSING / Cloud Properties The underlying principle for infrared retrieval is based on the sensitivity of the brightness temperature or the cloud emissivity (related to blackbody or graybody emission) to cloud optical thickness and particle size. The brightness temperature is the temperature that, when applied to the calculation of Planck function for blackbody radiation, gives the same value as the satellite-measured infrared radiance. The cloud emissivity can be calculated as follows: e ð lÞ ¼ RðBÞ  R RðBÞ  RðCÞ ½6 where R is the upwelling radiance at the cloud top, RðBÞ the upwelling radiance at the cloud bottom, and RðCÞ the upwelling blackbody radiance corresponding to the cloud temperature. In practice, for a given scene, the radiance at cloud base can be obtained by the noncloudy (i.e., clear-sky) pixels. Future Challenges in Cloud Property Retrieval Current efforts to derive a global cloud climatology from satellite data generally do not account properly for multiple cloud layers in pixel-level imager data. To date, operational algorithms are designed to infer cloud properties for each imager pixel under the assumption that only one cloud layer is present. Climatologies of retrieved cloud properties do not address the effect of an optically thin upper cloud layer, such as cirrus, which may overlie a lower cloud layer such as a cumuliform cloud deck. Surface observations show that clouds often occur in multiple layers simultaneously in a vertical column, i.e., cloud layers often overlap. Multiple cloud layers occur in about half of all cloud observations and are generally present in the vicinity of midlatitude fronts and in the tropics where cirrus anvils may spread hundreds of kilometers from the center of convective activity. When multilayered clouds are present, the retrieval algorithms will generally place the cloud layer at a height between the two (or more) actual layers present in the field of view. Satellite cloud climatologies currently available provide a horizontal distribution of clouds, but need improvement in the description of the vertical distribution of clouds. At this point, no reliable method has been developed for the retrieval of microphysical cloud properties (optical thickness, cloud thermodynamic phase, effective particle size) when multilayered, overlapping clouds are present. Even for a single-layered cloud, satellite retrieval algorithms do not account for the effect of a likely vertical variation of cloud microphysical properties, which in turn will decrease the ability of radiative transfer calculations to accurately simulate the cloud. It is unlikely that cloud particles are homogeneously distributed throughout any given cloud. For example, for midlatitude cirrus ice crystal size and habit are typically quite different at cloud top and cloud base. A common assumption in satellite-imager-based cirrus retrieval algorithms is that the radiative properties of a cirrus cloud may be represented by those associated with a specific ice crystal shape (or habit) and a single particle size distribution. However, observations of cirrus clouds have shown that pristine small ice crystals with hexagonal shapes having an aspect ratio close to unity (length and width are approximately equal) are predominant in top layers. The middle layers of cirrus are often composed of well-defined columns and plates, while irregular polycrystals or aggregates are dominant near cloud base. Another interesting area of complexity in satellite remote sensing is caused by mixed-phase clouds. Single-layered clouds composed of mixtures of supercooled water droplets and ice particles have been observed frequently during various field campaigns. Recent analyses of these data and MODIS satellite cloud property retrievals highlight the difficulty of ascertaining phase. If mixed-phase clouds are present in the data, one might expect larger errors in retrieved properties such as optical thickness and particle size than clouds that are primarily of a single phase. From the perspective of satellite remote sensing, the working assumption is that any imager pixel contains either ice or water, but not a mixture. There is no rigorous method available for determining the single-scattering properties of mixed-phase clouds. From the microphysical cloud process perspective that is important for developing cloud model parameterizations, the presence of both ice particles and supercooled water droplets will affect cloud lifetime. Why? It is likely that the ice particles will grow much more quickly from vapor deposition than the water droplets, as the environment may be supersaturated with respect to ice. The result of this process is that the ice particles will rime, grow quickly in size, and fall through the cloud, and the available water will be depleted quickly. The process of glaciation is very important for modelers because the water–ice conversion rates affect cloud lifetime. Details of cloud microphysics, such as cloud water amount, cloud ice amount, snow, graupel, and hail are important for improving cloud retrieval. While approaches exist to retrieve a variety of cloud properties from satellite imager data, it is not an easy problem to compare the satellite retrievals with ground-based measurements of the same cloud. Comparisons are often attempted between surface-based measurements at a fixed location over a long time and satellite measurements that are instantaneous over a SATELLITE REMOTE SENSING / GPS Meteorology 1965 wide area. While these are difficult and often require inventiveness, some confidence in retrievals is often gained through such painstaking efforts. For some cloud properties, it may be possible to compare measurements derived from two or more different satellite instruments; this will be one of the more active areas in future research. See also Aerosols: Role in Cloud Physics. Cloud Chemistry. Cloud Microphysics. Clouds: Classification; Climatology; Measurement Techniques In Situ. Convective Cloud Systems: Modelling. Mesoscale Meteorology: Cloud and Precipitation Bands. Noctilucent Clouds. Parameterization of Physical Processes: Clouds. Further Reading Kidder SQ and Vonder Haar TH (1995) Satellite Meteorology: An Introduction. San Diego, CA: Academic Press. Liou KN (1992) Radiation and Cloud Processes in the Atmosphere. Oxford: Oxford University Press. Mishchenko MI, Hovenier JW and Travis LD (eds) (1999) Light Scattering by Nonspherical Particles: Theory, Measurements, and Geophysical Applications. San Diego, CA: Academic Press. Stephens GL (1994) Remote Sensing of the Lower Atmosphere. Oxford: Oxford University Press. Thomas GE and Stamnes K (1999) Radiative Transfer in the Atmosphere and Ocean. Cambridge Atmospheric and Space Science Series. Cambridge: Cambridge University Press. GPS Meteorology S B Healy, Met Office, Bracknell, Berkshire, UK The Global Positioning System Copyright 2003 Elsevier Science Ltd. All Rights Reserved. Introduction This article outlines two measurement types that are made with the Global Positioning System (GPS) satellite constellation. Radio occultation is a satellite-to-satellite, limb sounding technique that is based on measuring how radio signals are refracted when propagating through the Earth’s atmosphere. It can be shown that, assuming spherical symmetry, this information can be inverted with an Abel transform to yield a refractive index profile. When the atmosphere is dry the refractive index profile can subsequently be used to solve the hydrostatic equation, and temperature profiles with an accuracy of B1 K and a vertical resolution of B1 km can be derived. More generally, if water vapor cannot be ignored, statistically optimal retrieval techniques can be used to extract information on both temperature and water vapor from the measurement. Typical horizontal and vertical scale lengths are derived and the inherent limitations and possible applications of the measurements are discussed. Ground-based GPS measurements are also described. The atmosphere reduces the velocity of the GPS signal, and the additional time taken for a signal to propagate a known distance between a transmitter and a receiver near the surface of the Earth can be inverted to provide an estimate of the total integrated water vapor in a vertical column above a receiver. The accuracy of the measurement is B1–2 kg m  2. The GPS is a constellation of 24 satellites in six orbital planes about the globe. Each satellite has a circular orbit with an inclination of 551, a period of 12 hours, and an altitude of 20 200 km. It transmits radio signals at two frequencies, L1 5 1575.42 MHz and L2 5 1227.6 MHz. Although the primary purpose of the GPS is a tool to aid precise positioning and navigation, it also provides the opportunity for some novel approaches to remote sensing of the Earth’s atmosphere. These arise because the velocity of a GPS signal propagating through the Earth’s atmosphere is reduced because of the refractive index of the medium and the ray paths are curved as a result of refractive index gradients. These two factors increase the time taken for signals to propagate between GPS satellites and receivers a known distances apart. In broad terms, GPS meteorology is concerned with inverting this additional time delay in order to retrieve useful atmospheric information. The two main branches of GPS meteorology deal with radio occultation (RO) with GPS and ‘groundbased GPS’ measurements. The former is a satellite-tosatellite, limb sounding technique based on measuring how the ray paths of GPS radio signals are bent by atmospheric refractive index gradients. It can be shown that, by making certain assumptions such as spherical symmetry, the variation of bending angle with height can be inverted to yield a refractive index profile and subsequently a temperature profile. The technique was pioneered within the astronomical 1966 SATELLITE REMOTE SENSING / GPS Meteorology community and has been used by NASA to study the atmospheres and ionospheres of other planets since the 1960s. It was suggested as early as 1965 that the measurement technique could be applied to sounding the Earth’s atmosphere, but it became a feasible proposition only after deployment of the GPS satellite constellation. The ‘proof of principle’ GPS Meteorology (GPS/MET) experiment – which began producing data in 1995 – was a success, demonstrating that RO measurements using GPS could provide globally distributed temperature profiles with an all-weather capability to an accuracy of B1 K and a vertical resolution of B1–1.5 km. Ground-based GPS measurements provide a means of estimating the total integrated water vapor in the vertical column above a receiver on the Earth’s surface with an accuracy of B1–2 kg m  2. Many of the ideas in this area originate from work by geodesists and geophysicists who spent a great deal of effort trying to estimate and remove atmospheric ‘noise’ when attempting precise positioning or range finding measurements with GPS. Basic Principles of Radio Occultation The geometrical optics approximation describes the physical basis of the RO measurement. When electromagnetic radiation propagates through a medium of varying refractive index, n, its phase velocity becomes c=n and the direction of its path is modified in order to satisfy Snell’s Law, n sin g ¼ constant Sounded region Receiver Transmitter Earth Atmosphere Figure 1 Schematic illustrating the radio occultation concept. A radio signal transmitted by the GPS satellite passes through the atmosphere and is received at the LEO satellite. The motion of the satellites, principally the LEO, enables a region of space to be probed. a expðz=HÞ, where H is approximately the atmospheric density scale height. The magnitude of a bending angle near the surface is typically of order 11. The time required to sample tangent points between an altitude of 80 km and the surface is of order 1 minute. The bending angle and tangent point height are not measured directly. The receiver on the LEO satellite measures the phase of the GPS signal. An excess phase delay, Dc, is introduced because the ray path is curved, rather than the straight line connecting the satellites. This can be written as ½1 where g is the angle between the ray direction and the local refractive index gradient vector. The ray path through the medium satisfies Fermat’s principle of least time and is referred to as ‘the path of stationary phase’. In RO the electromagnetic radiation is produced by a GPS transmitter and the refracting medium is the Earth’s atmosphere. The geometry of an RO measurement with GPS is illustrated in Figures 1 and 2. The radio signal is transmitted by the GPS satellite, passes through the atmosphere, and is received at a low Earth-orbiting (LEO) satellite. The ray is refracted, or bent, in the direction of increasing refractive index, which is usually directed towards the surface. The motion of the satellites, principally the LEO, enables the different levels in the atmosphere to be sounded. The variation of the bending angle, a, with tangent point height, where the ray is parallel to the Earth’s surface, can then be determined. To a first approximation, the bending angle decreases exponentially with increasing tangent point height, i.e. 1 Dc ¼ l Z nðsÞds  d  ½2 Tangent point D Gs I LEO a GPS rR rT T Figure 2 Geometry of an individual RO bending angle measurement. The magnitude bending angle has been exaggerated for illustration purposes. The bending angle, a, is the angle between the ray tangent vectors at the satellites. The impact parameter, a, would be the distance of closest approach in the absence of any bending. r T and r R are the radius values of the GPS and LEO satellites, respectively. SATELLITE REMOTE SENSING / GPS Meteorology 1967 where the integral is along the ray path s; l is the wavelength of the signal, n the refractive index, and d the straight line distance between the satellites. Both satellites are in motion, so the frequency of the received signal is Doppler-shifted. The time derivative of the excess phase, Dc, gives the additional Doppler shift introduced by the atmosphere, Df , which arises because the bending modifies the angle of intersection at the satellites relative to the straight line path. This additional Doppler shift can be written as 1 Df ¼ ðVT . kT  VR . kR  ðVT  VR Þ . kÞ l ½3 where VT and VR are the transmitter and receiver velocity vectors respectively, kT and kR are the unit vectors of the ray at the transmitter and receiver, and k is the unit vector of the straight line connecting the transmitter and receiver. The total bending angle is given by a ¼ cos1 ðkT . kR Þ, so Df and a are clearly related. However, deriving a from Df is an ill-posed problem because there is an infinite number of kT and kR pairs that are consistent with Df . The problem is made well-posed assuming the refractive index field is spherically symmetric and using Bouguer’s formula. This defines a quantity known as the impact parameter, a. a ¼ nr sin f ¼ constant If the assumption of spherical symmetry is valid, and the impact parameter is constant along the ray path then the corrected bending angle as a function of the impact parameter, aðaÞ, can be written as a ¼ 2a Z 1 a d ln n dx ðx2  a2 Þ1=2 dx ½5 where x ¼ nr. This integral equation can be inverted with an Abel transform to recover the refractive index profile, ! Z 1 1 aðaÞ nðxÞ ¼ exp da ½6 p x ða2  x2 Þ1=2 which can be integrated numerically. A useful substitution for removing the singularity in this integral is a ¼ x cosh y. Note that the upper limit of the Abel integral is 1, but in practice signal-to-noise limitations mean that the altitude of the highest bending angle is B80 km. It is therefore necessary to extrapolate the measured bending angle profile a further 100 km above the uppermost measurement. This extrapolation is usually based on some form of climatological bending angle profile. Deriving Temperature Information ½4 from Refractive Index Profiles where n is the refractive index value, r the radius, and f the angle between the ray vector and the radius vector. Bouguer’s formula for a spherically symmetric refracting medium is analogous to the conservation of angular momentum of a classical particle moving under the influence of a central force. Geometrically, the impact parameter, a, represents the distance of closest approach the ray would have had in the absence of any bending. If the position and velocity vectors of the satellites are known accurately then the bending angle and impact parameter values can be found by simultaneously solving eqns [3] and [4] at the satellite locations, with an iterative calculation. This is usually performed assuming the refractive index is unity at the satellites. Note that the ionosphere also causes bending of the ray path paths (in fact, GPS/MET RO measurements have been used to derive electron density profiles in the ionosphere), but fortunately it is dispersive, meaning that the ionospheric component of the refractive index is frequency-dependent. The GPS satellites transmit at two frequencies – L1 and L2 – and the ionospheric signal can be removed or corrected to first order by taking a linear combination of these bending angle values. In the neutral atmosphere the refractive index is related to the total pressure (in hPa), temperature (in Kelvins) and water vapor pressure (in hPa) P, T, and Pw through   c2 P w 6 c1 P þ 2 n ¼ 1 þ 10 ½7 T T where c1 ð¼ 77:6 K hPaÞ and c2 ð¼ 3:73105 K2 hPaÞ are known constants. The refractive index is often rewritten as n ¼ 1 þ 106 N, where N is referred to as the refractivity. A refractivity value calculated using c1 and c2 is accurate to within 0.1% under normal atmospheric conditions. Near the Earth’s surface the refractivity is typically NB330. In regions where the atmosphere is dry ðPw ¼ 0Þ; N is directly proportional to density, r, and the refractivity profile can be used to integrate the hydrostatic equation dP ¼ rg dz ½8 to determine the pressure as a function of height. This calculation is usually performed integrating downwards, towards the surface of the Earth, and it requires an a priori estimate for the temperature at an upper 1968 SATELLITE REMOTE SENSING / GPS Meteorology these will be mapped into the solution if they are not accounted for within the inversion method. For example, a temperature profile can be found by rearranging eqn [7] and solving Pressure (hPa) 10 TðzÞ ¼ 100 GPS/MET 220 240 260 Temperature (K) Figure 3 A comparison of a GPS/MET ‘dry’ temperature retrieval (provided by the Jet Propulsion Laboratory) with the colocated NWP analysis. The measurement was taken at 01.33 UT on 5 May 1995 at 68.81 N, 81.31 W. The cold bias near the surface apparent in the GPS/MET profile is a result of the ‘dry’ atmosphere approximation. level boundary, where the bending angle signal to noise is low. This is typically 260 K at 50 km. The vertical temperature profile can be calculated using the ideal gas law P ¼ rRT ¼ NT c1 ½10 iteratively, but the magnitude of typical uncertainties in the a priori water vapor estimate, Pw ðzÞ, usually mean that the derived temperatures have large errors, often exceeding 5 K, and are of little practical use. Deriving a water vapor profile using a priori temperature information, Ta ðzÞ, with NWP 1000   1 c2 Pw ðzÞ c1 P þ NðzÞ TðzÞ ½9 where R (5 287 J kg  1 K  1) is the gas constant for dry air. An example of a GPS/MET temperature retrieval (data provided by the Jet Propulsion Laboratory) is shown in Figure 3, along with the colocated NWP profile from the Met Office global forecast model. More generally, comparisons between temperature profiles derived from RO and colocated radiosonde and numerical weather predictions (NWP) data have demonstrated that the root-mean-square differences are typically B1.5 K between altitudes of 5–30 km. The Water Vapor Ambiguity The dry atmosphere approximation is reasonable in the stratosphere and upper troposphere, but closer to the surface water vapor makes a significant contribution to the refractivity and this leads to the water vapor ambiguity. This refers to the fact that it is only possible to derive a temperature profile from the measurement given independent, a priori (or background) water vapor information, or conversely derive a water vapor profile using a priori temperature information. It is relatively straightforward to adapt the ‘dry’ processing method to incorporate a priori information, but it should be recognized that the a priori data, derived from climatology or NWP models, contains errors and Pw ðzÞ ¼ NðzÞTa2 ðzÞ  c1 PTa ðzÞ c2 ½11 is more reasonable because the fractional errors in the a priori temperature tend to be smaller. The requirement of a priori information to solve the water vapor ambiguity has led to application of a statistically optimal retrieval technique to the RO problem. This accounts for measurement and a priori errors, and enables the simultaneous retrieval of temperature and humidity. It has been widely used in the processing satellite sounder radiance measurements, and the technique has been used to show that RO measurements potentially contain significant surface pressure information. The method is based on a Bayesian approach for finding the most probable atmospheric state, given an a priori estimate and the measurement data. It requires solving the forward problem, mapping a priori temperature, humidity, and surface pressure information into measurement space, which could be, for example, bending angle or refractivity as a function height. In simple terms, the solution is found by adjusting the a priori information in a way consistent with the estimated errors, in order to produce simulated measurement values that fit the observations to within their expected errors. The ^, is an optimal fit to both background solution vector, x ðxb Þ and measurement ðyo Þ information. For Gaussian ^ is found by minimizing a cost error distributions, x function JðxÞ given by JðxÞ ¼ 12ðx  xb ÞT B1 ðx  xb Þ þ 12ðyo  HðxÞÞT ðE þ FÞ1 ðyo  HðxÞÞ ½12 where B is the expected background error covariance matrix, HðxÞ is the forward model, mapping the atmospheric information into measurement space, and E and F are the expected error covariances of measurements and forward model respectively. The SATELLITE REMOTE SENSING / GPS Meteorology 1969 's 10 Straight line path Pressure (hPa) R ' 100 Statistically optimal Background 1000 200 (A) 220 240 Temperature (K) Figure 5 Horizontal and vertical scale sizes ðDsÞ and ðDzÞ of an individual bending angle measurement. For an atmosphere where refractivity falls exponentially with height, with scale height H, around 68% of the bending occurs over section of path of length Ds ’ 2ðHRÞ1=2 and in a layer Dz ¼ H=2 above the tangent point. 260 Pressure (hPa) 100 are related to the refractivity scale height H. If the refractivity falls exponentially with height ðNðzÞ ¼ Nð0Þ expðz=HÞÞ and the ray is assumed to follow a straight line it can be shown that   ! da 1 y  yT 2 / exp  ½13 dy 2 Dy Statistically optimal Background 1000 0.0 (B) R  'z 0.5 1.0 1.5 2.0 Specific humidity (g kg1) Figure 4 Statistically optimal temperature (A) and humidity (B) profiles derived from the measurement used in Figure 3. superscripts T and  1 denote matrix transpose and inverse. Figure 4 shows the temperature and humidity profiles derived simultaneously with the statistically optimal approach outlined above, for the RO measurement used in Figure 3. The a priori or background information is obtained from the Met Office global NWP model. The solution profiles are found by adjusting background data in order to fit the measured refractivity profile to within its expected error. The temperature profile reproduces the structure around the tropopause shown in Figure 3, but the cold bias near the surface, which arises as a result of the dry atmosphere approximation, is removed. Scale Lengths and Resolution The straight-line distance between the satellites is around 28 000 km, but most of the bending occurs over a relatively short section of the ray path, Ds, centered on the tangent point. The length of Ds and Dz, the corresponding width of the vertical layer over which most of the bending takes place (see Figure 5), where yT is the y (see Figure 2) value at the tangent point, Dy ’ ðH=RÞ1=2 , and R is the radius value at the tangent point. This is a Gaussian function, so around 68% of the bending occurs within yT
Dy. Since Dy is small (B0.035 rd), the corresponding horizontal scale size is Ds ’ 2R Dy ¼ 2ðHRÞ1=2 (see Figure 5), where the factor 2 arises from the symmetry either side of the tangent point. For a refractivity scale height of H ¼ 8 km using R  Re (radius of the Earth) 5 6371 km Ds ’ 450 km. At yT
Dy the ray will be entering and exiting a layer Dz ¼ H=2ð¼ 4 kmÞ above the tangent point level (using Dz ¼ Dy2 =ð2RÞ. The values Dz and Ds give a useful indication of the vertical and horizontal resolution of an individual bending angle measurement. However, the vertical resolution of the retrieved refractivity or temperature profiles can be significantly better because they are derived from a series of closely spaced (in the vertical) bending angle values. In fact, the achievable vertical resolution for an inversion based on geometrical optics is limited by diffraction effects. The geometrical optics picture of the dimensionless ray path between the satellites is only an approximation to the actual propagation of the radio wave, which is valid as the wavelength approaches zero. In reality, the signal measured at the receiver originates from a region of space near the tangent point, with dimensions related to the diameter of the first Fresnel zone, F0 . Consequently, F0 gives a better estimate of the achievable 1970 SATELLITE REMOTE SENSING / GPS Meteorology vertical resolution for an inversion scheme based on geometrical optics. This is around 1.5 km in the upper stratosphere, falling to less than 0.5 km nearer the surface, as a result of the ray bending. Alternative inversion techniques based on wave optics can, in principle, provide profiles each with a vertical resolution superior to the limits imposed by Fresnel diffraction theory. However, in practice the assumption of spherical symmetry limits the achievable vertical resolution, since horizontal structures cause errors in the vertical profile. This is because in most cases is it unlikely that the fine-scale vertical structure (hundreds of meters in the vertical) will have a horizontal scale length of order hundreds of kilometers and be comparable to that of the measurement. Limitations and Errors Sources RO using GPS is undoubtedly an extremely promising new source of data for atmospheric science, but it nevertheless has some limitations of which the user should be aware. For example, the assumption of spherical symmetry is inherent in most techniques used to process RO data, because the impact parameter, a, is assumed to be constant along the ray path. This suggests that measurement errors will be largest in regions of strong horizontal gradients, such as weather fronts, suggesting that more advanced methods, entailing more careful interpretation of the signal, will be required in these areas. In addition, sharp vertical gradients of refractive index, particularly near the surface, can give rise to problems with ‘atmospheric multipath’. This situation occurs when the signal measured at the receiver is effectively composed of multiple rays, which have quite distinct paths between the satellites. It is often associated with sharp gradients in the humidity profile. However, more sophisticated retrieval techniques based on wave optics can be employed to disentangle the multiple rays associated with atmospheric multipath and reduce the errors in the derived profiles. At higher altitudes, around 50 km, the retrievals are limited by instrumental noise and residual ionospheric bending (the signal that is not removed with the first order ionospheric correction). These factors lead to bending angle errors of order 1–3 mrad which can produce spot temperature errors of around 10 K near 50 km. Uses of RO Measurements The success of the proof-of-principle GPS/MET experiment has led to considerable interest in the technique in many areas of meteorology and atmospheric science. For example, the high vertical resolution of the GPS/MET data has recently been exploited to investigate power spectra gravity waves between heights of 15 and 30 km and vertical wavelengths of 2–10 km. The data are also of interest to the numerical weather prediction (NWP) community. A single LEO provides around 500 globally distributed occultations per day, and a number of LEO constellations of 6–8 satellites have been proposed. This quantity of data could be used to provide improved initial conditions for forecast runs with a process known as data assimilation. Data assimilation is concerned with correcting NWP forecast errors by merging the forecast with observations in a statistically optimal way, thereby providing improved initial conditions for the next forecast run. Satellite data is playing an increasingly important role in this area and a new source of high-quality, globally distributed data is of interest to the NWP community. As a result, considerable amount of research has been undertaken in order to establish how best to assimilate RO data into an NWP model. The direct assimilation of either bending angle or refractivity are currently considered the best options. The RO data may also prove useful in the validation NWP models, particularly in the stratosphere. The measurement of geopotential height of fixed pressure levels is a useful approach for detecting warming associated with climate change. In a global warming scenario the thermal expansion of the troposphere will increase the geopotential height of the fixed levels. GPS RO data have a number of useful properties which make them particularly suitable for these purposes. First, unlike radiosondes, which are primarily in the Northern Hemisphere and are over land, the RO data are globally distributed. Since the solution of the hydrostatic equation is top-down, the geopotential height ‘measurements’ depend only on the bending angles above that height, and therefore do not require a surface pressure estimate. In addition, the measurement is based on an excess phase or time delay, rather than the measurement of radiances. As a result, it is less sensitive to calibration or instrumental drift issues common to other satellite-based radiometric measurements. RO has good vertical resolution compared wtih other satellite measurements, and derivation of the geopotential heights is essentially a linear problem. Comparisons between GPS/MET RO measurements and NWP analyses have indicated rootmean-square differences of B20 m for fixed pressure levels in the upper troposphere and lower stratosphere. This provides an upper limit for the measurement error, since the NWP data also contain errors. However, in the context of climate monitoring, it is important to be aware that RO measurements will also contain some level of a priori information. For example, the extrapolation of the bending angle SATELLITE REMOTE SENSING / GPS Meteorology 1971 profile used in the Abel transform and the assumed temperature used to initiate the solution of the hydrostatic equation are usually based on a climatology. Ground-based GPS Measurements GPS satellites are used to derive the total integrated water vapor (IWV) in the vertical column above a receiver on the Earth’s surface. As with RO, the measurement can described within a geometrical optics framework. The geometry of such a measurement is illustrated in Figure 6. The path of the radio signal is slightly curved and the wave velocity is reduced to c=n. The additional transit time required for the ray to propagate between transmitter and receiver is equivalent to an excess path length DL. This has a magnitude of around 2.5 m for a GPS transmitter at zenith, and to a first approximation it increases as 1= sin e where e is the elevation angle. The excess path length, DL, is given by Z DL ¼ nðsÞds  d ½14 s where the integral is taken over the actual path s and d is the straight-line distance between the transmitter and receiver. This can be rewritten in terms of the refractivity, N, as Z 6 DL ¼ 10 NðsÞ ds þ ðs  dÞ ½15 s where s is the total length of the ray path. In practice for elevation angles above 151 the bending of the ray path is small and the magnitude of js  dj is only a centimeter or less, so the term can be ignored. In ground-based GPS it is necessary to use a more accurate expression for the refractive index (or refractivity) than that given earlier. Neglecting nonideal gas effects, the formula usually adopted is N¼ c 1 Pd c 2 Pw c 3 P w þ þ 2 T T T ½16 GPS Straight line path Elevation angle Ray path Receiver H Figure 6 Geometry of the ground-based GPS measurement. The velocity of the signal is reduced because of the refractive index of the atmosphere and the ray path will be curved. However, for elevation angles above 151 the ray bending angle is small, and a straight-line path is a reasonable approximation. where c1 (5 77.6 K hPa  1), c2 (5 64.8 K hPa  1), c3 (5 3.776  105 K2 hPa  1), Pd is the partial pressure of dry air and Pw is the water vapor partial pressure. This can be rearranged by defining a new constant c02 ¼ c2  ðRd =Rw Þc1 , where Rd and Rw are the gas constants for dry air and water vapor respectively, leading to N ¼ c1 Rd r þ c02 Pw c3 Pw þ 2 T T ½17 Hence it can be shown that the total atmospheric delay, DL, is composed of a quantity which is referred to as the ‘hydrostatic delay’, which is the first term on the right-hand side of eqn [17], and the ‘wet delay’ corresponding to the second and third terms. The ‘hydrostatic delay’ typically represents 90% of the total value. In general, the GPS satellite will be at an elevation angle e and will not be at zenith. However, it is possible to use ‘mapping functions’, MðeÞ, to relate the measured total delay, DL, to the total delay at zenith, DLz . To a first approximation the mapping function is of the form DLz ’ sin eDL, but more sophisticated and accurate formulations have been derived. Having mapped the total delay to zenith, the hydrostatic component of the zenith delay (ZHD) can be removed because it can be evaluated to within a few millimeters or better given a measurement of the total pressure at the receiver. This leaves the zenith wet delay terms (ZWD), which can be equated to Z 1 0  Z 1 c2 Pw c3 P w 6 ZWD ¼ 10 dz þ dz ½18 T T2 0 0 The total integrated water vapor (IWV) is given by Z 1 Pw ZWD ½19 IWV ¼ dz ¼ k Rw T 0 where k ¼ 106 ðc3 =Tm þ c02 ÞRw . Tm is a weighted mean temperature which is defined as R 1 Pw dz Tm ¼ R 01 PTw ½20 0 T 2 dz Note that the definition of Tm means that strictly k itself is a function of the water vapor profile at the receiver at the time of the measurement. However, it is usually sufficient to estimate a value Tm from a linear regression of a climatology, using the measured surface temperature at the receiver Ts. An alternative approach is to estimate Tm from the output of an NWP forecast and/or use statistically optimal retrieval techniques to extract the water vapor information from the measurement. Note that GPS IWV estimates 1972 SATELLITE REMOTE SENSING / Precipitation are generally within 1–2 kg m  2 of measurements with water vapor radiometers and radiosondes on typical IWV values of 20–40 kg m  2. See also Satellites: Orbits; Research (Atmospheric Science). Weather Prediction: Data Assimilation. Further Reading Bevis M, Businger S, Herring T, et al. (1992) GPS meteorology: remote sensing of atmospheric water vapor using the global positioning system. Journal of Geophysical Research 97(D14): 15787–15801. Born M and Wolf E (1986) Principles of Optics. London: Pergamon Press. Fishbach FF (1965) A satellite method for temperature and pressure below 24 km. Bulletin of the American Meteorological Society 9: 528–532. Fjeldbo G and Eshlemann VR (1968) The atmosphere of Mars analyzed by integral inversion of the Mariner IV occultation data. Planetary and Space Science 16: 1035–1059. Leick A (1990) GPS Satellite Surveying. New York: Wiley. Melbourne WG, Davis ES, Duncan CB, et al. (1994) The Application of Spaceborne GPS to Atmospheric Limb Sounding and Global Change Monitoring. Jet Propulsion Laboratory JPL. Publication 94-18. Rodgers CD (2000) Inverse Methods for Atmospheric Sounding: Theory and Practice. World Scientific Publishing. Terrestrial, Atmospheric and Oceanic Sciences (2000) Special issue for Applications of the Constellation Observing System for Meteorology, Ionosphere and Climate (COSMIC). 11(1). Ware RH, Fulker DW, Stein SA, et al. (2000) Suominet: A Real-Time National GPS Network for Atmospheric Research and Education. Bulletin of the American Meteorological Society 81(4): 677–693. Precipitation Guosheng Liu, Florida State University, Tallahassee, FL, USA Copyright 2003 Elsevier Science Ltd. All Rights Reserved. Introduction Precipitation is one of the least well measured atmospheric parameters, especially over the vast oceanic regions of the globe. Two major obstacles contribute to the lack of comprehensive global precipitation measurements. First, there are few surface-based observations over the oceanic areas, which cover about two-thirds of the Earth’s surface. Second, precipitation is highly variable in both time and space compared to atmospheric variables such as temperature and pressure. Rainfall measured by a rain gauge at a given location can be significantly different from that measured just a couple of hundred meters away; similarly, rainfall measured at a given time can be significantly different from that measured just minutes earlier or later. As a result, a high-density rain gauge network is required in order to reasonably measure rainfall over a given area if one attempts to derive the area rain total from rain gauge observations alone. Such a high-density rain gauge network is generally not available even in well-developed countries and regions. Although ground-based radars can provide better spatial and temporal coverage than rain gauges, well-calibrated radars are only available in limited land regions in developed countries. These problems related to surface-based measurements make satellite remote sensing of precipitation indispensable. Satellite remote sensing of precipitation is based on the radiative intensities emitted or reflected by cloud and precipitating hydrometeors. For infrared and microwave wavelengths, the radiative intensity is often expressed in terms of brightness temperature, defined as the temperature that is required to match the measured intensity to the Planck blackbody function. Brightness temperature in the infrared often represents the physical temperature of the cloud top because most clouds are optically thick for infrared radiation. Microwave radiation, on the other hand, can penetrate through cloud and rain layers, and its intensity reflects the integrated contribution by all water drops and ice particles in the atmospheric column. In the visible spectrum, the measured radiative intensity is due to the reflection of sunlight by clouds and surface features. The dependence on sunlight limits the utility of visible sensing to daylight hours. Although visible radiation has a deeper penetration than infrared radiation, visible reflectivity still represents only the top portion of clouds. Methods for passive satellite remote sensing of precipitation may be divided into the following three categories, based on whether the information received by a satellite represents the physical properties near the SATELLITE REMOTE SENSING / Precipitation 1973 cloud top or over the whole atmospheric column: (1) sensing by visible/infrared radiation, (2) sensing by microwave radiation, and (3) sensing by combination of visible, infrared and microwave radiation. Visible and infrared methods are physically indirect because precipitation is derived from the radiative properties near the cloud top. In comparison, microwave techniques use more direct information on the vertical distribution of hydrometeors. In the following sections, the precipitation signatures of different wavelengths in the electromagnetic spectra are first discussed using an actual satellite observation. The principles of satellite remote sensing under each of the aforementioned categories are then explained, omitting specific details of any particular retrieval algorithm. It should also be noted that active sensing by spaceborne radars is another important development for satellite remote sensing of precipitation. However, this topic is not covered in the following for two reasons. First, the basic principle of radar sensing from space is essentially the same as for sensing from ground, i.e., it employs the radar reflectivity–rainfall rate relationship. Additionally, there has been only one precipitation radar in space so far, the Tropical Rainfall Measuring Mission Precipitation Radar. Radar sensing from space is still in an early stage. The Radiative Signatures of Precipitation The radiative signatures of precipitation may be understood by examining the observations of a hurricane shown in Figures 1 and 2, which display data simultaneously collected by different sensors on the Tropical Rainfall Measuring Mission satellite. Figure 1 is the thermal infrared image over south-eastern Pacific Ocean. The image covers an area approximately 720 km wide by 3050 km long. Figure 2 shows the observed radiative properties at satellite nadir along line A–B (see Figure 1), which crosses the outer cloud band of the hurricane. The parameters shown in Figure 2 include the space radar-derived rainfall rate cross-section, the near-surface rainfall rate (also derived from the space radar), visible reflectivity (0.63 mm), and brightness temperatures at infrared (11 mm), and microwave (19 and 85 GHz, horizontal polarization) wavelengths. The radar data indicate that the major rainfall area corresponds to the deep clouds to the right of the hurricane’s eye (Figure 2). Compared to those in the cloud-free area near point B, radiometric properties for the rainy areas show the following features: high visible reflectivities, low infrared brightness temperatures, high brightness Figure 1 Thermal infrared image of a hurricane over the southeastern Pacific Ocean observed by an infrared radiometer on the Tropical Rainfall Measuring Mission satellite. The image covers an area approximately 720 km wide by 3050 km long on 5 January 1998. temperatures at 19 GHz, and low brightness temperatures at 85 GHz. Most of the areas covered by the spiral cloud are actually not associated with rain, although clouds in those areas have low infrared brightness temperature and high visible reflectivity. It is the microwave brightness temperatures that most closely follow the radar-observed rainfall variation. The radiative properties shown here are the fundamental basis for satellite remote sensing of precipitation. They are explained theoretically below. Visible Reflectivity In the visible spectrum, reflectivity increases with cloud optical depth, which is approximately proportional 1974 SATELLITE REMOTE SENSING / Precipitation 0 Rainfall rate (mm h-1) Altitude (km) 15 1 2 3 4 5 6 7 8 9 10 Radar rain cross-section mm h-1 Cloud droplets are very absorptive in the thermal infrared spectrum. A consequence of this high absorption is that the cloud top may be viewed as the surface of a blackbody having a temperature the same as the air temperature at the level of the cloud top. Therefore, infrared brightness temperature indicates cloud top temperature. At least in the tropics, most rainfall is associated with well-developed convective systems that have tall cloud tops. Statistically, colder infrared brightness temperatures often indicate higher rainfall rates at the surface. However, shallow convection with warm cloud tops does sometimes produce substantial rainfall. On the other hand, nonprecipitating cirrus clouds do not produce rainfall although they also have cold cloud-top temperatures. Problems associated with the infrared sensing of rainfall are more serious in the midlatitudes where most precipitation is produced by frontal stratiform clouds, for which cloud top temperature and precipitation are less correlated than for deep tropical convection. 10 5 0 Near-surface rain 10 1 Reflectivity 0.9 0.6 0.3 Visible T B (K) Infrared 275 250 225 200 Microwave Brightness Temperature 85 GHz T B (K) Brightness temperature 0.0 300 250 200 150 19 GHz Microwave A Infrared Brightness Temperature B Figure 2 Radiative properties of hurricane clouds along line A–B shown in Figure 1, including distance–height cross-section of rainfall rate from radar, near-surface rainfall rate, reflectivity in visible, and brightness temperatures at 11 mm (infrared) and 19 GHz and 85 GHz (microwave). to liquid water path (vertically integrated liquid water) if the effective particle size remains constant. Therefore, clouds with higher values of liquid water path are more reflective. Generally, these clouds are also more likely to be associated with precipitation. This is the underlying principle for sensing precipitation by visible reflectivity. However, the sensitivity of visible reflectivity to liquid water path decreases with the increase of optical depth, and becomes virtually insensitive when optical depth is larger than 100, a value that a raining cloud easily exceeds. Consequently, instead of sensing the entire vertical column, reflected radiation at a visible wavelength reflects the microphysical properties only near the top portion of a cloud. Therefore, the relation between the visible reflectivity and rainfall rate at the surface is rather indirect. Microwave radiation observed by satellite measures the integrated radiative effects of the surface, atmospheric gases, and hydrometeors. Microwave brightness temperatures may either increase or decrease with increasing rainfall rate, depending on the frequency and the cloud microphysical properties. To understand the microwave signatures, consider an idealized rain cloud that contains raindrops below the freezing level and ice particles above. Although accurate estimation of satellite-received radiation requires solving a radiative transfer model including absorption and multiple scattering, the primary radiative signature may be understood by examining the approximation of eqn [1]. TB  Ts ½1  e2tw ð1  es Þeti ½1 In eqn [1], TB is the brightness temperature received by the satellite, Ts is the surface temperature, tw and ti are the optical depths for the raindrops and ice particles, respectively, and es is the surface emissivity. For simplicity, emission from atmospheric gases is not included in this equation although its contribution is important, particularly near water vapor and oxygen absorb frequencies (e.g., 22 GHz and 60 GHz). The emissivities for land and water surface are roughly 1 and 0.5, respectively, for microwave frequencies commonly used for precipitation retrievals. Consider the following two situations: Low-frequency (o20 GHz) microwave radiation If the frequency is sufficiently low, scattering by ice SATELLITE REMOTE SENSING / Precipitation 1975 particles aloft becomes negligible, i.e., ti  0: Equation [1] then becomes eqn [2]. TB  Ts over land TB  Ts ½1  e 2tw ð1  es Þ over ocean ½2 It is seen that rainfall cannot be detected over land by low-frequency microwaves because of the high surface emissivity. Since ocean surface temperature and emissivity generally do not vary dramatically, the small spatial scale or short temporal scale variation of brightness temperature in eqn [2] can be attributed to the change in optical depth of raindrops, tw , which is approximately proportional to integrated total rainwater amount. Because total rainwater is closely related to the surface rain, low-frequency microwave brightness temperature over the ocean provides a relatively direct representation of rainfall rate. This positive correlation between rainfall rate and brightness temperature is shown in Figure 2 for 19 GHz. Because the increase of brightness temperature is due to the emission by raindrops, the rainfall signature at low microwave frequencies is called an emission signature. Figure 3 depicts brightness temperatures calculated by a radiative transfer model for nadir viewing at 19 GHz for various assumed freezing levels. Brightness temperature increases with rainfall rate until reaching a maximum that indicates the saturation of microwave radiation. If rainfall rate further increases beyond the saturation point, the brightness temperature starts to decrease. The saturation problem prevents higher rainfall rates from being retrieved using microwave emission signatures. Brightness temperature (K) High-frequency (480 GHz) microwave radiation For high-frequency microwaves, scattering by ice particles aloft is no longer negligible; rather it becomes the dominant signature of the rain cloud. The optical depth due to raindrops, tw , usually is so large at high frequencies that e2tw  0. Equation [1] then becomes eqn [3]. TB  Ts eti ½3 Brightness temperature decreases with increasing optical depth of ice particles. The low value of the imaginary part of the ice dielectric constant determines that scattering is the dominant process for the interaction between ice particles and microwave radiation. Since the scattering cross-section is approximately proportional to the sixth power of the particle diameter, large and dense ice particles contribute the most to ti . Therefore, lower brightness temperatures at high microwave frequencies indicate more large ice particles aloft, which is commonly an indication of heavier rainfall rates at the surface. This relationship is seen in Figure 2 for 85 GHz microwave observations. Since the decrease of brightness temperature is caused by ice scattering, the rainfall signature at high microwave frequencies is called the scattering signature. Compared to the microwave emission signature, the scattering signature is a relatively indirect indication of surface rainfall. In Figure 4, the model-simulated brightness temperatures at 92 GHz are shown for a 451 viewing angle. The brightness temperature falls more than 100 K for rainfall rate increasing from 1 to 15 mm h  1 given a 1 km ice layer. It must be cautioned that the magnitude of the brightness temperature depression due to ice scattering depends greatly upon ice particle size and density, for which there have not been sufficient observations so far. Sensing by Visible and Infrared Measurements 250 200 5 km 4 km 150 0.1 3 km 2 km 1 km 1 10 100 1000 _ Rainfall rate (mm h 1) Figure 3 Brightness temperature for nadir viewing over an ocean surface at 19 GHz for various assumed freezing levels calculated by a radiative transfer model. (Adapted with permission from Wilheit (1986).) Since visible reflectivity and infrared brightness temperature are physically indirect indicators of surface rainfall, satellite retrieval techniques based on visible and infrared measurements are generally based on regression. That is, surface rainfall data, measured by rain gauges or radars, or both, are considered to be the true values; co-located satellite-measured radiative properties are regressed against true values to derive a statistical expression relating surface rainfall to satellite measurements. The true rainfall data are usually available only for limited regions and periods. As a result, most of these algorithms are subject to significant errors in regions where the climatological conditions are different. In addition, because of the statistical nature of these algorithms, retrieval 1976 SATELLITE REMOTE SENSING / Precipitation Mean particle radius (Pm) 100 200 0.1 0.5 Particle density 0.9 _ (g m 3) 0 km Brightness temperature (K) 250 200 0.5 km integral ATI (eqn [4]). Z ATI ¼ AðX > Xth Þ dt ½4 where X is the measured radiative property, AðX > Xth Þ is the area with X exceeding a threshold Xth , and the integration is over time t. The most utilized, infrared algorithm that uses this principle is the GOES Precipitation Index ðGPIÞ which gives the rain total over an area of 2.51 latitude by 2.51 longitude and for a time period of Dt (eqn [5]). GPI ¼ R0 AðTB oTB0 Þ Dt ½5 1 1 km 150 5 km 3 km 100 0 5 10 15 _ Rainfall rate (mm h 1) Figure 4 Brightness temperature calculations at 92 GHz (horizontal polarization, 451 view angle) for various thicknesses of the ice layer as a function of rain rate (lower abscissa). The two upper abscissae give the mean particle radius and the particle density corresponding to the rain rate through Marshall–Palmer size distribution. (Adapted with permission from Wilheit (1986).) accuracy generally increases with the increase of averaging area and time. Simple Regression The notion that colder cloud tops usually correspond to heavier rainfall leads to the most straightforward approach: simply regressing rainfall rate against infrared radiation. This type of approach has mostly been done using broadband outgoing long-wave radiation, instead of narrowband brightness temperature within the atmospheric window (8–12 mm). Avariety of equations relating rainfall rate to outgoing long-wave radiation have been derived. Nonlinear functions are commonly used to account for the nonlinearity of the relation between the two parameters. Area Time Integral Techniques Studies of surface radar and rain gauge observations have shown that the total volume of rain falling over a sufficiently large area and for a long enough time period can be well predicted by the so-called area time R0 ¼3 mm h and TB0 ¼235 K are determined by comparing satellite measurements with ground radar observations over the tropical Atlantic Ocean. In essence, eqn [5] states that only clouds with top temperature colder than 235 K produce rain and their average rainfall rate is 3 mm h  1. This algorithm works reasonably well within the tropical belt of 301 S to 301 N for monthly rain total. Error increases dramatically toward high latitudes, particularly during cold seasons. A number of other techniques have been similarly developed, but take into account rain types and storm development stages. It is well established that deep convection more often produces heavy rainfall than stratiform clouds. The most common technique for determining cloud types relies on horizontal texture information of satellite images, such as identifying a local minimum in infrared brightness temperature imagery as the convective center and the surrounding relatively smooth portion as stratiform. If we divide observations over an area into several types, the rain total may be expressed by eqn [6], where Ri is the average rainfall rate for type i, which covers an area fraction of Ai and a time duration of Dti . X Ri Ai Dti ½6 R¼ i This method is known as the ‘cloud indexing’ technique. Observations also show that for thunderstorms, rainfall rate peaks while the cloud area is growing rapidly, and rainfall is much reduced at the time of maximum cloud area. In techniques that includes cloud life history, a different average rainfall rate Ri will be assigned for different development stages. This method is known as the ‘life-history’ technique. Bispectral Techniques Infrared and visible measurements both have important deficiencies in detecting rainfall. For example, stratus clouds are highly reflective but do not rain as SATELLITE REMOTE SENSING / Precipitation 1977 much, or as often, as cumulonimbus clouds. On the other hand, cirrus cloud tops are cold but do not produce rainfall. Bispectral techniques seek to combine information from visible and infrared measurements to obtain the optimal rainfall retrieval. In one such method, two lookup tables are first generated using coincident satellite and ground truth data. One of the tables is the probability of rain, pi;j , determined by the number ratio of raining cases to all cases in the infrared brightness temperature bin i and visible reflectivity bin j. Another table gives the mean rainfall rate, ri;j , derived only from raining cases in the same two-dimensional bin. The variation of rain probability and mean rainfall rate in the reflectivity–brightness temperature space is shown schematically in Figure 5. For a given pixel whose brightness temperature falls in the ith bin and reflectivity falls in the jth bin, rainfall rate may be determined by eqn [7]. R ¼ pi;j ri;j ½7 Most bispectral methods attempt to retrieve instantaneous rainfall rate by constantly updating the lookup tables using radar, rain gauge, or even satellite microwave observations as truth. At least in theory, the bispectral methods should be superior to infraredonly or visible-only methods. However, this superiority has not been convincingly demonstrated, partially because visible data are only available for a fraction of the day, and partially because many other uncertainties still remain, such as the quality of truth data and the co-location of satellite and surface data. r+2dr r +dr r Probability of rain Visible reflectivity Mean rainfall rate p+2dp p+dp p Infrared brightness temperature Figure 5 Schematic illustration of probability of rain ðpÞ and mean rainfall rate ðr Þ in the two-dimensional diagram of visible reflectivity and infrared brightness temperature. dp and dr are positive increments. Sensing by Microwave Measurements Owing to its physical directness, microwave sensing of precipitation has drawn particular attention since the late 1970s. Except for a few pure regression-type algorithms, a characteristic of the microwave methods is that they rely on radiative transfer models either at the algorithm development stage or during the retrieval computation. Through a radiative transfer model, microwave brightness temperatures are directly connected to the amount and distribution of precipitating hydrometeors. The microwave algorithms may be grouped into the categories of emission-based, scattering-based, combined emission and scattering, and radiative transfer model-based profiling techniques. Emission-based Techniques The emission signature provides the most direct physical relation between rainfall and brightness temperature. Data from frequencies under 20 GHz are primarily used for this type of algorithm, although higher frequencies are sometimes included to minimize atmospheric water vapor and/or surface effects. The relation between brightness temperature and rainfall rate may be derived from radiative transfer model calculations by specifying atmospheric temperature and humidity profiles, cloud liquid water content, rain-layer thickness, and size distribution of raindrops. The most commonly used raindrop size distribution is the so-called Marshall–Palmer distribution, in which the number concentration decreases exponentially with drop size. Rain-layer depth may be assumed to be the freezing level height, although the validity of this assumption deserves further investigation. There are several problems associated with emission-based algorithms. (1) They may only be applied over ocean; the high land surface emissivity prevents emission signatures from being detected by a low-frequency microwave radiometer. (2) Brightness temperature saturates for heavy rain. This problem is particularly serious for tropical regions where the rainlayer is deep. (3) Nonuniform rain rate across the beam causes underestimation of rainfall rate. As shown in Figure 3, the brightness temperature versus rainfall rate relation is highly nonlinear. The spatial resolution of a satellite pixel for microwave radiometers is on the order of several tens of kilometers; the rain field within one satellite pixel is generally inhomogeneous. If R ¼ RðTB Þ is the theoretical relation between brightness temperature TB and rainfall rate R for a homogeneous rain field, the retrieval resulting from the field-of-view-averaged brightness tempera B Þ, does not equal the field-of-view-averaged ture, RðT rainfall rate, RðTB Þ. Instead, in the case of microwave 1978 SATELLITE REMOTE SENSING / Precipitation 2.5 300 Scattering-based Techniques The scattering signature is physically less directly related to precipitation than the emission signature because it is an indication of the ice amount above freezing level. Frequencies higher than 80 GHz are primarily used for scattering-based algorithms. Algorithms have been developed either based on statistically regressing brightness temperatures to surface rainfall measurements or using results of radiative transfer models. The advantage of scattering-based algorithms is that they can be applied over both ocean and land. However, this type of algorithm has greater error for regions where warm rain has a significant contribution. 2.0 250 1.5 200 1.0 19 GHz 150 0.5 MWI 0 100 Techniques Using Both Emission and Scattering 0.1 For rain associated with a shallow rain-layer, scattering-based techniques fail to work because of lack of ice scattering. For heavy rainfall with deep rain-layers, emission-based techniques cannot correctly determine rainfall rate because brightness temperature saturates. A better solution is to take advantage of both emission and scattering signatures by combining them in a single algorithm. One such algorithm uses a ‘microwave index’ (MWI) defined for the Special Sensor Microwave/Imager according to eqn [8]. MWI ¼ ð1  D=D0 Þ þ 2ð1  PCT=PCT0 Þ Brightness temperature (K) 85 GHz ½8 Here D ¼ TB19V  TB19H , is the depolarization at 19 GHz, and PCT ¼ 1:818TB85V  0:818TB85H , is the polarization-corrected brightness temperature at 85 GHz. In brightness temperature TBnp , the subscript n depicts frequency and the subscript p depicts polarization (V for vertical and H for horizontal). D0 and PCT0 are the rain threshold values for D and PCT. The first term in eqn [8] is the emission signature and the second term is the scattering signature. Because D decreases monotonically with the increase of rainfall rate, it represents the emission signature better than 19 GHz brightness temperatures themselves. Figure 6 depicts the 19 GHz and 85 GHz brightness temperatures and the microwave index for a viewing angle of 531. The results are calculated from a radiative transfer model assuming a typical tropical profile of hydrometeors for deep convection agreed. The microwave index relates to rainfall rate monotonically without saturation. An alternative way to combine the two signatures is to use the emission signature until brightness temperature at low frequency saturates, then to use the scattering signature at higher rainfall rates. Microwave index, MWI emission, it is always true that RðT B ÞoRðTB Þ, i.e., the technique underestimates rain rate. 1 10 50 _ Rainfall rate (mm h 1) Figure 6 Radiative transfer model calculated brightness temperatures at 19 and 85 GHz, and microwave index (MWI) over ocean as a function of rainfall rate for tropical convective rains. A viewing angle of 531 is assumed and the brightness temperatures shown are for horizontal polarization. Radiative Transfer Model-based Techniques If the surface emissivity and vertical distributions of atmospheric temperature, humidity, and hydrometeors are known, brightness temperatures for any given set of frequencies can be calculated with a radiative transfer model. Inversely, if brightness temperatures observed at several frequencies match well with those calculated by a radiative transfer model, it will be very likely that the profiles assumed in the model are the same as those in the actual rain clouds. Radiative transfer model-based techniques use this logic, and generally consist of the following retrieval procedures. First, a large database of vertical profiles of hydrometeors must be prepared. This database should include all possible profiles that occur in nature. Because of the lack of observational data, this database is usually constructed with simulated results from numerical cloud models. Radiative transfer model calculations are performed using the hydrometeor profiles in the database, which result in many sets of calculated brightness temperatures. The set that best matches the satellite-observed brightness temperatures is selected, and the hydrometeor profile used to produce the best match is determined to be the retrieval. The retrieval gives not only rainfall rate at the surface but also its vertical distribution. Model-based techniques have the advantage of fully using physical relations between cloud microphysics and microwave radiation. With SATELLITE REMOTE SENSING / Surface Wind more observational data becoming available in the future to build the database of hydrometeor profiles, this approach is expected to play a more significant role in satellite remote sensing of precipitation. There are two major problems with this technique. First, the retrieval depends heavily on the preconstructed database, which, at present, relies on numerical cloud models because observational data are insufficient. Any cloud model deficiency could directly affect the quality of the rainfall retrieval. The second problem arises from the ill-posed problem in finding the best match between the observed and the calculated brightness temperatures. The number of unknowns in the retrieval problem (all components that interacts with microwave radiation) is far greater than the information content (number of independent information in brightness temperatures). Several totally different hydrometeor profiles may result in a similar ‘good’ match, causing nonuniqueness for the solution. This is usually dealt with by averaging the hydrometeor profiles of the closest brightness temperature matches. Combination of Multichannel and Multiplatform Observations Currently, it is practical to put microwave radiometers only on low-altitude, polar-orbiting satellites to ensure useful spatial resolution. The frequency of observation by a single satellite of a certain area on the Earth is unacceptably low (1 to 2 times a day) for determining rainfall accumulation. As a result, although microwave techniques work better for instantaneous rainfall rate, they do not outperform visible/infrared techniques on daily or monthly time scales, because visible/infrared measurements are more frequent. Therefore, combining measurements from multiple wave bands and multiple platforms has been proposed. One proposed approach is to increase the number of satellites that carry microwave sensors, so that local sampling frequency will be increased to an acceptable level. With increasing international collabo- 1979 ration, this proposal is expected to become reality in the near future. The current solution has been to combine visible, infrared, and microwave measurements from available satellites. Visible/infrared measurements have the advantage of ample coverage, while microwave measurements have the advantage of physical directness. The combined techniques use microwave retrievals as truth to constantly train visible/infrared algorithms, while the trained visible/ infrared algorithms are used to fill the gap left by microwave measurements. See also Optics, Atmospheric: Optical Remote Sensing Instruments. Radar: Incoherent Scatter Radar; Precipitation Radar. Further Reading Arkin PA and Ardanuy PE (1989) Estimating climatic-scale precipitation from space: a review. Journal of Climate 2: 1229–1238. Barrett EC and Martin DW (1981) The Use of Satellite Data in Rainfall Monitoring. London: Academic Press. Grody NC (1993) Remote sensing of the atmosphere from satellites using microwave radiometry. In: Janssen MA (ed.) Atmospheric Remote Sensing by Microwave Radiometry, pp. 259–314. New York: Wiley. Kidder SQ and Vonder Haar TH (1995) Satellite Meteorology. London: Academic Press. Smith EA, Kummerow C and Mugnai A (1994) The emergence of inversion-type precipitation profile algorithms for estimation of precipitation from satellite microwave measurements. Remote Sensing Review 11: 211–242. Stephens GL (1994) Remote Sensing of the Lower Atmosphere: An Introduction. New York: Oxford University Press. Wilheit TT (1986) Some comments on passive microwave measurement of rain. Bulletin of the American Meteorological Society 67: 1226–1232. Surface Wind W T Liu, California Institute of Technology, Pasadena, CA, USA Copyright 2003 Elsevier Science Ltd. All Rights Reserved. Introduction Just a few decades ago, almost all ocean wind measurements came from merchant ships. The quality and geographical distribution of such reports are however uneven. Today, many citizens believe that operational numerical weather prediction (NWP) will give us all the wind information we need, until a hurricane suddenly intensifies and changes course, or the unexpected delay of a monsoon brings drought, or the Pacific trade wind collapses along the Equator before an El Niño. When prediction fails and disaster hits, then we remember that NWP depends on models 1980 SATELLITE REMOTE SENSING / Surface Wind that are limited by our knowledge of the physical processes and the availability of data. Sailors understand both the importance and the difficulty in getting information on wind over oceans. Textbooks still describe global ocean wind distribution in sailor’s terms: the calms of the doldrums and horse latitudes, the steady trade winds, and the ferocity of the roaring forties; these features are clearly visible in Figure 1, which is derived from one day of observations by a space-based scatterometer, QuikSCAT. Wind is a vector quantity. Spaceborne microwave scatterometers are the only proven instruments that will give us measurements of both wind speed and direction over ocean, under clear and cloudy conditions, day and night. Scatterometers provide not only a near-synoptic global view, but also provide details not possible using NWP models. Typhoon Paul is observed south of Japan, tropical depression Rachael is forming south-east of Taiwan, and tropical depression Eugene is visible as closed circulation in the eastern Pacific off Central America – such coverage and resolution are crucial to understanding and predicting the changes of weather and climate. The principles of scatterometry and scatterometer missions are summarized in the next two sections. After that, examples of the scientific impact of the space-based scatterometer are given under ‘Major Applications’. The primary functions of radar altimeter, synthetic aperture radar, and microwave radiometer are not wind measurement, but they provide wind speed as a secondary product. Wind speeds are important in their own right, and wind speed from these sensors can be applied with directional information derived from other means. The measuring capability of these sensors are described under ‘Wind Speed Measurements’. A brief discussion of future missions and technology then concludes the article. Figure 1 Over the ocean, white streamlines indicating wind direction are superimposed on the color image of wind speed at 00Z for 6 August 1999, derived from objective interpolation. Typical average backscatter coefficients over land and Antarctica are also added. The data are all based on observations by the scatterometer QuikSCAT. SATELLITE REMOTE SENSING / Surface Wind Principles of Scatterometry During the Second World War, marine radar operators observed disturbances on their radar screens that obscured small boats and low-flying aircraft. They termed this noise ‘sea clutter’; it was the backscatter of the radar pulses by small waves on the ocean’s surface. Not until some decades later did this backscatter come to have important applications. The scatterometer sends microwave pulses to the earth’s surface and measures the power backscattered from the surface roughness. Over land, the roughness may describe the characteristics of polar ice or vegetation. Over the ocean, which covers over threequarters of the earth’s surface, the backscatter is due largely to the small waves a few centimeters high on the surface. The idea of the remote sensing of ocean surface winds was based on the belief that these surface ripples are in equilibrium with the local wind stress. At incident angles greater than 201, the radar return is governed by Bragg scattering, and the backscatter increases with wind speed. The backscatter is governed by the in-phase reflections from surface waves. For a smooth surface, the radar receives no return when viewing is at an angle. But as the surface roughness increases, backscatter occurs as scattering from periodic structures in the surface roughness constructively interferes. The backscatter depends not only on the magnitude of the wind stress but also on the wind direction relative to the direction of the radar beam (the azimuth angle). The capability of measuring both wind speed and direction is the major unique characteristic of the scatterometer. Because the backscatter is symmetric about the mean wind direction, observations at many azimuth angles are needed to resolve the directional ambiguity. A scatterometer that measures only at two orthogonal azimuth angles, such as Seasat (see next section), will always include wind solutions of nearly equal magnitude and 1801 apart. Because of uncertainties in the wind retrieval algorithm and the noise in the backscatter measurements, the problem with directional ambiguity was not entirely eliminated with additional azimuthal looks in the scatterometers launched after Seasat. A median filter iteration technique initialized by the wind direction solution closest to the NWP wind field has been commonly used to remove the directional ambiguity. There is a long history of theoretical studies of the relationship between wind and backscatter, based on laboratory data. However, these theoretical or dynamic-based relationships were not sufficient for operational wind retrieval in open oceans. The geophysical model function, from which ocean surface 1981 wind vectors are retrieved from the observed backscatter, is based largely on empirical fits of data. Because the capillary waves, which determine backscatter, are governed by stress, an approach has been developed whereby the backscatter observations are related directly to measurements of surface stress. The definition of the geophysical data product of scatterometer as the equivalent neutral wind is based on the same reasoning. The backscatter has also been related to the pressure gradient or to geostrophic winds, which may be more coherent over the scatterometer footprint than surface winds. While wind is the primary factor in the changes of backscatter measured by a scatterometer, other secondary factors – such as sea surface temperature (SST), rain, surface film, atmospheric stability, sea state, and surface currentFmay also affect scatterometer measurement, and may cause errors in wind retrieval. With the increasing accuracy of scatterometer wind measurement, understanding and quantifying such effects are becoming increasingly important and have become scientific fields in their own right. Scatterometer Missions Historically, scatterometers of the European Space Agency (ESA) used the C-band (5 GHz), but the National Aeronautics and Space Administration (NASA) prefers the Ku-band (14 GHz). A higher frequency is more sensitive to shorter surface waves. The Ku-band is more sensitive to wind variation at low winds, but is more subjective to atmospheric effects and rain contamination. Five scatterometers have been launched on polar orbiting satellites, and their major characteristics are summarized in Figure 2. NASA launched a scatterometer on the Seasat Mission in June 1978. Four fan-beam, dual-polarized antennas, oriented at 451 and 1351 to the spacecraft subtrack, illuminated two 500 km swaths, one on each side of the spacecraft, providing wind vectors at 50 km resolution. However, only one side was in operation most of the time, covering less than 40% of the global ocean daily. The incident angle varied from 251 to 551. The accuracy of the backscatter was about 0.7 db. The two orthogonal azimuth angles were not able to resolve the wind direction unambiguously. Seasat failed in October 1978. A scatterometer was launched by ESA on the first European Remote Sensing (ERS-1) satellite in August, 1991, and it was followed by an identical instrument on the ERS-2, launched in April 1995 and put into operation in 1996. The ERS scatterometers scan a 500 km swath on one side of the satellite, and measure at three azimuth angles, 451, 901, and 1151, with 1982 SATELLITE REMOTE SENSING / Surface Wind Seasat ERS−1/2 NSCAT QuikSCAT 14.6 GHz 5.3GHz 13.995 GHz 13.402 GHz Polarization V-H, V-H V ONLY V, V-H, V V, H Inc. angle 22°−55° 18°−47°, 24°−57° 18°− 57°, 22°−63° 46°, 54° RANGE GATE Variable Doppler Frequency Scan pattern Beam resolution Resolution Fixed Doppler Spot 50 km 25 km 25 km 500 km 500km 500 km 600 km 600km 1800 km Variable 41% 77% 93% 6/78–10/78 8/91−1/01 8/96– 6/97 6/99 + 50 km Swath Daily coverage Dates Figure 2 Characteristics of spaceborne scatterometers. vertical polarization only. They provided winds over only 41% of the global ocean daily. The incident angle varies from 241 to 571 for the fore and aft beams and from 181 to 471 for the mid beam. The backscatters have 50 km spatial resolution but are sampled at 25 km. The NASA Scatterometer (NSCAT) was launched in August 1996 on the first Japanese Advanced Earth Observing Satellite (ADEOS), later renamed Midori. The six fan-beam antennas provide 600 km swaths on both sides of the spacecraft, covering 77% of the global ocean at 25 km resolution daily. The accuracy of backscatter is 0.2 db. The antennas made observations at 451, 1151, and 1351 azimuth angles. The fore and aft beams measure only at vertical polarization, with incident angle varies from 221 to 631, while the midbeam measures at both vertical and horizontal polarization, with incident angle varying from 181 to 511. The unexpected destruction of the solar array caused the early demise of NSCAT, after it had returned 9 months of data. NASA launched QuikSCAT, a Ku-band scatterometer with a new design, in 1999. It uses pencil-beam antennas in a conical scan and has a continuous 1800 km swath that covers 93% of the global ocean in a single day. The standard wind product has 25 km spatial resolution, but special products with 12.5 km resolution have been produced for selected regions. It measures horizontally and vertically polarized backscatter at incident angles of 461 and 541 respectively. Major Applications One of the basic applications of scatterometer wind measurement is in predicting weather. Although the ERS-1 scatterometer was launched in 1991, the data were not assimilated operationally into NWP until 1994. All major weather forecast centers in Europe, Japan, and the USA implemented the assimilation of ERS scatterometer winds between 1994 and 1997. NSCAT had only a short life span; the spacecraft failed before any NWP center could set up the system to assimilate its data. A recent comprehensive impact study of NSCAT revealed an approximately 1-day extension of useful forecast skill in the Southern Hemisphere. The impact of assimilation of NSCAT data to regional weather forecast has also been demonstrated. The European Center for Medium Range Weather Forecasting (ECMWF) in the United Kingdom and the National Center of Environmental Prediction in the United States began operational assimilation of the QuikSCAT data in January 2002. ECMWF reported a robust improvement in its forecasts of atmospheric conditions over the Southern Hemisphere and in the upper atmosphere after SATELLITE REMOTE SENSING / Surface Wind assimilating these data. Its ability to forecast the tracks of tropical cyclones is also enhanced. Besides the potential use in four-dimensional assimilation by operational NWP, scatterometer data have been widely used by marine weather and hurricane centers in analyzing and predicting marine storms. For most of the Atlantic hurricanes in 1999, closed circulations with intensity meeting the criteria of a tropical depression were observed by QuikSCAT up to a few days before their identifications by the National Hurricane Center. QuikSCAT data were used to track the surface vortex of hurricane Floyd all the way back to the African coast five days before it was identified as a tropical depression east of the West Indies. Such vortices in their early stages are too small to be resolved by operational NWP products, and their convection not strong enough to produce organized cloud signals. Hence the scatterometer, with its high spatial resolution, is the best means to study these early vortices, their tracks across the Atlantic, and their evolution into full-blown hurricanes. Oceanographers, who were in great need of information on wind forcing of ocean circulation, were the first group to support space-based scatterometer missions. One of the applications is to use scatterometer winds to force ocean general circulation models. Many studies show that winds from scatterometers are superior in forcing more realistic oceanic responses in the models than NWP winds. Since scatterometer winds have become continuously available, they have been used in studies of seasonal phenomena like the monsoons and interannual signals like El Niño. A monsoon is the seasonal change of wind forced by the temperature contrast between the continent and the ocean. Scatterometer winds have been used to study oceanic responses to the changes of monsoons in the South China Sea, the Arabian Sea, and the Atlantic Ocean. They have been used to study the influence of moisture advection on continental precipitation in China, Africa, and South America. El Niño Southern Oscillation (ENSO), the strongest interannual climatic signal, is believed to be associated with the collapse of the Pacific trade winds near the Equator. Scatterometers have revealed, with unprecedented resolution, the evolution of the tropical wind systems associated with ENSO. Model initialization with scatterometer winds have been shown to improve El Niño forecasts. Scatterometer winds have been used to link the ocean warming in the equatorial Pacific during an El Niño to the intraseasonal wind surge in the Western Pacific and the modification of decadal phenomena in the North Pacific. The high resolution of scatterometer data allows studies of small coastal jets and eddies and derivative 1983 parameters, such as atmospheric convergence. Scatterometer winds were used to study ocean response to the wind jets coming out of the mountain gaps near Vladivostok and in Central America. For the first time, the cyclonic circulation of the small Catalina Eddy, which brings the ocean-cooling effect to Los Angeles, was visualized by scatterometer winds. A convergence zone south of the Equator, running east from Brazil, is also revealed unambiguously for the first time with scatterometer data. Besides the strong tropical convergence zones driven by deep convection, scatterometers also help to identify weak convergence zones caused by different mechanisms. The broad coverage reveals new phenomena in data-poor tropical and southern oceans. By combining observations of QuikSCAT and Tropical Rain Measuring Mission (TRMM) a narrow break in the westward Trade winds and North Equatorial Currents system was found, stretching over 2000 miles from the Hawaii Islands to the Western Pacific. This break consists of eastward current, warm water, atmospheric convergence, and positive curl of wind stress; the system was revealed as a whole for the first time. The system is postulated to be triggered by the Hawaii Islands but sustained by positive ocean–atmosphere feedback. The use of QuikSCAT and TRMM enabled the study of the coherent and in-phase propagation of sea surface temperature and wind vectors in the tropical instability waves in the eastern equatorial Pacific. Measurements from merchant ships and weather stations are extremely sparse in the hostile environment around Antarctica, where strong winds circulate around the globe over open oceans. Scatterometer data have been used to study wind forcing of the circumpolar current. Scatterometers are also capable of monitoring both the Antarctic sea ice extent (SIE) and the wind field over adjacent oceans at the same time, making it possible to characterize the joint variabilities of both wind and ice. Scatterometers observe a wavenumber-3 pattern in the wind, which coincides with three SIE maxima. The wind and ice patterns move eastward together during the winter season. The SIE maxima also provide favorable conditions for storm generation over the ocean, which has interannual variabilites linked to ENSO. Wind shear facilitates the turbulent transfer of heat, moisture, and greenhouse gases between the ocean and the atmosphere. The transport is parameterized mostly in terms of wind speed, but there are suggestions that, in additional to wind speed, the backscatter measured by the scatterometers contains information on secondary factors (e.g., small-scale wave fields) affecting ocean–atmosphere gas transfer. The unique contribution of the scatterometer in ocean–atmosphere exchanges is likely to be in estimating the 1984 SATELLITE REMOTE SENSING / Surface Wind transport terms in the conservation equation, whether it is the curl of wind stress in oceanic biological pumping or the atmospheric moisture advection in the atmospheric hydrologic balance. Wind Speed Measurements Both the microwave altimeter and the synthetic aperture radar (SAR) are similar to the scatterometer, in the sense that all three are active sensors that send microwave pulses to the Earth’s surface and measure the backscattered power. The altimeters are designed to measure the dynamic topography of the ocean. While the scatterometer views at oblique angles, the altimeters view at nadir (very small incident angles). At nadir, the backscattered energy is a result of specular reflection (the wavelets serve as small mirrors), and the backscatter is not sensitive to wind direction. Because the instrument is not scanning, data are available only at a very narrow repeated ground tracks. The coverages of all the past altimeters are poor compared with the scatterometer and the microwave radiometers. Altimeters were flown on the Seasat and ERS spacecraft described above. Geosat, which was in operation between 1985 and 1989, and Topex-Poseidon, launched in 1992, are two missions dedicated to the altimeter. The same model function used to retrieve winds from scatterometer can be used for SAR. However, an SAR looks perpendicular to aircraft path only at one azimuth angle, and cannot resolve wind direction. The main objective of SAR is to provide high-resolution imaging of the Earth’s surface. SAR has spatial resolutions much better than those of scatterometers, but the high resolution also introduces higher uncertainties in accuracy caused by secondary effects that affect surface roughness. The instrument and the data processing procedure are much more complicated than with the scatterometer and there have been severe calibration problems. Both the SAR on Seasat and ERS have spatial resolution of 30 m and a swath width of 100 km. The narrow swath width and the sporadic operation prevent global monitoring of ocean surface wind. Radarsat-1, launched in 1995, can operate in the scanning mode with a spatial resolution of 100 m and a 500 km wide swath; this instrument is the closest to providing continuous global coverage. Ocean surface wind speed can also be derived from the radiance observed by microwave radiometer. It is generally believed that wind speed affects the measured radiance indirectly through the generation of ocean waves and foam and the change of the surface emissivity. Radiometers designed to observe the ocean surface operate primarily at window frequencies, where atmospheric absorption is low. To correct for the slight interference by tropospheric water vapor, clouds, and rainfall, and, to some extent, the effect of sea surface temperature, radiances at frequencies sensitive to sea surface temperature, atmospheric water vapor, and liquid water are also measured. Microwave radiometry has a much longer history than the active microwave sensors. Ocean surface wind speeds were derived from the Scanning Multichannel Microwave Radiometer (SMMR) on Seasat and Nimbus-7, which were launched in 1978. A major improvement in wind speed availability was made by the Special Sensor Microwave/Imager (SSM/I), the first of which was launched in 1987 on the spacecraft Defense Meteorological Satellite Program (DMSP). Several DMSP satellites with SSM/I on board have been in orbit at the same time, providing continuous, global coverage since July 1987. Future Missions and Technology Quikscat will be followed by an identical scatterometer on ADEOS-2 scheduled to be launched in November 2002. If there is sufficient overlap between the operations of the two identical scatterometers, the importance of high frequency and high-wave-number wind forcing on the ocean can be demonstrated. ESA is scheduled to launch a series of C-band dual-swath advanced scatterometers (ASCAT) on its operational platform METOP, starting in December 2005. NASA is planning to launch a scatterometer on the Japanese Global Change Observation Mission (GCOM), so that two wide-swath scatterometers will provide continuous time series of high-frequency wind forcing. Conventional microwave radiometer measures surface radiances at horizonal and vertical polarizations, which independently do not give wind direction. Preliminary studies indicate that measurement of the coherence between vertically and horizontally polarized radiances will provide directional information on surface winds. The Naval Research Laboratory is scheduled to launch the Windsat mission to test the capability of a polarimetric microwave radiometer in measuring ocean surface wind vectors in 2003. One of the drawbacks to scatterometry is the wind direction ambiguity. The backscatter is a cosine function of the azimuth angle. In a recent experiment, it was demonstrated that the correlation between copolarized and cross-polarized backscatter is a sine function of azimuth angle. By adding a receiver of cross-polarized backscatter to the scatterometers (like QuikSCAT), the directional ambiguity problem can be mitigated. A polarized scatterometer has been proposed for GCOM. SATELLITE REMOTE SENSING / Temperature Soundings 1985 Acknowledgment Further Reading This article was prepared at the Jet Propulsion Laboratory, California Institute of Technology, under contract with the National Aeronautics and Space Administration. Greernaert GL and Plant WJ (1990) Surface Waves and Fluxes, vol. II. Remote Sensing. Dordrecht: Kluwer. Ikeda M and Dobson FW (1995) Oceanographic Applications of Remote Sensing. Boca Raton, FL: CRC Press. Liu WT (2002) Progress in scatterometer application. Journal of Oceanography 58: 121–136. Simpson R, Garstang M and Anthes R (2002) Coping with Hurricanes. Washington, DC: American Geophysical Union. Siedler G, Church J and Gould J (2001) Ocean Circulaton and Climate. San Diego, CA: Academic Press. Stewart RH (1985) Methods of Satellite Oceanography. Berkeley, CA: University of California Press. See also Air–Sea Interaction: Surface Waves. Coastal Meteorology. Cyclones, Extra Tropical. El Niño and the Southern Oscillation: Observation. Land–Atmosphere Interactions: Overview. Monsoon: Overview. Ocean Circulation: Surface–Wind Driven Circulation. Sea Ice. Weather Prediction: Severe Weather Forecasting. Temperature Soundings Introduction Temperature plays a key role in radiative, dynamical, and chemical processes in the atmosphere. However, compared with most other parameters, atmospheric temperature has a relatively low variability: typically
20 K, or about
10% of the absolute value, at any altitude (Figure 1). This low variability imposes correspondingly tight constraints on the useful accuracy of any measurements. Nevertheless, remote sounding has now developed to the point where temperature can be retrieved with accuracies of 2 K or better, comparable with the quality of measurements made in situ with radiosondes. The main impetus for this development has come from the meteorological community: although radiosondes provide good coverage over populated land areas, accurate weather forecasting requires global temperature fields, which can only be obtained from satellites. ‘Operational’ temperature sounders have been flown on the NOAA series of polar orbiting satellites since 1972. These are nadir viewing instruments, measuring emission in the infrared and microwave regions of the spectrum. By selecting channels sensitive to emissions from different depths into the atmosphere, the vertical temperature structure can be determined. Nowadays, the more sophisticated forecast models assimilate directly the satellite radiance measurements themselves, bypassing the need for any explicit temperature profile retrieval. Temperature can also be retrieved using emission measurements from the atmospheric limb, i.e., viewing the atmosphere tangentially rather than vertically. Limb sounding allows temperature to be retrieved to higher altitudes and with improved vertical resolution compared with nadir sounding, but at the expense of reduced horizontal resolution and signal-to-noise 120 110 10− 4 Thermosphere 100 Mesopause 90 10− 2 80 70 Mesosphere 60 Stratopause 100 50 Approx. altitude (km) Copyright 2003 Elsevier Science Ltd. All Rights Reserved. Pressure (hPa) A Dudhia, Oxford University, Oxford, UK 40 Stratosphere Tropopause 102 180 20 10 Troposphere 160 30 200 220 240 260 280 0 300 Temperature (K) Figure 1 Typical atmospheric temperature profiles. Solid line: US standard atmosphere (midlatitudes); dotted line: an equatorial profile; dashed line: a polar winter profile. 1986 SATELLITE REMOTE SENSING / Temperature Soundings along the path, conveniently expressed in terms of an optical thickness w: t ¼ exp ðwÞ w¼ Atmospheric temperature and density affect electromagnetic radiation through absorption, emission, refraction and scattering. All of these mechanisms can be exploited in order to retrieve temperature using remote sensing techniques. 0 1 dt B ds þ I1 t1 ds ½1 where B is the Planck function, t the transmittance from space to a point at distance s along the path, t1 represents the attenuation of any emission source I1 beyond the atmosphere. The atmospheric contribution to this radiance is therefore a spatially weighted average of the Planck function along the path, dt=ds being the ‘weighting function’. Since B is a known function of temperature and wavelength, determining BðTðsÞÞ from the above relationship is equivalent to retrieving the temperature profile. However, to use eqn [1] it is also necessary to know the transmittance tðsÞ ½3 The Planck function (Figure 2) is given by: 2hc2 l5 ðexp ðhc=lkTÞ  1Þ ½4 where h is Planck’s constant, c the speed of light, k Boltzmann’s constant, l the wavelength, and T the temperature. This has a maximum at lmax ¼ 2:9103 =T m (Wien’s displacement law), which for Wavenumber (cm 1) 10 1000 100 10 000 100 270 K Radiance (W m 2 sr  1 (cm 1) 1) I¼ Z vsr ds 0 Planck Function Thermal Radiation Most techniques for temperature sounding rely on measurements of thermally emitted radiation in either the infrared or the microwave region of the spectrum. The monochromatic radiance, I, from a line of sight through a non-scattering atmosphere in local thermodynamic equilibrium can be represented by s where v is the absorber volume mixing ratio, r the (molar) air density, and s the absorption coefficient. At thermal wavelengths, s is a function of the concentrations of various absorbing species, pressure and temperature. Composition v can be eliminated as an unknown by selecting spectral regions where the absorption is primarily from a well-mixed species, usually the 15 mm or 4:3 mm CO2 bands in the infrared, or the 60 GHz O2 band in the microwave. Pressure is either implicit in the retrieval coordinates (nadir sounding) or retrieved simultaneously with temperature (limb sounding). B¼ Physical Mechanisms Z ½2 1 100 210 K Solar 10 5 10 10 10 n 10 15 10 20 10 6 10 5 10 4 10 3 Temperature exponent n ratio. A third possibility for temperature sounding is solar occultation: viewing the Sun as it rises or sets beyond the atmospheric limb and determining temperature through its effect on the atmospheric absorption. Such measurements all rely on modeling molecular absorption spectra, which largely determine radiative transfer at infrared and microwave wavelengths. However, since temperature and density are linked via the hydrostatic equation, the temperature profile may also be inferred from measurements which are more directly related to the atmospheric density profile, such as scattering in visible and UV wavelengths, or refraction at radio frequencies. Molecular (Rayleigh) scattering is routinely used to determine atmospheric temperature from ground-based lidars, and atmospheric composition from space, but has had only limited application in temperature sounding from space. Radio occultation techniques for determining density via refraction were originally developed for sounding the atmospheres of other planets, but the introduction of the GPS network of navigational satellites offers a new possibility for routine temperature sounding of the Earth’s atmosphere. 1 10 2 Wavelength (m) Figure 2 Solid lines: Planck function for 210 and 270 K, representing the typical range of atmospheric temperatures. Dotted curve: Planck function for 5800 K, scaled to give an integrated irradiance of 1370 W m  2 (solar constant) representing the maximum diffuse solar contribution. Dashed curve (right axis): sensitivity n fitting B  T n , evaluated at T ¼ 240 K. SATELLITE REMOTE SENSING / Temperature Soundings 1987 Molecular Absorption The molecules of many atmospheric species exhibit absorption bands at infrared wavelengths corresponding to transitions between quantized vibrational energy levels (molecules of nitrogen and oxygen being notable exceptions, having no permanent dipole moment). Superimposed on these vibrational states are rotational states which have a finer quantization. Changes in vibrational level are often accompanied by changes in the rotational quantum number J, giving the typical band structure shown in Figure 3. The central peak (Q-branch) corresponds to a pure vibrational transition ðDJ ¼ 0Þ, and the envelope of lines at lower (P-branch) and higher (R-branch) wavenumbers correspond to DJ ¼
1, the spread in these envelopes reflecting the rotational energy dependence ER  JðJ þ 1Þ. Pure rotational transitions also occur, leading to absorption features in the microwave spectrum (Figure 4). The overall strength of a particular absorption band is determined by the population of the two vibrational levels involved. In local thermodynamic equilibrium, populations of each vibrational energy level EV follow the Boltzmann distribution  expðEV =kTÞ.
(m2 mole1) 8000 6000 4000 2000 0 620 640 660 680 700 720 700 720 Wavenumber (cm1) (A) '
(m2 mole1) 150 100 50 0 50 100 150 620 640 680 660 Wavenumber (cm1) (B) Figure 3 (A) The absorption coefficient s of the CO2 15 mm fundamental n2 vibration band at 10 hPa total pressure, 240 K, and (B) the change Ds resulting from a 10 K increase in temperature. (The atmospheric CO2 column amount is around 120 moles m  2.) For T ¼ 240 K; kT  EV ð167 cm1, compared with 1000 cm1 ), so usually only ‘fundamental’ transitions between the ground state and the first excited level are significant. The shape of an absorption band, given by the envelope encompassing the P and R branches, depends on the population distribution over the rotational energy levels ER . This also follows Boltzmann statistics, but here kT  ER . The additional degeneracy factor ðJ þ 1Þ, representing the increased number of states available at higher rotational quantum 1.0 0.0020 1.5 Wavenumber (cm1) 2.0 2.5 3.0 70 80 0.0015
(m2 mole1) typical atmospheric temperatures lies between 10–15 mm. The curve falls off as 1=l5 at longer wavelengths, and as exp ðhc=lkTÞ at shorter wavelengths. The rapid decay at short wavelengths means that thermal emission is negligible below a few micrometers and, in any case, below 4 mm scattered or reflected solar radiation becomes significant during daytime. The accuracy of a temperature retrieval depends not only on the radiance signal-to-noise ratio (SNR) but also on the sensitivity of radiance to temperature, dB=dT, which reduces with increasing wavelength. At 10 mm; B  T 6 but in the microwave region (1–10 mm wavelengths) B / T (in fact, ‘radiance’ and ‘temperature’ are often used interchangeably in referring to microwave measurements). For a given SNR, this means that temperature can be retrieved more accurately at shorter wavelengths. Conversely, this also means that at shorter wavelengths retrievals of other species are more sensitive to any given temperature error. For example, a number of constituents ðH2 O; CH4 ; N2 O; NO2 Þ are often retrieved using bands in the 6–8 mm region. At these wavelengths, a 1 K temperature overestimate results in a 4% increase in predicted radiance, and therefore a roughly equivalent underestimate in retrieved concentrations. Hence the need for accurate temperature retrievals for all remote sensing experiments based on infrared emission. 0.0010 0.0005 0.0000 40 50 60 Frequency (GHz) Figure 4 The absorption coefficient s of the O2 rotational band at 60 GHz, at 1000 hPa (smooth curve) 100 hPa (irregular curve). 1988 SATELLITE REMOTE SENSING / Temperature Soundings Hydrostatic Equation In most situations, the profiles of atmospheric pressure p, molar density r, and temperature T can be assumed to be in hydrostatic balance: dp gM ¼ dz p RT ½5 Refraction The speed of electromagnetic radiation is reduced in air owing to the polarizability of air molecules. The refractive index n is conveniently expressed in terms of a refractivity N ¼ n  1, which is usually assumed to be proportional to air density. For dry air at 15 C and standard pressure, Edlén’s dispersion relation models the wavelength dependence of refractivity from 0.2 to 2 mm (Figure 5): N106 ¼ 64:328 þ N106 ¼ 77:6 ½7 ðp  eÞ e þ 3:73105 2 T T ½8 where p; e are the total pressure and partial pressure of water vapor in hectopascals and T is the temperature in kelvins. At 15 C and standard pressure, this gives values of refractivity varying from N ¼ 2:7104 for dry air to N ¼ 3:4104 for saturated air (1.6% water vapor by volume). Refraction introduces a curvature into limb paths, lowering the tangent point and increasing the pathintegrated air mass compared with the straight-line path. For a circularly symmetric atmosphere it can be shown that the tangent height correction is dz ’ Na ½6 where g is the gravitational acceleration, M the molar weight of air, R the gas constant, and z the altitude. In addition to retrieving temperature, it is usually necessary to know the pressure in order to evaluate terms in eqn [3], although absolute altitude is not critical. If TðpÞ is retrieved directly (e.g., nadir sounding) then this is no problem and eqn [6] can be integrated to obtain layer thicknesses Dz. If pðzÞ is retrieved (e.g., microwave limb sounding) then eqn [6] can be used to obtain temperature. If TðzÞ is retrieved (e.g., infrared limb sounding) it is necessary also to retrieve pressure at least at one altitude, pðzref Þ, and obtain pðzÞ by integrating eqn [6]. If rðzÞ is retrieved (e.g., occultation) it may be adequate to assume some climatological pressure at high altitude and integrate eqn [5] downwards to obtain pðzÞ, hence TðzÞ, with 29 498:1 255:4 þ 146  m2 41  m2 where m is the wavelength expressed in mm. For wavelengths longer than 1 mm, refractivity is essentially independent of wavelength. However, at radio frequencies ðo20 GHzÞ the dipole moment of water vapor has a significant effect which can be modeled as: Refractivity u 106 dp ¼ gMr dz any error dp in the climatological assumption decreasing in significance further down the profile. ½9 320 10.0000 310 1.0000 300 0.1000 Scattering 290 280 270 0.0100 Refractivity 0.5 1.0 0.0010 1.5 Scattering extinction numbers, ensures that the most probable rotational quantum number is not J ¼ 0 but some higher number, and increases with increasing temperature (indicated by the outward displacement of the peaks of the P- and R-branches in Figure 3). As well as the line strength and the shape of bands, the temperature–pressure profile also affects the width of individual lines through Doppler broadening (high altitudes) and pressure broadening (low altitudes, Figure 4). The above assumes ‘local thermodynamic equilibrium’ (LTE), i.e., that the populations of the vibrational and rotational energy levels are characterized by the same temperature as the mean kinetic energy. In practice, this means that collisions between molecules occur sufficiently frequently to ensure that the internal energy levels are redistributed according to the local kinetic temperature. At high altitudes, other processes may dominate, leading to non-thermal population distributions of the vibrational states (so called ‘nonLTE’ effects). These often limit the upper altitude of practical infrared retrieval schemes. Non-LTE effects are usually negligible in the microwave region, owing to the small energy difference between rotational levels; instead, when sounding the mid-stratosphere or higher altitudes, complications are introduced by having to model the Zeeman splitting of lines in the Earth’s magnetic field. 0.0001 2.0 Wavelength (Pm) Figure 5 Variation of refractivity (left axis) and Rayleigh scattering cross-section (right axis) in the visible and near-infrared region of the spectrum. The scattering extinction is scaled to the number of molecules in a vertical column of atmosphere. SATELLITE REMOTE SENSING / Temperature Soundings 1989 where a is the radius of the Earth. For a tangent height of 25 km; dz ’ 70 m for infrared wavelengths, equivalent to a 1% increase in the tangent point pressure and a similar increase in integrated air mass. Since the effect is proportional to density ð/ NÞ it doubles for every 5 km decrease in tangent height. from space to pressure level p for a well-mixed absorber with volume mixing ratio v and constant absorption coefficient s is given by: Scattering where a ¼ vs=g is a constant. The weighting function is then given by sRa ¼ 32p3 N2 3NA r2 l4 ½10 where NA is Avogadro’s number. Since refractivity N / r, the scattering cross-section per mole of air depends only on wavelength, predominantly through the 1=l4 term (Figure 5). For optically thin paths, single-scattering can be assumed so that measurements of extinction or scattered radiation can be simply related to the air density. However, this assumption breaks down at higher pressures, when multiple scattering and/or Mie scattering (by particles of radius comparable to the wavelength, e.g., aerosols) have to be considered. Nadir Sounding The earliest satellite temperature sounders viewed downwards, measuring the radiance emerging from the top of the atmosphere in a range of spectral bands. The different transmission characteristics of each band can be used to derive information on temperature from different optical depths into the atmosphere. This is the basis of the operational temperature sounders used today. Weighting Functions Adapting eqn [1], the radiance I emerging from the top of the atmosphere above a non-reflective surface is given by: Z 1 dt dZ ½11 B I ¼ B0 t0 þ dZ 0 where Z ¼  ln ðp=p0 Þ is a height-like coordinate, and subscripts 0 indicate surface values. For nadir viewing, it is convenient to use a pressure-based coordinate such as Z, since the transmittance, and therefore weighting functions, are themselves mostly pressuredependent. Using the hydrostatic equation (eqn [5]) to adapt eqn [3] to pressure coordinates, the optical depth w Z 0 p vs dp ¼ ap g ½12 dt dt ¼ p ¼ ap expðapÞ dZ dp ½13 It can be shown that this has a maximum where the optical depth w ¼ ap ¼ 1, and a width at half maximum of DZ ’ 2:5 scale heights ð15 kmÞ. By suitable placement of filters within the band it is possible to select weighting functions peaking at different pressures (Figures 6 and 7). Vertical Resolution From eqn [13] and Figure 7, it can be seen that the width of nadir sounding weighting functions is comparable to the thickness of the entire troposphere. The weighting function width does not fundamentally limit the vertical resolution of the retrieval, but the large overlap means that in order to retrieve a profile at, say, seven levels corresponding to the weighting function peaks, most of the information will come from the difference between radiance measurements in adjacent channels rather than from the absolute values, hence reducing the effective SNR. 1 1 2 3 Pressure (hPa) For visible and shorter wavelengths, ‘Rayleigh’ scattering by air molecules becomes significant. The Rayleigh scattering cross-section sRa ðm2 mole1 ; cf. s in eqn [3]) can be computed theoretically: w¼ 10 4 5 6 100 1000 600 7 650 700 Wavenumber 750 (cm1) Figure 6 The CO2 15 mm absorption band showing the pressure level for which optical depth 5 1 (i.e., from which transmittance to the top of the atmosphere is e 1 ). The spectrum is averaged over 1 cm  1 intervals. Also shown are the positions of HIRS/3 channels 1–7. 1990 SATELLITE REMOTE SENSING / Temperature Soundings Pressure (hPa) 1 10 1 2 3 100 4 1000 0.0 5 6 7 0.2 0.4 0.6 d/dZ 0.8 1.0 Figure 7 The weighting functions for HIRS/3 channels 1–7 (see Figure 6). These are normalized so that the maximum atmospheric contribution is 1. To produce narrower weighting functions requires finding spectral regions where the optical depth w increases more rapidly with pressure than w / p (eqn [12]). One method is to select spectral regions where emission is predominantly from the wings of pressurebroadened lines (giving s / p, hence w / p2 ). Another method is to target emission from a gas whose concentration increases with pressure, such as tropospheric water vapor. Since the absorber is no longer well mixed, additional channels are required in order to retrieve its concentration, and the weighting function peaks are no longer at fixed pressures. A different method of improving the vertical resolution is to scan at an angle y to nadir, increasing the optical path to approximately w ¼ ap sec y, reducing both the peak pressure and width by a factor cos y. Taken to its extreme, this is, of course, the basis of limb sounding. Scanning to
50 across the orbit track is commonly employed for nadir sounders, but this is done in order to cover the atmosphere between adjacent orbit tracks, and a ‘correction’ applied in order to remove the resulting variation of the weighting functions. Nadir Sounding Instruments Table 1 lists nadir viewing instruments that have been used for temperature sounding. The first such instruments were infrared filter radiometers targeting various parts of the CO2 15 mm band, a simple technique still in use on operational satellites (Figure 6). However, such filters are limited to a minimum width of several wavenumbers, which does not allow much Table 1 Satellite nadir sounding temperature sensors Launch Satellite Instrument Technique a 1969, 1970 Nimbus 3, 4 1970, 1972 1972–1976 1972 Nimbus 4, 5 NOAA 2–5 Nimbus 5 1975 Nimbus 6 FR MI GC FR FR MW FR GC MW 1974–1994 NOAA 6–14 SIRS Satellite Infrared Spectrometer IRIS Infrared Interferometer Spectrometer SCR Selective Chopper Radiometer VTPR bVertical Temperature Profile Radiometer ITPR Infrared Temperature Profile Radiometer NEMS Nimbus E Microwave Spectrometer HIRS High Resolution Infrared Radiation Sounder PMR Pressure Modulated Radiometer SCAMS Scanning Microwave Spectrometer TOVS bTIROS Operational Vertical Sounder, comprising: HIRS/2 High Resolution Infrared Radiation Sounder/2 SSU Stratospheric Sounding Unit MSU Microwave Sounding Unit SSM/T Special Sensor Microwave/Temperature VAS VISSR Atmospheric Sounder (sounders with similar channels to HIRS) ATOVS bAdvanced TOVS comprising: HIRS/3 High Resolution Infrared Radiation Sounder/3 AMSU Advanced Microwave Sounding Unit AIRS Atmospheric Infrared Sounder TES cTropospheric Emission Spectrometer IASI Infrared Atmospheric Sounding Interferometer ATOVS (as above) 1977– 1980–1996 1994– 1998– 2001 2003 2005 DMSP GOES 4–7 GOES 8– NOAA 15– Aqua Aura Metop FR GC MW MW FR FR FR MW GS MI MI FR/MW a FR 5 filter radiometer, MI 5 Michelson interferometer, GC 5 gas correlation radiometer, MW 5 microwave radiometer, GS 5 grating spectrometer. b See also Table 2. c See also Table 3. SATELLITE REMOTE SENSING / Temperature Soundings 1991 scope for improving vertical resolution or extending coverage to higher altitudes. More recently, instruments have been developed to measure the full infrared spectrum at high resolution: 1 cm1 for AIRS (grating spectrometer), 0:25 cm1 for IASI (interferometer), 0:1 cm1 for TES (interferometer, in nadir-viewing mode). The increased spectral resolution, together with the large number of potential channels represented by the complete spectrum, allows combined temperature–water-vapor retrievals to be performed under a variety of atmospheric conditions, hence improved (tropospheric) vertical resolution. Figure 6 suggests that 4 mb is about the highest level that can be sounded using the 15 mm band with 1 cm1 resolution. However, Figure 8 demonstrates that emissions can be detected from higher levels, but in order to discriminate these it is necessary to resolve individual lines. Doppler-broadened line widths at the stratopause are of the order of 0:001 cm1 , well beyond the resolution obtainable using spaceborne interferometry, and even were such a resolution attainable the reduced photon flux from such a narrow bandwidth would lead to SNR problems. Gas correlation radiometry is one technique which has been used to extend the altitude range of infrared nadir sounding. By passing the signal through a pressure-modulated cell of CO2 , a synchronous component of the signal can be extracted corresponding to 0.10 Operational Temperature Sounders 0.08 250 240 0.06 230 220 0.04 210 200 190 180 0.02 Equiv. BB temperature (K) 260 Radiance (W m2 sr1 (cm1)1) emission in just the modulated regions of the cell transmittance spectrum, i.e., the CO2 line wings. The response can be tuned to maximum sensitivity at particular parts of line wings by adjusting the mean cell pressure. Since this also integrates the signal over all lines within the filter band, it gives an improved SNR compared with a single narrow-bandwidth measurement. This was the principle used in the SSU that provided the stratospheric sounding channels for the TOVS instruments. Microwave sounders have a major advantage over infrared sounders, since clouds are transparent at millimeter wavelengths. Spectral selection for microwave instruments is achieved by radio, rather than optical, techniques. Heterodyne mixing is used to combine the atmospheric microwave signal with a local oscillator (LO) at some central frequency ðGHzÞ. Since the mixing process is nonlinear, an ‘intermediate-frequency’ ðMHzÞ signal is produced corresponding to the difference between the two input signals. The result is to convert the atmospheric spectrum immediately above the LO frequency from microwave to radio frequencies, with the mirror image of the atmospheric spectrum below the LO frequency also superimposed. The spectral features can then be resolved with radio-frequency filters. The technology has now developed to the point where it is possible to achieve adequate SNR in bandwidths comparable to a stratospheric line width, allowing sounding up to the stratopause (Figures 9 and 10). 0.00 666 668 670 Wavenumber 672 674 (cm1) Figure 8 Atmospheric radiance spectrum (nadir view) calculated near the center of the CO2 15 mm band. The smooth envelope around 220 K corresponds to emission from the pressure-broadened line wings in the lower stratosphere (Figure 1), the upward spikes from Doppler-broadened lines at the stratopause (260 K), and the downward spikes from centres of strong lines near the mesopause (190 K). The National Oceanic and Atmospheric Administration (NOAA) began routine atmospheric temperature sounding measurements (Table 2) with the Vertical Temperature Profile Radiometer (VTPR) instruments on board the NOAA 2–5 satellites which operated from 1972 to 1979. These were infrared radiometers with six temperature sounding channels from 13–15 mm, plus a water vapor channel at 18 mm and another channel in the 11 mm atmospheric window. The VTPR was superseded by the TOVS (TIROS Operational Vertical Sounder) suite, first flown on the TIROS-N satellite in 1978 and subsequently on the NOAA 6–14 satellites. TOVS consisted of three instruments: HIRS/2 (High-resolution Infrared Radiation Sounder), a development of the HIRS instrument originally flown on Nimbus 6. This was a 20channel infrared radiometer with 12 temperature sounding channels covering both the 15 mm and the 4:3 mm CO2 bands, in addition to water vapor, ozone and atmospheric windows. 1992 SATELLITE REMOTE SENSING / Temperature Soundings 1 0.01 14 1.00 10.00 13 9 8 7 6 5 3 Pressure (hPa) Pressure (hPa) 0.10 4 12 11 10 9 10 8 7 6 5 4 100 100.00 1000.00 45 50 (A) 55 60 65 Frequency (GHz) 70 75 1000 0.0 3 2 0.2 0.01 14 13 12 11 10 Pressure (hPa) 0.10 1.00 1000.00 56.8 (B) 9 57.0 57.2 0.6 d/dZ 0.8 1.0 Figure 10 The AMSU/A weighting functions for channels 2–14 (channel 2 lies at 31.4 GHz; see Figure 9 for the other channels). These are normalized so that the maximum atmospheric contribution is 1. 10.00 100.00 0.4 57.4 57.6 Frequency (GHz) Figure 9 (A) O2 60 GHz absorption band showing the pressure at which the optical depth equals 1, and (B) close-up of the region around the AMSU/A 57.29 GHz local oscillator frequency. Also shown are the positions of AMSU/A channels 3–14. Note the use of both sidebands for channels 5 and 10, and four-sideband combinations for channels 11–14, distributed about combinations of two LO frequencies (57.29 GHz70.32 GHz). This superimposes similar spectral features into the intermediate frequency signal in order to improve the SNR. The resulting weighting functions are shown in Figure 10. SSU (Stratospheric Sounding Unit), a development of the PMR instrument also flown on Nimbus 6. This measured CO2 emission at 669 cm1 and used three different pressure modulator cells (at 1.5, 5, and 15 hPa) for stratospheric temperature sounding. MSU (Microwave Sounding Unit), a four-channel microwave radiometer sounding the O2 band at 60 GHz. ATOVS (Advanced TOVS) was first flown on NOAA 15, launched in 1998, and consists of two instruments: HIRS/3, a 20-channel infrared radiometer with similar spectral channels to HIRS/2 (Figures 6 and 7). AMSU (Advanced Microwave Sounding Unit), a 20-channel microwave radiometer designed for temperature and water sounding (Figures 9 and 10). This replaces the MSU and SSU with a single microwave instrument. Under the current Polar Operational Environmental Satellites (POES) program, NOAA aims to operate Table 2 The NOAA operational temperature sounders Satellite Launch VTPR NOAA-2 Oct. 72 NOAA-3 Nov. 73 NOAA-4 Nov. 74 NOAA-5 Jul. 76 TOVS (HIRS/2, MSU, SSU) TIROS-N Oct. 78 NOAA-6 Jun. 79 NOAA-7 Jun. 81 NOAA-8 Mar. 83 NOAA-9 Dec. 84 NOAA-10 Sep. 86 NOAA-11 Sep. 88 NOAA-12 May 91 NOAA-13 Aug. 93 NOAA-14 Dec. 94 ATOVS (HIRS/3, AMSU) NOAA-15 May 98 NOAA-16 Sep. 00 NOAA-M Mar. 02 NOAA-N Dec. 03 NOAA-N 0 Mar. 08 Deactivated Orbit Jan. 75 Aug. 76 Jun. 86 Jul. 79 a.m. a.m. a.m. a.m. Jan. 80 Mar. 87 Jun. 86 Dec. 85 Feb. 98 Sep. 91 Mar. 95 Dec. 98 Aug. 93 p.m. a.m. p.m. a.m. p.m. a.m. p.m. a.m. p.m. p.m. a.m. p.m. a.m. p.m. p.m. SATELLITE REMOTE SENSING / Temperature Soundings 1993 two satellites at any one time, each in a Sun-synchronous orbit with either a southward Equator crossing at around 7.30 a.m. local time (AM orbit) or a northward Equator crossing at around 2.30 p.m. (PM orbit), so that coverage of any point is repeated every 6 hours. Eventually the European Organisation for the Exploitation of Meteorological Satellites (EUMETSAT) will take over responsibility for the AM orbit with its Metop satellites containing additional instruments such as IASI, while the NOAA program merges with the military Defense Meteorological Satellite Program (DMSP) to form the National Polarorbiting Operational Environmental Satellite System (NPOESS) which will provide satellites in two other orbits. Limb Sounding Viewing tangentially through the atmospheric limb means the background is cold space, so that semitransparent optical paths can be used in preference to opaque paths, and consequently the weighting functions are determined by geometry rather than optical thickness. Compared with nadir viewing, limb viewing generally allows better vertical resolution and coverage to higher altitudes. However, a fundamental problem with the limb viewing geometry is that the ray paths traverse significant horizontal distances in the atmosphere ð200 km in the 1 km thick layer above the tangent point), which limits the scale of horizontal structures which can be resolved. Also, tropospheric limb views are more likely to be obscured by clouds than nadir views, restricting low-altitude coverage using the infrared. For these reasons, limb viewing is particularly suited to temperature sounding in the stratosphere and mesosphere, while nadir sounding is used for the troposphere. For a typical polar orbiting satellite at 700 km altitude the tangent point is some 3000 km away, so that 3 km at the tangent point subtends only 0.001 rad, or approximately 30 of arc. The narrow field of view reduces the radiance flux, so that SNR becomes a significant problem. Diffraction is also a limiting factor for microwave instruments: angular resolution varies approximately as the ratio of antenna width to wavelength, so to resolve 0.001 rad at 2 cm (60 GHz) would require a 5 m antenna. Weighting Functions Neglecting the background term due to cold space, eqn [1] becomes I¼ Z 1 1 B dt dx dx ½14 where x is now the distance along the tangent path, with x ¼ 0 at the tangent point and x ’ þ1 at the satellite. Ignoring refraction and assuming a spherical Earth of radius a, distance x and altitude z ð aÞ along a path are related by x2 ’ 2aðz  zt Þ ½15 where zt is the altitude of the tangent point. For an isothermal atmosphere at temperature T, the molar density of air, r, varies with altitude as:
zz  pt t r¼ exp  ½16 RT H where pt is the pressure at the tangent point, and H ð¼ RT=gM, from eqn [6]) the atmospheric scale height. Integrating eqn [3] along the path, assuming constant absorber volume mixing ratio v and absorption coefficient s, and converting to altitude coordinate z0 ¼ z  zt , Z 1 w ¼ vs r dx ½17 1 pt pffiffiffiffiffiffi 2a ¼ vs RT ¼ vs Z 1 0 pt pffiffiffiffiffiffiffiffiffiffi p 2aH RT  0 1 z pffiffiffiffi exp  dz0 0 H z ½18 ½19 If the absorption is weak then I ’ Bw and dt=dz  dw=dz, so the characteristic shape of limb viewing weighting functions is given by  0 dt 1 z / pffiffiffi0ffi exp  ½20 dz H z In practice, the width of the peak is limited by the layering assumed for the retrieval. Examples, converted to temperature ðdI=dTðzÞ rather than dI=dBðzÞÞ, are plotted in Figure 11. Note that at lower altitudes, where this spectral region becomes opaque, the weighting functions resemble those of a nadir sounder (Figure 7). The vertical profile can be sounded either using a single detector and scanning in elevation, or by using a detector array to view the different elevation angles simultaneously. However, in practice, at least two spectral channels are required in order to retrieve pressure profile information as well. Pressure Determination To model the atmospheric transmittance (eqn [14]) it is also necessary to know the pressure, which is usually retrieved simultaneously with temperature (the problem does not arise in nadir sounding, since the 1994 SATELLITE REMOTE SENSING / Temperature Soundings 60 50 km 45 km 40 Altitude (km) 40 km 35 km 30 km 25 km 20 20 km 15 km 10 km 0 0.0000 0.0010 0.0020 0.0030 0.0040 0.0050 0.0060 dI/dT (Wm2sr1(cm1)1K1) Figure 11 Temperature weighting functions for a limb-viewing instrument using a 610–640 cm  1 filter (HIRDLS channel 3). Curves show response in radiance at the labeled tangent heights to a 1 K temperature perturbation in 1 km layers at altitudes indicated by the left axis. equivalent radiative transfer equation, eqn [11], is formulated in pressure coordinates). This requires radiance measurements in at least two spectral channels with different pressure–temperature characteristics. Differences in temperature sensitivity arise from the spectral dependence of the Planck function or variations in the absorption coefficient, e.g., the changing shape of the rotational band structure (Figure 3). Some variation in pressure sensitivity may also arise from the absorption coefficient, but the biggest effect is the nonlinear variation of radiance with optical thickness: I ’ Bð1  exp ðwÞÞ (from eqn [14]), with w / p (eqn [19]). Figure 12 illustrates the problem graphically, with good pressure–temperature discrimination corresponding to conditions where radiance contours from any two channels intersect at a large angle. At high pressure, all radiances are independent of tangent point pressure (I ’ B, opaque limit). At low pressure, all radiances vary linearly with pressure (I  w, transparent limit). The best pressure– temperature discrimination occurs between 100 and 10 hPa (15–35 km in the atmosphere) as the channels undergo different transitions from the opaque to transparent limits. 0.001 0.010 Pressure (hPa) 0.100 1.000 10.000 100.000 1000.000 180 200 220 240 260 280 300 Temperature (K) Figure 12 Radiance contours for the four HIRDLS CO2 channels (Figure 13) for a range of pressures and temperatures simulating different tangent point conditions. The thick line shows the T ðpÞ profile of the US standard atmosphere. In order of increasing wavenumber and absorption, HIRDLS channels 2–5 are indicated by solid (2), dotted (3), dashed (4), and dot–dash lines (5). SATELLITE REMOTE SENSING / Temperature Soundings 1995 For microwave measurements, radiances can be almost independent of temperature. Assuming I ’ Bw (optically thin), and considering only the pressure and temperature dependent components of eqn [19], I / Bs p pffiffiffiffiffi H T ½21 pffiffiffiffi For the wings of Lorentz-broadened lines, s / p= T , and since the scale height H and, in the microwave region, B are both proportional to T, the temperature dependences all cancel out, leaving I / p2 . Since it is generally not possible to retrieve both temperature and pressure independently over all altitudes, it is usual to include some knowledge of the vertical distance between tangent points, either using the fixed geometry of a detector array or else pointing information from the elevation scan, and assume hydrostatic balance (eqn [6]) to constrain the problem. Limb Emission Sounders Table 3 lists limb emission instruments used for temperature retrievals. The various techniques for nadir-viewing instruments all have their parallels in limb sounding, although with modifications driven by the different viewing geometry. As with nadir sounding, the earliest measurements were made with infrared radiometers using the CO2 15 mm band. The good SNR performance achievable with broad filters means that this technique continues to be used for high spatial resolution measurements (Figure 13). Although the precise choice of filter position does not have as crucial an influence on the sounding characteristics as in the nadir viewing case, it is nevertheless desirable to choose spectral regions of intermediate absorption. Too opaque, and little information is obtained from the tangent point; too transparent, and the SNR is unnecessarily reduced. Apart from requiring at least two channels in order to retrieve both pressure and temperature, additional channels can be used in order to optimize the transmission characteristics for different altitude ranges. Gas correlation and spectrally resolving instruments are usually employed in limb sounding in order to discriminate between lines of the target molecule and those of other species. However, foreign lines have only a small influence in the 15 mm CO2 band so, for temperature sounding, the main advantage of these instruments is in improving the pressure–temperature discrimination. The main drawback is the reduced SNR associated with the narrower bandwidth and, in the case of Table 3 Satellite limb sounding temperature sensors Launch Satellite Instrument Technique a 1976 1978 Nimbus 6 Nimbus 7 1981 1991 SME UARS FR FR GC RS GC FP FP MI MW 1994, 1997 Shuttle 2001 Odin 2001 Envisat 2001 TIMED LRIR Limb Radiance Inversion Radiometer LIMS Limb Infrared Monitoring of the Stratosphere SAMS Stratospheric and Mesospheric Sounder Solar Mesosphere Explorer ISAMS Improved Stratospheric and Mesospheric Sounder CLAES Cryogenic Limb Array Etalon Spectrometer HRDI High Resolution Doppler Imager WINDII Wind Imaging Interferometer MLS Microwave Limb Sounder CRISTA Cryogenic Infrared Spectrometers and Telescopes for the Atmosphere OSIRIS Odin Spectrometer and IR Imaging System SMR Sub-Millimeter Radiometer MIPAS Michelson Interferometer for Passive Atmospheric Sounding SCIAMACHY cScanning Imaging Absorption Spectrometer for Atmospheric Chartography SABER Sounding of the Atmosphere using Broadband Emission Radiometry TIDI TIMED Doppler Interferometer HIRDLS High Resolution Dynamics Limb Sounder MLS Microwave Limb Sounder TES bTropospheric Emission Spectrometer 2003 Aura GS RS MW MI GS FR FP FR MW MI a FR 5 filter radiometer, GC 5 gas correlation radiometer, FP 5 Fabry–Perot spectrometer, MI 5 Michelson interferometer, MW 5 microwave radiometer, GS 5 grating spectrometer, RS 5 Rayleigh scattering. b See also Table 1. c See also Table 4. 1996 SATELLITE REMOTE SENSING / Temperature Soundings 2 3 5 4 1.0 0.8 Absorption 0.6 0.4 0.2 0.0 600 650 700 750 1 Wavenumber (cm ) Figure 13 Absorption (1-transmittance) for limb paths through 20 (solid), 30 (dotted), and 40 (dashed) km tangent heights across the CO2 15 mm band. Also shown are the spectral positions of the four HIRDLS temperature sounding channels (channels 2–5). gas correlation, the need to view perpendicularly to the orbital motion in order to avoid introducing Doppler shifts between the atmospheric lines and those in the onboard cell. Microwave instruments have an advantage over infrared instruments in being insensitive to cloud at low altitude and to non-LTE effects at high altitude. However, vertical resolution is limited by diffraction, which can cause problems in resolving the tropopause and at high altitudes it becomes necessary to model the Zeeman splitting of lines which varies with the Earth’s magnetic field along the line of sight. Since the radiances are almost independent of temperature, these instruments effectively retrieve a pressure profile, with temperature information coming mostly from hydrostatic balance. Concomitantly, the impact of any temperature errors on the constituent retrievals is also reduced. These measurements all rely on thermal emission from the atmosphere, but, at shorter wavelengths, Rayleigh-scattered sunlight can also be detected. Since the scattering is proportional to air density, measurements of the scattering profile can be used to determine the temperature profile. The technique is usable only during daytime, and for relatively high altitudes where Rayleigh single scattering can be assumed, but has been used with SME and OSIRIS measurements (both grating spectrometers). Occultation Measurements Occultation measurements use the Sun, Moon, stars, or other satellites as the source of the detected radiation and monitor the change as the source rises or sets beyond the atmospheric limb. While the geometry is the same as that of limb emission measurements, the location of the tangent point is defined by ephemeris data (i.e., knowledge of the positions of the satellite, Earth and source), which is usually more accurate than using the satellite attitude/ pointing data which defines the tangent point for limb emission measurements. Potentially, occultation retrievals can therefore be performed on an absolute height scale. However, the relative motion of the source often means that the locus of tangent points is far from vertical, leading to ‘slanted’ profiles, extending over several hundred kilometers horizontally. Table 4 lists the instruments used for occultation measurements of temperature. Solar Occultation The Sun can be viewed through the atmospheric limb as a satellite passes between the day and night hemispheres, i.e., twice an orbit, or about 30 times in 24 hours for a polar orbiting satellite. Solar radiation, equivalent to that of a 6000 K blackbody, has a peak at visible wavelengths (Figure 2), but for temperature pressure sounding it is necessary SATELLITE REMOTE SENSING / Temperature Soundings 1997 Table 4 Satellite occulation temperature sensors Launch Satellite Instrument Technique a 1984 1985 1991 1993 1995 1996 1998 2001 ERBS Spacelab 3 UARS SPOT-3 Microlab-1 ADEOS SPOT-4 Envisat FR MI GC/FR FR RO GS FR GS 2001 2002 2005 Meteor-3 M SCISAT Metop SAGE II Stratospheric Aerosol Gas Experiment II ATMOS Atmospheric Trace Molecule Spectroscopy HALOE Halogen Occultation Experiment POAM II Polar Ozone and Aerosol Measurement II GPS/MET Global Positioning System Meterology ILAS Improved Limb Atmospheric Spectrometer POAM III Polar Ozone and Aerosol Measurement III GOMOS Global Ozone Monitoring by Occultation of Stars SCIAMACHY bScanning Imaging Absorption Spectrometer for Atmospheric Chartography SAGE III Stratospheric Aerosol Gas Experiment III ACE Atmospheric Chemistry Explorer GRAS GPS Receiver for Atmospheric Sounding GS GS MI RO a FR 5 filter radiometer, MI 5 Michelson interferometer, GC 5 gas correlation radiometer, RO 5 radio occulation, GS 5 grating spectrometer. b See also Table 3. I ¼ Bsun t ½22 Taking the ratio with the high-altitude radiance Isun ¼ Bsun gives the atmospheric transmittance t, with weighting functions dt=dz as for limb sounding (eqn [20]). The main advantage of solar occultation over limb emission measurements is the high SNR, allowing increased vertical and spectral resolutions, and sounding to higher altitudes. Also, since the measurements are of transmittance rather than emission, non-LTE effects are generally less significant and, by taking the ratio of radiances, the measurements are selfcalibrating. Since the absolute altitude of the tangent point is known (in practice this depends on the ability of the solar tracker to keep locked onto the same part of the solar disk), it is possible to retrieve temperature using only a single channel. Pressure information at the lowest altitude pðzÞ can be obtained from meteorological fields and integrated upwards. Alternatively, transmittance spectra can be acquired using an interferometer. Pressure and temperature can be retrieved from the strength and shape of a band, but if the resolution can be made high enough (e.g., 0:01 cm1 for ATMOS, Figure 14) then pressure information can also be retrieved from the individual line widths. At shorter wavelengths, it is possible to determine the temperature profile (via air density) by measuring the attenuation due to Rayleigh scattering. This has been applied to SAGE II data, but measurements of molecular absorption are generally preferred. Whichever technique is used, the main disadvantage of solar occultation is that only around 30 profiles a day can be obtained, with the sunrise and sunset profiles confined to two narrow latitude bands and no information obtainable during nighttime or the polar winter. SAGE III and SCIAMACHY can also use lunar occultation (although not for temperature retrievals), which can double the number of occultation events per orbit and extend coverage outside sunrise or sunset conditions. GOMOS uses stellar occultation: observing any of 100 bright stars as they rise or set through the atmosphere gives near global coverage, although at the expense of muchreduced SNR. 1.0 0.8 Transmittance to use molecular features in the near infrared such as the O2 A-band at 0:76 mm or the CO2 bands at 2:7 mm and 4:3 mm. With the Sun in the line of sight, thermal emissions from the atmosphere are negligible, so eqn [1] can be simply integrated: 0.6 0.4 0.2 0.0 2370 2372 2374 2376 2378 2380 Wavenumber (cm1) Figure 14 Transmittance of part of the CO2 4:3 mm band at 0.01 cm  1 resolution, calculated for a limb path tangent height of 50 km. 1998 SATELLITE REMOTE SENSING / Temperature Soundings GPS Sounding In the Global Positioning System (GPS) a network of 24 satellites each continuously emits precise timing and location information. By comparing the received signals from at least four satellites, it is possible to fix the three-dimensional coordinates of any point in space, the fourth satellite being required to establish the time offset. Although intended as a navigation aid, the signals can also be used to determine the atmospheric density profile. Relative to a receiver placed in low Earth orbit, GPS satellites rise or set beyond the horizon several hundred times each day. As the signals pass through the atmosphere, refraction introduces a time delay which, if measured relative to a reference clock, can be used to determine the refractive index profile, hence density and temperature. It is conventional to describe the GPS clock delay in terms of the refraction angle e (Figure 15). For a spherically symmetric atmosphere, Snell’s law gives the following relationship along the refracted path: nr sin y ¼ constant ¼ q ½23 where n is the refractive index, r the distance from the centre of curvature, and y the angle of the ray to the local horizontal (y ¼ 0 at the tangent point). The constant q is the tangent distance of the unrefracted ray, sometimes known as the ‘impact parameter’ from the analogy with nuclear physics. Defining a refractive radius x ¼ nr, it can be shown that the total deflection of the ray e is given by an integral of the refractive index profile nðxÞ: eðqÞ ¼ 2q Z þ1 q   q ln nðxÞ ðx2  q2 Þ1=2 dx ½24 qx During the occultation, a set of measurements eðqÞ is acquired. The Abel transform can then be used to invert this relationship to recover the refractive index profile (hence density, from eqn [8]): Z 1 þ1 eðqÞðq2  x2 Þ1=2 dq ½25 ln nðxÞ ¼ p z The maximum altitude, limited by the minimum phase shift which can be detected, is around 50 km and the vertical resolution is of the order of 1 km. The main problems with using GPS occultation as a standalone temperature sounder are the assumptions of horizontal uniformity and the effect of tropospheric water vapor (eqn [8]). GPS measures only one quantity – refractive index – so there is no means of separating the effects of air density and water vapor concentration or determining the horizontal gradients unless external information is used. This is less of a disadvantage when used in conjunction with a forecast model which can provide prior estimates of water vapor and temperature fields. With advances in data assimilation techniques, the refraction itself may eventually be directly modeled, so even occultations extending over large horizontal distances could be used. GPS receivers are relatively cheap and lightweight ð1 kgÞ, and free of calibration errors (time being the only measured variable), so it seems likely that GPS occultation measurements, either inverted or directly assimilated, will be an important source of atmospheric temperature information in the future. See also Radiative Transfer: Absorption and Thermal Emission; Non-local Thermodynamic Equilibrium. Satellites: Research (Atmospheric Science).  Further Reading Receiver q r q GPS emitter Figure 15 Temperature sounding using GPS signals. The receiver, in low-Earth orbit, measures the refraction angle e of the signal as the GPS satellite rises or sets beyond the horizon. Barnett JJ (1987) Satellite-borne measurements of middleatmosphere temperature. Philosophical Transactions of the Royal Society of London, Series A 323: 527–544. Houghton JT, Taylor FW and Rodgers CD (1984) Remote Sounding of Atmospheres. Cambridge: Cambridge University Press. Liou KN (1992) Radiation and Cloud Processes in the Atmosphere. Oxford: Oxford University Press. Stephens GL (1994) Remote Sensing of the Lower Atmosphere, An Introduction. Oxford: Oxford University Press. SATELLITE REMOTE SENSING / TOMS Ozone 1999 TOMS Ozone R S Stolarski and R D McPeters, NASA Goddard Space Flight Center, Greenbelt, MD, USA Copyright 2003 Elsevier Science Ltd. All Rights Reserved. Introduction and History The Total Ozone Mapping Spectrometer (TOMS) is a satellite-borne instrument that measures ozone by measuring the ultraviolet light scattered from the atmosphere. These measurements are used to determine the total column amount of ozone in the atmosphere. The idea that ozone could be measured quantitatively from a satellite was first put forward in 1957 by Singer and Wentworth. During the 1960s, a number of satellites were launched with instruments that could be used to deduce the concentration of ozone as a function of altitude. In 1967, Dave and Matter published a theory for the derivation of the total column amount of ozone from a satellite backscatter instrument. This theory was used for the interpretation of the data from the first instrument, to measure the total column amount, of ozone from space, the backscatter ultraviolet (BUV) instrument. The BUV instrument was launched on the Nimbus 4 satellite in 1970. It made nearly-global measurements for two years and then operated more sporadically for an additional five years. The modern global data record starts with the launch of the Nimbus 7 satellite in 1978. The satellite carried two ozone-measuring instruments, the Solar Backscatter Ultraviolet (SBUV), and the Total Ozone Mapping Spectrometer (TOMS). The SBUV instrument was a double monochromator designed to measure backscattered radiation at 12 wavelengths from 255 nm to 340 nm. These were used to deduce the upper-stratospheric concentration profile of ozone and the total column amount of ozone along the nadir track of the satellite. The TOMS instrument measured the backscattered radiation at six wavelengths from 312.5 nm to 380 nm so that the total column amount of ozone could be deduced. TOMS was a single monochromator with a scanning mirror that allowed the instrument to make measurements at 35 scan angles from left to right across the ground track of the satellite. TOMS was thus able to measure over the entire sunlit portion of the globe each day. The TOMS instrument on Nimbus 7 made measurements for more than 14 years. The instrument finally failed in May 1993. A second TOMS instru- ment was launched on the Russian Meteor 3 satellite in 1991. This instrument made total ozone measurements until the end of 1994. A third TOMS instrument was launched on the Japanese ADEOS satellite in 1996. The satellite power array failed after seven months of operation. The fourth instrument was launched on the Earth Probe satellite, also in 1996. Designated EP-TOMS, it was still taking data in late 2002. A fifth instrument, Quik-TOMS, was launched in September, 2001 but the launch vehicle’s second stage did not fire correctly and the instrument did not reach orbit. The TOMS measurements are best known for images of the ozone hole. When Farman’s paper on low ozone in the Antarctic was published in 1985, TOMS images revealed that this was a continent-wide phenomenon and not local. TOMS now maps the development of the ozone hole each Antarctic spring. Theory of TOMS Measurement Rayleigh Scattering Light from the Sun penetrates into the atmosphere, with most of the visible light reaching the ground. Light is scattered by the molecules that make up the atmosphere in a process called Rayleigh scattering, named after Lord Rayleigh who first described it in the late nineteenth century. The probability of Rayleigh scattering depends inversely on to the fourth power of the wavelength (l4 ). Thus, an ultraviolet photon of 300 nm wavelength is 16 times more likely to be scattered than a visible photon of wavelength 600 nm. This is why the sky is blue; when we look at the sky away from the Sun, blue light is much more likely to be scattered towards us than is red light. Ozone Absorption Sunlight can be absorbed in the atmosphere by a variety of molecules. The principal absorber of ultraviolet light in the Earth’s atmosphere is ozone. Absorption in the UV by ozone is strong enough that a few parts per million of ozone remove all of the sunlight at wavelengths shorter than about 300 nm before they can reach the ground. We are thus provided with a shield from the high-energy radiation that could break important DNA bonds in living cells. The absorption of UV by ozone is the property that we generally use to measure the amount of ozone in the atmosphere. From the ground, we can look upward and measure how much radiation reaches us at a 2000 SATELLITE REMOTE SENSING / TOMS Ozone wavelength that is absorbed by ozone. We can compare this to the radiation received at a nearby wavelength that is not absorbed by ozone to determine the amount of ozone that must be between us and the Sun. From satellites, we can look down and measure the radiation that is scattered back out of the atmosphere and again compare the amount at an absorbed wavelength with an unabsorbed (or less absorbed) wavelength. Surface Reflection Radiation that does reach the ground can be absorbed or reflected by the surface. The probability of reflection depends on the nature of the surface. In the ultraviolet, the Earth is a very poor reflector. The UV reflectivity of the ocean in the 300–350 nm region of the spectrum is only about 4%. Most land surfaces have similarly low reflectivities, no more than 5% except in desert areas. Areas covered by ice and snow have very high reflectivities, reaching 90% in the Antarctic. Clouds When clouds are present, radiation reaching them is reflected back to space with high efficiency. Cloud reflectivities can reach 80–90% for thick clouds. Solar radiation that is reflected by clouds does not pass through the part of the atmosphere below the cloud and has no opportunity to be absorbed by the ozone below the clouds. The TOMS measurement is thus a measurement of the ozone above the cloud layer. Fortunately, this is a small effect since 90% of the ozone in the atmosphere is in the stratosphere. Only part of the 10% in the troposphere will be below the clouds. This amount can be estimated from climatology, so that the measurement can be transformed into a fairly accurate estimate of the total column of ozone in the atmosphere. Aerosols Aerosols (dust particles in the atmosphere) also scatter radiation, further adding to the atmosphere’s overall reflectivity. Their scattering does not follow the l4 dependence of Rayleigh scattering but is close to a l1 dependence. When dust is in the atmosphere, the sky appears more nearly white. Aerosols do not affect our ability to measure ozone. However, the multiple reflectivity wavelengths can be used to deduce some information about the properties of aerosols. Measurement deviations from the expected result for a Rayleigh scattering atmosphere can be used to determine an aerosol index (see results section below). Description of the Retrieval Algorithm TOMS measures ultraviolet light scattered from the atmosphere and the Earth and clouds. An algorithm is needed to infer ozone from these measurements. The instrument looks downward at the Earth and also uses a diffuser plate to look at direct sunlight. The basic measured quantity is the ratio of the direct solar irradiance to the backscattered radiance. This is usually expressed as the N-value, or logarithm of the ratio (eqn [1]). N ¼ 100 logðI0 =FÞ ½1 F is the solar irradiance at the particular wavelength and I0 is the Earth’s backscattered radiance. Using the ratio of direct solar to backscattered radiation cancels some of the main instrumental errors; that is, the instrument throughput is the same for each measurement. However, a diffuser plate is used to reflect the sunlight into the instrument. The reflectivity of the diffuser plate affects the solar irradiance measurement, but not the backscattered radiance measurement. If a pair of wavelengths is used in the analysis, then the diffuser reflectivity can be canceled out in the ratio if that reflectivity is the same for both wavelengths. Thus we form the pair N-value as in eqn [2]. Np ¼ N ðl1 Þ  N ðl1 Þ ¼ logðI01 =F1 Þ  logðI02 =F2 Þ ¼ log ðI01 =I02 Þ ½2 ðF2 =F1 Þ These N-values reflect the effects of scattering, reflection, and absorption. Figure 1 illustrates the dependence of N-value on wavelength, clearly showing that an ozone signal can be derived from the data. The actual algorithm used for the TOMS retrieval uses a radiative transfer code based on the early work of Dave. Forward calculations are carried out for a matrix of parameters including total ozone. These then form a lookup table that is interpolated to derive total ozone. Description of the TOMS Instrument Instrument The TOMS instruments are single, fixed monochromators with exit slits at six near-UV wavelengths. The slit functions are triangular with a nominal 1 nm bandwidth. The order of individual measurements is determined by a chopper wheel. As it rotates, openings at different distances from the center of the wheel pass over the exit slits, allowing measurements at the different wavelengths. The order was not one of SATELLITE REMOTE SENSING / TOMS Ozone 2001 This is a qualitative description of the orbits. Actually, for the purpose of orbit stability, the satellite does not pass exactly over the pole. For the three spacecraft above, the orbital inclination was approximately 981, which gave a maximum poleward latitude of 801. From this orbit TOMS could see the pole itself by scanning to the far right or left. The Meteor 3 spacecraft was in a polar orbit but was not Sun synchronous. Its Equator-crossing time drifted from near noon to near sunset and back to near noon in a 220-day cycle. 200 Measured N -value = 100 log10 (I/F ) N -value 150 100 Extrapolation from reflectivity wavelengths 50 300 Geometry and Timing 320 340 360 Wavelength (nm) 380 400 Figure 1 Illustration of the dependence of N-value on wavelength. The N-values for all of the TOMS measurements for one day (1 January 1985) within one degree of latitude of 351 N were averaged to make the plot. The linear straight line is fitted to the three longest wavelengths to illustrate an extrapolation to shorter wavelengths. The actual TOMS algorithm uses a full radiative transfer code to determine this extrapolation. The difference between the short-wavelength N-values and the extrapolation represents the absorption by ozone. monotonically increasing or decreasing wavelength; instead, the wavelengths were interleaved to minimize the effect of scene changes on the ozone retrieval. A ground aluminum diffuser plate was deployed to reflect sunlight into the instrument for measurement of the solar irradiance. This diffuser plate was shared with the Solar Backscatter Ultraviolet (SBUV) experiment on the Nimbus 7 satellite. It was normally deployed once a week for TOMS solar irradiance measurements, in addition to the SBUV deployments. Orbit The Nimbus 7, ADEOS, and Earth-Probe satellites were in Sun-synchronous polar orbits. The nearly circular orbit is oriented perpendicular to the plane of the Earth’s orbit around the Sun such that the satellite comes over the south pole of the Earth toward the Equator; crosses the Equator near local noon; and then passes over the north pole onto the nightside of the Earth. The satellite crosses the Equator again on the nightside at near midnight local time. By the time the satellite comes back onto the dayside, the Earth has rotated for approximately 90 minutes and the satellite passes over a point at the Equator that is 27 degrees of latitude to the west of the previous orbit, again at local noon. In this way, the satellite orbits l5 times per day, fixed relative to the Sun, and the Earth rotates underneath so that the satellite sees the whole of the surface of the Earth within a 24-hour period. The instrument field of view for TOMS is 3  3 degrees. At an altitude of 950 km for Nimbus 7, this projects to a nadir spot size on the surface of 50 km by 50 km. Earth-Probe was launched initially into a 500 km orbit. This resulted in a nadir spot size of 26 km. In December of 1997, it was boosted to an altitude of 740 km, increasing the nadir spot size to 40 km. For each of the TOMS instruments, a mirror scans perpendicular to the orbital plane in 35 steps of 31. The scan angles range from 511 on the right side of spacecraft nadir to 511 on the left (relative to the direction of flight). At the end of the scan, the mirror returns to the first position and begins another scan. For Nimbus 7, the cross-track scans from consecutive orbits overlapped, creating a completely filled global map of the sunlit part of the Earth each day. The lower altitude of the Earth-Probe TOMS results in small areas between orbits near the Equator where no measurements are made. The location of these gaps shifts from day to day so that no place fails to be measured over the span of a few days. During the cross-track scan, each of the 35 measurement locations is observed for 200 ms. The total duration time for a single scan is 7.8 s, during which time the satellite travels approximately 40 km. One orbit consists of nearly 400 cross-track scans or 13 000 measurements. Fifteen orbits result in about 190 000 measurements of total ozone every day. Wavelengths TOMS makes measurements at six wavelengths. These are selected by slits cut into a chopper wheel that rotates at 5 revolutions per second. The wavelengths for the Nimbus 7 TOMS and the Earth Probe TOMS are shown in Table 1, along with the absorption and scattering coefficients averaged over the slit function of the spectrometer. The basic ozone-measuring wavelengths are at 312.5 nm and 317.5 nm. These are sufficiently absorbed by ozone to get a signal and sufficiently 2002 SATELLITE REMOTE SENSING / TOMS Ozone Table 1 Effective absorption and scattering coefficients Vacuum wavelength Effective ozone (nm) absorption coefficient (atm cm  1) at 273 K (C0 ) Nimbus 7 TOMS 312.34 317.35 331.06 339.66 359.96 380.01 EP-TOMS 308.65 312.56 317.57 322.37 331.29 360.40 80 80 Rayleigh scattering coefficient (atm  1) 60 80 40 1.9000 0.9915 0.1703 0.0390 1(  8) 1(  8) 1.022 0.954 0.797 0.715 0.560 0.446 3.23 1.83 0.973 0.536 0.165 1(  8) 1.077 1.020 0.953 0.894 0.795 0.557 80 20 40 60 Some Results from TOMS Measurements Reflectivity The basic TOMS measurement is of reflected radiation at six wavelengths. The longer of these wavelengths are not affected by ozone absorption and are thus a measure of the reflectivity of the atmosphere in the ultraviolet. The algorithm calculates the expected backscattered radiation from a pure Rayleigh-scattering atmosphere. Deviations from this expectation are driven primarily by clouds and secondarily by aerosols. The deviation caused by clouds can be represented as a percentage reflectivity (see Figure 2). A major feature of nadir remote sensing in the ultraviolet is that the surface is relatively dark. Typical minimum reflectivities off the surface of the ocean are about 4%. While the surface is dark, Rayleigh scattering is strong in the UV. The Rayleigh scattering cross-section varies as the inverse fourth power of the wavelength (l4 ). Thus, only about 30% of the radiation at 350 nm reaches the ground for typical mid-latitude conditions. The return signals to TOMS are generated mostly in the lowest part of the troposphere. This has implications for the derivation of total ozone; TOMS does not see pollution in the boundary layer very well. 40 60 20 20 20 40 20 40 60 40 60 60 20 20 40 transmitted to reach near the surface. On Nimbus 7 TOMS, the 360 nm and 380 nm wavelengths measure the reflectivity of the surface/atmosphere. 80 40 80 40 20 60 40 Figure 2 Single-day (1 March 1982) reflectivity map at a wavelength of 360 nm over North America. Gray shaded area indicates where reflectivity is greater than 60%, indicating the presence of clouds. High reflectivity over Northern Canada may be clouds or snow/ice. Ozone Maps The original selling point for the TOMS was the capability to map the total ozone content on a daily basis to help understand its relationship to changes in the meteorology of the atmosphere. The problem of the relationship of total ozone to meteorology goes back to Dobson in the 1920s. Dobson had six of his spectrophotometers built and distributed throughout Europe to examine this problem. He found that when a high-pressure system was present, ozone was low; and when a low-pressure system was present, ozone was high. TOMS can make a map of the entire sunlit portion of the globe in a single day (see Figure 3). 500 450 350 400 300 275 Figure 3 Single-day (1 March 1982) ozone map over North America. Gray shaded area indicates where total ozone amount is greater than 400 DU. SATELLITE REMOTE SENSING / TOMS Ozone 2003 contrast in the seasonal variation of total ozone over the two poles. These maps clearly demonstrate the day-to-day and year-to-year variability of ozone over the Arctic. Ozone Trends Figure 4 Single-day (5 October 2000) ozone map over the Antarctic. Dark blue to purple shades near pole indicate where total ozone amount is less than 220 DU, a common definition for the region of the Antarctic ozone hole. The map is a polar orthographic projection with the south pole at the center and the equator at the outer boundary. Zero longitude is to the right. When the discovery of the ozone hole was announced in 1985, TOMS was immediately used to map the extent of the ozone-depleted region (Figure 4). Using TOMS, the daily progress of the hole could be followed. These maps demonstrated how the depleted region rotated around the pole, was distorted by the meteorology, and was finally broken up by a series of wave events that eroded the polar vortex. TOMS also can produce maps of ozone over the Arctic polar region (not shown). These show the The Nimbus 7 TOMS instrument was originally designed to map ozone on a daily basis as a study of day-to-day variability in total ozone. TOMS is now used as a part of a satellite-based measurement system for detecting long-term trends in stratospheric ozone. A number of features in the TOMS measurements have made it possible to detect calibration drifts of the instrument well enough that a data record now exists for more than 20 years that is estimated to be good to nearly 1% per decade (2s). An important feature of the TOMS measurement is the redundancy of having more than one ozoneabsorbing wavelength. In the algorithm, the use of pairs of wavelengths to calculate albedo cancels out many of the potential instrument errors. Drift errors that remain have a tendency to be larger with larger separation in wavelength of the pair. The redundant pairs of wavelengths can be used in a ‘pair justification’ to remove drift errors that are linearly proportional to wavelength. TOMS (along with SBUV) has become an instrument that provides a long-term calibrated data record for trend detection. These data have been used in standard statistical analyses for trends (Figures 5 and 6). These statistical analyses fit the time-series to terms for mean, seasonal variation, linear trend, 11-year sunspot cycle, and 26-month quasi-biennial oscillation. The continuation of this data set will be used in the search for the expected turnaround in ozone as the provisions of the Montreal Protocol begin reducing the amount of ozone-depleting chlorine in the atmosphere. 2.5 2.0 2.0 1.0 1.0 1.5 0.5 0.5 0 0 0.5 0 0 1.0 2.0 2.5 0.5 0.5 1.0 1.5 3.0 2.0 2.5 Figure 5 Linear trends (in % per decade) calculated from the 1979–2000 data from TOMS and SBUV instruments as a function of longitude and latitude. Gray shaded areas indicate where trends are more negative than 1.5% per decade. 2004 SATELLITE REMOTE SENSING / TOMS Ozone North mid-latitude (25qN–60qN) 20 Deviation (DU) Deviation (DU) Global (60qS–60qN) 4 2 0 2 4 6 8 10 1980 1985 1990 Year 1995 10 0 10 20 30 40 2000 1990 Year 1995 2000 10 5 Deviation (DU) Deviation (DU) 1985 South mid-latitude (60qS–25qS) Tropical (10qS–10qN) 10 0 5 10 1980 1980 1985 1990 Year 1995 2000 5 0 5 10 15 20 1980 1985 1990 Year 1995 2000 Figure 6 Time-series of total ozone measurements averaged over broad bands of latitude. The mean seasonal variation within the band has been removed before plotting, to emphasize the year-to-year variability. Aerosols TOMS measures the reflectivity of the Earth–atmosphere system at several wavelengths not absorbed by ozone. If the atmosphere were perfectly clean, the backscattered radiation received by the satellite could be determined from a Rayleigh scattering calculation that would predict a specific ratio of radiation between two wavelengths. Aerosols disturb this ratio in a predictable manner; one direction for absorbing aerosols, the opposite for nonabsorbing aerosols. Using these facts, the TOMS data has been used to determine an ‘aerosol index’. Reasonable assumptions about the nature of the aerosols lead to global maps of the spread of dust from deserts and smoke from biomass burning in Africa and South America. There are now 20 years’ worth of such data from the TOMS instruments. UV at Surface Many of the effects of ozone depletion are related to the dose of ultraviolet radiation received at the surface of the Earth. TOMS measures the outgoing, absorbed UV radiation and the reflectivity due to clouds and aerosols. These data can be combined with a radiative transfer modal to estimate the UV flux at the surface daily over the globe. Again, TOMS now has 20 years’ worth of daily UV flux maps available. Tropical Tropospheric Ozone TOMS measures the total column amount of ozone with some adjustments for the inefficiency of the penetration of UV sunlight into the boundary layer. In the Tropics, most of the variability of total ozone around a circle of constant latitude is in the troposphere rather than the stratosphere. Several schemes have been developed for taking advantage of this property of the total ozone measurements to derive tropical tropospheric ozone column amounts. The first of these combined the TOMS measurements with concurrent measurements from the SAGE (Stratospheric Aerosol and Gas Experiment) occultation measurements of the stratospheric amount. Subsequently, several techniques have been developed that use the assumption of constant stratospheric ozone around a circle of constant latitude on a given day. The amount of that stratospheric ozone can be estimated by using the difference between the tropospheric ozone amount measured by a sonde and the concurrent TOMS total ozone measurement. Alternatively, the stratospheric ozone amount can be estimated directly from TOMS measurements above the location of the highest clouds. That amount can be subtracted from the total ozone, yielding a tropical map of column tropospheric ozone. Application of these techniques for deriving tropical tropospheric ozone gives maps showing the ozone SATELLITE REMOTE SENSING / Water Vapor generated by the products of biomass burning. The ozone development and transport can be seen far downwind from the burning source. Summary and Future of TOMS Measurements We now have more than 20 years’ worth of global total ozone data from TOMS instruments and the related SBUV instruments. The Earth Probe TOMS is the last in the scheduled series. Future global total ozone measurements will be made by a continuing series of SBUV/2 instruments on the NOAA polar-orbiting satellites and by a new generation of ozonemapping instruments. The new-generation instruments will use charge-coupled-device arrays to image the Earth at a large number of wavelengths. These include the Global Ozone Measurement Experiment (GOME), its successor GOME II, the SCHIAMACHY instrument on Envisat, all launched by the European Space Agency, and the Ozone Measuring Instrument (OMI) on the EOS-Aura satellite. These instruments will transition into the Ozone Mapper Profiler Suite (OMPS) scheduled to fly on the NOAA/NASA/DOD National Polar Orbiting Environmental Satellite System (NPOESS) beginning in about 2010. See also Observations for Chemistry (In Situ): Ozone Sondes. Observations for Chemistry (Remote Sensing): 2005 IR/FIR; Microwave. Stratospheric Chemistry and Composition: Overview. Further Reading Dave JV and Mateer CL (1967) A preliminary study on the possibility of estimating total atmospheric ozone from satellite measurements. J. Atmos. Sci. 24: 414–427. Farman JC, Gardiner BG and Shanklin JD (1985) Large losses of total ozone in Antarctica reveal seasonal ClOx/ NOx interaction. Nature 315: 207–210. Hilsenrath E and Schlesinger BM (1981) Total ozone seasonal and interannual variations derived from the 7 year Nimbus-4 BUV data set. J. Geophys. Res. 86: 2086–2096. McPeters RD, Bhartia PK, Krueger AJ and Herman JR Earth Probe Total Ozone Mapping Spectrometer (TOMS) Data Products User’s Guide, NASA Technical Publication 1998–206895, National Aeronautics and Space Administration, Goddard Space Flight Center, Greenbelt, MD, 20771, 1998. Also available in PDF format at http://toms.gsfc.nasa.gov/eptoms/ epsat.html. McPeters RD and Labow GJ (1996) An assessment of the accuracy of 14.5 years of Nimbus 7 TOMS version 7 ozone data by comparison with the Dobson network. Geophys. Res. Lett. 23: 3695–3698. Singer SF and Wentworth RC (1957) A method for the determination of the vertical ozone distribution from a satellite. J. Geophys. Res. 62: 299–308. Stolarski RS, Krueger AJ, Schoeberl MR, McPeters RD, Newman PA and Alpert JC (1986) Nimbus-7 satellite measurements of the springtime Antarctic ozone decrease. Nature 322: 808–811. Water Vapor J E Harries, Imperial College of Science, Technology and Medicine, London, UK Copyright 2003 Elsevier Science Ltd. All Rights Reserved. Introduction The measurement of water vapor from space now has an extensive history going back to 1978, when the NASA Nimbus 7 spacecraft was launched carrying two sensors, the Limb Infrared Monitor of the Stratosphere (LIMS), and the Stratospheric and Mesospheric Sounder (SAMS). Both carried channels at 6 mm wavelength in the infrared, to detect thermal emissions from the atmosphere as the spacecraft orbited the Earth in a polar orbit. Shortly afterwards, a new generation of operational meteorological sounders was launched by NOAA, also in the USA, carrying the TIROS Operational Vertical Sounder (TOVS) package; part of TOVS was the High Resolution Infrared Sounder (HIRS), which made measurements of water vapor in the troposphere. Since that time satellite instruments operating in the infrared, the visible, and the microwave regions of the electromagnetic spectrum have operated in space, and a long series of measurements in both troposphere and stratosphere have been made. This article reviews these measurements. 2006 SATELLITE REMOTE SENSING / Water Vapor Principles and Overview In this section we discuss some of the general principles involved in remote sensing of atmospheric constituents, before we present some specific examples of satellite missions which have measured atmospheric water vapor. Remote sensing operates by detection of electromagnetic (e.m.) radiation from the Earth by an instrument on an orbiting spacecraft. There are two principal sources of e.m. radiation that have been most widely used in measurements of water vapor to date. These are the thermal emission from water molecules, and absorption of the visible/near ultraviolet radiation from the Sun. There are other techniques that could be mentioned, such as measurement of solar scattering, or the use of artificial radiation sources, such as lasers, but since these have not been used very widely for water vapor measurements, we will not consider them further in this work. Thermal Emission The Planck radiation law describes the maximum thermal radiation that can be emitted as a function of wavelength from any object at a temperature, T. This law shows that the intensity of radiation is a smooth curve with a single maximum at a wavelength in the infrared, lm , defined by Wien’s displacement law ½1 where the wavelength is expressed in microns, mm. Therefore, as an example, at a temperature of T ¼ 250 K (which is a typical temperature in the midtroposphere), the maximum intensity occurs at about 11.6 mm. In practice, the intensity of radiation is also dependent on the optical thickness of the atmospheric path under consideration: this, in turn, depends on the density or concentration of the absorbing molecules along that path. If this is written in terms of the Planck function, Bðl; TÞ, and the density of water vapor as a function of height, rw ðzÞ, then Iðl; yÞ ¼ Z z Absorption In the absorption case, the controlling equation, say for the case of looking at the Sun through the atmosphere at the limb of the atmosphere, is Iðl; yÞ ¼ I0 ðlÞ tðzg ; zt Þ Emission or Absorption lm T ¼ 2898 mm K measurements, then the dependence of Iðl; yÞ on tðz; zt Þ and rw ðzÞ allows density to be determined. zt Bðl; TðzÞÞ tðz; zt Þ al rw ðzÞ f ðyÞ dz ½2 where al is the absorption coefficient expressing the strength of electromagnetic coupling to theR radiation zt field at wavelength l, tðz; zt Þ ¼ exp ð z al rw ðzÞ f ðyÞdzÞ is the transmittance from z to zt (the top of the atmosphere), and f ðyÞ allows for non-vertical transmission paths ð¼ 1= cos y for yo  60 Þ. If the temperature of the atmosphere is known from other ½3 where zg is the grazing height, or the minimum height of the solar beam as it traverses the atmosphere, and tðzg ; zt Þ represents the transmittance of the path from the edge of the atmosphere on the sunward side, via the grazing height, and on to the edge of the atmosphere on the satellite side. Knowing the extraatmospheric solar intensity, I0 ðlÞ, allows the transmittance and therefore the density to be determined. The principal differences from emission sensing are that absorption does not depend (at least to first order) on atmospheric temperature, and that emission measurements may be made at any time of day or night, whereas solar-dependent methods can obviously be applied only when the Sun is in the right location. Limb or Nadir Sounding We have already indicated that the direction in which we view the atmosphere is important. There are two principal techniques in use. In the first, nadir sounding, widely used for sounding the lower atmosphere or troposphere, the satellite sensor is directed towards the nadir, i.e., downwards from the spacecraft. This is used most, often in emission mode, and the upwelling thermal radiation from the atmosphere is observed. It proves possible to detect emission from broad but distinguishable layers of the atmosphere because the term known as the ‘weighting function’, WðzÞ, defined as WðzÞ ¼ dt ¼ tðz; zt Þ al rw ðzÞ f ðyÞdz dz ½4 displays a single peak at a height dependent on the values of the separate terms, which can be arranged to be at heights between the surface and the tropopause. Because of this property, the density of water vapor in the troposphere can be sounded as a function of height. The technique of nadir sounding using atmospheric emission is used widely in meteorological sounding, including the measurement of humidity. In limb sounding, the limb of the atmosphere, just above the horizon, is viewed with a sensor with a narrow field of view. Either emission or solar absorption may be employed. By geometrically limiting the field of view, the vertical profile of emission or SATELLITE REMOTE SENSING / Water Vapor absorption, and therefore the vertical profile of water vapor density, may be scanned. Choice of Wavelength In principle, any wavelength at which there is thermal emission or solar absorption, and at which the water vapor in the atmosphere is spectrally active, may be employed. In practice, thermal emission from the infrared bands or microwave bands of water vapor have been used most frequently for the detection of atmospheric humidity. Summary of Missions and Instruments We have space here to only review a relatively few examples of satellite remote sensing of water vapor. Table 1 lists some of the more successful attempts to do so, organized into the three categories of infrared emission, microwave emission, and infrared and visible solar absorption (occultation). Examples of Satellite Experiments to Measure Water Vapor in the Atmosphere Infrared Thermal Emission Measurements Troposphere The principal satellite instrument for measuring tropospheric water vapor over the past few decades has been the High Resolution Infrared Sounder (HIRS), one of the suite of instruments making up the TIROS Operational Vertical Sounder (TOVS) package. ‘TIROS’ in this nested acronym stands for the Television and Infra Red Operational Sounder, the original name for the NOAA weather monitoring system. HIRS is an infrared sounder, viewing the upwelling radiation from the atmosphere 2007 below the spacecraft. HIRS has 20 channels, formed using interference pass-band filters, throughout the thermal infrared, including 3 that measure around 6.3 mm wavelength in the v2 emission band of water vapor. The ‘footprint’ or spatial resolution of HIRS is about 25 km in the nadir, and about 40 km at each end of the sideways scan that it uses to maximize coverage. Table 2 gives some of the important parameters of channel 12 of HIRS, which detects upper-tropospheric water vapor. The upwelling IR radiation comes from a restricted range of altitudes, defined by the strength of water vapor absorption at each wavelength. This idea is captured by the definition of the weighting function, WðzÞ , introduced above, which measures the relative amount of radiant energy reaching the spacecraft flying above the atmosphere from each layer within the atmosphere. WðzÞ is actually equal to the vertical derivative of transmittance at the wavelength in question, as shown in eqn [4]. An example of the typical shape of the HIRS 12 weighting function is shown in Figure 1. At wavelengths where the spectral absorption coefficient is larger, the curve shown is higher in the atmosphere, and vice versa for a wavelength at which the absorption coefficient is smaller. In order to develop a qualitative understanding of the shape of the weighting function, and how its height depends on absorption coefficient, consider the following. From the highest layers of the atmosphere, the density of water vapor is very low, so that the emission signal reaching our spacecraft instrument is very low from these layers. If we imagine moving deeper into the atmosphere, the emission signal seen at the spacecraft increases, as the density of water vapor increases. At some point, defined by the magnitude of the absorption coefficient at the wavelength in Table 1 Satellite instruments for measurements of water vapor: a sample Instrument Latitude Period of assessment Vertical range 1. IR emission TOVS (nadir sounder) LIMS (limb sounder) global 641 S–841 N 1979–1999 (daily) Oct. 1978–May 1979 (daily) 200 hPa to 500 hPa 1 hPa to 100 hPa 2. Microwave emission (limb) MLS stratospheric MLS UTH 341 N–801 S or 341 S–801 N See MLS stratospheric 0.01 hPa to 100 hPa 147, 215, 316, 464 hPa Near-global range of latitudes per Shuttle mission Sep. 1991–Apr. 1993 Sep. 1991–Sep. 1994; intermittent thereafter Mar/Apr. 1992, Apr. 1993, Nov. 1994 701 S–701 N 571 N–731 N, 641 S–881 S 541 N–711 N, 631 S–881 S Oct. 1991–Sep. 1999 Nov. 1996–Jun. 1997 Mar. 1998–Sep. 1999 0.01 hPa to 200 hPa 0.1 hPa to cloud tops 3 hPa to cloud tops MAS 3. Solar occultation (limb) HALOE ILAS POAM III (solar occultation) 0.01 hPa to 50 hPa 2008 SATELLITE REMOTE SENSING / Water Vapor Table 2 Characteristics of the HIRS channel 12 Parameter Description Method Measure upwelling clear-sky IR emission near 6.3 mm. Accuracy 1.11 K (5 observations per month); 0.11 K (100 observations per month); 0.1 K (global/interannual). Precision 0.01 K. Time/space resolution Monthly averages; 2.51 latitude and longitude. Altitude range 200–500 hPa (depends on water vapor amount). Calibration Onboard black body and cold space view. question, another effect begins to come into play: because the emitting layers are now overlain by quite a lot of absorbing water vapor, the IR radiation emitted from these layers towards space is actually partly absorbed before it emerges from the top of the atmosphere. Moving deeper still into the atmosphere means that this reabsorption effect increases, until, for the deepest layers (depending on the absorption coefficient, and therefore the wavelength), the energy actually reaching the top of the atmosphere can fall to near zero. Hence the general shape of the weighting function. If the wavelength is changed to a more absorbing one, the height of the peak in WðzÞ rises, and vice versa. In this way, though with modest vertical resolution, the water vapor density of the troposphere can be mapped in three dimensions as the satellite orbits over the greater part of the globe. Data from HIRS/TOVS have been available since 1979, with minimal gaps, and provide an important source of information about the distribution and variability of water vapor in the troposphere, used by scientists interested in both short-term weather and long-term climate. Figure 2 illustrates some data from HIRS. An instrument called MOPITT (Measurements Of Pollution In The Troposphere) was launched on board the Terra satellite in December 1999, to provide measurements of CO in the troposphere, with relatively coarse vertical resolution, but with good horizontal sampling; these allow maps of CO distribution to be produced, and MOPITT techniques are also Wet tropical profile Dry tropical profile 0 PW1000−500 = 19.927 kg m−2 PW500−300 = 0.043 kg m−2 PW300−100 = 0.005 kg m−2 PW1000−500 = 38.198 kg m−2 PW500−300 = 1.534 kg m−2 PW300−100 = 0.129 kg m−2 27.75 21.19 17.46 13.71 200 10.97 8.72 6.31 600 Height (km) Pressure (hPa) 400 3.75 SSM/T2 channel 2 HIRS channel 12 GMS-5 channel 4 800 1.53 1000 0.00 0.0 (A) 0.2 0.4 0.6 0.8 Normalized weighting function 1.0 0.0 0.2 0.4 0.6 0.8 (B) Normalized weighting function 1.0 Figure 1 Examples of weighting functions for HIRS 12 in the TOVS package. The HIRS 12 weighting function for two different profiles of water vapor, expressed as the precipitable water vapor in the layers 1000–500, 500–300, and 300–100 hPa, are shown, compared with weighting functions for two other instruments, SSM/T2, and GMS-5, a geostationary instrument. SATELLITE REMOTE SENSING / Water Vapor UTH DJF (1980−1997) 10 21 32 43 54 UTH MAM (1980−1997) 65 76 10 21 32 % 21 32 43 43 54 65 76 65 76 % UTH JJA (1980−1997) 10 2009 54 UTH SON (1980−1997) 65 76 % 10 21 32 43 54 % Figure 2 Measurements of upper-tropospheric humidity (UTH) in the layer 500–200 hPa measured by HIRS, expressed as maps for the four seasons DJF, MAM, JJA, and SON, averaged over the years 1980–1997. usable in principle to measure water vapor. In addition, in 2003 the EOS Aura satellite will carry the TES (Tropospheric Emission Spectrometer) instrument, which will make a variety of very valuable measurements of tropospheric constituents, including water vapor, with excellent spatial and temporal resolution. TES is based on the pedigree established by the Fourier transform experiment called ATMOS, which flew on the Space Shuttle, to measure the highresolution spectrum and composition of the stratosphere. Table 3 shows some of the basic characteristics of TES. Stratosphere In many ways, the remote sounding of the stratosphere is simpler than that of the troposphere. This is because in the stratosphere there are Table 3 TES characteristics Maximum sampling time of 16 s with a signal-to-noise ratio of up to 600:1 Limb mode: altitude coverage 5 0–34 km Nadir and limb viewing (fully targetable) Spectral region: 3.2 to 15.4 mm Swath: 5.3  8.5 km Spatial resolution: 0.53  5.3 km Mass: 385 kg (allocation) Power: 334 W (allocation) Data rate: 6.2 Mbps (peak); 4.9 Mbps (average) Physical size: 140  130  135 cm (stowed); 304  130  135 cm (deployed) virtually no clouds to interrupt the line of sight and complicate the observed signal; the densities are lower, so that spectral lines are narrow and better separated than in the troposphere; and techniques like limb sounding may be used, in which the long path to the horizon maximizes the signal from trace molecules, and provides a cold background of space against which to make measurements. These advantages are considerable, and have led to the development of many stratospheric remote sounding techniques, while tropospheric sounding is still in its infancy. As an example, we take the Limb Infrared Monitor of the Stratosphere (LIMS), which flew on the Nimbus 7 satellite, and which operated from 25 October 1978 to 28 May 1979. LIMS was a cryogenically cooled broadband filter radiometer, which included a water vapor channel at 6.9 mm. The 7-month life was limited, by design, by the lifetime of the solid cryogens used. LIMS provided the first comprehensive global view of stratospheric water vapor. The new data were used, along with measurements of CH4 from the UK Stratospheric and Mesospheric Sounder (SAMS), to study the budget of water vapor and hydrogen in the stratosphere. Precision, or relative accuracy, varied from 710% for pressures greater than about 10 hPa to 15% between 5 and 10 hPa. Much was learned about the operation and characteristics of an advanced IR sounder during this experiment, experience which was valuable in developing later generations of instruments. 2010 SATELLITE REMOTE SENSING / Water Vapor The LIMS experiment has been followed by other IR sensors, employing a variety of techniques, too many to be reviewed fully here. The reader is referred to the survey contained in the book Earthwatch listed under Further Reading. Microwave Thermal Emission Measurements of the Stratosphere In the extreme far infrared, or microwave region, at wavelengths beyond about 1 mm or so, a fundamental change in techniques is necessary. Thermal radiation is still emitted by all objects, but the intensity is so weak that it may be detected with any certainty only by using the radio techniques of superheterodyne detection. Thus, systems involving mixers and local oscillators (LO) have been developed (the mixers/detectors are related to the old crystal set whiskers) and have been used to detect the sidebands formed by mixing of the incoming signal from the atmosphere with the LO. Such systems have been developed over many years to achieve the high sensitivity in brightness temperature that is necessary to detect emission lines from atmospheric gases such as H2O, O3 , CO, ClO, and others. The principal advantages of the microwave technique are that the spectral resolution of the system is extremely high, so that atmospheric line shapes can be resolved, while at these long wavelengths, scattering due to particles and droplets is small (cf. the l4 dependence of Rayleigh scattering on wavelength), so that clouds and dust are less of a problem to remote sounding. Problems can arise in the calibration of such systems, for example because standing waves are set up in apparatus which can change if a component is physically rotated or moved, and which can give rise to a varying background signal. Also, because of the long wavelength, diffraction limits the field of view that 0.01 70 Solar occultation, uses the Sun as a source of radiation and measures the change in signal as the Sun rises or sets behind the limb of the atmosphere. Solar occultation has been used with great success in a number of experiments. We note the valuable data produced by the Stratospheric Aerosol and Gas Experiment (SAGE). However, we use as an example the most highly successful stratospheric experiment perhaps of recent times, the Halogen Occultation Experiment (HALOE). This is a thermal IR solar occultation device with a number of spectral channels, which uses gas correlation techniques as well as broadband filters. In gas correlation spectroscopy, a sample of the gas to be detected is carried in a cell on board, and provides a natural filter to atmospheric radiation specifically from that gas. HALOE has worked now for just on 10 years, and has provided a quite unprecedented set of data on the stratosphere over this long period. Given 0.01 80 70 0.10 60 50 1.00 40 0.10 60 50 1.00 40 10.00 30 20 100.00 2 (A) Solar Occultation Measurements 3 4 5 6 7 H2O mixing ratio (ppmv) 10.00 30 20 Pressure (hPa) Approx Height (km) 80 can be achieved, unless very large antennae are used. Nevertheless, very useful measurements have been made, relatively uncontaminated by cloud or dust. The accuracy quoted for latest version of MLS water vapor measurements is about 70.3 ppmv at 20 km, 0.2 ppmv at 40 km, and about 0.5 ppmv at 70 km. The Microwave Limb Sounder (MLS) which flew on the Upper Atmosphere Research Satellite (UARS) demonstrated the true power of a limb sounding microwave radiometer for making global measurements for the first time. A more advanced instrument is being developed for the NASA EOS Aura mission, and is currently scheduled to be flown in 2003. Some examples of MLS data, in a comparison with HALOE measurements, are shown in Figure 3. 100.00 −20 0 20 40 (B) Percentage difference (MLS-HALOE) Figure 3 (A) Examples of midlatitude water vapor mixing ratios for the period Oct. 1991–Apr. 1993, measured by MLS (version 0104, solid lines) and HALOE (version 19, broken lines). In each case the darker line is the mean and the lighter lines the standard deviation. (B) Differences between MLS and HALOE: solid 5 mean difference; dotted 5 r.m.s. difference; dashed 5 absolute mean difference; dot–dashed 5 combined instrumental uncertainties. SATELLITE REMOTE SENSING / Water Vapor 90 beta sunrise sunset 60 Latitude 30 0 30 60 90 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec 1995 Figure 4 HALOE sampling pattern for the year of 1995. Circles mark the tangent points for sunrise observations, crosses those for sunset observations. The tracks are due to a combination of orbit, season, and the reversal of the spacecraft for thermal balance reasons. that more and more scientific interest is coming to be centered on long-term changes in our climate, the continued operation of this extremely valuable instrument is most important. 2011 Among the advantages of this experiment are the well-known ones due to limbsounding (see earlier). In addition, however, the solar occultation method allows an absolute calibration against the Sun outside the atmosphere at each limb scan. This has proven an extremely important advantage when the data have come to be used for long-term trend and variability studies. The main disadvantage of the solar occultation technique is that since the measurement is made only during local sunrise and sunset, only two measurements are made per orbit: for a 15 orbit day, that is 30 observations per day, with a quite widely spaced horizontal sampling (see Figure 4). This should be compared with the closely spaced, uniform sampling possible from a nadir sounder like HIRS. HALOE measures not only H2O but also CH4 , O3 , NO, NO2 , and temperature. The combination of water vapor and methane measured simultaneously by one sensor has been exploited by a number of groups to study the hydrogen budget and water vapor trends of the stratosphere: trends of the order of 50–150 ppbv year  1 occurred during the first half of the 1990s, less in the second half. Such changes produce significant change in the water vapor amount in the stratosphere, which may influence the radiative balance due to this species. Figure 5 shows an estimate from HALOE data of the trends over the periods January 1992–December 1996 and January 1992–April 1999, demonstrating the change in trends detected over these 0.1 Jan 1992−Dec 1996 Jan 1992−Apr 1999 Pressure (hPa) 1.0 10.0 100.0 −150 −100 −50 0 50 100 150 Trend (ppbv y−1) Figure 5 Trends in stratospheric water vapor at levels between 120 and 0.15 hPa for two periods, Jan. 1992–Dec. 1996 (dotted) and Jan. 1992–Apr. 1999 (solid). Bars show standard deviation at each level. 2012 SATELLITE REMOTE SENSING / Wind, Middle Atmosphere two periods. These trends are still not well understood, but it has been shown that they have a significant effect on the Earth’s radiative energy balance. See also Climate: Overview. Global Change: Upper Atmospheric Change. Methane. Observations for Chemistry (In Situ): Gas Chromatography; Resonance Fluorescence; Water Vapor Sondes. Observations for Chemistry (Remote Sensing): IR/FIR; Microwave. Satellite Remote Sensing: Water Vapor. Stratospheric Chemistry and Composition: Hydrogen Budget. Stratospheric Water Vapor. Further Reading For more comprehensive information about missions, instruments and data, the reader is particularly directed towards the following two publications: Harries JE (1995) Earthwatch: The Climate from Space. Wiley-Praxis. SPARC Assessment of Upper Tropospheric and Stratospheric Water Vapour (2000) World Climate Research Programme Report No. 113, World Meteorological Organisation, Geneva. (The author is particularly indebted to the authors of this report, which has provided very valuable background in writing this article.) Also, the NASA web site gives details of many instruments and missions: http://www.earth.nasa.gov/ Wind, Middle Atmosphere P B Hays and W R Skinner, University of Michigan, Ann Arbor, MI, USA Copyright 2003 Elsevier Science Ltd. All Rights Reserved. Historical Perspective The middle atmosphere is that portion of our atmosphere above the troposphere where we live and below the high altitudes where the atmosphere is exposed to the extreme ultraviolet radiation from the Sun. Motion of the air in this middle region of the atmosphere has traditionally been difficult to observe owing to its remoteness. Historically, what we understand about the dynamics of the middle atmosphere has been based on theoretical prediction. It is known that this region is strongly influenced by the absorption of sunlight by ozone and by the upward propagation of wave energy generated by heating of the Earth’s surface and by weather systems in the troposphere. However, direct observations of the motions resulting from these sources of energy were extremely difficult to obtain. In the period following World War II, direct observations of winds in these high regions were first obtained by tracking smoke and vapor trails generated from payloads carried on small sounding rockets. These observations were extended by using other rocket-borne means of observing the winds, including timing the arrival on the ground of sound bursts generated by a series of grenades, tracking lightweight inflated Mylar spheres by radar as they descended through the moving atmosphere, and observing more exotic vapor trails from several ground stations. However, the most important information came with the introduction of high-frequency sounding radar systems that observed the motion of irregularities in the upper regions of the atmosphere. These important observations allowed the time variation of the winds over a ground station to be followed throughout the day and over long time periods to reveal the diurnal and seasonal behavior of the middle atmosphere. These mixed techniques provided a glimpse of what the atmosphere was doing in a series of places on Earth, but many mysteries remained concerning the global behavior of the atmosphere. Satellite Observations In the early 1980s a series of satellites, called the Dynamics Explorers, were launched to examine the behavior of the thermosphere by observing a large number of the variables that defined the state of that region of the atmosphere. One of these satellites contained an instrument that was designed to measure the winds in the upper atmosphere by determining the Doppler shift of isolated spectral features in the spectrum of light emitted from the thermosphere. This instrument, called the Fabry–Perot Interferometer, was able to detect the slight shift in wavelength of light caused by the Doppler effect when the lightemitting gas was in motion relative to the spacecraft. This instrument was very small and relatively unsophisticated but was able to define on a global scale the relatively large dynamic motions of the thermosphere. The Dynamics Explorer Fabry–Perot Interferometer illustrated that the winds could be observed remotely from a satellite by using an instrument that could detect the Doppler shift of sharp spectral features in the spectrum of light emitted or scattered SATELLITE REMOTE SENSING / Wind, Middle Atmosphere 2013 by the atmosphere. In the early 1980s when the Dynamics Explorers were still in orbit, construction of a new satellite was begun. This satellite, called the Upper Atmosphere Research Satellite (UARS), was designed to observe the entire middle atmosphere with a multiplicity of instruments in order to define the dynamic and chemical state of this important region. Again it was recognized that the dynamics of the middle atmosphere needed to be defined if the chemistry and thermodynamic state of the middle atmosphere were to be understood. Two instruments were selected for this purpose, one called the High Resolution Doppler Imager (HRDI) and the other the Wind Imaging Interferometer (WINDII). HRDI was designed to remotely sense the winds in the middle atmosphere, and WINDII to sense the winds in the thermosphere and upper reaches of the middle atmosphere. These two instruments have vastly increased the knowledge of the dynamic state of the entire upper regions of the atmosphere above the tropopause. The instruments are similar in concept, but quite different in detail. Thus, a discussion of their true natures will be given to introduce the observations that have been made by these two unique sensors. High Resolution Doppler Imager (HRDI) The HRDI has at its heart a very sensitive Fabry–Perot interferometer to detect the small wavelength shifts that are caused by the Doppler shift of absorption and emission lines in the Earth’s spectrum. HRDI must look at both absorption and emission lines, and thus is somewhat more complex than WINDII, which views only isolated emission lines. HRDI consists of three components: a telescope for viewing the atmosphere, an interferometer, and a microprocessor and electronics. Light from the atmosphere is collected by a fully gimbaled telescope that allows observations on either side of the spacecraft to be obtained in rapid sequence. The HRDI telescope is mounted on the bottom of the UARS spacecraft and is able to view most of the hemisphere below the satellite. This allows the instrument to examine the horizon of the atmosphere all the way around the UARS spacecraft. An illustration of the HRDI telescope is shown in Figure 1. During the launch phase of the mission the aperture is covered by the caging device. After the satellite reaches orbit, the cover is pulled away and the telescope zenith and azimuth drives are actuated to view points on the atmospheric horizon. Light enters the interferometer either from the telescope or from calibration sources. The light beam is expanded and passes through broadband filters contained in one of two filter wheels. The beam is then Zenith drive Telescope Yoke Azimuth drive Cover/caging device Adapter Optical interface Figure 1 A schematic illustration of the HRDI telescope. further expanded and enters the spectrally sensitive interferometer, which contains three progressively higher-resolution spectral filters called etalons. The spectrally dispersed beam is focused onto a multichannel concentric ring image plane detector that spatially scans the ring-shaped wavelength pattern. A schematic illustration of the interferometer is shown in Figure 2. This diagram shows the light rod that transmits the light from the telescope to the interferometer, the filter wheels, followed by the three spectrally sensitive etalons (HRE, MRE, LRE), and finally the image plane detector (IPD) where the spectrum is examined. Doppler shifts of monochromatic emission lines appear on the detector as changes in the diameter of the ring of light that is imaged by the etalons onto the detector. Figure 3 shows the signal observed by an emission line on the image plane detector with and without a Doppler shift. Figure 3 clearly demonstrates that the shift has a very small effect at the line center: the most notable change is seen on the sides of the line. The Doppler shift causes the signal on one side of the line to increase and the signal on the other side to decrease. Careful measurement of this signal difference allows the Doppler shift of the line to be determined. The design of HRDI allows for versatile programming of operational modes, which are stored in the computer’s memory as lookup tables. A single mode is described to illustrate the sequence of operations used by the instrument to determine the wind as the satellite moves along its orbit. A stratospheric daytime wind mode is described in which the atmosphere from 10 to 40 km in altitude is observed with a horizontal resolution of about 500 km along the orbit. The telescope initially looks forward at an azimuth of about 451 from the spacecraft velocity vector, and vertically scans the atmosphere in 2.5 km steps by pointing at the horizon. After this sequence is performed, the telescope slews to look backward, at an azimuth of 1351 to the satellite velocity vector. The altitude scan is repeated to provide the second 2014 SATELLITE REMOTE SENSING / Wind, Middle Atmosphere Figure 2 A schematic illustration of the HRDI interferometer. HRE, MRE, LRE are high, medium-, and low-resolution etalons; IPD is the image plane detector. component of the wind at the same series of altitudes as done in the forward scan. This cycle is continued until a new mode of operation is commanded by the instrument’s computer. A diagram of these operations is shown in Figure 4. Lines of Doppler shift are observed in the two orthogonal directions and the selected altitudes and converted to line-of-sight winds. These data are then subjected to an inversion process that yields the horizontal component of wind velocity at a set of fixed altitudes at specified points along the satellite orbit. The final step in processing the information is to interpolate the data onto a common UARS temporal and spatial grid for use by the scientific community. The High Resolution Doppler Imager has been in operation since early November 1991 and is still providing the scientific community with detailed observations of the winds in the middle atmosphere as at this writing early in 2002. A diagram showing the latitude coverage by HRDI over the first five years of its history is shown in Figure 5, where the vertical bars show where observation were made. In most cases, each observation set consists of either stratospheric or mesospheric altitude profiles of wind vectors. Wind Imaging Interferometer (WINDII) The WINDII senses temperatures and winds in the upper mesosphere and lower thermosphere by meas- uring both Doppler widths and shifts of isolated spectral lines emitted by airglow and auroras in the visible and near-infrared portion of the spectrum. The instrument views the atmospheric line simultaneously in two directions, 451 and 1351 from the velocity vector and, owing to the spacecraft motion, the same atmospheric region is viewed by each with a separation of a few minutes. This provides both horizontal components of the neutral wind. An imaging detector provides simultaneous measurements of temperature and wind profiles over the instrument’s entire altitude range. The instrument consists essentially of a chargecoupled device (CCD) camera viewing the limb of the Earth through a field-widened Michelson interferometer. It takes four to eight images with the interferometer optical path difference changed by 1/4 or 1/8 of a wavelength between images. It views a number of emission lines in order to retrieve spectra from 70 to 315 km altitude. These emissions are summarized in Table 1. The WINDII instrument began operation in the fall of 1991 and is still capable of taking measurements in early 2002. The basic approach of a Michelson interferometer is to split an optical beam into two beams, insert a phase delay between them, recombine them, and measure the resulting signal. The phase delay is introduced by forcing the beams to traverse separate paths of unequal length. The signal as a function of the path length difference is the Fourier transform of the input spectrum. For a good representation of a general input Signal level (arbitrary) SATELLITE REMOTE SENSING / Wind, Middle Atmosphere 2015 110 100 90 80 70 Unshifted line Shifted line 60 50 40 30 20 10 0 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 Channel (A) 2.0 Difference 1.0 0.5 0 _ 0.5 _ 1.0 _ 2.0 (B) 0 4 8 20 12 16 Channel number 24 28 32 Figure 3 Illustration of the effect of Doppler shifts on the spectrum observed by HRDI. (A) Spectra with and without a small Doppler shift. (B) The difference between the two spectra. The center and wings are unaffected, but the sides show a dramatic change. spectrum to be obtained, the path separations must vary by a significant distance, up to several centimeters. However, a significant simplification is possible for the type of observations conducted by WINDII. The spectral region examined is very small, containing at most two or three emission lines of a known spectral shape and a small continuum background that can be assumed to be spectrally flat. For the fundamental quantities of the lines (position, which provides the wind; width, which provides the temperature; and brightness, which provides the volume emission rate) and the background to be obtained, the amount of movement between the two paths need only be about a wavelength of light as long as the paths have a mean separation of about 4.5 cm. The WINDII optics are shown in Figure 6. The Michelson interferometer optics consist of a cemented glass hexagonal beam splitter, a glass block with a deposited mirror in one arm of the interferometer, and a glass block combined with an air gap and a piezoelectrically driven mirror in the other arm. The mirror position is controlled through capacitive sensing to provide stability and accurate step size. The CCD camera consists of a fast camera lens and an RCA 501E CCD cooled to 501C. The imaging area has 320  256 pixels, with a corresponding storage area into which the image is shifted after the exposure. During readout, binning and windowing techniques select desired altitude ranges, and tailor the image to the available telemetry rate. Interference filters are mounted in a temperature-controlled filter wheel assembly to isolate specific spectral lines. The two telescope inputs provide two fixed views of the atmosphere at 451 and 1351 to the flight directions. They are designed to minimize stray light, to transform the field of view to the desired value, and to provide a suitable location for the beam combiner. The WINDII mechanical configuration is shown in Figure 7. The two orthogonal input beams enter through the two ports on the left in the outer baffle assembly (each 1 m w Sle 45 ° Warm side 585 km Vertical scan 2500 km 135° Cold side Satellite ground track Figure 4 Illustration of the telescope viewing during a normal science mode. Actual modes will look on just one side of the spacecraft or the other. Typical HRDI operations look on the warm side (toward the Sun) to maximize the signal and coverage. 2016 SATELLITE REMOTE SENSING / Wind, Middle Atmosphere 90° N Latitude 60° N 30° N 0 30° S 60° S 90° S Oct. Jan. 1992 Apr. Jul. Oct. Jul. Oct. Jan. Apr. Jul. Oct. Jan. 1994 Oct. Jan. 1996 Apr. Jul. 1993 Apr. 90° N Latitude 60° N 30° N 0 30° S 60° S 90° S Apr. Jan. 1995 Apr. Jul. Figure 5 Latitude coverage of the HRDI instrument from the beginning of the operations in October 1992 through July 1996. A vertical line is drawn covering the latitude extent for each day the instrument was switched on. long) and cross over before entering the inner baffle assembly. The two inputs are combined into a single beam that passes through the filter wheel and the rear telescope before entering the Michelson interferometer in its thermal enclosure. The optical train also contains the filter wheel, a mirror for calibration sources, and an aperture stopdown to provide lower scattered light levels for daytime viewing. The CCD camera is located immediately behind the Michelson interferometer. A separate calibration box contains spectral lamps, a tungsten lamp that acts as a continuum source, and a laser. An internal microprocessor controls all of the instrument functions including camera control, filter wheel control, thermal control, and control of all other mechanisms. Buffer memory exists for additional onboard processing of the image data before the data are sent to the telemetry stream. The instrument parameters are summarized in Table 2. The data undergo several transformations that convert the raw telemetry into profiles of geophysical quantities. The first reads the telemetry and generates a level one file that corrects for instrument effects such as dark count and background subtraction, and for observatory effects such as roll offsets and pixel location, and then converts the signal to intensity units (Rayleighs). The Rayleigh is a convenient unit of brightness that was invented by Lord Rayleigh and is a brightness of 106 photons per square centimeter per second. These data are then inverted into vertical profiles of emission rate, temperature, and horizontal wind components for both fields of view. These profiles from the two fields of view are then combined to obtain meridional and zonal wind profiles. This level 2 file is then mapped onto a specific grid (called level 3) that is common for all UARS instruments. Satellite Observation of Middle Atmosphere Winds During the lifetime of the Upper Atmosphere Research Satellite, the HRDI and the WINDII have revolutionized our knowledge of the winds in the middle Table 1 Emission characteristics of spectral features observed by WINDII Emission Wavelength (nm) Filter bandwidth FWHM (nm) Height range (km) O(1S) green line O(1D) red line OH (8–3) band O2 atmospheric (0–0) band O1(2D) 557.7 630.0 730.0 762.0 732.0 1.6 1.6 1.2 0.09 1.4 80–110, 150–300 150–300 80–110 80–110 200–300 SATELLITE REMOTE SENSING / Wind, Middle Atmosphere 2017 Front telescope Field stop Filter wheel Michelson interferometer Field combiner Rear telescaope Collector optics Baffle CCD Aperture Limb pointing mirror Figure 6 A schematic diagram of the WINDII optical system. atmosphere. The information that has been provided by the two instruments can be divided into classes by considering the short-term, or daily, variations, and the longer-term (seasonal and interannual) variations of the winds on a global basis. Mesosphere The instantaneous picture of the winds in the upper regions of the middle atmosphere exhibits the most dramatic variations. In the upper mesosphere, the energy being propagated upward from the diurnal oscillation of solar heating at the surface and in the lower regions of the atmosphere results in a very dramatic diurnal solar tide that grows exponentially with altitude. The first maps of the wind fields taken by HRDI shortly after launch, shown here in Figure 8, exhibit extreme oscillations in the 80–100 km altitude region. The equatorial winds at 80 km observed by the satellite as it passed through local noon are directed Michelson interferometer and thermal enclosure Thermal radiator plate Rear telescope CCD camera Filter wheel Thermal interface tube Split field telescope Inner baffle assembly Outer thermal enclosure Attachment points to UARS (1 of 3) Electrical and fiberoptics cable Outer baffle assembly Electrical unit Figure 7 The WINDII instrument configuration. 2018 SATELLITE REMOTE SENSING / Wind, Middle Atmosphere Table 2 WINDII instrument parameter summary Parameter Description Type of instrument Geophysical parameters determined Field-widened Michelson interferometer Atmospheric temperature, horizontal wind vector, volume emission rate 550–780 nm Limb view at 451 and 1351 to spacecraft velocity vector, maximum latitude viewed in 741 61,corresponding to 70–315 km altitude 4 km at limb (nominal) 20 km at limb B8 s 1.78  10  7 sr pixel  1 32.7 cm2 (night) 3.60 cm2 (day) 30 electrons s  1 pixel  1 100 electrons 75 electrons count  1 B5  10  3 counts s  1 Rayleigh  1 (night) B5  10  4 counts s  1 Rayleigh  1 (day) Wavelength coverage Viewing geometry Vertical field of view Vertical resolution Horizontal resolution Time to perform single measurement Field of view Aperture area Detector thermal noise Read noise Digitization Responsivity strongly toward the west with magnitudes in excess of 75 m s  1. At the 90 km level, the situation is very different, with little or no wind seen near the Equator. As the satellite orbital plane rotates through local time, the pattern of winds shows strong local time variations, reversing every 12 h in a sinusoidal Wind field at 80.0 km Wind field at 85.0 km Latitude 50° N 0° 50° S 75 m s−1 (A) 75 m s−1 (B) Wind field at 90.0 km Wind field at 95.0 km Latitude 50° N 0° 50° S 75 m s−1 75 m s−1 180° W 135° W 90° W (C) 45° W 0° Longitude 45° E 90° E 135° E 180° E 180° W 135° W 90° W (D) 45° W 0° 45° E 90° E 135° E 180° E Longitude Figure 8 Vector wind fields for 25 March 1992 (day 92085) showing cuts of the data at (A) 80, (B) 85, (C) 90, and (D) 95 km.There are about 15 orbits every day and all the orbits follow almost the same local time path. The alternating converging/diverging flows at lower latitudes are a clear signature of the diurnal migrating tide. SATELLITE REMOTE SENSING / Wind, Middle Atmosphere 2019 Major long-term variations of the tidal oscillations have been discovered over the past eight years with a clear pattern of large amplitudes in the tides at the equinoxes and minimal amplitudes at times of the solstice. This is further illustrated in Figure 10, which shows the seasonal variation of the tidal amplitude for 1992 through 1996. Notice here that the tidal amplitude reaches a maximum in the spring and fall, while reaching minimum values in the solstices. It is also interesting that over longer periods the pattern of tidal amplitude variations exhibits large interannual variability that has recently been shown to be correlated with the semiannual oscillation that is seen in the stratosphere. local time pattern. This pattern of strong altitude variations at a particular location and local time is characteristic of a global gravity wave or tidal oscillations of the atmosphere and was predicted theoretically much earlier. The winds increase in magnitude with altitude in an inverse relation to the square root of the density (see Figure 9) until dissipation starts to damp the waves near the 100 km level. Figure 9 shows the amplitude and phase of the diurnal tide for two seasons, with the spring having a much larger tide than the summer. In this figure a comparison between the HRDI observations and model simulations from the Global-Scale Wave Model (GSWM) is shown. April July 120 Altitude (km) 100 80 60 0 40 20 60 _ Amplitude (m s 1) 80 0 40 20 60 _ Amplitude (m s 1) April 80 July 120 Altitude (km) 100 80 60 _12 _6 0 Phase (h) 6 12 _12 _6 0 Phase (h) 6 12 Figure 9 Comparison between HRDI meridional wind diurnal tidal amplitudes and phases (solid line) with the GSWM results (dashed line) as a function of altitude at a latitude of 201 for the two months April and July. 2020 SATELLITE REMOTE SENSING / Wind, Middle Atmosphere Amplitude (m s_1) 100 80 60 40 20 (A) 0 Oct. Jan. Apr. Jul. Oct. Jan. Apr. Jul. Oct. Jan. Apr. Jul. Oct. Jan. 1992 1994 1993 1995 12 8 Phase (h) 4 0 _4 _8 _ 12 Jan Apr. Jan Apr. Apr Jul. Jan Apr. Jul Oct. Apr Jul. Jan Jul Oct. Apr Jul. Jul Oct. Oct Jan. Oct. Oct Jan. Oct Jan. Oct Jan. 1992 1994 1993 1995 (B) Figure 10 Daily estimates of the (1,1) diurnal component of the meridional wind obtained from HRDI data for an altitude of 95 km and a latitude of 201. Panel (A) shows the derived amplitudes and panel (B) the phases, defined as the local time of the maximum positive (northward) wind at a latitude of 201 N. The solid line is a 10-day running average, which serves to highlight the long-term variations. The upper regions of the middle atmosphere also oscillate on shorter and longer time scales. A semi-diurnal tide is observed in the middle and high latitudes, again showing regular seasonal variations, and significant interhemispheric asymmetries have been identified. Global oscillations with 3-day and 5-day periods are observed. These longer-period waves appear for very short times and appear to be generated by instabilities in the atmosphere. In addition to these short-term oscillations of the middle atmosphere, there are longer-term variations in the mean winds at altitude. The long-term variation in winds in the middle atmosphere is shown in Figure 11, where the mean zonal wind is presented as a function of altitude at the Equator, and as a function of latitude at a fixed altitude of 82.5 km. The background zonal wind observed by the HRDI has an interesting behavior exhibiting regular direction changes with the winds shifting from easterly to westerly in a very regular semiannual pattern in the altitude range from 70 to 90 km. Note that the easterly wind maxima occur at the equinoxes, while the weaker westerly maxima occur near the solstices. These oscillations in the zonal wind are called the mesospheric semiannual oscillation. The semiannual oscillation in the zonal wind field is in accord with the tidal amplitude variations that were noted earlier. In addition to the semiannual variation in both the zonal wind and the tides, there are longer-term modulations of the amplitudes of the mesospheric semiannual oscillation and of the tidal amplitudes that correlate with the quasi-biannual oscillation in the stratosphere. SATELLITE REMOTE SENSING / Wind, Middle Atmosphere 2021 100 Altitude (km) 80 60 40 20 Oc (A) 60° N 40° N Latitude 20° N 0° 20° S 40° S 60° S (B) Oct. Jan. Apr. Jul. Oct. Jan. Apr. Jul. Oct. Jan. Apr. Jul. Oct. Jan. 1992 1993 1994 1995 Figure 11 Background zonal wind observed by the HRDI (A) as a function of altitude and time at the Equator and (B) as a function of latitude and time at an altitude of 82.5 km.The contour levels are given every 10 m s  1. Stratosphere The HRDI on the Upper Atmosphere Research Satellite is the first instrument capable of performing direct global-scale measurement of the stratospheric wind field. When observing the winds in the stratosphere, HRDI detects the Doppler shift of absorption features caused when light is transmitted through the air in the stratosphere. The fact that HRDI can observe a global-scale wind field extending from about 15 to 40 km in altitude in a single day, coupled with the length of the data set, has facilitated the investigation of a range of temporal effects from shortterm (days) to long-term (years). Figure 12 is an example of a stratospheric wind map obtained on a single day. The measurements shown here are distributed along paths that are parallel to the satellite orbit tracks. To suppress noise generated in the inversion process, the HRDI wind analysis employs a sequential estimation technique. The operation of the sequential estimation procedure is such as to generate a vector for both the upward and downward limb scans and for the forward and backward line-of-sight measurements for approximately the same geographic location. Thus, clusters of four vectors are generated, the members of which are not independent of each other. Strong westerly flows are seen in the Northern (winter) Hemisphere, while in the Southern (summer) Hemisphere, the winds are weaker and predominantly easterly. The strong southward and northward flows over Canada and Siberia, respectively, are indicative of planetary wave activity. Figure 13 is a chronological sequence of HRDI daily stratospheric wind maps that document the breakup of the Antarctic polar vortex in the austral spring of 1993. Here the dramatic decay of the polar vortex is seen, the breakup of the cold polar air mass occurring 2022 SATELLITE REMOTE SENSING / Wind, Middle Atmosphere 50 m s_1 Figure 12 Global wind field obtained by HRDI on 15 February 1993 at an altitude of 25 km.Such maps can be obtained on a daily basis for altitudes from 10 to 40 km, at intervals of 2.5 km. as the Sun begins to heat the polar air in the polar spring. Longer time scales are represented by the time– altitude sections of longitudinally and monthly averaged zonal winds at the Equator as measured by HRDI and given by the UK Meteorological Office assimilation model, shown in Figure 14. Very close agreement between HRDI and the model are apparent. Both HRDI and the model reveal the well-known structure of the quasi-biennial oscillation (QBO) in the middle 8 Sep. 1993 31 Oct. 1993 50 m s 19 Nov. 1993 _1 29 Nov. 1993 Figure 13 Sequence of four HRDI Southern Hemisphere wind maps for an altitude of 25 km, documenting the breakup of the polar vortex during the Antarctic spring of 1993. These data, together with simultaneous constituent measurements obtained by other UARS instruments, provide important information on how the ozone hole is formed and how it ultimately contributes to global ozone depletion. SATELLITE REMOTE SENSING / Wind, Middle Atmosphere 2023 40 40 0 25 _ 20 20 _ 40 5 15 Jan. 1993 (A) Jul. Jul. Jan. 1994 40 Jan. 1995 40 5 30 20 _1 Altitude (km) 35 5 0 25 _ 20 20 15 (B) ms 30 _1 20 ms Altitude (km) 35 _ 40 Jan. 1993 Jul. Jan. 1994 Jul. Jan. 1995 Figure 14 Time–altitude sections of monthly zonally averaged zonal winds at the Equator (A) as measured by HRDI and (B) as predicted by the UK Meteorological Office. Contours are at 5 m s  1 intervals. stratosphere, with downward propagation of successive easterly and westerly flow regimes. Above 35 km the principal feature in both sets of results is the semiannual oscillation of the upper stratosphere. In the discussion of the mesosphere, a clear relation between the QBO and the mesospheric zonal winds and the tides was seen. Conclusion The pattern of winds observed in the middle atmosphere clearly shows that the atmosphere is a tightly coupled system. Both small-scale and global-scale waves, like the diurnal tides, are generated in the lower atmosphere and grow in amplitude as they propagate upward into the less-dense regions of the middle atmosphere, where they dissipate or break, releasing energy into the surroundings. The energy and momentum released by these waves cause the middle atmosphere to exhibit a behavior that is both unique and related to the sources of disturbances from below. Since both the sources of energy in the lower reaches of the atmosphere and the transmission of this energy upward are dependent on the season, we find that the middle atmosphere is an extremely interesting region of our atmosphere. In many ways the middle atmosphere is like the beach of a great ocean, responding to events both near and remote, but always a place of great fluctuation. This interesting complexity has been revealed by the direct observation of winds from space by the HRDI and WINDII instruments, which have been in orbit for over 10 years on the Upper Atmosphere Research Satellite. These instruments are pioneers and will eventually be replaced by newer and more sensitive detectors of atmospheric winds. The new instruments will be both passive, possibly using the infrared region of the spectrum, and active, using large lasers to stimulate the atmosphere in a much more controlled fashion for observations of the motions in a very small volume of space. This new generation of observations will increase the detail of what we know about the winds, but will never provide the degree of excitement provided by the first pioneering observations of winds from space. See also Middle Atmosphere: Gravity Waves; Polar Vortex; Quasi-Biennial Oscillation; Semiannual Oscillation; Zonal Mean Climatology. Further Reading Andrews DG, Holton JR and Leony CB (1987) Middle Atmosphere Dynamics. New York: Academic Press. Bern M and Wolf E (1975) Principles of Optics. Oxford: Pergamon. Hernandez G (1986) Fabry–Perot Interferometers. New York: Cambridge University Press. 2024 SATELLITES / Orbits Holton JR (1992) Introduction to Dynamic Meteorology. New York: Academic Press. Ortland DA (1995) A sequential estimation technique for recovering atmospheric data from orbiting satellites. In: Johnson RM and Killeen TL (eds) The Upper Mesosphere and Lower Thermosphere: A Review of Experiment and Theory, pp. 329–338. Washington, DC: American Geophysical Union. Shepherd GG (2002) Spectral Imaging of the Atmosphere. New York: Academic Press. Steel WH (1983) Interferometry, 2nd edn. New York: Cambridge University Press. Vaughan JM (1989) The Fabry–Perot Interferometer: History, Theory, Practice and Applications. Philadelphia: Adam Hilger. SATELLITES Contents Orbits Research (Atmospheric Science) Orbits S Q Kidder, Colorado State University, Fort Collins, CO, USA Copyright 2003 Elsevier Science Ltd. All Rights Reserved. Introduction To fully understand and use data from meteorological satellites, it is necessary to understand the orbits in which satellites are constrained to move and the geometry with which they view the Earth. This article begins with a review of basic physical principles, which reveal the shape of a satellite orbit and how to orient the orbital plane in space. This knowledge allows us to calculate the position of a satellite at any time. Orbit perturbations and their effects on satellite orbits are then discussed. Finally, the geometry of satellite tracking and Earth location of the measurements made from the satellites are explored. Newton’s Laws Isaac Newton discovered the basic principles that govern the motions of satellites and other heavenly bodies. Newton’s Laws of Motion 1. Every body will continue in its state of rest or of uniform motion in a straight line except insofar as it is compelled to change that state by an impressed force. 2. The rate of change of momentum is proportional to the impressed force and takes place in the line in which the force acts. 3. Action and reaction are equal and opposite. Since momentum is the product of the mass of a body and its velocity, Newton’s second law is the familiar eqn [1], where F is force, m is mass, a is acceleration, v is velocity, and t is time. F ¼ ma ¼ m dv dt ½1 In addition, Newton gave us the functional form of the force that determines satellite motion in the law of gravitation. Newton’s Law of Universal Gravitation The force of attraction between two point masses m1 and m2 separated by a distance r is given by eqn [2], where G is the Newtonian (or universal) gravitation constant (6.67259  10  11 N m2 kg  2). F¼ Gm1 m2 r2 ½2 Consider the simple circular orbit shown in Figure 1. Assuming that the Earth is a sphere, we can treat it as a point mass. The centripetal force required to keep the satellite in a circular orbit is mv2 =r, where v is the orbital velocity of the satellite. The force of gravity that supplies this centripetal force is Gme m=r2 , where me is the mass of the Earth (5.97370  1024 kg) and m SATELLITES / Orbits 2025 Therefore, the required radius for a geosynchronous orbit is about 42 164 km, or about 35 786 km above the Earth’s surface. r Satellite Keplerian Orbits Although a circular orbit is the goal for most meteorological satellites, in general satellites do not travel in perfect circles. The exact form of a satellite’s orbit may be derived from Newton’s laws of motion and the law of universal gravitation. The results of this derivation are neatly summarized in Kepler’s laws and in Kepler’s equation. F Earth Kepler’s Laws Figure 1 A circular satellite orbit. is the mass of the satellite. Equating the two forces gives eqn [3]. mv2 Gme m ¼ r2 r ½3 Division by m eliminates the mass of the satellite from the equation, which means that the orbit of a satellite is independent of its mass. The period of the satellite is the circumference of the orbit divided by the velocity (eqn [4]). 2pr ½4 T¼ v Substituting eqn [4] in eqn [3] gives eqn [5] for the period. 4p2 3 T2 ¼ r ½5 Gme A typical weather satellite orbits 833 km above the Earth’s surface. Since the equatorial radius of the Earth is 6378.137 km, the orbit radius is about 7211 km. Substituting in eqn [5] yields a period of about 6094 s or 102 min. As a second example, we calculate the radius required for a satellite in geosynchronous orbit, that is, an orbit in which the satellite has the same angular velocity as the Earth (7.29211510  5 rad s  1). The angular velocity of a satellite is given by eqn [6]. x¼ 2p T ½6 Substituting eqn [6] in eqn [5] gives eqn [7] for the radius. r3 ¼ Gme x2 ½7 Johannes Kepler died 12 years before Newton was born and, thus, did not have the advantage of Newton’s work. Kepler formulated his laws by analyzing data on the position of the planets. This task was complicated by the rotation of the Earth and the motion of the Earth about the Sun, which make planetary motions seem very complex. In modern form, Kepler’s laws may be stated as follows. 1. All planets travel in elliptical paths with the Sun at one focus. 2. The radius vector from the Sun to a planet sweeps out equal areas in equal times. 3. The ratio of the square of the period of revolution of a planet to the cube of its semimajor axis is the same for all planets revolving around the Sun. The same laws apply if we substitute satellite for planet and Earth for Sun. Equation [5] is a statement of Kepler’s third law for the special case of a circular orbit. Ellipse Geometry The parameters that are used to specify satellite orbits are based in part on geometric terminology. Figure 2 illustrates the geometry of an elliptical orbit. The point where the satellite most closely approaches the Earth is termed the perigee, or more generally the perifocus. The point where the satellite is farthest from the Earth is called the apogee or apofocus. The distance from the center of the ellipse to the perigee (or apogee) is the semimajor axis (denoted by the symbol a). The distance from the center of the ellipse to one focus (to the center of the Earth) divided by the semimajor axis is the eccentricity ðeÞ. For an ellipse, the eccentricity is a number between zero and 1 ð0oeo1Þ. A circle is an ellipse with zero eccentricity. 2026 SATELLITES / Orbits Satellite a l_2 r Earth Focus Apogee (apofocus)  a a a Perigee (perifocus) a Figure 2 Elliptical orbit geometry. The equation for the ellipse, that is, the path that the satellite follows, is given in polar coordinates with the center of the Earth as origin by eqn [8]. að1  e2 Þ r¼ 1 þ e cos y ½8 The angle y (see Figure 3) is the ‘true anomaly’ and is always measured counterclockwise (the direction of satellite motion) from the perigee. Kepler’s Equation A satellite in a circular orbit has uniform angular velocity. By Kepler’s second law, however, a satellite in an elliptical orbit cannot have uniform angular velocity; it must travel faster when it is closer to the Earth. The position of the satellite as a function of time can be found by applying Kepler’s equation Satellite M ¼ nðt  tp Þ ¼ M0 þ nðt  t0 Þ ¼ e  e sin e ½9 Here M is the mean anomaly, an angle that increases linearly in time at the rate n, called the mean motion constant, given by eqn [10]. rffiffiffiffiffiffiffiffiffiffi 2p Gme ¼ n¼ ½10 T a3 By definition, M is zero when the satellite is at perigee; therefore, tp is the time of perigeal passage. Time t0 is called the epoch time. M0 is called the mean anomaly true of epoch, that is, the mean anomaly at the epoch time t0 . The angle e is the eccentric anomaly. It is geometrically related to the true anomaly (Figure 3) through eqns [11a] and [11b]. cos y ¼ cos e  e 1  e cos e ½11a cos e ¼ cos y þ e 1 þ e cos y ½11b Given a, e, and tp (or M0 and t0 ), one can calculate r and y at any time t using eqns [8]–[11].  e Earth Elliptical orbit as eqn [9]. Perigee Circumscribed circle Figure 3 The geometric relationship between true anomaly (y) and eccentric anomaly (e). Orientation in Space By calculating r and y at time t, we have positioned the satellite in the plane of its orbit. Now we must position the orbital plane in space. To do this requires the definition of a coordinate system. This coordinate system must be an inertial coordinate system, that is, a nonaccelerating system in which Newton’s laws of motion are valid. A coordinate system fixed to the rotating Earth is not such a system. We will adopt an astronomical coordinate system called the right SATELLITES / Orbits 2027 (north) Earthcs spin axis z Autumnal equinox Earth n Su 23.45q Winter solstice Sun Vernal equinox cs ap l Celestia p ar Summer solstice t en eq u pa ato th y r x Figure 4 The right ascension–declination coordinate system. ascension–declination coordinate system.1 In this system (Figure 4), the z-axis is aligned with the Earth’s spin axis. The x-axis is chosen such that it points from the center of the Earth to the Sun at the moment of the vernal equinox, when the sun is crossing the equatorial plane from the Southern Hemisphere to the Northern Hemisphere.2 The y-axis is chosen so as to make it a right-handed coordinate system. In this system, the declination of a point in space is its angular displacement measured northward from the equatorial plane, and the right ascension is the angular displacement, measured counterclockwise from the x-axis, of the projection of the point in the equatorial plane (Figure 5). Three angles are used to position an elliptical orbit in the right ascension–declination coordinate system: the inclination angle, the right ascension of ascending node, and the argument of perigee (Figure 6). The inclination angle (i) is the angle between the equatorial plane and the orbital plane. By convention, the inclination angle is zero if the orbital plane coincides with the equatorial plane and if the satellite rotates in the same direction as the Earth. If the two planes coincide but the satellite rotates opposite to the Earth, the inclination angle is 1801. Prograde orbits are those with inclination angles less than 901; retrograde orbits are those with i greater than 901. z r 1  Because the origin of this coordinate system moves about the Sun with the Earth, it is not truly inertial. However, the Sun’s gravity causes the satellite to rotate around the Sun as does the Earth. Therefore, the satellite acts as if the right ascension–declination coordinate system were inertial. y : 2 This x-axis is also referred to as the First Point of Aries because it used to point at the constellation Aries. Because of the influence of the Sun and Moon on the nonspherical Earth, the Earth’s spin axis precesses like a top with a period of 25 781 years. This causes the vernal equinox to change. Today, the x-axis points to the constellation Pisces, but it is still referred to as the First Point of Aries. x Figure 5 Coordinates used in the right ascension–declination coordinate system: right ascension (O), declination (d), and radius (r ). 2028 SATELLITES / Orbits Earthcs spin axis z Perigee Orbit Center of Earth  : i nal x Ver inox q e u Ascending node y Equatorial plane Figure 6 Angles used to orient an orbit in space. The ascending node is the point where the satellite crosses the equatorial plane going north (ascends). The right ascension of this point is the right ascension of ascending node ðOÞ. It is measured in the equatorial plane from the x-axis (vernal equinox) to the ascending node. In practice, the right ascension of ascending node has a more general meaning. It is the right ascension of the intersection of the orbital plane with the equatorial plane; thus, it is always defined, not just when the satellite is actually at an ascending node. Finally, the argument of perigee ðoÞ is the angle measured in the orbital plane between the ascending node (equatorial plane) and the perigee. anomaly. Also, in less formal descriptions of satellite orbits, the height of the satellite above the Earth’s surface is substituted for the semimajor axis. Since the Earth is not round, the height of a satellite varies according to its position in the orbit. Such heights are converted into semimajor axis by adding the equatorial radius of the Earth. Orbits in which the classical orbital elements (except M) are constant are called Keplerian orbits. Viewed from space, Keplerian orbits are simple. The satellite moves in an elliptical path with the center of the Earth at one focus. The ellipse maintains a constant size, shape, and orientation with respect to the stars (Figure 7). Perhaps surprisingly, the only effect of the Sun’s gravity on the satellite is to move the focus of the Orbital Elements The parameters just discussed for locating a satellite in space are collectively known as the classical orbital elements; they are summarized in Table 1. These parameters may be determined by optical, radar, or radio ranging observations. They are carefully determined by various agencies and are available over the Internet for most satellites. There is some variation in how the orbital elements are specified. Some agencies, for example, substitute true anomaly for mean Sun Table 1 Classical orbital elements Element Symbol Semimajor axis Eccentricity Inclination Argument of perigee Right ascension of ascending node Mean anomaly Epoch time a e i o0 O0 M0 t0 Figure 7 The change with season of a Keplerian orbit. SATELLITES / Orbits ellipse (the Earth) in an elliptical path around the Sun (the Earth’s orbit). Viewed from the earth, Keplerian orbits appear complicated because the Earth rotates on its axis as the satellite orbits the Earth (Figure 8). The rotation of the Earth beneath a fixed orbit results in two daily passes of the satellite near a point on the Earth (assuming that the period is substantially less than a day and that the inclination angle is greater than the latitude of the point). One pass occurs during the ascending portion of the orbit; the other occurs during the descending portion of the orbit. This usually means that one pass occurs during daylight and one during darkness. Orbit Perturbations Although satellites travel in nearly Keplerian orbits, these orbits are perturbed by a variety of forces (Table 2). Forces arising from the last five processes are small and can be viewed as causing essentially random perturbations in the orbital elements. Operationally they are dealt with simply by periodically (1) observing the orbital elements and (2) adjusting the orbit with on-board thrusters. Forces due to the nonsphericity of the Earth cause secular (linear with time) changes in some of the orbital elements. These changes can be predicted theoretically and indeed are useful. The gravitational potential of the Earth is a complicated function of the Earth’s shape, the distribution of land and ocean, and the density of 2029 Table 2 Orbit-perturbing forces Force Source Nonspherical gravitational field Nonspherical, nonhomogeneous earth Moon, planets Gravitational attraction of auxiliary bodies Radiation pressure Particle flux Lift and drag Electromagnetic forces Sun’s radiation Solar wind Residual atmosphere Interaction of electrical currents in the satellite with the Earth’s magnetic field crustal material. As a first-order correction to a spherical shape, we may treat the Earth as an oblate spheroid of revolution. In cross-section, the Earth is approximately elliptical. The distance from the center of the Earth to the Equator is, on average, 6378.137 km, whereas the distance to the poles is 6356.752 km.The gravitational potential of the Earth is given approximately by eqn [12], where ree is the equatorial radius of the earth, d is the declination angle, and J2 (1.082 63  10  3) is the coefficient of the quadrupole term. U¼ Gme r  1
ree 2  1  3 sin2 d þ  1 þ J2 2 r ½12 The higher-order terms are more than two orders of magnitude smaller than the quadrupole term and will not be considered here, although they are necessary for very accurate calculations. A satellite travels at a slightly different speed in this gravitationally perturbed orbit. The time rate of change of the mean anomaly is given by the mean motion constant ðnÞ in the unperturbed orbit and by the anomalistic mean motion constant ð nÞ in a perturbed orbit. Considering only the quadrupole term we have eqn [13]. dM  ¼n dt    3
ree 2 3 ¼ n 1 þ J2 ð1  e2 Þ3=2 1  sin2 i 2 2 a Figure 8 The orbit of a representative satellite as viewed from a point rotating with the Earth. ½13 When the inclination angle is less than 54.71 or  is greater than n, and the greater than 125.31, n satellite orbits faster than it would in an unperturbed orbit. For inclination angles between 54.71 and 125.31, the satellite orbits more slowly than it otherwise would. 2030 SATELLITES / Orbits The rate of change of the right ascension of ascending node is given by eqn [14].  dO 3
ree 2 ð1  e2 Þ2 cos i ½14 ¼  n J2 dt 2 a The rate of change of the argument of perigee is given by eqn [15].    do 3
ree 2 5 2 2 2  J2 ¼n ð1  e Þ 2  sin i ½15 dt 2 2 a The other three orbital elements, a, e, and i, undergo small, oscillatory changes that may be neglected. The anomalistic period of a perturbed orbit is simply that given by eqn [16].  ¼ 2p T  n ½16 However, because M is measured from perigee, the anomalistic period is the time for the satellite to travel from perigee to moving perigee. Of more use is the ~ , which is the time for the synodic or nodal period, T satellite to travel from one ascending node to the next ~ must be ascending node. An exact value of T calculated numerically; however, eqn [17] holds to very good approximation. ~¼ T 2p  þ ðdo=dtÞ n ½17 In summary, the first-order effects of the nonspherical gravitational potential of the Earth consist of a slow, linear change in two of the classical orbital elements, the right ascension of ascending node and the argument of perigee, and a small change in the mean motion constant. Table 3 shows orbital elements for some representative satellites. Meteorological Satellite Orbits Nearly all meteorological satellites are in one of two orbits, Sun-synchronous or geostationary, but other orbits are also useful. Sun-Synchronous Orbits As shown in Figure 7, in a Keplerian orbit the angle between the Sun and the plane of a satellite’s orbit changes because the orbital plane is fixed while the Earth rotates around the Sun. This causes the satellite to pass over an area at different times of the day. For example, if the satellite passes over near noon (1200) and midnight (2400) in the spring, it will pass over near 0600 and 1800 in the winter. Fortunately, the perturbations caused by the nonspherical earth can be employed to keep the Sun–Earth–satellite angle nearly constant. The Earth makes one complete revolution about the Sun (2p radians) in one tropical year (31 556 925.9747 s). Thus, the right ascension of the Sun changes at the average rate of 1.991064  10  7 rad s  1 (0.98564731 day  1). If the inclination of the satellite is correctly chosen, the right ascension of its ascending node can be made to precess at this rate. An orbit that is so synchronized with the Sun is called a Sun-synchronous orbit. For a satellite with a semimajor axis of 7221 km and zero eccentricity, eqn [14] requires an inclination of 98.751 to be Sun-synchronous. Figure 9 shows the change with season of a Sun-synchronous orbit. The subsatellite point is the point on the Earth’s surface that is directly between the satellite and the center of the Earth. The ground track of a satellite is the path of the subsatellite point. Figure 10 shows the ground track for three orbits of the Sun-synchronous NOAA 11 satellite. Geostationary Orbits Earlier we calculated the radius of a geosynchronous orbit to be 42 164 km. Perturbations due to the nonspherical Earth, however, require a slight adjustment in this figure. The adjustment is small because the radius of geosynchronous orbit is about 6.6 Earth radii and the correction terms are inversely proportional to the square of this ratio. For a geosynchronous orbit with zero eccentricity and zero inclination, eqns [6], [13], [15], and [17] require a semimajor axis of 42 166.3 km. The terms geosynchronous and geostationary are often used interchangeably. In fact, they are not the same. Geosynchronous means that the satellite orbits with the same angular velocity as the Earth. A geostationary orbit is geosynchronous, but it is also required to have zero inclination angle and zero eccentricity. Geostationary satellites, therefore, remain essentially motionless above a point on the Equator. They are classified by the longitude of their subsatellite point. Second-order perturbations cause a geostationary satellite to drift from the desired orbit. Periodic maneuvers, performed as frequently as once a week, are required to correct the orbit. These maneuvers keep operational geostationary satellites very close to the desired orbit. Figure 11 shows the ground track of a typical geostationary satellite. Other Orbits Geostationary and Sun-synchronous are only two of an infinity of possible orbits. Others have been and will become useful for meteorological satellites. Table 3 Orbital elements of representative satellites Satellite Semimajor axis (km) Name ID INSAT 3B GOES 8 GOES 10 METEOSAT 7 GMS 5 FENGYUN 2B ELEKTRO TRMM UARS ERBS MOLNIYA 3-4 METEOR 3-6 TERRA QUIKSCAT NOAA 15 FENGYUN 1C 00016B 94022A 97019A 97049B 95011B 00032A 94069A 97074A 91063B 84108B 98040A 94003A 99068A 99034A 98030A 99025A 42 165.44 42 164.66 42 166.53 42 164.70 42 166.75 42 167.40 42 171.69 6 729.00 6 948.65 6 953.02 26 554.87 7 572.34 7 077.71 7 180.38 7 189.40 7 233.57 Inclination (deg) 0.08 0.16 0.25 0.54 0.58 0.94 3.25 34.98 56.98 57.00 63.08 82.56 98.18 98.63 98.63 98.73 Eccentricity 4.846  10  4 3.691  10  4 3.304  10  4 4.900  10  5 1.647  10  4 9.130  10  5 5.438  10  4 1.923  10  4 5.552  10  4 8.553  10  4 7.285  10  1 1.542  10  3 3.067  10  4 3.750  10  5 1.168  10  3 1.495  10  3 Right ascension of ascending Argument of perigee node Mean anomaly Nodal period (min) Value (deg) Motion (deg day 1) Value (deg) Motion (deg day 1) Value (deg) Motion (deg day 1) 288.81 104.85 276.33 296.54 65.29 264.49 81.41 6.60 45.52 330.37 126.09 252.61 326.40 73.99 277.24 289.04  0.0134  0.0134  0.0134  0.0134  0.0134  0.0134  0.0134  6.7736  4.0224  4.0122  0.1391  0.7073 0.9847 0.9868 0.9828 0.9733 233.08 61.66 248.95 68.78 259.00 164.52 138.84 293.05 102.59 94.57 279.98 350.18 104.03 0.98 44.01 30.07 0.0268 0.0268 0.0268 0.0268 0.0268 0.0268 0.0267 9.7410 1.7883 1.7807 0.0038  2.5018  3.1085  2.9189  2.9058  2.8357 266.58 104.11 44.58 339.61 161.45 20.73 202.39 159.10 87.93 59.04 221.47 335.10 207.77 264.67 223.53 135.06 360.98 360.99 360.97 360.99 360.97 360.96 360.90 5666.31 5395.39 5390.29 722.21 4740.49 5245.62 5133.63 5123.98 5077.16 1435.97 1435.93 1436.03 1435.94 1436.04 1436.07 1436.29 91.33 96.05 96.14 717.79 109.41 98.88 101.04 101.23 102.16 Epoch time 5 0000 UTC 6 September 2000. SATELLITES / Orbits 2031 2032 SATELLITES / Orbits Sun Figure 9 The change with season of a Sun-synchronous orbit. The Earth Radiation Budget Satellite (ERBS) was launched from the Space Shuttle and orbits at an altitude of 600 km with an inclination angle of 571. It was placed in this orbit so that it would precess with respect to the Sun and sample all local times over the course of a month. Meteor satellites fly in low Earth orbit with inclination angles of about 821. Molniya communications satellites fly in highly elliptical orbits. It has been suggested that this orbit would be useful for meteorological observations of the high latitudes. The Molniya orbit has an inclination angle of 63.41, at which the argument of perigee is motionless; thus, the apogee, from which measurements are made, stays at a given latitude. The semimajor axis is chosen such that the satellite makes two orbits while the Earth turns once with respect to the plane of the orbit. The eccentricity is made as large as possible so that the satellite will stay near apogee longer. However, the eccentricity must not be so large that the satellite encounters significant atmospheric drag at perigee. A semimajor axis of 26 554 km and an eccentricity of 0.72 result in a perigee of 7378 km (1000 km above the Equator), an apogee of 45 730 km (39 352 km above the Equator), and a period of 717.8 min.The attractiveness of this orbit is that it functions as a highlatitude, part-time, nearly geostationary satellite. For about 8 h centered on apogee, the satellite is synchronized with the Earth so that it is nearly stationary in the sky. For a meteorological satellite in a Molniya orbit, the rapid imaging capability, which is so useful from geostationary orbit, would be available in the high latitudes. As meteorological satellite instruments become more specialized, more custom orbits are likely to be used. Satellite Positioning, Tracking, and Navigation It is important to be able to calculate the position of a satellite in space, to track it from Earth, and to know where its instruments are pointing. These topics are discussed in turn in this section. Positioning in Space To locate a satellite in a perturbed orbit at time t, one needs current values of the orbital elements. The three constant elements, a, e, and i, are taken directly from a recent bulletin. Such bulletins are available from a variety of sources, and many are available on the Internet. The other three, M, O, and o, are calculated Figure 10 The ground track of a typical Sun-synchronous satellite (three orbits of NOAA 15). SATELLITES / Orbits argument of perigee (Figure 12B). This rotation is conveniently accomplished by multiplying the vector by a rotation matrix (eqn [20]). 0 01 0 10 1 x cos o  sin o 0 x B 0C B CB C ¼ sin o cos o 0 y @ A @ A@ y A 0 0 0 1 z z 0 1 x cos o  y sin o B C ¼@ x sin o þ y cos o A ½20 z 0.3qN 0.2qN 0.1qN Latitude 2033 0q _0.1qN _0.2qN 75.1qW 75.0q 74.9q 74.8q 74.7q Longitude 74.6q 74.5qW Figure 11 The ground track of a typical geostationary satellite (ten orbits of GOES 8). Second, the vector is rotated about the x-axis through the inclination angle (Figure 12C) as in eqn [21]. 10 0 1 0 00 1 0 1 0 0 x x CB 0 C B 00 C B @ y A ¼@ 0 cos i  sin i A@ y A z0 0 dM ðt  t0 Þ dt ½18a O ¼ O0 þ dO ðt  t0 Þ dt ½18b o ¼ o0 þ do ðt  t0 Þ dt ½18c Next, the satellite is located in the plane of its orbit; that is, the true anomaly y and the radius r are calculated. This is done by (1) solving for e using Kepler’s equation [9]; (2) calculating y using eqn [11a]; and (3) calculating r using eqn [8]. (For a circular orbit, this step is simplified because the mean anomaly, the eccentric anomaly, and the true anomaly are identical, and r is constant.) A vector can now be constructed that points from the center of the Earth to the satellite in the right ascension–declination coordinate system. The Cartesian coordinates of this vector are given by eqn [19]. 1 0 1 0 r cos y x @ y A ¼ @ r sin y A ½19 0 z At this point, the orbital ellipse is assumed to lie in the xy plane (the equatorial plane) with the perigee on the positive x-axis (Figure 12A). In the next three steps, the vector is rotated so that the orbital plane is properly oriented in space. First, the vector is rotated about the z-axis through the z0 cos i 1 x0 B C ¼@ y 0 cos i  z 0 sin i A y 0 sin i þ z 0 cos i 0 according to eqns [18a], [18b], and [18c] M ¼ M0 þ sin i 0 ½21 Third, the vector is rotated about the z-axis through the right ascension of the ascending node (Figure 12D) as in eqn [22]. 0 000 1 0 10 0 0 1 x cos O  sin O 0 x CB 0 0 C B 000 C B @ y A ¼@ sin O cos O 0 A@ y A z0 0 0 0 0 z0 0 1 x 0 0 cos O  y 0 0 sin O B 00 C ¼@ x sin O þ y 0 0 cos O A z0 0 0 1 ½22 The vector ðx 0 0 0 ; y 0 0 0 ; z 0 0 0 Þ is the location of the satellite in the right ascension–declination coordinate system at time t. This vector may be converted into the radius, declination, and right ascension of the satellite through eqns [23a], [23b], and [23c]. pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi rS ¼ x 0 0 02 þ y 0 0 02 þ z 0 0 02 ¼ r ½23a dS ¼ sin 1 OS ¼ tan 1  0 0 0 z rS ½23b y0 0 0 x0 0 0 ½23c   Finally, it is useful to calculate the latitude and longitude of the subsatellite point. Assuming that the Earth is a sphere, the latitude (known as the geocentric latitude) is simply equal to the declination (eqn [24]). Geocentric latitude ¼ jS ¼ dS ½24 2034 SATELLITES / Orbits y (A) y (B)  x (C) z x y (D) y y  Descending node i  i : Ascending node x x Figure 12 Rotations used to position a satellite in its orbit: (A) the satellite in the plane of its orbit; (B) rotation about the z-axis through the argument of perigee (o); (C) rotation about the x -axis through the inclination angle ðiÞ; and (D) rotation about the z-axis through the right ascension of ascending node (O). For more precise measurements, the latitude corrected for the nonspherical shape of Earth (the geodetic latitude) is usually used (eqn [25]). Geodetic latitude ¼jg ¼ tan 1 " ree rep 2 tan dS # ½25 where rep is the polar radius of the Earth. The longitude of the subsatellite point ðls Þ is the difference between the right ascension of the satellite and the right ascension of the prime meridian (01 longitude), which passes through Greenwich, England (Figure 13) (eqn [26]). lS ¼ OS  OGreenwich The inverse problem of finding when a satellite passes over (or close to) a particular point is solved iteratively by (1) estimating the time, (2) calculating the position of the satellite, and (3) correcting the time estimate. Steps 2 and 3 are repeated until a satisfactory solution is found. The above method can be streamlined in two ways. First, some of the rotations can be combined. Start as above by updating the orbital elements and calculating rS and y at time t. Locate the satellite on the x-axis at distance rS from the Prime meridian ½26 OGreenwich ¼ 99:9643 þ 360:9856376 Dt ½27 Since the rotation rate changes very slightly, owing to the actions of the wind and ocean currents, eqn [27] must be updated periodically. Long itu The right ascension of Greenwich is given in some satellite bulletins, and it can be calculated from eqn [27], where Dt is the time difference in days from 0000 UTC 1 January 2000. Satellite de :Sat : Greenwich North pole Vernal equinox Figure 13 The relationship between Earth longitude and right ascension. SATELLITES / Orbits center of the Earth (eqn [28]). 0 1 0 1 rS x @yA ¼ @ 0 A z 0 ½28 Define G, the argument of latitude, to be the angle, measured in the orbital plane, from the ascending node to the satellite as in eqn [29], where y is the true anomaly and o is the argument of perigee. G¼yþo ½29 Rotate this vector about the z-axis through the argument of latitude. Rotate again about the x-axis through the inclination angle. Finally, rotate about the z-axis through an angle equal to the right ascension of the satellite less the right ascension of Greenwich. Equations [23] now yield rS , latitude jS , and longitude lS . This method is useful for the navigation problem discussed below. A second way to streamline these equations is to combine them, which results in eqns [30a], [30b], and [30c]. rS ¼ r ½30a jS ¼ dS ¼ sin1 ðsin G sin iÞ ½30b   sin G cos i þ O0  Oe ðt0 Þ cos G   dOe dO   ðt  t0 Þ dt dt lS ¼ tan1 ½30c radius of the earth). The Cartesian coordinates of the satellite are then given by eqn [31]. 0 1 0 1 xS rS cos jS cos lS ½31 rS ¼ @ yS A ¼ @ rS cos jS sin lS A zS rS sin jS The Cartesian coordinates of the antenna are given by eqn [32]. 0 1 0 1 re cos je cos le xe ½32 re ¼ @ ye A ¼ @ re cos je sin le A re sin je ze The difference vector rD rS  re points from the antenna to the satellite (Figure 14). Assuming a spherical earth, the vector re points to the local vertical (Figure 15). The cosine of the satellite’s zenith angle z (the complement of the elevation angle) is given by eqn [33]. cos z ¼ re rD jre jjrD j ½33 Finding the azimuth angle is a little more difficult. First, we need to find two vectors in the tangent plane at the antenna. The first points north (eqn [34]). 1 0 1 0  sin jN cos lN xN ½34 rN ¼ @ yN A ¼ @  sin jN sin lN A: cos jN zN The second is the horizontal projection of rD . We define unit vectors in the directions of re and rD as in eqns [35a] and [35b]. Here rs is the distance of the satellite calculated with eqn [8]; jS and lS are its latitude and longitude, respectively. Oe ðt0 Þ is the right ascension of Greenwich at the epoch time, and therefore, O0  Oe ðt0 Þ is the longitude of ascending node at the epoch time. The quantity ðdOe =dt  dO=dtÞ is the relative Earth rotation rate, that is, the rotation rate of the Earth relative to the orbital plane. Earth Spin axis z Tracking A list of time versus position of a celestial body is called an ‘ephemeris’ (plural: ephemerides). To track a satellite, one must be able to point one’s antenna at it. The elevation angle, measured from the local horizontal, and the azimuth angle, measured clockwise from the north, can be calculated from the ephemeris data as follows. Suppose the subsatellite point is at latitude jS and longitude lS , and that the satellite is at radius rS from the center of the Earth. Suppose also that the antenna is located at latitude je , longitude le , and radius re (the 2035 rO re rS x ich nw an e re di G eri M Figure 14 Satellite tracking geometry. y 2036 SATELLITES / Orbits the telescope is pointing is given by eqn [38]. 1 0 1 0 cos dT cos OT xT @ yT A ¼ @ cos dT sin OT A sin dT zT rD re   rH rN Figure 15 Definition of zenith angle (z) and azimuth angle (c). ^re re jre j ½35a ^rD rD jrD j ½35b The required horizontal vector is given by eqn [36]. rH ¼ rD  ð^re rD Þ ^re ¼ rD  jrD j cos z ^re ¼ jrD j ð^rD  cos z ^re Þ ½36 The azimuth angle c is then given by eqn [37]. cos c ¼ rN rH jrN j jrH j ½37 One must be careful when taking the inverse cosine. If the satellite is west of the antenna, c will be greater than 1801. It must also be noted that these equations assume a spherical Earth. Fortunately, most receiving antennas are insensitive to the slight errors that this assumption causes. ½38 Figure 16 shows that the ray from which the telescope receives radiation (that is, the line in space through the satellite and in the direction of the telescope) is given by eqn [39], 0 1 0 1 x xS þ sxT @ y A ¼ @ yS þ syT A ½39 z zS þ szT where s is the distance from the satellite. The location at which this ray strikes the spherical Earth is the solution of eqn [40]. ðxS þ sxT Þ2 þ ðyS þ syT Þ2 þ ðzS þ szT Þ2 ¼ r2e ½40 This is a quadratic equation in s that has no real roots if the ray does not intersect the Earth or two real roots if it does. The smaller root is to be chosen; the larger root represents the location from which the ray re-emerges from the opposite side of the Earth. When the ray is just tangent to the Earth, the two roots are equal. After a solution for s has been found, eqn [39] gives the Cartesian coordinates in the right ascension–declination coordinate system of the point on the Earth’s surface being viewed. One way to specify the telescope pointing vector is to use the pitch, roll, and yaw angles familiar from z Navigation In addition to knowing where a satellite is in its orbit, it is necessary to know the Earth coordinates (latitude, longitude) of the particular scene it is viewing. The problem of calculating the Earth coordinates is known as the navigation problem; fundamentally, it is a complex geometry problem. It requires an accurate knowledge of where the satellite is in its orbit, the orientation of the satellite (its attitude), and the scanning geometry of the instrument involved. Suppose that at a particular time a satellite is at position ðxS ; yS ; zS Þ with respect to the center of the Earth in the right ascension–declination coordinate system. Suppose further, that the telescope is pointing in a direction specified by declination dT and right ascension OT . A unit vector in the direction in which re SrT rS x rT Figure 16 Navigation geometry. y SATELLITES / Orbits aircraft flight. Position the satellite on the positive x-axis at distance rS from the center of the Earth. Let the satellite be traveling in the xy plane with the positive z-axis on the left; that is, the satellite is traveling eastward in the equatorial plane. Orient the satellite so that its ‘nose’ is pointing in y-direction (not parallel to the velocity vector), the ^ the left ‘wing’ is pointing in the ^z-direction, and ‘up’ ^-direction. Let the telescope begin by is in the x pointing straight down toward the center of the Earth, that is in the direction given by eqn [41]. 0 1 0 1 xT 1 @ yT A ¼ @ 0 A ½41 0 zT Pitch is defined as the angle of rotation about the zaxis; positive is in the sense of the nose pointing up.3 The matrix that accomplishes this rotation is given in eqn [42]. 0 1 cos P  sin P 0 Pitch rotation matrix ¼ @ sin P cos P 0 A 0 0 1 ½42 A positive pitch angle causes the telescope to point in the along-track direction ahead of the subsatellite point. Roll is defined as the angle of rotation about the yaxis; positive is in the sense of the left wing pointing up. The matrix that accomplishes this rotation is given in eqn [43]. 0 1 cos R 0  sin R Roll rotation matrix ¼ @ 0 1 0 A sin R 0 cos R ½43 A positive roll angle causes the telescope to point in the cross-track direction, to the left of the subsatellite point. Yaw is defined as the angle of rotation about the xaxis; positive is in the sense of the nose pointing right. The matrix that accomplishes this rotation is given in eqn [44]. 0 1 1 0 0 Yaw rotation matrix ¼ @ 0 cos Y  sin Y A 0 sin Y cos Y ½44 3 Note that the axes of rotation described here are dependent on the satellite being at the specified position and orientation in the right ascension–declination coordinate system. 2037 A nonzero yaw angle does nothing to a telescope pointing straight down; however, if the pitch or roll angles are nonzero, a positive yaw angle moves the telescope in a clockwise direction around the subsatellite point. Common scanning schemes can easily be described with these angles. A cross-track scanner can be described by a roll angle that increases linearly in time (right-to-left scanning) or decreases linearly in time (left-to-right scanning). A conical scanning instrument can be described by a constant pitch angle followed by a yaw angle that increases linearly in time (for clockwise scanning). Finally, a geostationary scanner can be described by a stepped roll angle followed by a pitch angle that increases (west-to-east scanning) or decreases (east-to-west scanning) with time. Corrections in the pitch, roll, and yaw angles need to be applied if the satellite is not aligned as indicated above. After the appropriate pitch, roll, and yaw rotations have been applied to the initial telescope pointing vector (eqn [41]), eqn [40] yields the distance to the point being observed, and eqn [39] yields the coordinates of the point in the right ascension– declination coordinate system. Now, both the observed point and the satellite can be positioned by (1) rotating the vectors about the z-axis through the argument of latitude, (2) rotating about the x-axis through the inclination angle, and (3) rotating about the z-axis through the right ascension of ascending node less the right ascension of Greenwich. Equations [23] yield the latitude and longitude of the point. Finally, corrections may need to be applied for the nonspherical Earth and for the height of the terrain or the object being observed. The inverse problem, that of finding which satellite datum corresponds to a selected latitude and longitude, is solved iteratively. Each observation has a time associated with it, which determines all of the above angles. First, a time of observation is estimated, and the latitude and longitude of the point being observed at that time are calculated. Then the time is incremented and a new point is calculated. This process is iterated until a satisfactory solution is found. Space–Time Sampling To select an orbit for a satellite or a scan pattern for a particular instrument, several questions must be answered: What areas will the orbit and scan pattern allow the instrument to observe? How often will an area be observed? At what local times will the observations be made? At what viewing zenith and azimuth angles will the observations 2038 SATELLITES / Research (Atmospheric Science) be made? What will be the solar zenith and azimuth angle when the area is being observed? These questions are all aspects of what is called space–time sampling. Using the equations in this article plus some easily acquired equations that describe the position of the Sun, these questions can be answered. See also Observations for Chemistry (Remote Sensing): IR/FIR; Microwave. Satellite Remote Sensing: Aerosol Measurements; Cloud Properties; GPS Meteorology; Precipitation; Surface Wind; TOMS Ozone; Temperature Soundings; Water Vapor; Wind, Middle Atmosphere. Further Reading Brouwer D and Clemence GM (1961) Methods of Celestial Mechanics. New York: Academic Press. Chen HS (1985) Space Remote Sensing Systems: An Introduction. San Diego: Academic Press. Dubyago AD (1961) The Determination of Orbits. New York: Macmillan. Escobal PR (1965) Methods of Orbit Determination. New York: Wiley. Goldstein H (1950) Classical Mechanics. Reading: AddisonWesley. Kidder SQ and Vonder Haar TH (1995) Satellite Meteorology: An Introduction. San Diego: Academic Press. Research (Atmospheric Science) M D King, NASA Goddard Space Flight Center, Greenbelt, MD, USA D D Herring, Science Systems and Applications Inc., Lanham, MD, USA Copyright 2003 Elsevier Science Ltd. All Rights Reserved. Introduction The atmosphere changes chemically and physically on widely varying time scales – ranging from minutes to decades – and is therefore a challenge to measure precisely over the entire globe. But with the National Aeronautics and Space Administration’s (NASA) 1960 launch of the Television Infrared Observation Satellite (TIROS), Earth scientists began a new mission to observe large-scale weather patterns from space. In the late 1970s, their mission expanded to include globalscale measurements that would help them understand the causes and effects of longer-term climate change. NASA and its affiliated agencies and research institutions collaborated to develop a series of research satellites that have enabled the testing of new remote sensing technologies that in turn have advanced scientific understanding of both chemical and physical changes in the atmosphere. (‘Remote sensing’ involves the use of devices other than our eyes to observe or measure things from a distance without disturbing the intervening medium.) The goal is to examine our world comprehensively to determine what dynamics drive Earth’s climate system and how climate change both affects our environment and is affected by it. Depending upon their measurement objectives, research satellites primarily fly in one of two orbits: (1) a near-polar, Sun-synchronous orbit to allow their sensors to observe the entire globe at the same solar time each day, or (2) a mid-inclination, precessing orbit to focus their sensors on the equator and lower latitudes where the observations are made at different times of day to better sample time-varying phenomena such as clouds. Some polar orbiting satellite sensors can observe any given place on the globe as often as every day, thus collecting data with high temporal (time) resolution. Other satellite sensors view any given place as infrequently as once every 16 days, thus having relatively low temporal resolution for a satellite sensor, but still far surpassing our ability to make these same measurements with surface-based or aircraft sensors. Satellite sensors with high spatial resolution (15 meters per pixel) can discern objects in the atmosphere or on the surface as small as, say, a football field, thus providing high spatial resolution. Other satellite sensors that are designed to measure continental and global-scale dynamics typically have only moderate (500 m per pixel) to low (20 km per pixel) spatial resolution. Satellite sensors carry specially designed detectors that are particularly sensitive to certain wavelengths of the electromagnetic spectrum, called spectral bands. The more precisely a remote sensor can measure narrow bands of radiant energy, and the greater the number of these discrete bands it can measure, the higher is its spectral resolution. The atmosphere interacts with solar radiation much like a venetian blind – selectively absorbing and reflecting certain wavelengths of solar energy while allowing others to pass through. Satellite remote sensors are designed to be particularly sensitive to those wavelengths that can be reflected or emitted back up through the atmosphere to space, thus enabling them to make their measurements. SATELLITES / Research (Atmospheric Science) 2039 Earth-orbiting remote sensors provide the best means of collecting the data needed in research because they can measure things on scales of time and space that otherwise would not be possible. Moreover, satellite sensors not only observe wavelengths of visible light; they also precisely measure wavelengths of radiant energy that the eye cannot see, such as microwaves, ultraviolet rays, or infrared light. If it is known how certain objects (like cirrus clouds or windblown dust) typically absorb, reflect, and emit particular wavelengths of radiant energy, then by using satellite sensors to precisely measure those specific bands of the electromagnetic spectrum, a lot can be learnt about the Earth’s atmosphere and surface. Remote sensors allow us to observe and quantify key climate and environmental vital signs such as temperature, ozone concentrations, carbon monoxide, and other pollutants, water vapor and other greenhouse gases, cloud types and total cloud cover, aerosol types and concentrations, radiant energy fluxes, and many more. Balancing Earth’s Radiant Energy Budget Climate is defined as the average state of the atmosphere, hydrosphere, and land over a given time. Thus, measurements of radiant energy within Earth’s atmosphere are at the heart of the climate change discussion. How climate changes is related directly to how the planet balances the amount of incoming sunlight with outgoing radiant energy. The measurement of all incoming and outgoing energy provides a sort of ledger of all the physical motions and interactions of our world’s climate system, showing whether, over the course of a year and over the entire globe, the Earth’s total energy budget is in balance, or not, and if not, whether it is heating up or cooling down. So if we are to understand climate and accurately predict future climate change, then we must determine what drives the changes within the Earth’s radiation balance (Figure 1). In 1978, NASA launched its Nimbus-7 satellite carrying a new sensor, called the Earth Radiation Budget (ERB) experiment, designed to measure direct solar irradiance, reflected short-wave radiation (visible light), and emitted long-wave radiation (heat) every day over the entire Earth. This was the first space-based sensor capable of self-calibrating so that its total solar irradiance measurements were accurate to within
0:5%. The Nimbus-7 ERB collected 9 years of global-scale data, upon which long-term climate studies were begun. In the interest of extending the ERB data set and improving upon its measurement capabilities, NASA launched three more Earth Radiation Budget Experiments (renamed ERBE) in the 1980s. In addition to total solar irradiance, ERBE measured the reflected solar and emitted thermal radiation from the Earth–atmosphere–ocean system. These observations revealed that over the course of a year the global radiation budget is in balance – the Earth reflects and emits roughly the same amount of energy back into space that it receives from the Sun. The data showed also that the average annual, global contribution by clouds is that they reflect 17 W m  2 more short-wave energy (visible light) than they trap as long-wave energy (heat). Yet, owing to calibration uncertainties, deficiencies in ERBE’s sampling method, and the limitations of existing angular dependence models, there still exists a significant uncertainty (about 75 W m  2) regarding our understanding of Earth’s radiation budget. Part of this uncertainty lies in our limited knowledge of the spatial distribution of clouds as well as the optical properties of these clouds over time. Moreover, we cannot be sure how the distribution and optical properties of clouds will change over time. The endeavor to address these issues began with the 1997 launch of the Clouds and the Earth’s Radiant Energy System (CERES) sensor aboard the joint NASA/NASDA Tropical Rainfall Measuring Mission (TRMM) satellite. Twin CERES instruments were also launched aboard NASA’s Terra satellite in December 1999, and the pair will again fly aboard NASA’s Aqua satellite in May 2002. Many of the sampling and accuracy limitations on ERBE were addressed in the design of CERES so that it could meet the same measurement objectives as those for ERBE but with better than twice the former sensor’s accuracy. Ultimately, it is anticipated that CERES will not only extend the ERBE data set but will also provide the first long-term global measurements of the radiative fluxes within the Earth’s atmosphere that will help us more accurately account for the effects of aerosols and clouds on climate. Dust in the Wind Aerosols are tiny particles suspended in the air (mostly in the troposphere). Some come from natural sources, such as volcanic eruptions, dust storms, forest and grassland fires, living vegetation, and sea spray. About 11% of the total emitted aerosols in our atmosphere come from human activities, such as the burning of vegetation and fossil fuels and changing the natural land surface cover, which again leads to windblown dust. Yet human-produced aerosols account for about half of the total effect of all aerosols on incoming 2040 SATELLITES / Research (Atmospheric Science) Reflected shortwave radiation (W m 2) 0 100 200 300 Outgoing longwave radiation (W m 2) 100 180 260 340 Figure 1 Global reflected shortwave radiation and emitted longwave radiation escaping the top of Earth’s atmosphere, measured by CERES on 25 May 2001. sunlight. From a satellite’s perspective, aerosols raise the Earth’s albedo, or make it appear brighter, by scattering and reflecting sunlight back to space. The overall effect of these tiny particles is to cool the surface by absorbing and reflecting incoming solar radiation. They also serve as cloud condensation nuclei, or ‘seeds’ for cloud formation, which again helps to cool the surface. In terms of their net influence on global climate, for scientists aerosols represent the greatest subject of uncertainty. Yet computer climate models estimate that over the last century human-produced aerosols have offset global warming due to greenhouse gases by about 40% (Figure 2). Through the 1980s and most of the 1990s, the NOAA Advanced Very High-Resolution Radiometer (AVHRR) was the most frequently used satellite sensor for measuring aerosol optical thickness. (Aerosol optical thickness is a measure of how much sunlight airborne particles prevent from traveling through a column of atmosphere.) However, AVHRR can only make such measurements over the ocean, as the sensor requires a relatively uniform and dark-colored background. Because TOMS is particularly sensitive to absorbing aerosols, over both land and ocean, this sensor has also been widely used to measure aerosol optical thickness. In April 1991, the European Space Agency launched a new type of multi-angle sensor, called the Along Track Scanning Radiometer (ATSR), aboard their first European Remote Sensing Satellite (ERS-1). The ATSR makes aerosol optical thickness measurements by remotely sensing visible and nearinfrared wavelengths at nadir and oblique forward scan angles (both within a 2-minute interval). A modified version of the sensor, called the Advanced Along Track Scanning Radiometer (AATSR), was launched in 1995 aboard ERS-2. While data from neither of these missions have yet been used to produce SATELLITES / Research (Atmospheric Science) 2041 0.8 0.6 0.4 0.2 0.0 Figure 2 Global aerosol optical thickness measured by MODIS in April 2001. global-scale aerosol measurements, this should be possible. In 1996, Japan launched the first in its series of Advanced Earth Observation Satellites (ADEOS), which carried a payload of two sensors – the Polarization and Directionality of the Earth’s Reflectances (POLDER) sensor, contributed by the French Space Agency, and the Ocean Color and Temperature Scanner (OCTS), provided by NASDA. Both sensors can retrieve aerosol measurements, but POLDER was the first satellite sensor designed specifically to measure aerosols, and it can make its measurements over both land and ocean. The sensor observes Earth targets from 12 directions that enable measurements of the bidirectionality and polarization of solar radiation reflected from within the atmosphere. Unfortunately, owing to its solar panel failing, the ADEOS mission ended prematurely after only 8 months in orbit. Three sensors aboard NASA’s Terra satellite are particularly well suited for studying the effects of aerosols on climate: CERES, MISR, and MODIS. The Global Imager (GLI) planned for launch aboard ADEOS II offers aerosol measurement capabilities similar to those of MODIS. Both these sensors have the capacity to measure not only aerosol optical thickness but also the sizes of aerosol particles over both ocean and land. Particle size is an indicator of the source of the aerosol particles and helps scientists distinguish aerosols of natural origin from those that are man-made. Moreover, with its nine different look angles, MISR is ideally designed to quantify the reflective properties. Again, CERES complements MODIS and MISR by providing measurements of the short-wave radiation that aerosols reflect back into space. Together, these sensors are providing new insights into the roles of clouds and aerosols in Earth’s total energy budget. Abstract Art or Arbiters of Energy? More than just the idle stuff of daydreams, clouds help control the flow of radiant energy around our world. Clouds are plentiful and widespread throughout Earth’s atmosphere – covering up to 75% of our planet at any given time – so they play a dominant role in determining how much sunlight reaches the surface, how much is reflected back into space, how and where warmth is spread around the globe, and how much heat escapes from the surface and atmosphere back into space. Clouds are also highly variable. Clouds’ myriad variations through time and space make them one of the greatest areas of uncertainty in the understanding and prediction of climate change. In short, they play a central role in the world climate system. Whereas thick, low-level stratocumulus clouds cool the Earth’s surface by reflecting incoming solar radiation, thin, high-level cirrus clouds exert a warming influence by allowing sunlight to pass through but then trapping the heat emitted by the surface. The question of whether warming or cooling has the greater effect over time has been answered only relatively recently. From ERBE satellite data collected in the 1980s, coupled with aircraft and surface-based measurements, it has been demonstrated that, globally, clouds cool the surface more than they warm it. So great is the cooling effect that it is as if clouds remove the heat of a 60-watt light bulb from every 2-meter square of the Earth’s surface. But will they continue to cool our planet over the next century if a greenhouse-gasdriven global warming scenario comes to pass? Or even, could the type and distribution of clouds change so that they primarily exert a warming influence? (Figure 3). Two new sensors flying aboard NASA’s Terra satellite, launched in December 1999, are designed to help scientists answer these questions. The Moderate-resolution Imaging Spectroradiometer (MODIS) 2042 SATELLITES / Research (Atmospheric Science) 2 20 15 10 5 0 Figure 3 Global cloud optical thickness measured by MODIS in April 2001. and the Multi-angle Imaging Spectroradiometer (MISR) give scientists new capabilities for measuring the structure and composition of clouds. MODIS observes the entire Earth almost every day in 36 spectral bands, ranging from visible to thermal infrared wavelengths. With spectral and spatial resolutions superior to that of AVHRR (its heritage sensor), MODIS can measure a wide suite of clouds’ physical and radiative properties. Specifically, MODIS can determine whether a cloud is composed of ice or water particles (or some combination of the two), it can measure the effective radius of the particles within a cloud, it can observe how much (or little) sunlight passes through a cloud, and it can measure the temperature and altitude of cloud tops. Moreover, with its unique 1.38 mm channel, MODIS observes thin cirrus clouds with unprecedented sensitivity. This channel not only enables scientists to quantify the impact of cirrus clouds on the radiation balance, but also permits image analysts to ‘correct’ for the presence of cirrus in remote-sensing scenes used to examine surface or lower-level features. Complementing MODIS, the MISR instrument ‘sees’ the Earth simultaneously in red, green, blue, and near-infrared wavelengths at 9 different angles – at 4 progressively more oblique angles ahead of Terra, 4 angles aft of the satellite, and 1 at nadir. Because it measures any given scene from multiple angles, MISR is ideally designed to help scientists better understand how clouds interact with radiant energy as a function of both their structure and their type. CERES complements MODIS and MISR by providing measurements of the short-wave and longwave radiant energy that clouds reflect and emit back into space. ESA’s next-generation satellite missions for comprehensively examining Earth’s climate system began with the February 2002 launch of its first Environmental Satellite (Envisat). Similar to MODIS in the scope of its research applications, Envisat carries the Medium-Resolution Imaging Spectrometer (MERIS). Like Terra’s MODIS, MERIS has a wide viewing swath (1500 km), with a morning equatorial crossing, and it can see the entire Earth within every 3 days. Scientists are using its data to derive cloud cover, cloud altitude, water vapor, and aerosol properties. Unlike MODIS (which uses a cross-track scan mirror), MERIS is a push-broom scanner based upon ChargeCoupled Device (CCD) technologies with gains and offset settings that can be optimized for observing specific targets. This is a similar technology to that used by MISR. Serendipity and Stratospheric Ozone In the early 1970s, as Earth scientists intensified their studies into the possible causes and effects of global warming, one group of man-made gases in particular elicited the attention of scientists – the chlorofluorocarbons (CFCs). Increasingly, CFCs were being used by industrial nations in the production of a variety of commercial products (e.g., refrigerants, aerosol sprays). The concern is twofold: CFCs are up to 200 times more efficient than carbon dioxide at trapping heat in the Earth’s atmosphere, and they tend to remain in the atmosphere up to 120 years once released. Then, in 1974, two scientists wrote of a new concern that CFCs could potentially reduce levels of ozone in the stratosphere, the layer of atmosphere from 10 to 50 km in altitude. In 1975 the US Congress asked NASA to develop a ‘comprehensive program of research, technology, and monitoring of phenomena of the upper atmosphere’. In particular, Congress’s intent was to ascertain the ‘health’ of the ozone layer (Figure 4). So, in addition to ERB, in 1978 Nimbus-7 carried two other new NASA sensors designed to measure the total amount of ozone in a given column of atmosphere over the entire globe – the Solar Backscatter SATELLITES / Research (Atmospheric Science) 2043 Dobson Units 100 200 September 1983 September 1987 300 400 500 September 1993 September 1997 Figure 4 Total ozone content from TOMS in the Southern Hemisphere in September during the years 1983, 1987, 1993, and 1997. Dobson Unit (DU) 5 2.69  1016 molecules cm  2. Ultraviolet (SBUV) instrument and the Total Ozone Mapping Spectrometer (TOMS). Sensitive to radiant energy in the ultraviolet region of the spectrum, these sensors took advantage of the fact that molecules and aerosol particles reflect certain wavelengths of ultraviolet rays while ozone absorbs others at different levels in the atmosphere. By analyzing the amount of ultraviolet energy reflected back up to the spacecraft, researchers could produce profiles of how thick or thin the ozone was at different altitudes and locations. Ironically, it wasn’t until October 1985 that a British team of scientists reported a significant reduction in ozone over Halley Bay, Antarctica. Using a groundbased Dobson ozone spectrophotometer, the team found that the amount of stratospheric ozone there was about 40% less than it had been the previous year. Their finding stunned the science community because it had been expecting anthropogenic ozone depletion to occur first at upper levels in the stratosphere (30 to 50 km), and so had anticipated that the initial signal of depletion in a total column of ozone would be weak. NASA researchers hastily reviewed their TOMS data and found that it too had detected a dramatic loss of ozone over all of Antarctica. Why hadn’t they discovered the phenomenon earlier? Actually, the TOMS Team had noted the instrument measured ozone levels of less than 180 Dobson Units over Antarctica, but because these values were so much lower than expected, they wanted to make sure these were not erroneous readings. They compared the TOMS measurements with data collected by a Dobson ozone spectrophotometer located at the South Pole. Unfortunately, the ground-based instrument had been set improperly and so it had erroneously recorded ozone levels at around 300 Dobson Units. This puzzled the TOMS Team for a while and delayed the public report of its findings until the team could verify that the satellite instrument was working fine. The TOMS Team was in the process of producing a report of their findings when the British Antarctic Survey report was released. In the years following the discovery of the ozone hole, NASA and ESA satellites recorded depleting ozone levels over Antarctica growing worse with each passing year. In response, in 1987, 43 nations signed the Montreal Protocol, in which they agreed to reduce the use of CFCs by 50% by the year 2000. This protocol was amended in 1990 to eliminate all CFC emissions by 2000. ESA’s second European Remote-Sensing Satellite (ERS-2) carries a sensor called the Global Ozone Monitoring Experiment (GOME). GOME is a nadirlooking sensor with four bands ranging from 240 to 790 nm for measuring backscattered visible and ultraviolet solar radiation. Since the summer of 1996, ESA has routinely produced 3-day global measurements of total ozone and nitrogen dioxide using GOME data. 2044 SATELLITES / Research (Atmospheric Science) As recently as 1998, both TOMS and GOME data showed that at its Austral spring low, Antarctic ozone concentrations had worsened to 80% less than early 1970s levels. Today there is some evidence that the amount of chlorine in the stratosphere is leveling off. Is this a scientific success story in the making? Will stratospheric ozone concentrations return to pre1970s levels as the abundance of stratospheric chlorine stabilizes? Only time and continued monitoring will tell. ESA launched its Environmental Satellite (Envisat) in February 2002 with a new sensor called Global Ozone Monitoring by Occultation of Stars (GOMOS). Chemistry of Earth’s Atmosphere Some satellite sensors allow scientists to determine the chemical content of the Earth’s upper atmosphere using a technique called solar occultation, in which a sensor is pointed toward the horizon at sunrise and sunset to measure the profile of the stratosphere and mesosphere about 30 times per day. In this way sensors, such as the Stratospheric Aerosol and Gas Experiment (SAGE), can determine the presence and abundance of gases and particulates by measuring precisely the visible and ultraviolet wavelengths that are absorbed within the upper atmosphere. Since the spectra of ozone, nitrogen dioxide, sulfur dioxide, and certain aerosols are well known, scientists can directly correlate SAGE’s readings with the presence of these substances within the stratosphere. The solar occultation technique is particularly effective because the sensor is self-calibrating – each occultation event looks ClO O3 30 Aug 96 5 (A) directly at the unattenuated Sun outside the Earth’s atmosphere just prior to sunset or just following sunrise. These observations are then compared with observations of the Sun obtained by looking through the atmosphere. The direct Sun observations establish an ongoing baseline of the sensor’s performance. Adapted from the Stratospheric Aerosol Mission (SAM II) that flew aboard Nimbus 7, the SAGE sensor is essentially a modified Sun photometer. This kind of sensor first flew in 1979 aboard NASA’s Applications Explorer Mission-2 (AEM-2). A subsequent version of SAGE (SAGE II) was launched aboard ERBS in 1984 and performed well throughout 2002, thus giving a 18-year continuous dataset of upper atmosphere profile measurements. In 1991, NASA launched the Upper Atmosphere Research Satellite (UARS) with a payload of 10 sensors for measuring a wide array of chemical and physical phenomena in the stratosphere and mesosphere (the layers of atmosphere from approximately 10 to 90 km in altitude). Not only did UARS extend our ability to monitor stratospheric ozone concentrations into the 1990s, but it also provided the first comprehensive picture of the photochemical processes involved in ozone destruction. The UARS Microwave Limb Sounder (MLS) demonstrated that there is a direct link between the presence of chlorine, the formation of chlorine monoxide during winter in the Southern Hemisphere, and the destruction of ozone (Figure 5). UARS carries the first two spaceborne remote wind sounders ever launched, called the High Resolution Doppler Imager (HRDI) and the Wind Imaging Interferometer (WINDII). These sensors measure 10 15 20 1018 molecules m2 25 140 (B) 180 220 260 300 340 DU above 100 hPa Figure 5 Chlorine monoxide (A) and ozone concentration (B) derived by MLS at approximately 18 km altitude on 30 August 1996. The high chlorine monoxide within the Antarctic polar vortex in the left-hand figure (reds and dark purple shades) is directly associated with, and leads to, a reduced ozone concentration shown in the right-hand figure (light blue and light purple shades). SATELLITES / Research (Atmospheric Science) 2045 winds in the mesosphere through detection of shifts in airglow emission lines. Additionally, HRDI can detect daytime stratospheric winds by observing Doppler shifts in oxygen absorption lines. WINDII and HRDI gave scientists the first complete global picture of mesospheric circulation. Together with the Halogen Occultation Experiment (HALOE) and MLS aboard UARS, the sensors have enabled researchers to track the upward transport of water vapor in the tropical stratosphere. Most atmospheric water vapor originates from the tropical oceans, where it rises high into the atmosphere to form towering thunderheads. Encircling the globe along the equator is an almost continuous band of thunderheads known as the Intertropical Convergence Zone (ITCZ), producing roughly three-quarters of the energy in our atmosphere that helps to drive its circulation patterns. Data from the sensors just mentioned showed that the tropical tropopause (the gateway from the troposphere to the stratosphere) air rises into the stratosphere through these thunderheads. Once in the stratosphere, this air moves slowly upward and outward toward the midlatitudes. Ozone begins to form as incoming ultraviolet radiation breaks oxygen molecules (O2) into free oxygen atoms (O) that quickly bond with other oxygen molecules to form ozone (O3). Because ozone strongly absorbs certain wavelengths of ultraviolet radiation, the air begins to warm, helping to perpetuate the upward movement of the air mass as well as helping to create temperature gradients for stratospheric winds. UARS data showed that it takes about 2 years for water vapor anomalies to move from the tropopause (at about 17 km) up to the mid-stratosphere (at about 30 km). A Canadian instrument launched in 1999 aboard NASA’s Terra satellite uses gas correlation spectroscopy to determine the abundance of methane and carbon monoxide in the troposphere. The Measurements Of Pollution In The Troposphere (MOPITT) sensor measures emitted and reflected radiance from the Earth in three spectral bands. As this light enters the sensor, it passes along two different paths through onboard containers of carbon monoxide and methane. The different paths absorb different amounts of energy, leading to small differences in the resulting signals that correlate directly with the presence of these gases in the atmosphere. Both methane and carbon monoxide are byproducts of burning vegetation as well as fossil fuels. Over the last two decades levels of methane in the atmosphere have risen at an average rate of about 1% per year. This is cause for concern because methane (CH4) is a greenhouse gas about 30 times more efficient than carbon dioxide at trapping heat near the surface. Scientific interest in carbon monoxide (CO) is twofold. First, the gas controls atmospheric concentrations of oxidants, thus affecting the ability of the atmosphere to clean itself from the ongoing generation of harmful tropospheric ozone from biomass burning and urban smog. Second, through chemical reactions within the lower atmosphere, carbon monoxide contributes to the production of harmful ozone. MOPITT is helping researchers to identify the main sources of these gases as well as to improve four-dimensional models of their transport through the atmosphere. ESA’s Envisat carries the Scanning Imaging Absorption Spectrometer for Atmospheric Cartography (SCIAMACHY), which is an advanced version of the GOME sensor flying aboard ERS-2. In addition to the same four spectral channels contained on GOME (from ultraviolet to visible wavelengths; 240–800 nm), SCIAMACHY has an additional four channels in the infrared region of the spectrum (800–2400 nm). While the sensor’s wide spectral sensitivity makes it useful for cloud and aerosol research, its ability to view both nadir and the Earth’s horizon makes it useful for determining the content and distribution of 16 different trace gases in the atmosphere. Where Storm Clouds Gather Rain clouds form when moisture-laden air is driven skyward by warm updrafts emitted from a Sunwarmed land or ocean surface; or when mountain slopes push moist air aloft; or when a wedge of colder, denser air plows warmer, moist air upward to higher elevations. Because cold air cannot hold as much water vapor as warm air, and because the atmosphere cools at higher elevations, water vapor condenses readily into liquid droplet or ice crystal form, in the presence of seed aerosol particles. Were there no aerosol particles in the Earth’s atmosphere, there would be no fog, no clouds, no mist, and probably no rain. When water evaporates at the surface, it absorbs energy from its surroundings and stores it as latent heat. When water vapor condenses back into liquid or ice form it releases its latent heat into its surroundings. Only about 25% of the energy contained within the atmosphere comes directly from the Sun’s rays; the remaining 75% comes from the release of latent heat contained in water vapor, most of which, as mentioned, is present in the towering thunderheads of the Intertropical Convergence Zone (Figure 6). We cannot measure the latent heat contained within clouds. We can, however, measure tropical rainfall. Currently, there is a 50% uncertainty in estimates of annual global rainfall. If we are to determine more accurately how much energy our atmosphere receives 2046 SATELLITES / Research (Atmospheric Science) Figure 6 Hurricane Bonnie as observed by the TRMM/PR on 22 August 1998. Red shows intense precipitation, green and yellow hues are intermediate values, and blues are low values. The eye of the storm reached to 16 km. from latent heat, then we must more accurately measure rainfall. In 1997, NASDA and NASA jointly developed and launched the Tropical Rainfall Measuring Mission (TRMM) into a mid-inclination (351) precessing orbit. It is estimated that about 60% of precipitation on Earth falls within the band between 301 N and 301 S of the Equator. TRMM carries three instruments designed to measure rainfall – the Precipitation Radar (PR), the TRMM Microwave Imager (TMI), and the Visible and Infrared Scanner (VIRS). Designed and built by NASDA, the Precipitation Radar is the first satellite sensor to provide threedimensional images of the internal structures of storm clouds. Its measurements show the intensity and distribution of rain within a storm, the total height of a storm, and the elevation at which ice crystals melt into raindrops. Most importantly, the Precipitation Radar can measure rain rates to within 0.7 mm per hour. Researchers who expected to use ground-based Doppler Radar stations to validate TRMM’s Precipitation Radar measurements found much to their pleasant surprise that the latter exceeds most groundbased measurements in accuracy and spatial resolution. The TMI is a ‘passive’ sensor designed to measure minute amounts of microwave energy emitted by the Earth’s surface and from within its atmosphere. (Whereas ‘active’ sensors send pulses of energy and then measure how much gets absorbed and reflected by the target, ‘passive’ sensors measure only energy originating from, or reflected by external sources.) These measurements allow TMI to quantify the amount of water vapor, cloud water, and rainfall intensity within the atmosphere. Based upon the design heritage of the Defense Meteorological Satellite Program’s Special Sensor Microwave/Imager (SSM/I), the TMI has a wider viewing swath (780 km) and finer spatial resolution than its predecessors. The TRMM VIRS detects radiant energy in five spectral bands, ranging from infrared to visible wavelengths (from 0.63 to 12 mm). Ideally designed to measure temperature, VIRS can precisely determine cloud top temperatures that scientists can then indirectly correlate with rainfall amounts. Conclusion As the preceding sections demonstrate, the Earth’s atmosphere changes both physically and chemically over a range of scales of time and space. The atmosphere’s chemical makeup affects its physical state, such as its radiative properties. As already mentioned, the gases and particles in the atmosphere function much like a venetian blind, selectively absorbing and reflecting certain wavelengths of solar radiation while allowing others to pass through relatively unhindered. In turn, physical processes in the atmosphere also help determine its chemical makeup. There was growing consensus through the 1970s and 1980s among Earth scientists that we needed to take a more holistic approach to global climate change studies. We saw that nature does not compartmentalize climate phenomena into discreet SEA ICE disciplines, and therefore we need to examine the variables of change as integral parts of the vast, interconnected web of cause and effect that is Earth’s climate system. In short, it is not enough to identify where and when changes occur; we need to understand how and why the mechanisms of change work. Satellite remote sensors offer the only viable means of conducting a comprehensive examination of our planet. See also Aerosols: Observations and Measurements. Radiative Transfer: Absorption and Thermal Emission; Scattering. Satellite Remote Sensing: Cloud Properties; Precipitation; TOMS Ozone. Satellites: Orbits. 2047 Further Reading Gurney RJ, Foster JL and Parkinson CL (1993) Atlas of Satellite Observations Related to Global Change. Cambridge: Cambridge University Press. King MD, Kaufman YJ, Tanré D and Nakajima T (1999) Remote sensing of tropospheric aerosols from space: past, present, and future. Bulletin of the American Meteorological Society 80: 2229–2259. Parkinson CL (1997) Earth From Above: Using Color-Coded Satellite Images to Examine the Global Environment. Sausalito: University Science Books. Ramanathan V, Barkstrom BR and Harrison EF (1989) Climate and the Earth’s Radiation Budget. Physics Today 42: 22–32. SEA ICE W F Weeks, University of Alaska, Fairbanks, AK, USA Copyright 2003 Elsevier Science Ltd. All Rights Reserved. Introduction Sea ice – any form of ice found at sea that originated from the freezing of sea water – has been among the least studied of all the phenomena that have a significant effect on the surface heat balance of the Earth. Fortunately this neglect has recently lessened as the result of improvements in observational and operational capabilities in the polar ocean areas. Thus considerable information is now available on the nature and behavior of sea ice as well as on its effect on the weather, the climate, and the oceanography of the polar regions and possibly of the planet as a whole. Extent Considering that the vast majority of Earth’s population have never seen sea ice, its areal extent is extremely impressive; 7% of the surface of the Earth is covered by it at some time of year. In the Northern Hemisphere the area varies between 8106 and 15106 km2, with the smaller number representing the area of multiyear (MY) ice remaining at the end of summer. In summer this corresponds roughly to the area of the contiguous United States and to twice that area in winter, or to between 5% and 10% of the surface of the Northern Hemisphere ocean. At maximum extent, the ice extends down the western side of the major ocean basins paralleling the pattern of cold currents and reaching the Gulf of St Lawrence (Atlantic) and the Okhotsk Sea off the north coast of Japan (Pacific). The most southerly site in the northern hemisphere where an extensive cover forms is the Gulf of Bo Hai, which is located off the east coast of China at 401 N. At the end of the summer the perennial MY ice pack of the Arctic is largely confined to the central Arctic Ocean, with minor extensions into the Canadian Arctic Archipelago and along the east coast of Greenland. In the Southern Hemisphere the sea ice area varies between 3106 and 20106 km2, covering between 1.5% and 10% of the ocean surface. The amount of MY ice in the Antarctic is appreciably less than in the Arctic, even though the total area affected in the Antarctic is approximately a third larger than in the Arctic. These differences are caused largely by differences in the spatial distributions of land and ocean. The Arctic Ocean is effectively landlocked to the south, with only one major exit located between Greenland and Svalbard. The Southern Ocean, on the other hand, is essentially completely unbounded to the north, which allows unrestricted drift of the ice in that direction and results in the summer melting of nearly all the previous season’s growth. Geophysical Importance In addition to its considerable extent, there are good reasons to be concerned with the health and behavior of the world’s sea ice covers. Sea ice serves as an insulative lid on the surface of the polar oceans. This suppresses the exchange of heat between the cold 2048 SEA ICE polar air above the ice and the relatively warm seawater below. Moreover, the snow cover the ice surface supports is an even better insulator than the ice itself. Also, when the sea ice forms with its attendant snow cover, it changes the surface albedo, a (i.e., the reflection coefficient for visible radiation) of the sea from that of open water (a ¼ 0:10) to that of newly formed snow (a ¼ 0:85). This results in a 75% decrease in the amount of incoming shortwave solar radiation absorbed. As a result, there are inherent positive feedbacks associated with the existence of a sea ice cover. For instance, if a climatic warming reduces both the extent and the thickness of the sea ice then these changes will, in turn, result in increases in the temperature of the atmosphere and of the sea, which will further reduce ice thickness and extent. This positive feedback is a major factor in producing the unusually large increases in arctic temperatures forecast by numerical models simulating the effect of the accumulation of greenhouse gases. The presence of an ice cover limits not only the flux of heat into the atmosphere but also the flux of moisture. This effect is revealed by the common presence of linear, local clouds associated with individual leads (cracks in the sea ice) that are covered with either open water or thinner ice. In fact, sea ice exerts a significant influence on the radiative energy balance of the complete atmosphere–sea ice–ocean system. For example, as the ice thickness increases in the range 0–70 cm, there is an increase in the radiation absorption in the ice and a decrease in the ocean. There is also a decrease in the radiation adsorption by the total atmosphere–ice–ocean system. It is now also known that the upper 10 cm of the ice can absorb over 50% of the total solar radiation, and that decreases in ice extent produce increases in atmospheric moisture or cloudiness, in turn altering the surface radiation budget and increasing the amount of precipitation. Furthermore, all the ultraviolet and infrared radiation is absorbed in the upper 50 cm of the ice. Only visible radiation penetrates into the lower portions of thicker ice and into the upper ocean beneath the ice. Significant changes in sea ice extent and/or thickness would clearly result in major changes in the climatology of the polar regions. For instance, recent computer simulations in which the ice extent in the southern hemisphere was held constant and the amount of open water (leads) within the pack was varied showed significant changes in storm frequencies, intensities and tracks, precipitation, cloudiness, and air temperature. However, there are even less obvious but perhaps equally important air–ice and ice–ocean interactions. Sea ice drastically reduces wave induced mixing in the upper ocean, thereby favoring the existence of a 25– 50 m thick, low-salinity surface layer in the Arctic Ocean that forms as the result of desalination processes associated with ice formation and the influx of fresh water from the great rivers of northern Siberia. This stable, low-density surface layer prevents the heat contained in the comparatively warm (temperatures of up to 131C) but more saline denser water beneath the surface layer from affecting the ice cover. As sea ice rejects roughly two-thirds of the salt initially present in the sea water from which the ice forms, the freezing process is equivalent to distillation producing both a low-salinity component (the ice layer itself) and a high salinity component (the rejected brine). Both components play important geophysical roles. Over shallow shelf seas the rejected brine, which is dense, cold, and rich in CO2 , sinks to the bottom, ultimately feeding the deep water and the bottom water layers of the world ocean. Such processes are particularly effective in regions where there are large polynyas (semipermanent open water and thin-ice areas at sites where climatically much thicker ice would be anticipated). In that this ‘salt pump’ removes CO2 from the atmosphere, it has been hypothesized that it is a process contributing to the decrease in CO2/air ratios observed in ice core samples deposited during times of maximum glacial advance (colder 5 more sea ice formation 5 more CO2 removed from the atmosphere). Certainly, whatever the effectiveness of this process, it will be less effective in removing CO2 if a climate warming causes less ice to form over the world’s shelf seas. Sea ice also has important biological effects at both ends of the marine food chain. It provides a substrate for a special category of marine life, the ice biota, consisting primarily of diatoms. These form a significant portion of the total primary production and, in turn, support specialized grazers and species at higher trophic levels, including amphipods, copepods, worms, fish, and birds. At the upper end of the food chain, seals and walruses use ice extensively as a platform on which to haul out and give birth to young. Polar bears use the ice as a platform while hunting. Also important is the fact that in shelf seas such as the Bering and Chukchi, which are well mixed in the winter, the melting of the ice cover in the spring lowers the surface salinity increasing the stability of the water column. The reduced mixing concentrates phytoplankton in the near-surface photic zone, thereby enhancing the overall intensity of the spring bloom. Finally there are the direct effects of sea ice on human activities. The most important of these are its barrier action in limiting the use of otherwise highly advantageous ocean routes between the northern Pacific regions and Europe and its contribution to the SEA ICE numerous operational difficulties that hinder the safe extraction of the presumed oil and gas resources of the polar shelf seas. Properties Because ice is a thermal insulator, the thicker it is the slower it grows, other conditions being equal. And because sea ice either ablates or stops growing during the summer, there is a maximum thickness of first-year (FY) ice that can form during a specific year. The exact value depends of course upon the local climate and oceanographic conditions, reaching slightly over 2 m in the Arctic and as much as almost 3 m at certain protected Antarctic sites. It is also clear that during the winter the heat flux from areas of open water into the polar atmosphere is significantly greater than the flux through even thin ice, and is as much as 200 times greater than the flux through MY ice. This means that even if open-water and thin-ice areas comprise less than 1–2% of the winter ice pack, lead areas must still be considered in order to obtain realistic estimates of ocean–atmosphere thermal interactions. If an ice floe survives a summer, during the second winter the thickness of the ice added to it is less than the thickness of nearby FY ice for two reasons: it starts to freeze later and it grows slower. Nevertheless, by the end of the winter, the second-year ice will be thicker than the nearby FY ice. Assuming the above process is repeated in subsequent years, some ice will be ablated away each summer (largely from the upper ice surface) and some added each winter (largely on the lower ice surface). As the years pass, the ice melted on top each summer will remain the same (assuming no change in the climate over the ice) while the ice forming on the bottom will become less and less as a result of the increased insulating effect of the thickening overlying ice. Ultimately a rough equilibrium will be reached, with the winter addition equalling the summer ablation. Such steady-state MY ice floes can be layer cakes of 10 or more annual layers with total thicknesses in the range of 3.5–4.5 m. Much of the uncertainty in estimating the equilibrium thickness of such floes is the result of uncertainties in the oceanic heat flux. However, in sheltered fiord sites in the Arctic where the oceanic heat flux is presumed to be near zero, MY fast ice with thicknesses of up to roughly 15–20 m is known to have occurred. Another important factor affecting MY ice thickness is the formation of melt ponds on the upper ice surface during the summer, in that the thicknesses and areal extent of these shallow surface water bodies is important in controlling the total absorption of shortwave radiation. For instance, a melt pond with a depth of only 5 cm can absorb 2049 nearly half the total energy absorbed by the whole system. The problem here is that good regional descriptive characterizations of these features are lacking as the result of the characteristic low clouds and fog that occur over the Arctic ice packs in the summer. Particularly lacking are field observations on melt pond depths as a function of environmental variables. Also needed are assessments of how much of the melt water remains ponded on the surface of the ice as contrasted with draining into the underlying seawater. Thermodynamically these are very different situations. Conditions in the Antarctic are, surprisingly, rather different. There, surface melt rates within the pack are small compared with those at the northern boundary of the pack. The stronger winds and lower humidities encountered over the pack also favor evaporation and minimize surface melting. The limited ablation that occurs appears to be controlled by heat transfer processes at the ice–water interface, so that the ice remains relatively cold throughout the summer. In any case, because most of the Antarctic pack is advected rapidly to the north, where it encounters warmer water at the Antarctic convergence and melts rapidly, only small amounts of MY ice remain at the end of summer. Sea ice properties are very different from those of lake or river ice. The reason for the difference is that when seawater freezes, roughly a third of the salt in the seawater is initially entrapped within the ice in the form of brine inclusions. As a result, initial ice salinities are typically in the range of 10–12 parts per thousand (ppt). At low temperatures (o  8.71C), solid hydrated salts also form within the ice. The composition of the brine in sea ice is a unique function of the temperature, with the brine composition becoming more saline as the temperature decreases. Therefore, the brine volume (the volumetric amount of liquid brine in the ice) is determined by the ice temperature and the bulk ice salinity. Not only does temperature vary with level in the ice sheet but salinity decreases as the ice ages, reaching a value of B3 ppt in MY ice. Brine volumes are usually lower in the colder upper portions of the ice and higher in the warmer lower portions. They are particularly low in the part of MY ice above sea level, from which the salt has drained almost completely. In fact, the upper layers of thick MY ice and of aged pressure ridges produce excellent drinking water when melted. As brine volume is the single most important parameter controlling the thermal, electrical, and mechanical properties of sea ice, these properties show associated large changes both vertically in the same ice sheet and between ice sheets of differing ages and histories. To add complexity to this situation, exactly how the 2050 SEA ICE Drift and Deformation If sea ice were motionless, ice thicknesses would be controlled completely by the thermal characteristics of the lower atmosphere and the upper ocean. Such ice sheets would presumably have thicknesses and physical properties that would change slowly and continuously from region to region. However, even a casual examination of an area of pack ice reveals striking local lateral changes in ice thicknesses and characteristics. These changes are invariably caused by ice movements produced by the forces exerted on the ice by winds and currents. Such motions are rarely uniform and lead to the buildup of stresses within ice sheets. If these stresses become large enough, cracks may form and widen, resulting in the formation of leads. Such features can vary in width from a few meters to several kilometers and in length from a few hundred meters to several hundred kilometers. As mentioned earlier, during much of the year in the polar regions, once a lead forms it is immediately covered with a thin skim of ice that thickens with time. This is an ever changing process associated with the movement of weather systems as one lead system becomes inactive and is replaced by another oriented in a different direction. Because lead formation occurs at varied intervals throughout the ice growth season, the end result is an ice cover composed of a variety of thicknesses of uniform sheet ice. However, real pack ice thickness distributions (Figure 1) show that there is a significant amount of ice thicker than the 4.5–5.0 m maximum that might be expected for steady-state MY ice floes. This thicker ice forms by the closing of leads, which commonly results 0.40 _1 Probability density (m ) brine is distributed within the sea ice also affects ice properties. There are several different structural types of sea ice, each with characteristic crystal sizes and preferred crystal orientations and property variations, the two commonest being ‘congelation’ and ‘frazil’. In congelation ice, large elongated crystals extend completely through the ice sheet, producing a structure similar to that of directionally solidified metals. In the Arctic, large areas of congelation ice show crystal orientations so similar that they cause the ice to have directionally dependent properties in the horizontal plane as if it were a giant single crystal. Frazil, on the other hand, is composed of small, randomly oriented, equiaxed crystals that are not vertically elongated. Congelation is more common in the Arctic while frazil is more common in the Antarctic, reflecting the more turbulent conditions characteristically found in the Southern Ocean. 0.20 0 0 2 4 6 8 10 Draft (m) Figure 1 The distribution of sea ice drafts expressed as probability density as determined via the use of upward-looking sonar along a 1400 km track taken in April 1976 in the Beaufort Sea. All ice thicker than B4 m is believed to be the result of deformation. The peak probablities in the range between 2.4 and 3.8 m represent the thicknesses of undeformed multiyear ice, while the values less than 1.2 m come from ice that formed more recently in leads. in the piling of broken ice fragments into long, irregular features or ‘pressure ridges’. There are many small ridges, and large ones are rare. Nevertheless, the large ridges are very impressive, the largest free-floating sail height and keel depth reported to date in the Arctic being 13 and 47 m (values not from the same ridge). Particularly heavily deformed ice commonly occurs in a B150 km band running between the north coast of Greenland and the Canadian Arctic Islands and the south coast of the Beaufort Sea. The limited data available on Antarctic ridges suggest they are generally smaller and less frequent than ridges in the Arctic Ocean. The general pattern of the ridging is also different in that the long, sinuous ridges characteristic of the Arctic Ocean are not observed. Instead, the deformation can be better described as irregular hummocking accompanied by extensive rafting of one floe over another. Floe sizes are also smaller as the result of the passage of large-amplitude swells through the ice. These are generated by the intense Southern Ocean storms that move to the north of the ice edge and result in the fracturing of the larger floes, while the large vertical motions facilitate the rafting process. Pressure ridges are of considerable importance for four reasons. First, they change the surface roughness at the air–ice and water–ice interfaces, thereby altering the effective surface tractions exerted by winds and currents. Second, they act as plows, forming gouges in the sea floor up to 8 m deep when they ground and are SEA ICE pushed along by the ungrounded pack as it drifts over the shallower (o60 m) regions of the polar continental shelves. Third, as the thickest sea ice masses, they are a major hazard that must be considered in the design of offshore structures. Finally, and most important, the ridging process provides a mechanical procedure for transferring the thinner ice in the leads directly and rapidly into the thickest ice categories. Considerable information on the drift and deformation of sea ice has recently become available through the combined use of data buoy and satellite observations. This shows that on average there are commonly two primary ice motion features in the Arctic Basin, namely the Beaufort Gyre, a large clockwise circulation located in the Beaufort Sea, and the Trans-Polar Drift Stream, which transports ice formed on the Siberian Shelf over the Pole to Fram Strait between Greenland and Svalbard. The time required for the ice to complete one circuit of the gyre averages 5 years, while the transit time for the Drift Stream is roughly 3 years with about 9% of the sea ice of the Arctic Basin (919 000 km2) moving south through the Fram Strait and out of the basin each year. There are many interesting features of the ice drift that exist over shorter time intervals. For instance, recent observations show that the Beaufort Gyre may run backwards (counterclockwise) over appreciable periods of time, particularly in the summer and fall. It has also been suggested that such reversals can occur on decadal time scales. Typical pack ice velocities range from 0 to 20 cm s  1 although extreme velocities of up to 220 cm s  1 (4.3 knots) have been recorded during storms. During winter, periods of zero ice motion are not rare. During summer, when considerable open water is present in the pack, the ice appears to be in continuous motion. The highest drift velocities are invariably observed near the edge of the pack. Not only are such locations commonly windy, but the floes are able to move toward the free edge with minimal inter-floe interference. Ice drift near the Antarctic continent is generally westerly, becoming easterly further north, but in all cases showing a consistent northerly diverging drift toward the free ice edge. Various sea ice formations are shown in Figure 2. Trends Considering the anticipated geophysical consequences of changes in sea ice extent, it is not surprising that there is considerable interest in the subject. Is sea ice expanding and thickening, heralding a new glacial age, or retreating and thinning before the onslaught of a greenhouse-gas-induced heat wave? One thing that 2051 should be clear from the preceding discussion is that the ice is both surprisingly thin and variable. Therefore small changes in meteorological and oceanographic forcings could result in significant changes to the extent and state of the ice cover. They could also produce feedbacks that might have significant and complex climatic consequences. Before we examine what is known about sea ice variations, let us first examine other related observations that have a direct bearing on the question of sea ice trends. Land station records for 1966–96 show that the air temperatures have increased, with the largest increases occurring during winter and spring over both north-west North America and Eurasia, a conclusion that is supported by increasing permafrost temperatures. In addition, meteorological observations collected on Russian pack ice drifting stations deployed in the Arctic Basin show significant warming trends for the spring and summer periods. It has also recently been suggested that when proxy temperature sources are considered, they indicate that the late 20th-century Arctic temperatures are the highest in the past 400 years. Recent oceanographic observations also relate to the above questions. In the late 1980s the balance between the Atlantic water entering the Arctic Basin and the Pacific water appears to have changed, resulting in an increase in the areal extent of the more saline, warmer Atlantic water. In addition the Atlantic water is shallower than in the past, resulting in temperature increases of as much as 21C and salinity increases of up to 2.5 ppt at depths of 200 m. The halocline, which isolates the cold near surface layer and the overlying sea ice cover from the underlying warmer water, also appears to be thinning; a fact that could profoundly affect the state of the sea ice cover and the surface energy budget in the Arctic. Recent changes as revealed by the motions of data buoys placed on the ice show that there has been a weakening of the Beaufort Sea Gyre and an associated increased divergence of the ice pack. There are also indications that the MY ice in the center of the Beaufort Gyre is less prevalent and thinner than in the past and that the amount of surface melt has increased from B0.8 m in the mid 1970s to B2 m in 1997. This conclusion is supported by the operational difficulties encountered by recent field programs that have attempted to maintain on-ice measurements. The increased melt also is in agreement with observed decreases in the salinity of the near surface water layer. It is currently believed that the above changes appear to be related to atmospheric changes in the Polar Basin, where the mean atmospheric surface pressure is decreasing and has been below the 1979–95 mean every year since 1988. Before B1988–99 the 2052 SEA ICE (A) (B) (C) (D) (E) (F) Figure 2 (A) Ice gouging along the coast of the Beaufort Sea. (B) Aerial photograph of an area of pack ice in the Arctic Ocean showing a recently refrozen large lead that has developed in the first year. The thinner newly formed ice is probably less than 10 cm thick. (C) A representative pressure ridge in the Arctic Ocean. (D) A rubble field of highly deformed first-year sea ice developed along the Alaskan coast of the Beaufort Sea. The tower in the far distance is located at a small research station on one of the numerous offshore islands located along this coast. (E) Deformed sea ice along the NW Passage, Canada. (F) Aerial photograph of pack ice in the Arctic Ocean. Beaufort High was usually centered over 1801 longitude. Since then the high has been both weaker and typically confined to more western longitudes, which may account for lighter ice conditions in the western Arctic. There also has been a recent pronounced increase in the frequency of cyclonic storms in the Arctic Basin. So, are there also direct measurements indicating decreases in ice extent and thickness? Historical data based on direct observations of sea ice extent are rare, although significant long-term records do exist for a few regions such as Iceland, where sea ice has an important effect on both fishing and transportation. In monitoring the health of the world’s sea ice covers the SEA ICE use of satellite remote sensing is essential because of the vast remote areas that must be surveyed. Unfortunately the satellite record is very brief. If data from only microwave remote sensing systems are considered, because of their all-weather capabilities, the record is even shorter, starting in 1973. As there was a 2-year data gap between 1976 and 1978, only 25 years of data are available to date. The imagery shows that there are definitely large seasonal, inter-annual and regional variations in ice extent. For instance, a decrease in ice extent in the Kara and Barents Seas contrasts with an increase in the Baffin Bay–Davis Strait region and out-of-phase fluctations occur between the Bering and the Okhotsk Seas. The most recent study, which examined passive microwave data through December 1996, concludes that the areal extent of Arctic sea ice has decreased by 2.970.4% per decade. In addition, record or near-record minimum areas of Arctic sea ice were observed in 1990, 1991, 1993, 1995, and 1997. A particularly extreme recession of the ice along the Beaufort coast was also noted in the fall of 1998. Russian ice reconnaissance maps also show that a significant reduction in ice extent and concentration has occurred over much of the Russian Arctic Shelf since 1987. Has a systematic variation also been observed in ice thickness? Unfortunately there is, at present, no satellite-borne remote sensing technique that can measure sea ice thicknesses effectively from above. There is also little optimism about the possibility of developing such techniques because the extremely lossy nature of sea ice limits penetration of electromagnetic signals. Current ice thickness information comes from two very different techniques: in situ drilling and upward-looking, submarine-mounted sonar. Although drilling is an impractical technique for regional studies, upward-looking sonar is an extremely effective procedure. The submarine passes under the ice at a known depth and the sonar determines the distance to the underside of the ice by measuring the travel times of the sound waves. The result is an accurate, well-resolved under-ice profile from which ice draft distributions can be determined and ice thickness distributions can be estimated based on the assumption of isostacy. Although there have been many under-ice cruises starting with the USS Nautilus in 1958, to date only a few studies have been published that examine temporal variations in ice thickness in the Arctic. The first compared the results of the cruise of the Nautilus in 1958 with those of the nearly identical cruise of USS Queenfish in 1970. Decreases in mean ice thickness were observed in the Canadian Basin (3.08–2.39 m) and in the Eurasian Basin (4.06–3.57 m). The second study compared the results of two Royal Navy cruises 2053 made in 1976 and 1987 and obtained a 15% decrease in mean ice thickness for a 300 000 km2 area north of Greenland. Although these studies show similar trends, the fact that they each utilized only 2 years of data caused many scientists to feel that a conclusive trend had not been established. However, a recent study has been able to examine this problem in more detail by comparing data from three submarine cruises made in the 1990s (1993, 1996, 1997) with the results of similar cruises made between 1958 and 1976. The area examined was the deep Arctic Basin and the comparisons used data only from the late summer and fall periods. It was found that the mean ice draft has decreased by about 1.3 m from 3.1 m in 1958–76 to 1.8 m in the 1990s, with a larger decrease occurring in the central and eastern Arctic than in the Beaufort and Chukchi Seas. This is a very large difference, indicating that the volume of ice in the region surveyed is down by some 40%. Furthermore, an examination of the data from the 1990s suggests that the thickness decrease is continuing at a rate of about 0.1 m yr  1. Off the Antarctic the situation is not as clear. One study has suggested a major retreat in maximum sea ice extent over the last century based on comparisons of current satellite data with the earlier positions of whaling ships reportedly operating along the ice edge. As it is very difficult to access exactly where the ice edge is located solely from shipboard observations, this claim has met with some skepticism. An examination of the satellite observations indicates a very slight increase in areal extent since 1973. As there is no upward-looking sonar data for the Antarctic Seas the thickness data base there is very small. However, limited drilling and airborne laser profiles of the upper surface of the ice indicate that in many areas the undeformed ice is very thin (60–80 cm) and that the amount of deformed ice is not only significantly less than in the Arctic but would only add roughly 10 cm to the mean ice thickness. What are we to make of all of this? It is obvious that, at least in the Arctic, a change appears to be underway that extends from the top of the atmosphere to depths below 1000 m in the ocean. In the middle of this is the sea ice cover which, as has been shown, is extremely sensitive to environmental changes. What is not known is whether these changes are part of some cycle or whether they represent a climatic regime shift in which the positive feedbacks associated with the presence of a sea ice cover play an important role. Also not understood are the interconnections between what is happening in the Arctic and other changes both inside and outside the Arctic. For instance, could changes in the Arctic system drive significant lower-latitude atmospheric and oceanographic changes, or are the Arctic changes driven by more 2054 SEVERE STORMS dynamic, lower-latitude processes? In the Antarctic the picture is even less clear, although changes are known to be under way, as is evidenced by the recent breakup of ice shelves along the eastern coast of the Antarctic Peninsula. Not surprisingly, the scientific community is currently devoting considerable energy to attempt to answer these questions. We could say that a cold subject is heating up. See also Arctic Climate. Snow (Surface). Further Reading Cavelieri DJ, Gloersen P, Parkinson CL, Comiso JC and Zwally HJ (1997) Observed hemispheric asymmetry in global sea ice changes. Science 278(5340): 1104–1106. Ebert EE and Curry JA (1993) An intermediate onedimensional thermodynamic sea ice model for investigating ice–atmosphere interactions. Journal of Geophysical Research 98(C6): 10085–10109. Jin Z, Stamnes K, Weeks WF and Tsay SC (1994) The effect of sea ice on the solar energy budget in the atmosphere-sea ice-ocean system: a model study. Journal of Geophysical Research 99(C12): 25281– 25294. Leppäranta M (ed.) (1998) Physics of Ice-covered Seas, 2 vols. Helsinki: Helsinki University Printing House. Rothrock DA, Yu Y and Maykut GA (1999) Thinning of the Arctic sea-ice cover. Geophysical Research Letters 22: 3469–3472. Untersteiner N (ed.) (1986) The Geophysics of Sea Ice. NATO Advanced Science Institutes Series B, Physics, vol. 146. New York: Plenum Press. Dyer I and Chryssostomidis C (eds) (1984) Arctic Technology and Policy. New York: Hemisphere. SEVERE STORMS C A Doswell III, University of Oklahoma, Norman, OK, USA Copyright 2003 Elsevier Science Ltd. All Rights Reserved. Introduction The word ‘storm’ implies a disturbance of some sort in the weather, but many different types of weather can result in an event called a storm. Thus, it is possible to have windstorms, dust storms (which also are windstorms), hailstorms, thunderstorms, winter storms, tropical storms, and so on. Generally speaking, events called storms are associated with cyclones; undisturbed weather is usually found with anticyclones. Similarly, the meaning of severity needs to be considered. The intensity of the event in question will be the basis for deciding on the severity of that particular storm. However, if storm intensity is to be the basis for categorizing a storm as severe, it has to be decided what measure is to be used for intensity. This also implies an arbitrary threshold for deciding the issue of severity. That is, weather events of a given type will be called severe when some measure of that event’s intensity meets or exceeds a threshold that is usually more or less arbitrary. A hailstorm might be severe when the hailstone diameters reach 2 cm or larger; a winter snowstorm might be called severe when the snowfall rate equals or exceeds 5 cm h 1. On the other hand, some storms of any intensity might be considered severe. A tornado is a ‘storm’ embedded within a thunderstorm; any tornado of any intensity is considered a severe storm. The difficulty with arbitrary definitions is that they imply a change in character whenever the threshold criterion is met. That is, if a hailstorm produces hailstones 1.9 cm in diameter, a threshold of 2 cm means that such a storm is not severe. Is it reasonable, from most people’s viewpoint, to try to distinguish a storm producing 1.9 cm diameter hailstones from one producing 2 cm hailstones? In the majority of cases within the science of meteorology, there is no obvious way to distinguish events with this sort of precision. A small quantitative change in some intensity measurement is not necessarily associated with a qualitative change in the character of a storm. It is near the threshold (wherever that threshold is chosen) that it becomes challenging to analyze and predict storm ‘severity’. This will be elaborated on in dealing with the specific events described below. However, the challenge of defining severity should be kept in mind in the following discussion, as we consider various types of severe storms. Severe Mid-latitude Storms The tropics are defined formally as lying equatorward of 23.51 latitude in the Northern and Southern Hemispheres: the ‘Tropics’ of Cancer and Capricorn, respectively. Poleward of these latitudes and equatorward of 601 latitude lie the so-called mid-latitudes. SEVERE STORMS There are important distinctions between the weather of mid-latitudes and that of the tropics. Notably, in mid-latitudes, the Coriolis force is an important part of the meteorology, whereas in the tropics, its impact on the large-scale weather is less. Synoptic-Scale Storms Cyclones in mid-latitudes that are thousands of kilometers in horizontal extent are known as synoptic-scale systems. These are the familiar rotating weather systems (Figure 1) shown routinely in newspapers and on television. Such storms serve an important function in the global circulation, helping to carry warm air poleward from the tropics and cold air from the polar regions equatorward. This process keeps the imbalance of solar heating from creating too extreme a temperature contrast between the poles and the Equator. In association with these synoptic-scale cyclones, intense temperature contrasts can develop (Figure 2), called fronts, which are the leading edges of the cold air flowing equatorward and the warm 2055 air flowing poleward. These mid-latitude cyclones are part of the normal progression of weather systems, typically bringing clouds and precipitation with them. In some situations, when the hemispheric weather patterns have become slow-moving, these cyclones can bring prolonged periods of heat or cold to some regions. When extreme temperatures (hot or cold) are reached, these can be hazardous to humans for a variety of reasons, but would not generally be considered ‘storms’. Synoptic-scale cyclones can become particularly intense, and the pressures at their cores can become quite low in comparison to the average. Often, in the process of intensification, the pressures can fall quite rapidly as the result of the dynamic processes operating to cause the intensification. Cyclones with fastfalling pressures are sometimes called ‘bombs’ and, whereas they can be considered storms in their own right, the cyclones may be responsible for several different types of stormy weather. Figure 1 False-color-enhanced satellite image of a synoptic-scale cyclone on the afternoon of 10 November 1998, showing the center of the cyclone near the spiral of clouds in south-eastern Minnesota. This cyclone was producing severe thunderstorms in and near the Gulf of Mexico, as well as snow and high winds on the northern plains, in North and South Dakota. (Source: NOAA.) 2056 SEVERE STORMS Figure 2 Map of surface temperatures at about the same time as Figure 1, showing the strong contrast in temperature along the cold front, with subfreezing temperatures in North and South Dakota at the same time that quite warm temperatures are present over the Gulf of Mexico. Many subsynoptic-scale features also can be seen in mountainous regions; for example, in the Appalachian and Rocky Mountains. (Source: NOAA.) Rapidly falling pressures create strong winds over a wide region. These windstorms resulting from synoptic-scale cyclones can produce considerable damage and associated casualties; recent examples occurred in France during December 1999 and along the east coast of the United States in February 2000 and March 1993 (the so-called ‘Superstorm’ of 1993). Another wellpublicized example hit the United Kingdom in October 1987. The damaging winds can extend over many hundreds of kilometers and last in any one place for a full day or more. The result of such widespread damaging wind can be overwhelming to emergency services, and power outages alone can last for days in some places simply because of the sheer size of the affected area. At sea, strong winds from intense synoptic-scale cyclones produce large waves that represent hazards to ships of all sorts. While still at sea, the winds from intense cyclones can cause serious damage, including beach erosion, when they affect coastal areas. In addition, intense synoptic-scale cyclones can produce a full spectrum of hazardous precipitation. The time of such storms runs from Fall through Spring, and so the cyclones are capable of producing paralyzing snowstorms, ice storms, heavy rainstorms, and even severe thunderstorms. Accumulations of ice and snow during winter storms of this type are potentially hazardous to ships and aircraft. Depending on the circumstances, two or more of these different severe weather types could be happening at the same time in different places. A given location might experience all of them in the course of a single day during the passage of a synoptic-scale cyclone. In other situations, only one form of severe weather occurs within such a cyclone. Synoptic-scale cyclones are important in creating the conditions for the development of smaller-scale storms. It is a general principle in meteorology that, as the size of a weather system decreases, the maximum intensity of the weather it can create increases. Although synoptic-scale systems certainly can produce widespread damage, it is usually not of the most extreme intensity. However, the conditions within such storms can result in smaller concentrations of severe weather that become even more potentially hazardous. Mesoscale Storms Whereas synoptic-scale weather happens on scales of several thousand kilometers, mesoscale weather is in the range of hundreds of kilometers. Synoptic-scale weather processes go on essentially all the time (although the really intense events are generally rare), whereas mesoscale storms are intermittent. That is, they arise only occasionally in any given location and then only when the conditions for their formation are produced by the processes operating on the synoptic scale. There are two general classes of mesoscale storm systems: those that arise from interactions between the atmosphere and the underlying surface, and those that occur even in regions of uniform conditions at the surface. Those systems that depend on the underlying surface cover a wide range of phenomena. There are many atmospheric circulations, like land–sea breezes, SEVERE STORMS that are more or less routine processes, driven by the underlying topographic conditions; in the case of the land–sea breeze, it is the temperature contrast between the land and the sea that drives the flow. During the day, the land is warmer and air tends to rise over land, to be replaced by cooler air flowing in from the sea. At night, the opposite happens. Of course, most of these circulations would not be considered ‘storms’ in the sense that we have been using. However, such processes as land–sea breezes can be influential in the development of stormy weather, often in the form of thunderstorms that are initiated along them. Occasionally, the circumstances produced by the synoptic-scale flow as it interacts with the surface result in stormy conditions. A common example occurs when the air flow interacts with complex terrain, producing localized windstorms. There are examples of these mesoscale windstorms around the world, often given colorful names. Mesoscale windstorms such as the Chinook (in Alaska), the Foehn (in the European Alps), the Traumontana (in the western Mediterranean), and the Bora (in the Adriatic) have been recognized as important weather events for centuries. Windstorms in complex terrain arise in different circumstances; they are not all driven by the same mechanism. Some are simply cases where cool, stable air is being funneled through gaps in the terrain (e.g., the Traumontana); others develop when strong winds aloft are brought down to the surface by processes induced by air flow over the mountains (as in Boulder, Colorado). The situation creating the windstorms is created by the synoptic-scale flow, but the strongest winds are confined to a mesoscale area. Another class of mesoscale storms can arise when cold air flows over relatively warm waters. Storms of this sort, called ‘polar lows’, apparently arise through processes not unlike those of tropical cyclones, drawing energy from the ocean to develop their intense circulations. They occur when outbreaks of very cold polar air flows over relatively warm waters. Given their mesoscale size, they often are characterized by intense pressure gradients, leading to the occurrence of strong windstorms. Their size means that the weather they bring may only last for part of a day, but during the passage of the storm, winds can meet, and even exceed, the hurricane threshold of 33.5 m s  1. The windstorms associated with polar lows can be quite hazardous, especially when they occur in association with low temperatures (resulting in severe windchill conditions). In addition, polar lows can produce blinding snowstorms with snowfall rates of perhaps 200 mm h 1, leading to extremely dangerous blizzard conditions. On some occasions they can be associated with strong and possibly severe thunderstorms as well. 2057 Even when the underlying surface is more or less uniform, mesoscale storms can develop within synoptic-scale cyclones (Figure 3). These are usually tied to a disturbance in the middle or upper troposphere that encounters conditions favorable for its development. Such systems can produce unforeseen snow and ice storms in the winter, and severe thunderstorms during the warm season. There may not be a strong cyclone near the surface in such events. Severe Thunderstorms Severe thunderstorms typically produce weather events that cover a wide range of size scales, from a few hundred kilometers down to just a few kilometers or even smaller. This is because thunderstorms can occur as isolated events or in groups. In the United States, a thunderstorm-related event is considered severe when the wind gusts equal or exceed 25 m s 1, or when the hailstone diameters exceed 2 cm, or if a tornado is produced. A thunderstorm is composed of one or more cells, where a cell is the basic building block of a thunderstorm. Cells, in turn, are viewed as being made up of one group of air parcels being driven upward by positive buoyancy and another being driven downward by negative buoyancy, with the presence of precipitation in the air. Positive buoyancy arises in updrafts by the release of latent heat during the condensation of water vapor. This heat release acts like the burner of a hot air balloon, reducing the density of the air in which condensation is occurring and thereby causing the air to rise. As the process continues, Figure 3 An example of a polar low in the cold airstream behind a wintertime synoptic-scale cold front associated with a synopticscale cyclone (a low-pressure center). (Source: NOAA.) 2058 SEVERE STORMS precipitation can be formed in the updraft. This precipitation can produce downdrafts simply by its accumulating weight dragging downward on the surrounding air. Moreover, when precipitation falls into relatively dry air surrounding a developing storm, the evaporation of that precipitation chills the air because evaporation absorbs latent heat from the air in the same way that condensation releases that heat. When downdrafts caused by thunderstorms reach the surface, they are forced to spread out, like pancake batter poured onto a griddle. This creates an outflow at the surface (often called a downburst), with the outflow winds sometimes reaching the criterion for calling the thunderstorm severe. On some occasions, these outflow winds can exceed 40 m s 1. Under the right circumstances, notably when the updraft is particularly strong, the possibility of hail formation arises. Hailstones develop in the part of the storm where supercooled water and ice crystals are both present; liquid water is said to be supercooled when its temperature is below the melting point (01C) and the water is not yet frozen. Hailstones can become quite large, exceeding 5 cm diameters at times, and can be capable of penetrating roofs, shattering windows, and even creating human casualties. Even small hail can cause crop damage, of course. Occasionally, tornadoes form in association with severe thunderstorms. Tornadoes are intense lowpressure vortices that can produce the strongest winds of any storm: at their highest intensity, tornadic windspeeds can approach 140 m s 1. Most tornadoes, however, are not that intense. Tornadoes over bodies of water are called waterspouts. Tornadoes are created in thunderstorms when pretornadic, relatively weak circulations are intensified through conservation of angular momentum. Isolated Events The most intense form of thunderstorm is the so-called supercell thunderstorm, which typically is isolated from surrounding storms. Supercells are rotating thunderstorms that develop their rotation by tapping the vertical wind shear in the storm environment. The vast majority of supercells produce some sort of severe weather: hail, damaging straight-line winds, and/or tornadoes; only about 20% of them are tornadic. The most violent severe weather of all types is almost always associated with supercells (Figure 4), including the majority of strong and violent (F2–F5 on the Fujita Scale) tornadoes and giant hailstones (exceeding 5 cm in diameter). Although the typical thunderstorm cell has a lifetime of about 20–30 min, supercells can persist for many hours. This means that all forms of severe weather from supercells can be prolonged, sometimes leaving long, wide swaths of damage. The organized nature of a supercell, associated with its overall rotation, means that supercells produce a disproportionate share of the damage associated with thunderstorms. Perhaps only about 10% of all thunderstorms are supercells, but they are responsible for the majority of thunderstorm damage in areas where they occur. Because supercell updrafts are often intense, supercells can become prolific hail producers; a noteworthy example was a supercell that hit the Dallas–Fort Worth metroplex on the evening of 5 May 1995, with Figure 4 Supercell-associated tornado on 22 May 1981, near Alfalfa, Oklahoma. (Image r 2000 C. Doswell (used with permission).) SEVERE STORMS softball-sized hail and torrential rains. The damage from that one storm was estimated at $1 billion. Apart from supercells, isolated thunderstorms are usually not severe and typically do not last very long. On rare occasions, isolated thunderstorms can produce a brief ‘pulse’ of severe weather, usually hail or winds that are only marginally severe. Aggregations of Thunderstorms Thunderstorms do not typically occur as isolated events. Instead, they tend to form in groups, in either lines or in clusters of individual cells. The most common grouping is in lines, sometimes referred to as squall lines. When thunderstorm cells form in aggregations, the collection of storms can live for much longer than the individual cells (which usually retain their 20–30 min life cycles). This means that the hail and wind events produced by such groupings of thunderstorms are intermittent, rather than prolonged (as with supercells), as cells form and decay. Severe weather can go on in such cases for many hours in this intermittent fashion. The interactions between individual cells in lines and clusters of thunderstorm cells are often complicated and hard to predict, but those interactions are responsible for severe weather. A particularly dangerous form of thunderstorm aggregation arises when new cells are constantly forming in one place and tracking over the same region repeatedly, a situation called ‘training’ because the cells are like carriages in a train. This means that a particular area experiences rainfall from a succession of thunderstorm cells, which can result in extremely heavy rainfall. This is the process associated with the majority of flash flood events worldwide. In the United States, heavy rainfall is not considered to be a criterion for what is officially considered to be ‘severe’ despite the importance of such rainfall in flooding events. On the other hand, many other nations around the world consider heavy rainfall to be an important form of severe storm. 2059 peak sustained winds (i.e., not gusts) can approach 90 m s1 in extreme cases. The size of the region of damaging winds can vary considerably from one event to another, but winds exceeding ‘hurricane force’ (33.5 m s1) can be found within a circle on the order of 100 km or so in diameter. With such a large region of strong winds, damaging windspeeds can go on for many hours. Although they are well known for strong winds, tropical cyclones can pack a lethal combination of hazards: storm surge, heavy rainfalls, and even embedded tornadoes, as well as the better-known strong winds. Storm surge is created by a combination of strong winds and low pressure, resulting in an elevated sea level near the center of the storm. When this surge, which can be several meters high, makes landfall, low-lying coastal regions can be inundated. Nor is the rainfall component to be taken lightly. Hurricane Mitch (Figure 5) devastated parts of Nicaragua and Honduras in 1998, mostly from flash floods and landslides. There were more than 9000 fatalities, making it the worst weather disaster in this century in the Western Hemisphere. Tropical cyclones are usually several hundred kilometers in diameter and can last for tens of days. Their paths often take them out of the tropics into midlatitudes, where they can maintain their structure for a time before eventually dissipating or transforming into midlatitude cyclones. Tropical storms usually dissipate shortly after making landfall, because their energy source (warm sea water) is cut off. Nevertheless, dissipated tropical cyclonic storms can remain dangerous well after they lose their strong winds by creating an environment favorable for heavy rainproducing thunderstorms. Relatively little is known about other types of severe storms in the tropics. Severe thunderstorms, especially supercells, are uncommon in the tropics because of a general lack of vertical wind shear. Of course, heavy rain-producing tropical thunderstorms are relatively common in some parts of the tropics. Severe Tropical Storms The most obvious form of severe weather associated with the Tropics is the tropical cyclone. Tropical cyclones are known by different names in different parts of the world: hurricanes (in North America), typhoons (in the tropical Pacific), and cyclones (in the Indian Ocean and Australia), among others, but they are all the same phenomenon. Such storms arise when sea surface temperatures become warm, the vertical wind shear is weak, and tropical weather disturbances move through the easterly Trade winds of the tropics. They produce winds in excess of 33.5 m s1 and the Societal Impacts and Their Mitigation Severe storms in all their variety cause the loss of hundreds of lives and several billion dollars in property during the course of a year in the United States. It is worth noting that the United States can recover from such property damage because of its large, generally healthy economy. Economic losses from severe storms in the United States are typically much less than one percent of the gross domestic product (currently several trillion dollars), so by spreading out the impact of severe storms, the areas affected can recover and 2060 SEVERE STORMS Figure 5 View from the GOES-8 geostationary satellite of Hurricane Mitch near Honduras and Nicaragua. (Source: NOAA.) rebuild. On the other hand, when severe storms (like Hurricane Mitch) devastate less-developed nations with small economies, the damage to their infrastructure can be so large that it might take decades to recover. Forecasting severe storms has shown a slow increase in accuracy during the past several decades, as new technologies are leading to improved understanding and predictability. The accuracy of forecasts generally increases as the scale of the storm increases; it is possible to be more accurate with a synoptic-scale forecast than with a forecast on the scale of a single thunderstorm in most cases. There is more complete understanding of the synoptic-scale meteorology than that on scales smaller than synoptic. Furthermore, forecast accuracy generally decreases with the age of the forecast, at a rate that also depends on the scale. In general, the accuracy of a synoptic-scale forecast stays high for longer than does a short-range forecast of a thunderstorm-scale event. Mitigation of property damage depends mostly on making the right preparations for the storms that are possible in a given location, well in advance of the storms. Once the storms are under way, there tends to be relatively little that can be done to prevent property damage. For example, a home built on a barrier island Figure 6 Damage caused by the violent tornado that hit the city of Moore, Oklahoma on 3 May 1999. SNOW (SURFACE) that can be swept by landfalling tropical cyclones is unlikely to remain undamaged for more than a few decades, at most. Thus, some damage can be avoided by not building in vulnerable areas. As another example, there are several ways in which homes can be built to resist tornado damage (Figure 6), unless the homeowner is unlucky enough to be hit by the most intense winds in a violent tornado. Even within the whole violent tornado damage area, only a few places will actually experience the most violent winds; most of the rest of the structures will encounter winds that can be resisted through appropriate construction. Mitigation of casualties can also be a complex undertaking. In some instances, as with tropical cyclones, evacuation is possible and may be the best way to protect lives when it is feasible. For tornadoes, access to a suitable shelter is preferred; in situations where proper shelter is not available, the alternatives during tornadoes are not very good. In flooding situations, evacuation to higher ground is the appropriate way to prevent casualties, when time permits. Clearly, our ability to detect and predict severe storms is also important for casualty mitigation. In the United States, there has been a gradual reduction in weather-related fatalities with time, in part because there are fewer ‘surprise’ storms today and in part because education about severe storm hazards has led to improved public preparations. Nevertheless, we continue to be vulnerable to disasters caused by severe storms, and complacency can be a fatal error. See also Air–Sea Interaction: Storm Surges. Bow Echos and Derecho. Convective Storms: Overview. Cyclogene- 2061 sis. Cyclones, Extra Tropical. Downslope Winds. Flooding. Lake Effect Storms. Mesoscale Meteorology: Mesoscale Convective Systems. Microbursts. Orographic Effects: Lee Cyclogenesis. Polar Lows. Tornados. Waterspouts. Further Reading Agnone JC (ed.) (1995) Raging Forces: Earth in Upheaval. Washington DC: National Geographic Society. Anthes R (1982) Tropical Cyclones. Their Evolution, Structure and Effects. Boston: American Meteorological Society. Church C, Burgess D, Doswell C and Davies-Jones R (eds) (1993) The Tornado: Its Structure, Dynamics, Prediction, and Hazards. Washington, DC: American Geophysical Union. Doswell CA III (ed.) (2001) Severe Convective Storms. Boston: Amererican Meteorological Society. Foote GB and Knight CA (eds) (1977) Hail: A Review of Hail Science and Hail Suppression. Boston: American Meteorological Society. Hill CE (ed.) (1986) Nature on the Rampage: Our Violent Earth. Washington, DC: National Geographic Society. Junger S (1997) The Perfect Storm. New York: W.W. Norton. Lamb H (1991) Historic Storms of the North Sea, British Isles and Northwest Europe. Cambridge: Cambridge University Press. Lorenz EN (1993) The Essence of Chaos. Seattle: University of Washington Press. Ludlam FH (1980) Clouds and Storms. Philadelphia: Pennsylvania State University Press. Ray PS (ed.) (1986) Mesoscale Meteorology and Forecasting. Boston: Amererican Meteorological Society. Shapiro M and Grnås S (eds) (1999) The Life Cycles of Extratropical Cyclones. Boston: American Meteorological Society. SNOW (SURFACE) M Sturm, US Army Cold Regions Research & Engineering Laboratory-Alaska, Fort Wainwright, AL, USA Copyright 2003 Elsevier Science Ltd. All Rights Reserved. Introduction Snow blankets more than half of the Northern Hemisphere each winter, remaining in place for periods ranging from less than a month (typically south of 401 N) to more than 8 months of the year (typically north of 601 N). In the Southern Hemi- sphere, the coverage is less extensive, but still substantial. If the perennial snow covers of the Greenland and Antarctic ice sheets are included, along with the seasonal snow cover that forms on lake and sea ice, then the total percentage of the Earth’s surface covered by snow during some period of each year is considerable. This blanket of snow is a complex, layered material that can exhibit a high degree of spatial heterogeneity. Year-to-year variations in coverage and properties can be large and they have a direct and immediate impact on the Earth’s climate. In this article, the major types of snow cover are introduced 2062 SNOW (SURFACE) and the layered nature of the snow is discussed. The role of the snow in moderating the exchange of energy and mass with the atmosphere is also described. Snow Cover and Its Importance The term ‘snow cover’ is directly analogous to the term ‘formation’ when discussing layered sedimentary or metamorphic rocks. Both the sequence and character of the layers, and the lateral variation of each layer (facies changes), contribute to the overall properties of the formation. Similarly, the bulk physical and thermal properties of a snow cover, the properties that are of importance in moderating the exchange of energy and mass between the Earth and the atmosphere, are an aggregate of the properties of the individual layers. For each layer, these properties are the result of the conditions (snowfall, wind, temperature) that prevailed when the layer was deposited, and the postdepositional conditions (temperature, temperature gradients, snow overburden, liquid water percolation, solar radiation) to which the layer was subjected after deposition. Because both deposition and post-deposition conditions vary across the landscape, the layers themselves vary. In order to understand the role of snow cover in atmospheric processes, the layered nature and spatial variability of the material need to be considered. Much of the impact of snow on climate and atmospheric processes arises because of its high albedo and low thermal conductivity. Snow cover reflects up to 85% of incoming short-wave solar radiation, significantly reducing winter temperatures and retarding melting in the spring. At the same time, snow is an excellent insulator, so it can effectively lower the rate of heat loss from the ground or an underlying ice surface, thereby maintaining higher winter soil temperatures or retarding the rate of sea and lake ice growth. The total winter energy exchange across a snow cover is a complex balance between these two competing processes. Snow cover is also important because it traps aerosols and other atmospheric particulates like a filter, storing these until the snow melts, then releasing them abruptly. Snow can control, through thermal and physical means, the release of trace gases like CO2 from subnivian plants and soils during the winter, and it functions as a temporary storage reservoir of water, stockpiling winter precipitation then allowing it to run off in a much shorter period of time than it otherwise would have had it not fallen as snow. In some cases in higher latitudes where the snow lasts many months, as much as 80% of the annual river discharge can be from snow melt, and this discharge may occur in a period of less than two weeks. Perennial and Seasonal Snow Covers Because of their fundamentally different layered structures, it is customary to distinguish between perennial and seasonal snow covers. Seasonal snow covers are deposited in the fall and melt away completely each spring; therefore they never become very deep. Perennial snow covers form at higher levels on glaciers and ice sheets, where the combined decrease in temperature and increase in snowfall precipitation with altitude is sufficient to allow winter snow accumulation to survive the summer melt. Snowfall of the following winter is deposited on the residual snow of the previous year, forming a sequence of annual layers of snow that can be tens to hundreds of meters thick before compaction at depth converts the snow into glacier ice. Separate but related climate classification systems for perennial and seasonal snow covers have been suggested and are useful when thinking about both local and global variations in snow cover. For the perennial snow on glaciers and ice sheets, increasing elevation results in a decrease in melting. As a consequence, snow characteristics vary with elevation (Figure 1). At the lowest elevation, the melt removes all of the winter snow, and a seasonal rather than perennial snow pack forms each year. Higher, the snow pack survives the summer melt, but percolation of melt water into the snow pack and subsequent refreezing produce extensive icy features like the ice lenses and percolation columns. At the highest elevations, no melting takes place and the dry snow facies is observed. On a steep alpine glacier, the entire sequence is compressed into a distance of tens of kilometers. On ice sheets, sequence may spread over distances of hundreds of kilometers. For seasonal snow covers, local climate rather than elevation determines the prevailing snow cover characteristics, and this local climate can be represented by three simple binary variables: winter temperature, winter precipitation, and wind. High and low values for each of these variables (Figure 2) define eight possible types of seasonal snow covers, most of which have a counterpart in the glacier facies system shown in Figure 1. For example, under warmer, wetter winter conditions, a maritime snow cover will develop. This snow cover tends to be deep (41 m) and warm (near or at freezing temperatures), and exhibits similar icy features to those observed in the percolation facies on glaciers (compare Figure 3 to Figure 1). Similarly, alpine, tundra, and taiga snow cover classes exhibit features found in the dry snow facies on glaciers. The SNOW (SURFACE) 2063 Dry snow line (approx 2100 m) Saturation line (approx 1000 m) Firn line (approx 600 m) es faci ow n s  Dry es faci tion a l o c Per es faci ked a o S Summer surface of reference year Summer surface of the previous year tion Abla s e i c fa Glacier ice Figure 1 The glacier facies classification of Benson (1962), describing variations in the characteristics of the perennial snow cover found on glaciers and ice sheets. With increasing elevation, there is a decrease in the amount of melting and, as a consequence, a decrease in the amount of icy features in the winter snow pack. At the lowest level, all of the winter snow melts in the summer and the snow cover is essentially seasonal; at the highest level, no melting takes place and the snow has no features in it related to melting. (From Benson CS (1962) Stratigraphic studies in the snow and firn of the Greenland Ice Sheet. SIPRE Research Report 70, CRREL.) stratigraphic diagram and key in Figure 3 suggest the main snow cover characteristics associated with each climate class for seasonal snow. Te m pe ra tu re Pr ec ip ita tio n W in d sp ee d Rare (deep tundra snow) High Low Rare (deep taiga snow) High Low Taiga snow Low High Tundra snow Low High Very high Ephemeral snow Prairie snow High Low Low High Layer by Layer Development of a Snow Cover Snow cover builds up layer by layer. The initial characteristics of each layer are determined by how much solid precipitation falls, whether the precipitation is accompanied by wind, and the prevailing temperature at the time of deposition. After deposition, each layer is subjected to mechanical and thermal metamorphic processes that alter the layer characteristics. These vary in intensity and duration depending on when the layer was deposited, its height in the snow pack and the number of overlying layers, the prevailing conditions at the snow surface, and the temperature and temperature gradients in the snow pack as a whole. At any given time, the characteristics of each layer in the snow are a product of its initial deposition and post-depositional metamorphism. Alpine snow Maritime snow Low High Maritime snow Figure 2 A dichotomous classification of seasonal snow covers based on winter temperature, precipitation, and wind. In Figure 3, a typical snow stratigraphy for each class is shown. Broad similarities in snow characteristics exist between the seasonal snow classes and the glacier facies shown in Figure 1. (From Sturm M, Holmgren J, Liston G (1995) A seasonal snow cover classification system for local to global applications. Journal of Climate 8: 1261–1283.) Layer Deposition and Densification Almost 80 different types of falling snow crystals have been identified. The particular crystals that accumulate at the Earth’s surface in a snow storm are determined by the temperature and humidity in the layers of air through which the crystals fall and grow. However, crystal form is far less important than the rate of snowfall, the wind speed, and the temperature in determining the initial characteristics of a snow 2064 SNOW (SURFACE) 250 Snow depth (cm) 200 150 100 50 0 rie ai Pr al er m he Ep e m iti ar M ne pi Al a ig Ta ra nd Tu New snow Recent snow Wetted snow Fine-grained Wind slab Medium-grained Depth hoar Coarse-grained Ice Figure 3 Typical snow stratigraphy for the six seasonal classes listed in Figure 2. (From Sturm M, Holmgren J, Liston G (1995) A seasonal snow cover classification system for local to global applications. Journal of Climate 8: 1261–1283.) layer. In general, low temperatures, low wind, and low rates of snow fall produce the lowest-density layers of new snow (Table 1). Once deposited, new snow layers densify rapidly. Initially, much of this densification is a result of Table 1 The density of newly deposited snow Deposition conditions Density (g cm  3) No wind, low rate of snowfall, cold Low wind, low rate of snowfall Moderate wind, high rate of snowfall Moderate wind, low rate of snowfall High wind 0.02–0.05 0.05–0.10 0.20–0.35 0.35–0.40 0.40–0.55 thermodynamic instability. The sharp points and intricate branches of newly fallen snow crystals have high radii of curvature; the water vapor pressure over these highly curved surfaces is greater than elsewhere, so there is a net loss of water molecules from pointed areas to the air spaces in the snow, or to other areas on crystals that have lower degrees of curvature. The crystals rapidly break down and the resulting fragments become more rounded (Figure 4). The breakdown reduces the size of the crystals, increases the number of individual snow grains, and decreases the degree to which the crystals interlock. As a result, the entire snow layer settles. As additional new layers of snow are added to the snow pack, the overburden load (s) on buried layers SNOW (SURFACE) 0 1 2 3 5 12 14 16 19 23 49 57 2065 Figure 4 Changes in a snow flake held at a constant temperature of  11.51C for a total period of 57 days (indicated by small numbers). The snow flake grew in the atmosphere under conditions of supersaturation with respect to water vapor. Once deposited, the sharp points and thin branches were thermodynamically unstable and the snow flake metamorphosed, even in the absence of a temperature gradient or overburden stress. (From Bader H, Haefeli R, Bucher E, Neher J, Eckel O, Thams C (1939) Der Schnee und seine Metamorphose (Snow and its Metamorphism), US Army SIPRE Translation 14, 1954.) increases. For these layers, compaction due to vertical stresses begins to dominate the snow densification process. The response of the snow to these stresses has been modeled by assuming the snow layer behaves like a viscous fluid (eqn [1]).  1 dh 1 dr s ¼ ¼ h dt r dt Zc ½1 In eqn [1] h is the thickness of the layer (m), t is time (s), r is the layer density (kg m  3), and Zc is the compactive viscosity. Values of Zc (Pa s) have been determined from observations of the settlement of natural snow layers, from uniaxial strain compressive tests, and from depth–density profiles on glaciers and ice sheets. The combined results show wide scatter, but individual sets of data are usually fitted to the relation in eqn [2], where k is a factor that depends on the type of snow cover (Figures 1 through 3). Zc ¼ Z0 ekr ½2 The effective viscosity term incorporates a number of physical mechanisms including gravity-driven movement of snow grain centers of mass toward each other, vapor and volume diffusion, and sintering. 2066 SNOW (SURFACE) Table 2 Compactive viscosity factors for three classes of snow cover k-value (m3 kg  1) Maritime Alpine/taiga Tundra 18–22 35–60 470 Not surprisingly, viscosity factor values vary widely depending on the temperature, liquid water content, and grain characteristics of the snow – i.e., the snow cover class (Table 2). Colder, drier, finer-grained layers of snow tend to be more viscous than warmer, wetter, layers with larger grains, and therefore compact more slowly. In the absence of melting or the introduction of liquid water, snow layers will continue to densify until they reach a limiting density of about 0.6 g cm  3. By this time, the snow grains will have metamorphosed until they have become highly rounded, a shape that minimizes their surface free energy. The rounded grains will be in close contact with each other, and the grain arrangement will approximate that of hexagonal close-packing of ice spheres. Further densification will require actual deformation of the individual grains of snow, or the influx and refreezing of melt water in pore spaces. The overburden stresses required to achieve this further deformation are only realized in the deep perennial snow packs found on glaciers and ice sheets. Snow layers deposited during windy conditions (wind slabs) have much higher initial densities than other new snow layers. The wind tumbles snow crystals as it transports them, breaking the more fragile crystal junctions and pulverizing the crystals in general. The resulting grains are actually crystal fragments, often less than 0.1 mm in length, and these shardlike grains (Figure 5), when they come to rest, pack well and sinter together into a cohesive slablike layer. Initial densities for wind-transported layers of new snow range from 0.35 to 0.6. The upper limit occurs for the same physical reasons as discussed before. Due to their high initial densities and cohesiveness, wind slabs are highly resistant to compaction and often remaining at a fixed density after deposition. There has been much discussion and experimentation to determine the wind speed necessary to transport snow. The transport takes place through three mechanisms: creep, saltation, and suspension. Creep consists of the rolling movement of grains along the snow surface under the action of the wind. Saltation is the movement of grains along the surface by jumping and ricocheting after impact by other grains. Suspension is the movement of grains in the wind stream at some level above the snow surface. The threshold shear velocity, un , at which transport occurs is usually Figure 5 Wind-pulverized snow grains from Arctic Alaska, showing irregular shapes and thick bonds due to rapid sintering after deposition. estimated by assuming a logarithmic-shaped wind profile and projecting the 10-m high wind speed (u10 ) down to the snow surface (un ). In general the value of u10 is between 18 and 30 times the value of un . Experimental studies indicate that when u10 is greater than 6 m s  1 transport will occur if the snow has fallen recently. If the snow is new and falling while there is wind, transport will occur with wind of 5 or even 4 m s  1. If the snow is aged, was previously transported by the wind, or has undergone some melt– freeze processes, speeds in excess of 30 m s  1 may be needed before the snow will start to be tranported (Figure 6). u 10 (m s1) (approx.) 0 10 20 30 4 drifted Hardness (kg cm2) Snow cover type 3 2 1 new and recent 0 0.5 1.0 1.5 u * (m s1) Figure 6 The critical wind shear velocity (u n ) as a function of snow hardness, which is a good measure of the type of snow. Increasing hardness, common for wind slabs and layers of snow that have undergone melt–freeze, requires considerably higher winds to mobilize these types of snow. u10 is the wind speed measured at a standard height of 10 m. (From Kind RJ (1981) Snow drifting. In: Gray DM, Male DH (eds). Handbook of Snow, pp. 338–359. Toronto: Pergamon.) Transport rate (kg m1 s1) SNOW (SURFACE) 2067 0.04 Snow Metamorphism 0.03 In addition to compaction and densification, several other metamorphic processes can affect layers of snow. These processes result chiefly in changes in snow grain characteristics and bonding, which in turn affect the thermal conductivity, air permeability, and albedo of the snow. The processes are typically divided into ‘wet’ and ‘dry’ categories because different snow grain characteristics are produced depending on whether liquid water is present. Further metamorphic subdivisions are shown in Table 3. For wet snow metamorphism, the degree to which grains and a snow layer are changed is mainly a function of how much water is present. For low liquid contents (o5% by weight), the water in the snow exists as thin films and isolated pockets or veins around grains; continuous ice grain and air space pathways still exist through the snow layer. This is called the pendular regime. Under this regime, snow grains will rapidly round, and clusters of grains, looking much like bunches of grapes, will form as a result of the minimization of surface free energy. The clusters themselves are quite strong because the bonds between the spherical grains are still intact and substantial. The wet snow pack will have considerable bearing strength. Spring skiing, which can be excellent, takes advantage of these ball-bearing like grain clusters and the general strength and cohesiveness of this type of wet snow metamorphism. If the temperature of the snow drops and the grain clusters freeze, they will take on the slightly more amorphous shapes of melt–grain clusters (Figure 8), while at the same time the strength of the layer will increase dramatically as all the interstitial water freezes. For higher liquid water contents, snow grains and air spaces become surrounded and isolated by the liquid water present in the layer. This water begins to drain downward under the influence of gravity and is called the funicular regime. Once again, when surrounded by water, the snow grains will round, but now boundaries between grains will not be thermodynamically stable and will melt rapidly, creating a slush. The slush has little or no bearing strength, and can even flow like a fluid under certain conditions. The grains themselves, if surrounded by water at 01C for long enough (24–36 hours), will metamorphose into oblate spheroids (Figure 9). In the absence of liquid water, snow will metamorphose in one of two ways depending on the temperature gradient imposed on the snow. Water vapor density over ice is a strong positive function of temperature, so temperature gradients in the snow give rise to water vapor density gradients in the air spaces in the snow and a diffusive flow of vapor from warmer to colder grain surfaces. For convenience the 0.02 0.01 0 0.20 0.30 0.40 0.50 0.60 Wind-shear velocity, u * (m s1) Figure 7 Snow transport rates for saltation (solid curve) and suspension (broken curve) as a function of wind shear velocity (u n ). The wind speed at 10 m height is approximately 18–26 times u n . At u n ¼ 0:44 (10 m height wind speeds of 8–11 m s  1), suspension begins to transport the majority of the wind-borne flux of snow. (From Liston GE, Sturm M (1998) A snow-transport model for complex terrain. Journal of Glaciology 44: 498–516.) In similar fashion, the flux of snow transported by the wind is a strong function of the wind speed, with increasing speeds producing a marked increase in the total amount transported (Figure 7). For values of un between 0.2 and 0.44 m s  1, saltation dominates the transport, but for un values in excess of 0.44 m s  1, suspension exceeds saltation in transporting snow. One other consequence of wind transport of snow is the development of a wide range of drift deposit and erosion features at the snow surface. These features include ripple marks, dunes, barchans, and sastrugi. Surprisingly, little is known about the relationship between these features and the wind speed, the temperature, and the snow conditions necessary for their formation. The final, and most efficient, method for densifying a layer of snow is through the infiltration of melt or rain water into the snow cover, followed by subsequent refreezing. Water can infiltrate, surround grains as thin films or lie in veins along grain junctions, and refreeze to produce large multiparticle grains. Water can also percolate downward in pipelike structures called percolation columns, or spread out along stratigraphic boundaries (owing to variations in the hydraulic conductivity of the snow). When this water refreezes, ice lenses and layers are created. Frequently, a single infiltration event will produce ice layers at multiple levels in the snow pack. Densities in excess of 0.6 g cm  3, sometimes even as high as 0.9 g cm  3, can result. This mechanism is commonly observed in ephemeral and maritime seasonal snow covers (Figures 2 and 3), and in the percolation facies for perennial snow (Figure 1). 2068 SNOW (SURFACE) Table 3 Metamorphic processes that affect the snow cover Wet snow metamorphism Dry snow metamorphism Dry snow metamorphism – older terms Melt-grain clusters and melt–freeze particles Slush Equilibrium or rounded growth Kinetic or faceted growth Equi-temperature metamorphism (ET) Temperature-gradient metamorphism (TG) Figure 8 Melt-grain clusters showing the well-rounded grains and the high degree of contact between grains. Figure 9 Snow slush, showing the oblate spheroid shape of the grains and the complete lake of bonding. (From Colbeck SC (1986) Statistics of coarsening in water-saturated snow. Acta Metallargica 34, 347–352.) temperature gradient is often defined as the difference between the basal and surface temperatures of the snow cover, divided by the thickness of the snow (Figure 10), but in reality the actual temperature gradient varies continuously with both time and height in the snow. For example, rapid fluctuations in air temperature can produce very large temperature gradients near the snow surface, at least for short periods of time. Experimental work has shown that when the temperature gradient exceeds a magnitude of approximately 0.251C cm  1, kinetic crystal growth will occur. If the gradient is lower, equilibrium growth takes place. Not surprisingly, temperature gradients in thick perennial snow covers tend to be lower than those in the thinner seasonal snow covers, particularly thin taiga, tundra, and alpine seasonal classes that can be subjected to very low air temperatures in the winter. As a result, kinetic growth is common in seasonal snow covers but occurs infrequently (often only in autumn) in perennial snow covers. Equilibrium crystal growth, also widely known as ‘equi-temperature metamorphism’ (ET-metamorphism) occurs when temperature gradients in the snow pack are less than 0.251C cm  1. These low temperature gradients produce weak water vapor density gradients in the snow and low rates of vapor diffusion. The rates are so low that the supply of vapor to a growing crystal, rather than crystal growth dynamics, controls the growth. Rounded, well-bonded grains result. Kinetic growth, also widely known as ‘temperaturegradient metamorphism’ (TG-metamorphism) produces ornate, faceted crystals commonly referred to as ‘depth hoar’. In this case, temperature gradients imposed on the snow are of a large enough magnitude to produce a flux of water vapor that exceeds the rate at which the crystal can grow. Crystal growth dynamics, rather than vapor supply, control both the growth rate and the crystal form, producing crystals with distinct sharp-edged facets, welldefined interfacial angles, and surface striae (Figures 11 and 12). Unlike the case for equilibrium growth, intergrain bonds are weakened and reduced in number during kinetic growth, producing layers that tend to be brittle and weak. This has two important ramifications: the brittle layers can shear easily and often create failure planes that are responsible for the release of avalanches. Second, the poor bonding creates layers that have low thermal conductivity. In absence of air movement in the snow, these layers provide excellent insulation that contributes to the retention of heat in the ground or ice underlying the snow cover. Height in snow (cm) SNOW (SURFACE) 2069 Snow surface 60 40 20 10 Dec. 1997 12 Dec. 1997 21 Feb. 1998 29 Mar. 1998 0 30 25 (A) 15 10 Temperature (qC) Equilibrium growth Height in snow (cm) 20 60 Figure 11 The initial stages of kinetic growth metamorphism. The grains are starting to exhibit distinct faceting. Kinetic growth envisioned as occurring in a ‘hand-to-hand’ manner across pore spaces, with vapor diffusion from the warm side of snow grains balanced by vapor condensation on the colder side. Because the contributions of these three individual mechanisms are difficult if not impossible to separate, 40 20 0 0.1 (B) 0.2 0.3 0.4 0.5 Temperature gradient (qC cm1) 0.6 Figure 10 (A) Temperature profiles and (B) computed vertical temperature gradients from the snow cover on the ice of the Beaufort Sea north of Alaska. The temperature profiles are not linear, and as a consequence the temperature gradients vary in a complex way with height in the snow. Note that at some heights and times the gradient is below the critical magnitude of 0.251C cm  1 and kinetic growth will not occur. Energy and Mass Exchange across a Snow Cover It is beyond the scope of this article to address in full the mass and energy exchange over a snow cover, but a few points particular to snow are discussed. The reader should also see articles on surface energy balance, albedo, turbulence, boundary layer meteorology, surface roughness, and solar radiation for more details. Heat transfer across a snow cover occurs mainly by conduction through the ice network of grains, by conduction across the air-filled pore spaces in the snow, and by diffusion of vapor across the pore spaces. The thermal conductivity of ice is more than 100 times higher than that of air, so the conduction of heat across air spaces is thought to contribute relatively little to the total. The heat transported by vapor diffusion, in contrast, is thought to contribute as much as 40%, particularly at temperatures near freezing when the vapor flux is high. This diffusive vapor transport is Figure 12 At-depth hoar cup, shown in typical growth position. The hexagonal pyrimidal cup opens downward because the flow of water vapor is upward. Heavy striae can be seen on all crystal facets. This is the late stage of kinetic growth metamorphism. 2070 SNOW (SURFACE) in practice they are always lumped together by reporting an ‘effective’ thermal conductivity for the snow. Both solid body conduction through the ice network and vapor diffusion are driven by the temperature gradients in the snow, suggesting that a simple heat flow equation can be used to model the flux of heat across the snow (eqn [3]). q ¼ keff dT dz ½3 Here q is the vertical heat flow through the snow cover, dT=dz is the temperature gradient across the snow, and keff is the effective thermal conductivity of the snow. However, the driving temperature gradient in the ice network may be quite different from the gradient across pore spaces that drives vapor diffusion, in which case eqn [3] may be an oversimplification. Be that as it may, it is customary to describe the heat transfer using eqn [3] and assigning an appropriate value for keff. Figure 13 shows compilation of most measured values of keff as a function of density. As the density of 1 9 8 7 6 5 the snow increases, so in general does the value of keff . In many climate models, regression equations relating keff to density (often using the viscous snow compaction (eqns [1] and [2]) to determine the snow density) are used to set the thermal conductivity of the snow. However, as the figure shows, the scatter in keff at any given density is large and real. It is the result of differences in the bonding of the snow, and perhaps also due to variations in snow temperature. For a given density, higher temperatures and better bonding between grains lead to higher values of thermal conductivity. Given the scatter, care should be exercised when choosing a value of keff for modeling. The values should be consistent with the type of snow cover (Figures 1 through 3) as well as a keff –density relationship. For improved accuracy, a value of keff for each layer of snow should be determined; then the bulk value for the entire snow cover should be computed using a series-type solution. Convective heat transfer is also known to operate in snow and complicates the energy exchange across a snow cover. Two types of convection have been reported: buoyancy-driven convection, and convection Center of data 4 keff (W m1 K1) 3 2 (1889) Hjelstrom (1893) Abel's Jansson (1901) (1905, 08) Okada (1924) Ingersoll & Koepp Devaux (1933) (1939) Kuz'min (1949) Bracht (1954) Kondrat'eva (1954) Kondrat'eva (cited) (1954) de Quervain (1954) Yosida & lwai (1962, 65) Yen (1967) Pitman & Zuckerman (1970) Jaafar & Picot Weller & Schwerdtfeger (1971) (1975) Izumi & Huzioka Kuvaeva & others (1975) Voitkovsky & others (1975) Reimer (1980) (1985) Lange (1989) Murakami & Maeno (1991) Ostin & Andersson 0.1 9 8 7 6 5 4 3 2 0.01 0.0 0.1 0.2 0.3 0.4 Density (g cm3) 0.5 0.6 0.7 Figure 13 A compilation of most published values of the thermal conductivity of snow. There is nearly an order of magnitude scatter at any given density, and this scatter is real. It arises from differences in snow cover characteristics. (From Sturm M, Holmgren J, König M, Morris K (1997) The thermal conductivity of seasonal snow. Journal of Glaciology 43: 26–41.) 2071 SNOW (SURFACE) 150 DE PT HO H AR _ 1 Permeability 10 2 cm s _ dyne cm 3 forced by the wind (wind-pumping). The former has been documented only in a highly permeable snow covers like taiga snow. This snow cover often wholly comprises layers of large, poorly bonded kinetic growth crystals called depth hoar. The layers have extremely high values of air permeability and, owing to low winter air temperatures, are subjected to temperature gradients of high magnitude, both conditions favorable for buoyancy-driven convection. Convective air flow velocities of several millimeters per second have been computed based on observations of temperature fields in the snow, and these air flow speeds are sufficient to increase the heat transfer rate by a factor of 3. The prevalence of buoyancy convection in other types of snow covers may be low, but this has not been shown experimentally. Forced convection also probably occurs in some snow covers. Theory indicates that pressure differences arising when wind blows across surface irregularities like dunes and sastrugi are most likely to produce a flow of air that can move both heat and mass (in contrast to turbulence or other aspects of the wind over snow). Flow rates are probably on the order of a few millimeters per second and are likely to be confined to near-surface layers of snow. Observations of the mixing depth of aerosols and particulates in snow layers indicate that wind pumping is definitely effective in moving mass, but the magnitude of the effect of wind-pumping on heat transfer has yet to be demonstrated. In addition, it appears that near-surface and surface wind and melt crusts in the snow can effectively eliminate any wind-pumping by reducing the air permeability of the snow creating barriers in the form of impermeable wind on melt crusts that can effectively shut off all air movement. As neither wind-pumping nor buoyancy-driven convection are state properties of the snow, they pose difficulties when one is trying to model heat transfer in snow. Both processes depend on external conditions for their onset and strength, and they can transport anything from zero to several times the conductive heat flux, depending on the snow characteristics, the temperature structure in the snow, and the wind speed and direction. Water, water vapor, CO2 , methane, and aerosols and particulates are all transferred across a snow cover and the transfer process for each is complicated. In general, mass transfer is controlled by the air permeability of the snow, the surface topography of the snow cover (for wind-pumping), and the supply rate of particles, gases, or chemicals. As discussed previously, both diffusive and convective transport of air are possible, and the chemicals and gases move with the air. The air permeability of naturally occuring snow (Figure 14) ranges over two orders of magnitude. It is a 100 O LD SN O CO AR W SE GR AIN 50 MED IUM NEW FI SNO W NE G GRA IN RA I N WIND SLAB 0 0 0.1 0.2 1.0 0.9 0.8 Density 0.3 0.4 0.7 0.6 ; porosity  s 0.5 _ s (g cm 3) 0.5 H Figure 14 The air permeability of snow. Again, there is a greater variation by snow type than by density. (From Shimizu H (1970) Air permeability of deposited snow. Low Temperature Science, Series A, 1–32.) major control on deposition and transfer rates, which vary widely with chemical species and environmental conditions. For aerosols, when the residence time of the air in the snow is greater than 15 seconds, the filter efficiency of the snow can be almost 100%. See also Boundary Layers: Overview. Energy Balance Model, Surface. Land–Atmosphere Interactions: Canopy Processes; Overview; Trace Gas Exchange. Reflectance and Albedo, Surface. Solar Terrestrial Interactions. Further Reading Colbeck SC (1986) Classification of seasonal snow cover crystals. Water Resources Research 22(9): 59S–70S. Gray DM and Male DH (1981) Handbook of Snow. Toronto: Pergamon Press. LaChapelle ER (1969) Field Guide to Snow Crystals. Seattle: University of Washington Press. Magono C and Lee CW (1966) Meteorological classification of natural snow crystals. Journal of the Faculty of Science, Hokkaido University 2(4): 321–335. Seligman G (1936) Snow Structure and Ski Fields. (Reprinted by the International Glaciological Society, Cambridge, 1980.) 2072 SOLAR TERRESTRIAL INTERACTIONS Shimizu H (1970) Air permeability of deposited snow. Low Temperature Science Series A (22): 1–32. Sommerfeld RA (1970) The classification of snow metamorphism. Journal of Glaciology 9(55): 3–17. Sturm M, Holmgren J, et al. (1997) The thermal conductivity of seasonal snow. Journal of Glaciology 43(143): 26–41. Waddington ED and Harder SL (1996) The effects of snow ventilation on chemical concentrations. In: Wolff EWand Bales RC (eds) NATO ASI Series, vol. I–43, pp. 403–451. Berlin: Springer-Verlag. Warren SG (1982) Optical properties of snow. Reviews of Geophysics and Space Physics 20(1): 67–89. SOLAR TERRESTRIAL INTERACTIONS J D Haigh, Blackett Laboratory, Imperial College of Science, Technology and Medicine, London, UK Copyright 2003 Elsevier Science Ltd. All Rights Reserved. Introduction Many studies have shown an apparent response in weather or climate indicators to solar variability. At various locations temperature, rainfall, surface pressure, cloud cover, storms, and droughts, among other meteorological parameters, have been found to correlate with measures of solar activity over the 11-year solar cycle and over periods extending from decades to centuries and longer. Some of these studies do not stand up to rigorous statistical analysis and some only appear to hold only over limited time periods, but there is mounting evidence of solar influence on climate on many time scales. The radiant energy output of the Sun varies by about 0.1% over the 11-year solar cycle. If simple radiation balance estimates are used, this does not appear to be large enough to explain some of the apparent solar signals, in particular those in lower-atmosphere temperatures, although it is consistent with observed decadal variations in sea surface temperature of order 0.1 K. Comparison of solar activity, reconstructed back to the Maunder Minimum in sunspot numbers at the end of the seventeenth century, with estimates of Northern Hemisphere temperature suggests that the Sun has made a significant contribution to climate variability since that time but cannot alone account for the warming of the latter half of the twentieth century. The amount of solar radiation reaching the Earth is also modified by variations in the Earth’s orbit around the Sun. These variations take place over periods of tens to hundreds of thousands of years and may be responsible for the occurrence of ice ages. Variations in factors other than the total amount of solar radiant energy affect the atmosphere and possibly influence weather and climate. Solar ultraviolet emission varies by several percent over the solar cycle and influences ozone production and temperatures in the middle atmosphere. The resulting change in thermal structure of the stratosphere may influence the climate of the lower atmosphere through dynamical and radiative processes. The chemical structure of the stratosphere is also affected by high-energy protons and electrons emitted during solar flares and coronal mass ejections. Alterations in the solar magnetic field affect the flux of galactic cosmic rays reaching the Earth and thus the strength of the Earth’s electric field and ionization rates in the lower stratosphere. It is plausible, but unproven, that these result in changes in thunderstorm activity or cloud cover. Observations The effects on the atmosphere of varying solar insolation can be observed clearly in diurnal and seasonal variations. However, direct observation of the effect of changing solar activity on weather and climate is more difficult, so the detection of solar signals in meteorological records has usually relied on statistical analysis. One approach is to perform a spectral analysis of a time series of data to see if periodicities associated with solar variability (e.g., around 11, 22, or 80 years) are present. Another simple statistical approach is to estimate the degree of correlation between a meteorological parameter and some measure of solar activity. More sophisticated methods take into account other possible factors influencing the state of the atmosphere and also information on the variability of all the parameters involved. Solar 11-Year Cycle Cycles of 10–12-year periodicity have been isolated in many data records, including global surface temperature, surface temperature at many land stations across the globe, rainfall in the United States and Africa, surface pressure in the North Atlantic and North Pacific Oceans, North American forest fires, Atlantic tropical cyclones, tropical corals, and the Southern Oscillation. An example of a 10–12 year SOLAR TERRESTRIAL INTERACTIONS 2073 First EOF of Decadal SST Anomalies Time Sequence of Amplitude 1.0 1.0 SOLAR qK 0 _ 0.5 _ 1.0 _ 0.5 SST 1960 _2 0.5 1970 1980 1990 Wm 0.5 _ 1.0 Spatial Pattern (50%) 60qN -0.2 30q 0q 60qN 0.3 0.3 0 0.1 30q 0.2 0 0.1 0 0 -0.1 30q 60qS 0q 30q C.I. = 0.1 < 0 SHADED 60q 120q 180q 120q 60q 0q 60qS Figure 1 Time series (A) and spatial pattern (B) of the mode of variability which accounts for 50% of decadal variance in sea surface temperatures between 1955 and 1994. CI, contour interval. (Reproduced with permission from White WB, Lean J, Cayan DR, and Dettinger MD (1997) Response of global upper ocean temperature to changing solar irradiation. Journal of Geophysical Research 102: 3255–3266.) oscillation in sea surface temperatures can be seen in Figure 1, which shows the amplitude of an oscillation derived from data which have been spectrally filtered to isolate the decadal component. Because the 10–12-year period seen in these many records includes that of the solar 11-year cycle it is often assumed that the two are causally related. However, in any particular data set the amplitude of the 10–12-year component usually varies, sometimes disappearing for considerable periods. Perhaps even more seriously, the phase of the cycle often varies with respect to the supposed solar forcing. Another aspect of many of the analyses is considerable geographical variation in the response of any particular meteorological parameter, as seen in the sea surface temperature analysis in Figure 1, which shows warming over most of the oceans during solar maxima but bands of cooling in the subtropics. A high degree of correlation with solar activity has also been observed in lowerstratospheric geopotential heights, with the largest values found in midlatitudes. Figure 2A shows 3-year running means of 30 hPa geopotential height at 20– 401 N in July and August (solid line) and the 10.7 cm solar index (dashed line). The correlation coefficient, R, of the two curves is 0.74. Also shown in Figure 2 is the 3-year running mean of the mean temperature of the middle to upper troposphere (750–200 hPa) in July and August for (B) 20–401 N (R ¼ 0:54) and (C) the entire Northern Hemisphere (R ¼ 0:66). At certain locations and times of year the correlations of these parameters with the solar cycle may be enhanced significantly when the data are sorted according to the phase of the Quasi-Biennial Oscillation (QBO), although care has to be taken that the analysis remains statistically sound. Analysis of data from the International Satellite Cloud Climatology Project has shown a high correlation between global low cloud cover and cosmic ray flux (inversely related to solar activity) between 1983 and 1994. It remains to be seen if this correlation is maintained for a period longer than one solar cycle. A problem with deductions from time series such as the cloud data, or those shown in Figure 2, is that nonsolar forcing factors are not considered. For example, a major factor affecting cloud cover is the phase of the El Niño Southern Oscillation and this will undoubtedly have contributed to low cloud variation within the 1983–94 time period. Other factors, which may have affected the geopotential height and temperature variations shown in Figure 2, were the major volcanic eruptions of 1982 and 1991, which probably warmed the lower stratosphere and cooled the troposphere during the few years subsequent. (Note, however, that the cooling in Figure 2B and C apparently due to the 2074 SOLAR TERRESTRIAL INTERACTIONS 24200 2400 24180 2000 24160 1600 24140 1200 24120 800 24100 24080 1950 (A) 1960 1970 1980 Year 1990 255.80 2400 255.60 2000 255.40 1600 255.20 1200 255.00 800 254.80 254.60 1950 400 2000 1960 (B) 1970 1980 Year 1990 10.7 centimeters solar flux 2800 Degrees K 24220 VAve: (J+A)/2: [750 _ 200] 20 _ 40N: 0 _ 360 [3 RunAve] cc=0.54 2800 256.00 10.7 centimeters solar flux Geopotential 30 hPa 30: (J+A)/2: 20 _ 40N: 0 _ 360 [3 RunAve] cc=0.74 400 2000 250.20 2400 Degrees K 250.00 2000 249.80 1600 249.60 1200 249.40 800 249.20 249.00 1950 (C) 1960 1970 1980 Year 1990 10.7 centimeters solar flux VAve: (J+A)/2: [750 _ 200] 0 _90N: 0 _ 360 [3 RunAve] cc=0.66 2800 250.40 400 2000 Figure 2 Solid lines: (A) time series (solid curve) of the zonally averaged 3-year running means of 30 hPa heights, 20–401 N, July–August; (B) same as (A) but for the zonal mean temperature in the layer 750–200 hPa; (C) same as (B) but averaged the over whole of the Northern Hemisphere. The dashed line is the 10.7 cm solar flux. (Reproduced with permission from van Loon H and Shea D (1999) A probable signal of the 11-year solar cycle in the troposphere of the northern hemisphere. Geophysical Research Letters 26: 2893–2896.) solar effects starts before the dates of the two eruptions.) It is also possible that the natural (unforced) variability of the climate system includes components of the same periodicity, making it more difficult to diagnose any solar component. Clearly, any theory which seeks to explain solar–climate links will need to address all these factors. Decadal–Centennial Scale Variations associated with the 11-year cycle are interesting from the point of view of diagnosing local meteorological records and can also indicate potential solar forcing mechanisms, but, unless the effects are somehow accumulative, do not imply any long-term climate effects. However, longer-term (and largeramplitude) variations in solar activity may have significant impact on global climate. The well-known coincidences of the cool period (‘Little Ice Age’) from the late fifteenth century to the early nineteenth with the Spörer and Maunder solar minima, and of the warm period between about 900 and 1300 with the solar Medieval maximum, have often been cited as inferring a solar influence. However, it is possible that these temperature records are biased towards Northern Hemisphere observations and that the warming/ cooling was not so large on the global scale. Figure 3 shows (A) an estimate of total solar irradiance back to 1600 along with (D) a reconstruction of Northern Hemisphere surface temperatures. Visually, there are similarities between the shapes of the curves, and correlations between records such as these show values of between 0.5 and 0.8 depending on the particular data sets, and length of records, used. SOLAR TERRESTRIAL INTERACTIONS 2075 Solar total irradiance (estimated) 1368 Wm _2 1367 1366 1365 1364 (A) Volcanic dust veil index (global) 4000 3000 2000 1000 (B) 0 Anthropogenic gases (CO2) ppmv 360 330 300 270 (C) Surface temperature anomalies 0.3 qC 0.0 Northern hemisphere _ 0.3 _ 0.6 1600 Global 1700 (D) 1800 Year 1900 2000 Figure 3 Time series of annual averages of (A) estimated total solar irradiance, (B) volcanic aerosol loading, (C) greenhouse gas concentration (as equivalent CO2), and (D) surface temperature: global average from the instrumental record (fine line) and Northern Hemisphere values reconstructed from proxy data (bold line). (Reproduced with permission from Lean J and Rind D (1998) Climate forcing by changing solar radiation. Journal of Climate 11: 3069–3094.) If, instead of solar irradiance, the length of the solar cycle is used as the measure of solar activity, then the correlation rises even higher. However, as in the case of the 11-year cycle, such correlations fail to explain the effect or to recognize the existence of other possible forcing factors. Two of these are also shown in Figure 3: (B) volcanic dust veil index and (C) anthropogenic greenhouse gas concentrations (as equivalent CO2). Statistical studies in which the components of the temperature record due to these different factors are isolated suggest that about half of climate variability in the preindustrial era may be due to solar influences. Long Term Indicators of longer-term variation in atmospheric temperature may be derived from oxygen isotope ratios in glaciers and ice sheets, lake sediments, ocean sediments, and corals. Spectral analysis of these data sets shows that the dominant component in the record for the last 800 000 years has a period of around 100 000 years. This is on the same order as the periodicity of the eccentricity of the Earth’s elliptical orbit around the Sun, which could indicate evidence of a direct solar influence on climate. However, the changes in solar irradiance associated with the 2076 SOLAR TERRESTRIAL INTERACTIONS variations in eccentricity are small and much less than the latitudinal and seasonal deviations due to variations in other orbital parameters – in particular, the tilt of the Earth’s axis, which demonstrates a periodicity of 41 000 years and the precession of the equinoxes with periodicities of 19 000 and 23 000 years. These periods are less evidenced in the paleoclimate records, indicating that factors other than simply the total solar irradiance are responsible for long-term variations in climate. A key factor for the potential climate impact of orbital variations is the change in the seasonal intensity of solar radiation. For example, if increased summer irradiance is insufficient to melt an extended ice sheet resulting from colder winters, then the climate system may be projected into an ice age. Current GCMs are not able to reproduce ice sheet growth, suggesting either that understanding of the processes involved is incomplete (feedbacks involving vegetation cover and type may be an example) or that nonlinear interactions are not properly represented. Mechanisms for Solar–Climate Interaction The observational analyses described in the previous section provide evidence for influence of solar variability on climate but, without a clear understanding of the mechanism(s) through which such interactions take place, doubts may remain that the periodicities and correlations found are due to natural internal atmospheric variability or to other climate forcing factors. The most direct means whereby solar variability may affect climate is by modulation of the total solar radiative energy received by the Earth. Other factors may possibly play a role indirectly through modification of atmospheric chemical composition or the Earth’s electric field or cloud formation. Radiation Balance The total radiative power emitted by the Sun crossing unit area at the Earth’s distance from the Sun is approximately 1370 W m  2. Historically, this was referred to as the ‘solar constant’ because it was believed not to vary, but measurements made from satellites since the late 1970s have shown that total solar irradiance (TSI) changes by about 1.4 W m  2, or 0.1%, over the 11-year cycle with higher values corresponding to periods of greater solar activity as indicated by, e.g., sunspot numbers. Based on these measurements and historical observations of sunspot numbers (and other indicators of solar activity) TSI values back to around the year 1600 have been reconstructed, one such estimate being given in Figure 3. This shows particularly low values occurring during the second half of the seventeenth century (corresponding to the Maunder Minimum in solar activity) and high values at the end of the twentieth century. The difference in total solar irradiance between these two periods is estimated to be about 3.3 W m  2, or 0.24%. The equilibrium change in global average surface temperature, Ts , due to a change in the Earth’s radiation balance may be estimated from the expression DTs ¼ l DF, where DF is the imbalance in global average radiative fluxes and l a climate sensitivity parameter which represents the response of the surface temperature to applied radiative perturbations, taking into account atmospheric feedback mechanisms through, for example, changes in humidity or cloud cover. l is derived empirically from atmosphere–ocean GCMs to be in the range 0.3 to 1.0 K W  1 m2. An increase in solar irradiance of 2 W m  2, as shown in Figure 3, between the years 1700 and 1780, corresponds to DF ¼ 0:35 W m2 (taking account of global averaging and terrestrial albedo) which would suggest, using the expression above, a surface warming of 0.10–0.35 K over that period – a figure very similar to the observed increase. However, the same approach would indicate that the increase in solar irradiance of 3.3 W m  2 between 1700 and 2000 would give a warming of 0.17–0.58 K, considerably less than the observed magnitude of 0.7 K. This confirms the results of the correlation studies discussed above that factors other than solar variability influence variations in surface temperature, but that the Sun probably plays a significant role on decadal to century time scales. Using the same radiative forcing arguments, the 0.1% variation in solar irradiance over the 11-year solar cycle would imply an equilibrium global average surface temperature response of about 0.07–0.24 K. This is of similar magnitude to the variations of about 0.1 K observed in sea surface temperature (see Figure 1). The above discussion was based on the idea of global radiative equilibrium. However, studies with GCMs show a geographically nonuniform response to variations in solar input, with surface temperatures at some locations showing cooling during periods of higher solar activity, although the global average response is consistent with that derived from energy balance arguments. Observational studies also suggest regional variations, although there are insufficient data to make detailed analyses. Nevertheless, it is probable that dynamical adjustments within the atmosphere result in the existence of regions of preferred response to solar activity. Were the warming of 0.1 K in surface temperature, suggested for the 11-year cycle response both by radiation balance calculations and by sea surface SOLAR TERRESTRIAL INTERACTIONS measurements, to be extended throughout the atmosphere it would result in an increase of about 10 m in 30 hPa geopotential height. This is considerably less than the 50–100 m deduced to be due to solar influence from observations at low to midlatitudes (see Figure 2). Again, the difference may be due to dynamical adjustments taking place within the atmosphere, which preferentially enhance the response at particular locations; one possibility is that the strength and extent of the tropical Hadley cells are affected. Alternatively there may be other physical factors acting to amplify the response to solar activity – some of these are discussed below. A third explanation might be that the results of the data analysis are not robust; a reliable solution to this question will be realized only with longer-term stable data sets. Ultraviolet The variability in total solar irradiance of order a few tenths of 1% discussed above represents the changes integrated across the whole electromagnetic spectrum. In the ultraviolet (UV), however, the fractional amplitude of variability is much higher: measurements made by satellite instruments, and also the results of theoretical models, suggest an increase of 7% at 200–208 nm and of 3:5% at 250 nm from solar minimum to maximum of the 11-year cycle. This variation in the spectrum complicates the issue of where in the atmosphere–surface system the solar energy is deposited. Most of the visible–near-infrared radiation passes through the atmosphere unhindered to the tropopause and hence, neglecting scattering by cloud, to the surface, although water vapor bands in the near infrared cause some absorption in the lower troposphere. Shorter UV wavelengths, however, are absorbed in the middle atmosphere, where they cause local heating and ozone production. The increased ozone tends to mask the lower atmosphere from the enhanced incident UV, while the warmer stratosphere will cause increased emission of thermal infrared (TIR) radiation into the troposphere. Thus the nature of the changes in the UV and TIR radiation fields depends on the ozone response. However, the variation of ozone to solar activity is not well established. Two-dimensional photochemistry–transport models of the stratosphere predict the largest fractional changes in the middle–upper stratosphere with monotonically decreasing effects towards the tropopause. Multiple-regression analysis of ozone measurements made from satellites suggests largest changes in the upper stratosphere, zero, or even slightly negative, changes in the middle stratosphere, and significant positive changes in the lower stratosphere. However, as the data are available over only about one and a half 2077 solar cycles, and have large uncertainties, especially in the lower stratosphere, and may not properly have accounted for the effects of volcanic aerosol, the true nature of solar-induced changes in stratospheric ozone remains uncertain. Changes in stratospheric thermal structure may also affect the troposphere through dynamical interactions rather than through direct radiative forcing. GCM studies indicate that changes in stratospheric zonal wind structure, brought about by enhanced solar heating, could interact with vertically propagating planetary waves in the winter hemisphere to produce a particular mode of response. This mode, seen also in GCM studies of the response to heating in the lower stratosphere caused by injection of volcanic aerosol, shows dipole anomalies in zonal wind structure which propagate down, over the winter period, into the troposphere. Thus, modifications to the stratosphere may result in modulation of tropospheric modes such as the Arctic Oscillation. If this does take place, as seems possible but not proven, then there is scope for significant local variation in meteorological parameters in response to solar activity. Such a mechanism might contribute towards the apparently solarinduced changes in 30 hPa geopotential height in the winter hemisphere, but does not provide a simple explanation for the summer hemisphere response. Solar-Energetic Particles and Galactic Cosmic Rays Solar flares and coronal mass ejections occur more frequently during periods of higher solar activity. As a result of these events high-energy particles (protons, electrons, and alpha particles) are emitted which can enter the Earth’s atmosphere along the open magnetic field lines near the polar caps. Particles with energies of more than 10 MeV can penetrate the middle atmosphere where they cause an increase in the concentration of NO through ionization and dissociation of N2 and O2. Significant reduction in middle atmosphere ozone concentrations can ensue. Higher-energy particles can propagate deeper into the atmosphere and to lower latitudes. Furthermore, if a geomagnetic storm is also in progress this will tend to expand the polar cap, allowing further atmospheric exposure to the particles. Although individual solar particle events only last on the order of a few hours, the chemical perturbations may persist for several months, propagating downwards and equatorwards and possibly altering stratospheric dynamics. The climate impact of this is likely to be small, but has not been studied in detail. Galactic cosmic rays (GCRs) are particles formed outside the solar system and which bombard it from all directions. The flux of GCRs reaching the Earth is 2078 SOLAR WINDS modulated by interaction with magnetic structures advected with the solar wind such that at times of higher solar activity the GCR flux is reduced by about 20% with respect to periods of lower solar activity. The flux into the atmosphere is also affected by the Earth’s magnetic field such that it is greater at higher latitudes. The GCRs that do penetrate the atmosphere are a major source of ionization, particularly in the lower stratosphere. There are two main theories advanced as to the means whereby variations in GCR flux might impact climate. The first concerns modulation of the Earth’s electric field. Ionization of the atmosphere by GCRs, or indeed by solar energetic particles, will influence its conductivity and thus the current flow within the Earth’s electric circuit. It is plausible that the current flow into clouds may affect cloud microphysics, ice formation, and hence thunderstorm activity, although the details of this mechanism are not fully established. While GCRs are more prevalent during periods of minimum solar activity, solar energetic particles are more numerous at solar maximum, so the variation of this effect over the solar cycle is not clear. The second means proposed whereby GCRs may affect climate is through enhancing the production of cloud condensation nuclei through growth of aerosol on ionized air molecules. Evidence for the existence of this process has been obtained by a recent observational and modeling study, although not as yet in response to solar activity. The potential advantages of theories of solar– climate links which rely on GCRs over those based on electromagnetic radiation are twofold. First, they circumvent the argument that solar variations are energetically too small to produce the apparent effects. Second, the potential response time of the atmosphere is faster. However, much further research into the proposed mechanisms is required, alongside further careful observational studies, before their existence is established. Further Reading Board on Global Change (1994) Solar Influences on Global Change. Washington, DC: National Academy Press. Burroughs WJ (1992) Weather Cycles: Real or Imaginary? Cambridge: Cambridge University Press. Hoyt DV and Schatten KH (1997) The Role of the Sun in Climate Change. New York: Oxford University Press. Lean J and Rind D (1998) Climate forcing by changing solar radiation. Journal of Climate 11: 3069–3094. Nesme-Ribes E (ed.) (1994) The Solar Engine and its Influence on Terrestrial Atmosphere and Climate. Berlin: Springer-Verlag. SOLAR WINDS S T Suess, NASA Marshall Space Flight Center, Huntsville, AL, USA B T Tsurutani, Jet Propulsion Laboratory, Pasadena, CA, USA Copyright 2003 Elsevier Science Ltd. All Rights Reserved. Introduction The Sun is losing mass in form of the solar wind, which has affected its evolution from its birth and will continue to do so until its death. This is not unusual, in that nearly all stars are losing mass through stellar winds throughout a major portion of their lives. As far as the Earth is concerned, the solar wind blows against the Earth’s magnetosphere, causing auroras and geomagnetic storms. The solar wind forms in the corona and is caused by high pressure in the corona relative to the low pressure far from the Sun in the interstellar medium. This pressure gradient exerts an outward force against gravity and accelerates the wind from low speeds in the low corona up to supersonic speeds at 2–10 solar radii (RS ). To give a sense of scale, the Earth is 1.5 108 km 5 215RS from the Sun (defined as 1 astronomical unit or 1 AU). Typical solar wind speeds beyond 10RS are between 300 and 800 km s  1 so it takes the solar wind 2.2 to 5.8 days to reach Earth. The existence of the solar wind was inferred prior to the space age from the existence of auroras, disturbances to the Earth’s magnetosphere, and observations of comet tails. Today it is regularly observed with a number of spacecraft. Coronal Expansion The Corona The surface of the Sun is called the photosphere, above which lies the Sun’s atmosphere, known as the corona. Pressure in the corona is high because the temperature is high, more than 106 K, relative to the photospheric temperature of B5800 K. This is a sufficiently high SOLAR WINDS temperature that the corona emits copious X-rays. It is believed that the corona is heated to this high temperature as a by-product of magnetic field motions, interactions, and instabilities in the photosphere that directly transfer energy into the corona. This energy flux could be via direct heating, waves, jets of material, or other forms, but this is unknown and is the subject of research by several different observatories in space and a deep-space mission called Solar Probe that will travel to within 3RS of the photosphere. The corona is composed mainly of protons and electrons (ionized hydrogen), with minor amounts of silicon, carbon, iron, oxygen, and other elements. There is about 20% helium (by mass) that can be observed spectroscopically; this is how helium was first discovered – hence its name (after Helios, the Sun god of Greek mythology). All the components share in the expansion of the corona and can be measured in situ by spacecraft outside the Earth’s magnetosphere. The Sun’s magnetic field makes the solar wind far from a simple spherical expansion of a hot gas. The magnetic field is dipole-like but undergoes a reversal during every 11-year solar sunspot cycle. At sunspot minimum the field is aligned with the solar rotation axis, while at solar maximum many sunspots appear and the dipole field weakens and becomes irregular. From solar maximum to minimum the field again becomes dipolar, but is inclined relative to the rotation axis. These changes are reflected in changes in coronal structure, which can be seen during solar eclipses such as that shown in Figure 1. With the bright disk of the Sun being occulted by the moon, the faint corona becomes visible, primarily because of light coming 2079 from the photosphere being reflected off of electrons in the corona. The areas that are bright are regions of high density and are known as streamers. The dark regions at the top and bottom in Figure 1 are coronal holes. The streamers in Figure 1 lie over the magnetic equator and the density is higher because ions and electrons are trapped on closed loops of the dipolar magnetic field. The low-density coronal holes mark the locations of the north and south magnetic poles. Figure 1 was taken in 1994, just prior to solar minimum, so that the magnetic axis was only slightly inclined to the rotational axis, which is vertical in this image. Figure 2 is a schematic of the stages of coronal evolution over the 11-year sunspot cycle, starting at solar maximum on the left. Figure 1 is represented by the drawing in Figure 2C. The magnetic field loops in streamers are shown here to help suggest why the ions and electrons are trapped, just as iron filings tend to align along magnetic field lines around a bar magnet. Coronal holes are shown by the dark areas on the solar disk. This figure also indicates that fast solar wind originates from coronal holes and slow solar wind from above streamers. Fast solar wind has speeds above B600 km s  1 at 1 AU and slow solar wind has speeds below B500 km s  1. This division into fast and slow wind could be observed if one were able to pass around the Sun as shown in Figure 2C from south pole to north pole. It would then be possible to sample first fast wind from the south pole, slow wind from over the equatorial streamers, and then fast wind again from the north pole. The Ulysses spacecraft carried out this exercise in 1995–1996 at solar sunspot minimum Figure 1 Total solar eclipse as seen from Putre, Chile on 3 November 1994. (Photograph courtesy of High Altitude Observatory, National Center for Atmospheric Research.) 2080 SOLAR WINDS (A) (B) (C) N N N Slow solar wind st Fa r ola nd wi s Figure 2 Schematic illustration of the three stages in the 11-year solar sunspot cycle. (A) Solar maximum, when the corona is filled with streamers and there are few or no coronal holes. There is no well-defined large-scale field. (B) Declining phases when the large-scale field is dipole-like and inclined to the heliographic equator. (C) Solar minimum when the field is dipolar, aligned with the rotation axis, and when the polar coronal holes are largest. and a plot of the observed solar wind speed is shown in Figure 3 using what is called a dial plot. The dial plot indicates the solar latitude around the origin and the measured solar wind speed as distance from the origin. The fast wind in the north and south is very clearly divided from the slow wind above the equatorial 90q N 900 Proton velocity (km s1) 600 300 streamers in this plot. This demonstrates one of the major discoveries in recent years F that fast and slow solar wind represent two distinct states between which there is no continuous change. Fast wind comes from coronal holes and is rather smooth and uniform at 1 AU. Conversely, slow wind is relatively irregular and comes either from thin boundaries around streamers or leaks somehow from within streamers. Figure 4 shows profiles of how fast and slow wind vary with distance from the Sun, illustrating not only that the speeds are different but also that there are characteristic densities and temperatures differences. Te and Tp are the proton and electron temperatures in this plot. The distinct difference between the two solar wind states leads to important consequences because of solar rotation. Solar Rotation and the Magnetic Field in the Solar Wind 0q (Equator) 0 90q S Figure 3 Dial plot of solar wind speed, indicated by radial distance from the origin, as a function of heliographic latitude, measured around the origin of the plot. Data were collected by the Ulysses solar wind plasma instrument between September 1994 and July 1995, during which time Ulysses swept from 801 south latitude to 801 north latitude. Solar wind is an ionized gas made up primarily of protons and electrons with minor ions in amounts similar to those in the corona. The electrons and ions are very tightly bound to lines of magnetic flux, again like the coronal plasma in streamers. However, the magnetic field in the solar wind is relatively weak and thus is carried along by the solar wind. The rotation of the Sun results in the lines of magnetic flux in the solar wind being drawn into Archimedean spirals. This occurs because the Sun revolves once every B25.5 days while, as mentioned above, it takes solar wind several days to reach 1 AU. Therefore, the Sun revolves through a significant angle during the time it takes the solar wind to reach the Earth. For example, taking a typical speed of 400 km s  1, it takes solar wind 4.34 days to reach 1 AU. During the same time, the Sun will have revolved through about 601, or about 1/6 of a full rotation. The magnetic field in the solar wind, called 2081 SOLAR WINDS Flow speed (km s1) 1000 Density (cm3) 107 10 Temperature (K) 7 105 500 103 106 Tp Te 10 4 10 100 4 10 100 Tp 4 Te 10 100 Heliocentric distance (solar radii) Figure 4 Solar wind flow speed, density, and temperature between 2RS and 100RS , for coronal holes (yellow lines) and streamers (black lines). These are typical values, with the possible range around the interplanetary magnetic field, or IMF, is attached to the Sun at the point where the solar wind began. Thus, the point on the field line attached to the Sun is turned through an angle of 601 relative to the point on the magnetic field line that is at 1 AU. The field line between the Sun and 1 AU traces a continuous curve between these two points. Assuming the solar wind speed, v (km s  1), is independent of distance from the Sun, this curve is described by eqn [1]. r  r0 ¼ v ðf  f0 Þ O cos y ½1 In eqn [1], r is the distance from the center of the Sun in km, r0 5 6.96  105 km is the radius of the Sun, O 5 2.85  10  6 s  1 is the angular velocity of the sun, and (f  f0 ) is the difference in longitude (in radians) at the two points on the field line. y is solar latitude and the Earth lies in the range 7:25oyo7:25 degrees because the plane of the ecliptic is inclined to the solar equator by 7.251. The angle (f  f0 ) is also the angle between the magnetic field line and the radial direction at 1 AU, or wherever eqn [1] is evaluated. This is called the spiral angle. The geometry of the curved field line is precisely an Archimedean spiral when v is constant and this is one of the important predictions made by E. Parker when he developed his theory for the solar wind in the 1950s and 1960s. Figure 5 illustrates two spirals computed using eqn [1]. The tighter spiral above results from low speeds, o500 km s  1, and the spiral angle is 4451 at 1 AU. Conversely, the spiral angle at 1 AU is o451 for speeds 4500 km s  1. Parker predicted that (f  f0 )B451 (0.785 radians) at 1 AU and this is what has been measured for the average spiral angle by several different spacecraft. Figure 5 Diagram of spiraling interplanetary magnetic field (IMF) lines. The dependence on solar wind speed is illustrated by the more curved line at the top being for relatively slow wind and the less curved line at the bottom being for fast wind. Corotating Interaction Regions Solar rotation has an important effect on coronal expansion through the interaction of fast and slow wind. During the declining phases of the solar cycle, Figure 2B, regions on the Sun producing slow wind will sometimes face the Earth and at other times regions producing fast wind will face the Earth. Thus it will often be the case, especially during declining phases of the solar cycle, that slow wind will be followed by fast wind. This is just the example diagrammed in Figure 5. When this happens, fast wind overtakes slow wind, the gas in between becomes compressed, and eventually shocks form with forward shocks moving away from the Sun and reverse shocks moving toward the Sun in the frame of reference moving with the solar wind. This is called a corotating interaction region (CIR) because it appears stationary in the frame of reference rotating with the Sun. As the plasma between the fast and slow wind becomes compressed, the velocity profile is dynamically altered and the CIR becomes stronger and stronger with increasing distance until the shock forms. A simple 2082 SOLAR WINDS estimate for where the shocks will form can be made using eqn [2], where the same definitions are used as in eqn [1]. r  r0 ¼ v1 v2 ðf2  f1 Þ v2  v1 O cos y ½2 The quantity (f2  f1 ) is the difference in longitude of the source regions of fast and slow wind, v1 and v2 are the slow and fast wind speeds, respectively, and r is the estimated distance for shock formation. Taking (f2  f1 ) 5 0.53 radians 5 301, v1 5 400 km s  1, and v2 5 800 km s  1 gives r 5 1.5  108 km 5 1 AU. During the declining phases of the solar cycle it is observed that shocks generally form around 2 AU, which is consistent with eqn [2] since (f2  f1 ) is more nearly 1 radian than 0.5 radians at those times. Forward shocks are rarely observed at 1 AU and reverse shocks are only observed in B20% of CIRs at 1 AU. Equation [2] was derived simply by calculating when the two field lines shown in Figure 5 would cross. These field lines are the same as the streamlines in the frame of reference corotating with the Sun, and this is why eqn [2] looks closely related to eqn [1]. CIRs have a very distinctive character, as seen in the long series of CIRs observed by Ulysses in 1992 when it was near the Sun’s equator. About five solar rotations of the data are shown in Figure 6. At the time Ulysses was at B4 AU and fast wind had overtaken slow wind to form shocks where the speed is seen to abruptly jump upward as time progresses from left to right. CIRs have important consequences for the Earth since they can produce auroral activity and magnetic storms Flow speed (km s1) 1000 900 800 700 600 500 400 300 1992:01/08 24/08 16/09 9/10 1/11 24/11 17/12 Figure 6 Solar wind speed at Ulysses during August–December 1992 when Ulysses was near the heliographic equator and at B5 AU. Five corotating interaction regions (CIRs) are shown, occurring approximately every 25.5 days, or each solar rotation. Viewing the plot from left to right, each CIR is characterized by a sharp speed increases at forward and reverse shocks at the front of the CIR, followed by the speed maximum. The speed then decreases to a minimum before increasing in the next CIR. The very high speed on 10 November 1992 is due to a coronal mass ejection on top of the CIR. when they strike the Earth’s magnetosphere if the IMF is also directed southward so that it can easily merge with the Earth’s magnetic field. CIR-associated magnetic storms naturally tend to recur every solar rotation – 27 days as viewed from the Earth owing to the Earth’s motion around the Sun. This activity also has a distinct solar cycle signature as the Sun moves through the phases diagrammed in Figure 2. Thus, observation of coronal holes and streamers and the phase of the solar cycle provides a basic tool for the prediction of space weather and geomagnetic activity. A further consequence of CIRs is that the resulting shock waves produce large numbers of high-energy particles or cosmic rays. These particles affect the Earth’s ionosphere and the radio communications that depend on the ionosphere. Coronal Mass Ejections Up to this point, a picture of the solar wind has been drawn that depicts it as quasi-steady, changing only slowly over the 11-year solar sunspot cycle. This is not an accurate picture at any time, especially near solar maximum. There are many forms of solar activity, including flares and erupting prominences, but the most dramatic is the release of a coronal mass ejection, or CME. A picture of a CME is shown in Figure 7. This picture was taken from the SOHO spacecraft using a telescope called LASCO, which places an occulter over the solar disk so that the corona becomes visible, producing an artificial solar eclipse. The occulter is twice the size of the Sun, and the disk of the Sun is indicated by the white circle. Off to the lower right of the image is the CME. These are seen throughout the entire solar cycle, but they are 5–10 times more common near solar maximum, occurring at a rate of 3–4 per day. They occur in and near streamers, confined to low latitudes near solar minimum but reaching all latitudes at solar maximum. When an interplanetary CME (ICME) strikes the Earth, the consequences are similar to those of a large CIR. The magnetosphere is compressed, auroral activity increases, and a magnetic storm or substorm may occur if the IMF in the ICME is directed southward. Ionospheric activity is also affected. This is therefore a phenomenon that is actively monitored in the context of space weather. One CME is visible in the data shown in Figure 6. At about 10 November 1992 the solar wind speed increased to B1000 km s  1. This is above any speed for simple fast solar wind. Instead, what is seen here is a fast ICME that has overridden a CIR. This could have a doubly strong impact on the magnetosphere owing to the large speed enhancement. SOLAR WINDS 2083 Figure 7 A coronal mass ejection is seen in the lower right quadrant in this image from the LASCO coronagraph on SOHO. The Sun, which is covered by an occulter that is 4RS in diameter, is indicated by the white circle. ICMEs are another phenomenon in the solar wind that is only partially understood. The propagation of an ICME can be modeled fairly well using computers and a numerical solution of the equations of motion. However, the basic mechanisms causing the initiation of a CME are not known. CMEs are related to solar magnetic activity such as flares and erupting prominences, but that relationship is not so simple that one can predict a CME for anything except the very largest of these events. The Solar Wind over the Life of the Sun The IMF is not completely passive in the solar wind. Because it is attached to the Sun, and has a small, but finite strength, the IMF tends to cause the solar wind to rotate with the Sun out to some distance above the photosphere. In doing this, the IMF causes angular momentum of the Sun to be transferred to the solar wind. Generally this is a small effect, with the corotation distance being 10–20RS at most, or 0.1 AU. However, over the life of the Sun, the effect can be important. Calculations of the angular momentum transfer suggest that the present-day solar wind and IMF could easily have doubled the rotation period of the Sun, from 12.25 to 25.5 days, over the 4.5 billion year life of the Sun. Presently the solar wind carries away only a very small amount of mass from the Sun – so small that if assumed the same for 4.5 billion years it would have removed only B0.01% of the total mass of the Sun. However, the Sun changes over its life, as do all stars. The Sun probably had a very vigorous wind early in its life when the solar convection zone extended throughout the entire volume of the Sun. Later in its life the Sun will go through a red giant phase, expanding outward to envelop the Earth, and the wind may again become quite strong. If the Sun undergoes a catastrophic collapse to form a white dwarf then there may be one or several episodes of impulsive mass ejection called novae. However, the Sun is a relatively small and inactive star; other stars can have quite different and often far more intense winds. Winds from Other Stars Stellar winds are, as indicated above, common. One means by which they are detected and analyzed is through Doppler shifts in spectral lines. Another is to infer the presence of the wind through analysis of properties of the associated star. Stellar winds found this way are all far stronger than the solar wind, but the reader should be cautioned that this is an observational selection effect. The Sun’s wind would be invisible at stellar distances. If all stars were like the Sun, we would presently have no way to directly detect their winds. However, many stars are larger, hotter, denser, rotate faster, have stronger magnetic fields, are younger, or are older than the Sun and consequently have quite different kinds of winds. They fall into 2084 SOLAR WINDS several categories that are in addition to winds like the solar wind that are primarily driven by a thermal pressure gradient. Sound Wave-Driven Winds In stars with a convection zone just below the photosphere, the convective motions can generate acoustic waves that propagate upwards through the photosphere. The waves produce a wave pressure in the atmosphere that results in an additional force working against the stellar gravity. Cool stars have convection zones of this type but the phenomenon is normally important only for very low-gravity stars. To make a massive wind requires something else in addition to sound waves because sound waves will normally dissipate low in the stellar atmosphere. The dissipation of sound waves heats the atmosphere so that there can be some crossover between thermally driven winds and sound wave-driven winds. Dust-Driven Winds The outer atmospheres of luminous cool giant stars and early type stars can be driven outward by radiation coming from the photosphere of the star. In the case of cool stars, dust can condense out of the atmosphere and absorb photons over a broad range of wavelengths. The radiation pressure forces the grains outward, dragging ions along by viscous drag if the atmosphere is dense, thus forming a dust-driven wind. Alfvén Wave-Driven Winds Alfvén waves are waves dominated by fluctuations transverse to the magnetic field direction. The restoring force is the resistance of the magnetic field to forming a kink, as opposed to the resistance of a gas to being compressed in sound waves. These waves are more important for stars with stronger magnetic fields. The dissipation of energy and momentum associated with Alfvén waves can lead to the acceleration of a wind, just as in sound wave-driven winds. The waves are formed by motions in the photosphere causing the magnetic field line to be moved. Alfvén waves have been suggested to be one source of the energy flux driving the solar wind. However, it is not yet known whether this is the dominant energy source. The dissipation of Alfvén waves will heat the atmosphere and increase the thermal pressure so that there is also some crossover between thermally driven winds and Alfvén wave-driven winds. Radiation Pressure-Driven Winds In these winds, atoms in the atmosphere of the star resonantly absorb radiation coming from the photosphere of the star. As might be expected, these winds exist for stars that are brighter and hotter than the Sun. Instead of 10  4 solar masses being lost over the life of the star, these stars can lose 10  6 solar masses in a single year. The flow speeds are typically B2000 km s  1 and the density in these winds is many orders of magnitude higher than in the solar wind. The higher density means that the atmospheres of these stars are far more opaque than the solar corona. This is what enables them to absorb the radiation coming from the star. In this case the radiation pressure is the force that is working against the gravitational field of the star. The force ceases once the atmosphere becomes transparent as distance from the star increases. In red giant stars the radiation intensity is relatively weak, but the gravitational field is also weak and the stars are nevertheless observed to have radiatively driven winds. However, the strongest radiatively driven winds come from hot supergiants. Magnetic Rotator Winds In discussing the solar wind over the lifetime of the Sun we described how the magnetic field enhances the loss of angular momentum from the Sun by causing the ions and electrons to rotate together with the Sun as they move outward. At the same time, there is also a small outward centrifugal force, just as there is in a centrifuge. This force is completely negligible for the Sun, but one can imagine stars with stronger magnetic fields that might have centrifugally driven winds; these are called magnetic rotator winds. The most obvious example of a magnetic rotator wind is that from a neutron star. These stars have very strong magnetic fields and centrifugal forces fill the neutron star magnetosphere with charged particles. At some distance from the star, the azimuthal velocity of the charged particles, as they are carried around the star, reaches the speed of light. The surface at this distance is the ‘speed of light cylinder’ and somewhere in this region around the star the particles force the field lines to open and they are released. This is how a pulsar is formed, an extreme example of a magnetic rotator. Effects of Winds on Stellar Evolution and on the Surrounding Interstellar Medium Winds from stars are one way in which matter that has been processed in stellar interiors reaches the interstellar medium and becomes available for new star formation; the other way is via novae and supernovae. The composition of the wind reflects, but may not be identical to, the composition of the star. Primordial material will be processed and enriched in heavy elements in this process. SOLAR WINDS The solar wind serves as an example of this process, even though the wind is relatively weak. The wind moves outward to interact with interstellar material that is always present in the galaxy. There is a contact surface that divides interstellar material from solar wind and the volume inside this surface is known as the heliosphere – that volume dominated by the Sun. The solar system is moving through the local interstellar medium at B25 km s  1 – slow with respect to solar wind speeds – and the stand-off distance in the upstream direction is about 150 AU. Beyond this boundary lies pristine interstellar matter. In the downstream direction the solar wind flows into a heliotail that is analogous to the Earth’s magnetotail and is the path by which the solar wind escapes and mixes with interstellar matter. 2085 RS Radius of the Sun, 6.96  105 km. SOHO Solar-Heliospheric Observatory. A joint ESA and NASA spacecraft located at the L1 Lagrangian point between the Sun and the Earth, about 0.01 AU toward the Sun from the Earth. Solar Probe A future NASA mission to the Sun. Solar Probe is designed to go within 3RS of the photosphere. Ulysses A spacecraft in a near-polar 5-year orbit around the Sun. See also Convection: Convection in the Ocean. Electricity, Atmospheric: Global Electrical Circuit. Ionosphere. Magnetosphere. Radiation (Solar). Satellites: Orbits. Solar Terrestrial Interactions. Acknowledgement Portions of this work were performed at the Jet Propulsion Laboratory, California Institute of Technology, under contract with NASA. Glossary Alfvén wave A bending wave in a magnetic field in which the restoring force is due to the curvature of the magnetic field. Named after Hannes Alfvén, the Nobel prize-winning scientist who discovered the wave. AU Astronomical unit, the mean distance of the Earth from the Sun, 1.5  108 km. CIR Corotating interaction region. The dynamic interaction that occurs when fast solar wind catches up with and compresses preceding slow solar wind. See Figure 5. CME Coronal mass ejection. See Figure 7. Heliosphere The volume of space containing solar wind, as opposed to the interstellar medium, which is the Milky Way galaxy outside the heliosphere. IMF Interplanetary magnetic field. The magnetic field that is trapped in and carried along with the solar wind. LASCO Large Angle Spectroscopic Coronagraph. A coronagraph on the ESA/NASA SOHO mission. Photosphere The visible surface of the sun. Radian A measure of angular distance. There are 2p radians in a circle. Further Reading Fleck B, Noci G and Poletto G (eds) (1994) Mass Supply and Flow in the Solar Corona. Dordrecht: Kluwer. Habbal SR, Esser R, Hollweg JV and Isenberg PA (eds) (1999) Solar Wind Nine, AIP Conference Proceedings 471. NewYork: American Institute of Physics. Hundhausen AJ (1972) Coronal Expansion and the Solar Wind. New York: Springer-Verlag. Kivelson MG and Russell CT (1995) Introduction to Space Physics. Cambridge: Cambridge University Press. Lamers HJGLM and Cassinelli JP (1999) Introduction to Stellar Winds. Cambridge: Cambridge University Press. Marsden RG (ed.) (1986) The Sun and the Heliosphere in Three Dimensions. Dordrecht: Reidel. Parker EN (1963) Interplanetary Dynamical Processes. New York: Interscience/Wiley. Sturrock PA, Holzer TE, Mihalas DM and Ulrich RK (eds) (1980) Physics of the Sun, vols I, II, and III. Boston: Reidel. Suess ST and Tsurutani BT (eds) (1998) From the Sun: Auroras, Magnetic Storms, Solar Flares, Cosmic Rays. Washington DC: American Geophysical Union. Tsurutani BT, Gonzalez WD, Kamide Y and Arballo KK (eds) (1997) Magnetic Storms, Geophysical Monograph 98. Washington DC: American Geophysical Union. Ulmschneider P, Priest ER and Rosner R (eds) (1991) Mechanisms of Chromospheric and Coronal Heating. Berlin: Springer-Verlag. Winterhalter D, Gosling JT, Habbal SR and Kurth WS, Neugebauer M (eds) (1996) Solar Wind Eight. AIP Conference Proceedings 382. New York: American Institute of Physics. 2086 SOLITARY WAVES SOLITARY WAVES J P Boyd, University of Michigan, Ann Arbor, MI, USA Copyright 2003 Elsevier Science Ltd. All Rights Reserved. Definition In the narrowest sense, a solitary wave is a single, isolated wave crest which propagates steadily without either steepening or widening. However, the concept has been broadened by the discovery of many new species of similar phenomena. Also, nonpropagating ‘coherent structures’, especially vortices, have much in common with solitary waves. Thus, ‘solitary wave’ is no longer a phenomenon but a theme. The theme is nonlinear self-preservation of a crest or a vortex in the face of opposing, disruptive forces. and the slowest runners falling farther and farther behind at the rear. Wave dispersion is the same as track-and-field dispersion: When waves travel at different speeds, the disturbance must spread over time unless some other mechanism intervenes. One such mechanism is advective steepening. If the fluid velocity is proportional to height, then an initial bell shape will evolve a leading-edge front (Figure 1). As the fast-moving tip overtakes the lower, slower fluid, the trailing (left) edge is stretched while the leading edge steepens (‘frontogenesis’). In a solitary wave, dispersion and nonlinear steepening exactly balance so as to create a wave which propagates without change of shape. History of Solitary Waves Dispersion, Frontogenesis and the Bell Soliton The left curve in Figure 1 is a schematic of the simplest solitary wave. It is called a ‘‘bell soliton’’ because its shape resembles a church bell. Waves are said to be ‘dispersive’ if the propagation speed c of a sinusoidal wave varies with the wavelength l. It is possible to superimpose many sine waves of different wavelengths to make a bell shape, which is centered where all the crests are in phase. However, the bell shape rapidly disperses into an ever-widening patch of ever-shrinking ripples as illustrated in the upper right of Figure 1. In a marathon race, the runners are elbow-to-elbow at the start, but disperse into an ever-widening pack with the fastest runners in front Dispersion Bell soliton Advective steepening Figure 1 A bell-shaped crest (left) will dissolve into little ripples under pure wave dispersion; it will steepen and eventually break if advective steepening is unopposed by dispersion. In a solitary wave, dispersion and steepening exactly balance so that a bellshaped curve propagates steadily without change of shape. Solitary waves were discovered by the naval architect John Scott Russell in 1834. When a canal barge hit an underwater obstruction and stopped suddenly, Russell expected that the bow wave would dissolve into lots of little ripples through dispersion. Instead, a smooth, bell-shaped crest perhaps half a meter tall, independent of the cross-channel direction, emerged from the froth. On horseback, he followed the unchanging, steadily propagating crest for a couple of kilometers until he lost it ‘in the windings of the canal’. Russell later made solitary waves in a long, narrow water tank. Dropping a square block into the water produced a localized wave disturbance which speedily organized itself into one or more solitary waves followed by a few small dispersing ripples. Forty years later, Rayleigh and Boussinesq showed that Aðx; tÞ  constant B2 sech2 ðB½x  ctÞ ½1 where the phase speed c is proportional to B2 where B is a positive parameter that simultaneously controls the width, speed, and amplitude of the solitary wave. Korteweg and deVries (1895) assumed that the wave was independent of the cross-channel direction, as approximately true both on the canal and in Russell’s water tank, and also that the horizontal current was depth-independent (the shallow water approximation). Thus, the only nontrivial spatial coordinate is down the channel. The surface height is then proportional to the solution Aðx; tÞ of the Korteweg–de Vries (KdV) equation: At þ c0 Ax þ mAxxx þ nAAx ¼ 0 ½2 SOLITARY WAVES linear steepening can preserve vortices and other moving structures even in the absence of wave propagation. Monopoles, modons, and weakly nonlocal solitons, described below, are important quasisolitons. Solitary Vortices: the Vortex in a Strain Field Nonlinear self-preservation is not limited to waves. Blobs of fluid are teased out into long, stringy filaments by the ‘strain’ or ‘deformation’ field created by distant vortices, pulled apart like taffy candy. However, if the blob is a sufficiently strong vortex, its own selfinteraction will preserve it. A patch of uniform vorticity will distort into an elliptically shaped vortex with its long axis at an angle of p=4 to the axis of strain, as shown in 1971 by Saffman and Moore (Figure 2). The vortex in a strain field is not in any sense a wave. Nevertheless, a vortex is a coherent, isolated structure which is preserved by nonlinearity in the face of disruptive mechanisms. On a rotating Earth, a large vortex will drift westward because of the beta effect – a wave-like behavior. It is then impossible to speak of ‘solitary waves’ and ‘isolated vortices’ as separate species. The coherent structure is both wave and vortex. A History of Isolated Vortices Smoke rings were discovered by casual observation long before there was any science of fluids. Any smoker can make a ring merely by blowing out a 1 Steady vortex in strain field 0.5 y where the subscripts denote partial differentiation with respect to the subscripted coordinate, and where the coefficients depend on the water depth, channel width and the gravitational constant. Today, it is known that the KdV equation is generic in the sense that it can be rederived for weakly nonlinear long waves in a wide variety of physics and engineering contexts, not just water waves. For 70 years after Korteweg and deVries, however, solitary waves were only a curiosity, mentioned in textbooks out of duty rather than conviction. Zabusky and Kruskal in 1965 numerically integrated the KdV equation on a spatially periodic domain and discovered that a large-amplitude initial sine wave dissolved into a sequence of solitary waves. These peaks collided elastically (that is, without loss of energy from the colliding pair), as robust as if they were elementary particles. Zabusky and Kruskal therefore coined the name ‘soliton’ which is today used as a synonym for solitary wave. A couple of years later, Gardner, Green, Kruskal, and Miura discovered the ‘inverse scattering’ method. Although the KdV equation is nonlinear, the inverse scattering method is an exact algorithm which solves the general initial value problem for arbitrary time by solving a sequence of linear subproblems. The KdV equation has special solutions which consist of N solitary waves of different sizes where N is arbitrary. Tall solitons overrun shorter solitary waves, collide elastically, and then all rematerialize in their precollision amplitudes and widths. The general KdV solution has two parts: a finite number of solitary waves plus a dispersing wave train. Except for rather special conditions, such as an initial height which is nonpositive, at least one solitary wave is always generated, even from wildly nonsolitonic initial conditions, such as the bow wave of a canal barge. The flow spontaneously either steepens or disperses so as to evolve an exact balance between nonlinearity and dispersion in the solitary waves even when these two competing mechanisms are wildly imbalanced initially. The following 10 years were the Golden Age of Solitons. Envelope solitary waves, kink solitons and other new species were discovered, each solving an inverse-scattering-solvable generic partial differential equation in (usually) one space dimension. For a time, it seemed that the inverse scattering method was the Algorithm for Everything. Then chaos theory blossomed, and it became clear that inverse scattering fails for most physical systems including the three-dimensional hydrodynamic equations. The last quarter of the twentieth century has been the Golden Age of the Generalized Solitary Wave. Many species of coherent structures almost satisfy the classical definition of a solitary wave. Further, non- 2087 0  0.5 1 1  0.5 0 x 0.5 1 Figure 2 Thin arrows: a pure straining field. A patch of vorticityfree fluid would be irreversibly contracted towards the x -axis and stretched along the y -axis. If the fluid is a sufficiently strong vortex, the patch will be deformed into an ellipse (heavy curve) with its long axis oriented at 451 to the contraction and dilation axes of the strain field. The sense of rotation within the ellipse of uniform vorticity is shown by the heavy double-ended arrows. 2088 SOLITARY WAVES mouthful of smoke. The smoke is trapped in a torus of fluid which propagates away from the mouth under its own self-interaction. The propagating torus is a vortex ring, rotating about its narrow diameter. In the 1860s, Tait showed a trailing smoke ring can overtake and pass through another, then slow to be passed in its turn, as in the child’s game of leapfrog. This robust survival of leapfrogging rings is reminiscent of the durability of KdV solitons under collisions that was discovered by Zabusky and Kruskal a century later. Anticipating their analogy of coherent fluid structures with elementary particles, Lord Kelvin was inspired by Tait’s experiments to create a theory that atoms were vortex rings, and molecules were interlocking vortex rings. Around 1900, Chaplygin and Lamb independently discovered analytic solutions for a pair of contrarotating vortices, now usually called ‘modons’ or ‘Chaplygin–Lamb dipoles’. Three quarters of a century later, Stern, Larichev and Reznik generalized these solutions to incorporate the beta effect. Vortex pairs form spontaneously through random near-collisions of one-signed vortices in turbulence, through injection of river and estuary flows into the oceans, and a variety of other mechanisms. Boyd generalized modons to vortex pairs that straddle the equator in the early 1980s. For small amplitude, these modons are well described by the KdV equation and are classical ‘bell’ solitons in longitude with the usual structure of linear equatorial waves in latitude; most of the propagation is wavy, due to the Rossby beta effect. As the amplitude increases, equatorial modons develop pockets of recirculating fluid, just as in Chaplygin and Lamb’s solutions, and the westward propagation is more and more due to the mutual interaction of the two vortices. One can no more say that an equatorial modon is either a wave or a vortex pair than one can assert that the color purple is either red or blue. Periodic Generalizations of Solitary Waves: Cnoidal and Polycnoidal Waves The adjective ‘solitary’ is as misleading as ‘wave’. Korteweg-deVries showed that the KdV equation has an exact elliptic function solution they dubbed the ‘cnoidal wave’. This is spatially periodic with an arbitrary period. In the limit of small amplitude for fixed period, the cnoidal wave is an ordinary cosine function. The large-amplitude cnoidal wave has a single, narrow peak which is well approximated by the sech2 shape of the solitary wave (Figure 3). The soliton is just a limiting case of the cnoidal wave. Eighty years later, Toda proved that the cnoidal wave is the exact sum of an infinite number of copies of the solitary wave where the copies are evenly spaced, one centered on each spatial period. This sum-ofsolitons relationship is true even in the small-amplitude regime (foreground of Figure 3) where Aðx; tÞ is also well approximated by the cosine function. This KdV cnoidal wave 100 Soliton 80 A 60 40 20 0 20 10 a Sinusoidal 0 10 0 5 5 10 x Figure 3 The KdV cnoidal wave as a function of x (for fixed time) and amplitude a. For small amplitude (foreground), the cnoidal wave is sinusoidal. As the amplitude increases, the peaks become taller, narrower, and more soliton-like. SOLITARY WAVES nonlinear superposition principle has since been extended to many other wave equations. Similarly, the KdV equation has exact analytical N-soliton solutions on an unbounded spatial interval which have been extended to spatially periodic exact solutions. These generalizations, which are ratios of multidimensional theta functions, are called ‘N-polycnoidal’ waves where the cnoidal wave is the special case N ¼ 1. Polycnoidal waves depend on N independent phase speeds. It can be proved that the general solution to the KdV equation with periodic boundary conditions can be approximated to arbitrary accuracy by a polycnoidal wave of sufficiently large N and appropriate phase speed and amplitude parameters. Thus, solitary waves need not be solitary. This is true for solitary waves in general and not merely for KdV solutions. Because solitons usually decay exponentially with distance from the core of the structure, a pair of solitons can be rather close and yet have a negligible dynamic interaction. Weakly Nonlocal Solitary Waves If the phase speed c of a coherent structure is multivalued in the sense that there are infinitesimal amplitude waves of some wavenumber k that have the same phase speed as the structure, then the solitary wave will not decay to zero at large distances from its center, but will instead radiate waves of wavenumber k. In many cases, the amplitude of the radiation is exponentially weak so that the structure behaves very much like a classical solitary wave. Such structures are called ‘weakly nonlocal’ solitary waves (Figure 4). Ironically, water waves, the prototype of solitons, are weakly nonlocal. The solitary wave radiates Core Wing Wing Figure 4 The ‘core’ of a nonlocal solitary wave is similar to a bell soliton, but the wave decays to small amplitude sinusoidal ‘wings’ instead of to zero. 2089 capillary waves, but these were too small for J. S. Russell to observe. Weakly nonlocal solitary waves are found in all branches of physics and seem to be just as common as classical, decay-to-zero solitons. Baroclinic vortices and solitons, for example, are usually nonlocal through weak radiation in the barotropic vertical mode. A Bestiary of Solitary Waves and Coherent Vortices Figure 5 shows the diversity of solitary waves and coherent structures. The six species illustrated are only a set of interesting creatures from a much larger zoo. Bell solitons, such as those that solve the KdV equation, have been described above. An ‘envelope solitary wave’ is the product of a sinusoidal ‘carrier wave’ with a slowly varying amplitude factor called the ‘envelope’, which is dashed in the figure. The envelope solves the nonlinear Schrödinger (NLS) equation. ‘Breathers’ are solitary waves whose amplitude oscillates in time. The breather may be either stationary or propagating, but the period and amplitude of the ‘breathing’ oscillations never changes. The sineGordon, self-induced transparency (SIT), and f4 field theory equations have breathers. ‘Kinks’, also known in some contexts as ‘travelling shocks’, occur in both inviscid models (such as the sine-Gordon equation) and viscous equations, such as Burgers’ equation and the Kuramoto–Sivashinsky (KS) equation. Viscous shocks seem a paradox since mechanical energy is being damped and yet the shocks, like solitary waves, are independent of time except for a steady propagation. The plateaus, extending indefinitely away from the shock, act as limitless reservoirs of energy to sustain the shock. Real kinks do not extend indefinitely, but are consistent local approximations to coherent structures of finite width. Vortices, whether monopoles or modons, are not always identified as solitary waves. If the diameter of the vortex is sufficiently small compared to the radius of the Earth, then wave effects may be only a small correction to vortex dynamics. However, vortices often exhibit the same robustness and nonlinear selfpreservation as KdV solitons. Monopole vortices have vorticity which is everywhere of the same sign except perhaps for an annular ring surrounding the core. Modons are pairs of contrarotating vortices as described earlier. These have a strong nonlinear translation, indicated by the hollow arrow in the figure, which is augmented or opposed by westward Rossby wave propagation. 2090 SOLITARY WAVES Bell Breather Envelope Kink Monopole vortex Modon Figure 5 A selection of soliton species. Solitons and Coherent Vortices in the Ocean In the Andaman Sea, tidal flow triggers regular trains of internal gravity solitons. These are visible in satellite photographs as long parallel streaks and are well modelled by the KdV equation. When the Gulf Stream separates from the coast at Cape Hatteras, it develops unstable, amplifying meanders that eventually roll up into Gulf Stream rings. Most ‘cold core’ rings perish in a few months by reabsorption, but the few that drift south of the Gulf Stream live a couple of years in the Caribbean. This is an order of magnitude longer than the lifetime of a small-amplitude Rossby wave of the same initial size (roughly 200 km in diameter). Similar coherent, long-lived eddies split from the Aghulas Current off South Africa. The high evaporation of the Mediterranean Sea creates dense, salty water that flows out through the Straits of Gibraltar into the Atlantic Ocean. As it sinks to a depth of 1000 meters, the anomalously hot-andsalty water rolls up into anticyclonic vortices called ‘Meddies’. These spinning lens-shaded masses, perhaps sixty kilometers in diameter and a kilometer thick, have lifetimes of half a dozen years or more. Smaller coherent vortices, both monopoles and dipoles, are very common. Dipoles with long stems of vorticity are called ‘mushroom vortices’ from their shape. These are easily made in the laboratory merely by injecting a jet of fluid into a rotating tank. River outflows and melting at the edge of the icepack are prolific generators of such vortex pairs, a few kilometers in diameter. Solitary waves and isolated vortices and vortex pairs seem to be very important components of ocean dynamics. There is room here to catalog only a small subset of the rather wide range of observed species. Why Atmospheric Solitons are Vertically Trapped When a wave propagates upward into thinner and thinner air with weak or negligible dissipation, its SOLITARY WAVES amplitude u grows so that the energy flux remains constant even as the mass density decreases exponentially with altitude. A steady balance between nonlinearity and dispersion cannot occur because the nonlinearity is steadily increasing with height. In contrast, the dispersion depends only upon the wavelengths of the sinusoidal waves that comprise the wave pulse and thus does not change with height. However, some waves are reflected by wind shear or static stability variations at some level, thus being trapped below the reflection height. Only such ‘vertically trapped’ waves can form solitons. The ocean is a bounded fluid of almost constant density, so the difficulties of propagation to space and vertically increasing nonlinearity do not arise. This is one important reason why solitary waves are more readily observed in the ocean than in the atmosphere. Atmospheric Solitary Waves New species of atmospheric solitary waves and new applications of previously studied types are inevitable. The following three examples are representative of the diversity of ‘soliton thinking’ to date. Internal Gravity Waves: the Morning Glory A low-level temperature inversion can create a layer of very stable air in the lowest kilometer or two of the atmosphere. Internal gravity waves are vertically trapped, and then can be reshaped by nonlinearity into a sequence of solitary waves. This mechanism 2091 operates all over the world. In particular such gravity solitons have been detected by Doppler radar and surface networks in Oklahoma. On the shores of the Gulf of Carpentaria in Northern Australia, conditions are especially favorable to generate such soliton trains, and to further make the soliton crests visible through condensation. These trains of roll clouds are known as the Morning Glory (Figure 6). The waveguide is leaky, so these solitary waves are ‘weakly nonlocal’. Indeed, the upward leakage is so strong that recent articles have argued that convective forcing may be as important as soliton dynamics in sustaining the crests as they roll in from the Gulf. Great Red Spot of Jupiter The Great Red Spot (GRS) is an anticyclonic, eyeshaped vortex embedded in a shear zone between alternating East–West jets at about 201 S latitude on Jupiter (Figure 7). It has been spinning for at least three centuries with only minor fluctuations in amplitude and appearance. It is an isolated vortex in the sense that it is the only large feature in the shear zone. However, it cannibalizes smaller eddies that appear on the edges of the zone, and this may help to sustain the GRS against losses to viscosity and radiative damping. A KdV theory has produced plausible agreement with observations; the eddy is both a vortex and a Rossby wave. Numerical models by G. Williams and S. Marcus offer a vivid explanation of GRS genesis. Generically, shear instabilies roll up into a string of vortices. Such chains of same-sign vortices are Weakly z stable layer Top of waveguide Very stable layer Brunt–Vaisala frequency Figure 6 Schematic of the Australian Morning Glory. Left: the vertical profile of Brunt–Vaisala frequency after convection has created a very stable, well-mixed surface layer. When sea breezes collide at night over Cape Yorke Peninsula, this excites a gravity wave disturbance which is trapped in the bottom layer by reflection from the interlayer boundary (dashed line) where the stability changes abruptly. The disturbance spontaneously organizes into an undular bore as it propagates over the Gulf of Carpentaria. Each peak is a solitary wave, and its updrafts cause condensation (shaded). The roll clouds may extend for over 100 kilometers perpendicular to the direction of propagation, which is indicated by the large arrow. 2092 SOLITARY WAVES Figure 7 Schematic streamlines and velocity arrows of the Great Red Spot of Jupiter. unstable to vortex mergers, and eventually a single large quasisteady vortex emerges as the end-product of the instability. But Jupiter is banded with many alternating jets; why is there a strong vortex in only one of these, and only in the Southern Hemisphere? Atmospheric Blocking Atmospheric ‘blocking’ is the formation of a quasistationary vortex or vortex pair over mountains which is sufficiently strong to block the usual mid-latitude storm track, forcing weather systems to detour around the block. There are many conflicting theories of blocking. However, the block is certainly a finite amplitude, quasistationary coherent structure that propagates westward against the prevailing westerlies so as to remain fixed above the mountain range. Much theoretical work has explored the idea that mountain chains are able to excite coherent blocks because the forcing is resonant: the forced solutions are very strong because they are close to unforced, finite-amplitude vortex pairs (modons). Because the forcing is weak, the modon paradigm is much more useful for blocks than for strongly forced-and-damped vortices like hurricanes. Misconceptions 1. Solitary waves are necessarily waves. Vortices and travelling shocks display the same nonlinear self-preservation as KdV solitons, and may move under a mixture of advection and Rossby wave dynamics. 2. Solitary waves are solitary; periodic waves have nothing to do with solitons. Wave crests and coherent vortices may be very close geographically and yet have almost no dynamic interaction. Chains of crests that appear wavy or sinusoidal may in fact be weakly interacting solitary waves. 3. Solitary waves are small amplitude only. This misconception was created by the derivation of the KdV and other simplified wave equations, which usually employ expansions in powers of the amplitude. However, numerical solutions show that the solitary waves do not magically cease to exist above a tiny limiting amplitude. Instead, solitons persist as a continuous family of solutions to such large amplitude that the soliton contains entrained fluid that is trapped within the structure as it propagates. ‘Small amplitude’ is a restriction of the mathematics, not physics. 4. Solitary waves are one dimensional. The KdV equation has only a single space coordinate. However, KdV theories often multiply the KdV solution, Aðx; tÞ, by a function YðyÞ which is spatially confined because of Coriolis refraction, as true of equatorial solitary waves, or shear trapping, as in the Great Red Spot of Jupiter. 5. Solitary waves are unforced and inviscid. Travelling shocks of Burgers and the Kuramoto– Sivashinsky equations are solutions to viscous differential equations. Furthermore, weakly forced and damped nonlinear structures may be accurately approximated by unforced, undamped solitons. However, the soliton paradigm is not very useful when the forcing dominates the flow, as true of hurricanes. The ‘Leonardo–Kolmogorov Duality’ Leonardo da Vinci, who sketched turbulent streams and scribbled notes on what he called turbolenza in 1500, seems to have known that turbulence could only be described (or painted!) as a mixture of coherent and random motion. Science progresses through a willful blindness to some aspects of a phenomenon to think deeply about others. (In a language of Papua New Guinea, this is ‘mokita’, which means ‘things we all know but agree not to talk about’.) Kolmogorov in 1941 made the first great breakthrough in tubulence by willfully ignoring the coherent structures, and pretending that turbulence is purely random. The Voyager photographs of Jupiter showed instead what Frisch has called the ‘Leonardo–Kolmogorov duality’. The Jovian atmosphere is neither completely coherent and predictable nor completely random. Instead, the ‘Leonardian’ Great Red Spot, which is a solitary vortex, coexists with a seething Kolmogorovian sea of billowing, fluctuating, random-appearing turbulence. The mystery of this soliton/random duality challenges our understanding today as it challenged Leonardo’s pen five centuries ago. SOOT 2093 See also Downslope Winds. Hydraulic Flow. Further Reading Ball P (1999) The Self-Made Tapestry: Pattern Formation in Nature. New York: Oxford University Press. Boyd JP (1989) New directions in solitons and nonlinear periodic waves: Polycnoidal waves, imbricated solitons, weakly non-local solitary waves and numerical boundary value algorithms. In: Wu T-Y and Hutchinson JW (eds) Advances in Applied Mechanics, no. 27. New York: Academic Press, pp. 1–82. Boyd JP (1998) Weakly Nonlocal Solitary Waves and Beyond-All-Orders Asymptotics: Generalized Solitons and Hyperasymptotic Perturbation Theory, vol. 442 of Mathematics and Its Applications. Amsterdam: Kluwer. Boyd JP (1999) The devil’s invention: Asymptotics, superasymptotics and hyperasymptotics, Acta Applicandae 56: 1–98. Boyd JP and Haupt SE (1991) Polycnoidal waves: Spatially periodic generalizations of multiple solitary waves, In: Osborne AR (ed.) Nonlinear Topics of Ocean Physics: Fermi Summer School, Course LIX. Amsterdam: NorthHolland, pp. 827–856. Johnson RS (1997) A Modern Introduction to the Mathematical Theory of Water Waves. Cambridge: Cambridge University Press. Lugt HJ (1983) Vortex Flow in Nature and Technology. New York: John Wiley. Nezlin MV and Snezhkin EN (1993) Rossby Vortices, Spiral Structures, Solitons. New York: Springer-Verlag. Nihoul JCJ and Jamart BM (eds) (1989) International Liége Colloquium on Ocean Hydrodynamics, no. 20 in Liége Colloquium on Ocean Hydrodynamics. Amsterdam: Elsevier. Remoissenet M (1991) Waves Called Solitons: Concepts and Experiments, 3rd edn. New York: Springer-Verlag. Smith RK, Crook N and Roff G (1982) The Morning Glory: an extraordinary atmosphere undular bore. Quarterly Journal of the Royal Meteorological Society 108: 937– 956. Van Dyke M (1982) An Album of Fluid Motion, 2nd edn. Stanford: Parabolic Press. SOOT P Chylek, Dalhousie University, Nova Scotia, Canada S G Jennings, National University of Ireland, Galway, Ireland R Pinnick, US Army Research Laboratory, Adelphi, MD, USA Copyright 2003 Elsevier Science Ltd. All Rights Reserved. Introduction Soot (also often called black carbon, charcoal, elemental carbon, or graphitic carbon) is produced by incomplete combustion of carbonaceous materials. Soot is found everywhere on Earth, including the atmosphere, oceans, sediments, soil, and ice sheets. It is also found in meteorites, may be present in asteroids and comets, and is believed to be responsible for dark absorption bands observed in stellar spectra. Soot is even suspected of participating in the initiation of life processes. In the atmosphere, particularly in the boundary layer, soot is a major component of aerosols that strongly absorbs solar radiation. Soot particles, when combined with sulfates, nitrates, sea salt, and organic particulate carbon present in the atmosphere, can serve as cloud condensation nuclei. Soot particles inside cloud droplets increase the absorption of solar radiation by droplets and modify droplet size distribution. Soot also provides a suitable surface and serves as a catalyst for various atmospheric heterogeneous chemical reactions. Thus soot is an important constituent of the atmosphere that affects atmospheric chemical composition and atmospheric radiation balance through both its direct effects (absorption and scattering of solar radiation) and its indirect ones (modifying the formation and lifetime of clouds and the size distribution of droplets). Soot contributes to atmospheric pollution. It reduces visibility and is also blamed for a variety of adverse health effects including a long list of respiratory diseases and various cancers. The main sources of soot in the atmosphere are biomass burning and fossil fuel combustion. Soot is the only material suspended in the atmosphere with a long residence time (up to 10 days) that strongly absorbs electromagnetic radiation of all wavelengths. Other atmospheric aerosols have either a very low absorption (sulfates, nitrates, sea salt and organic particulate matter) or a moderate absorption (soil and mineral dust) at visible wavelengths. From this follows the unique role of soot in the atmosphere as the only component of the atmospheric aerosol which strongly absorbs the visible part of solar radiation. Carbon is a major component of all living material. Carbonaceous particles produced by biomass burning or fossil fuel combustion span a large range of sizes. Particles with diameter over 10 mm are subject to fast gravitational settling and are removed from the 2094 SOOT atmosphere within a short distance from their sources. On the other hand, submicrometer-sized particles remain suspended in the atmosphere for several days and are transported over long distances. Black carbon (soot) has been found at all places on the globe, including the most remote areas in Antarctica. Carbon has atomic number Z ¼ 6. There are two stable carbon isotopes, 12C and 13C, and four radioactive ones: 10C, 11C, 14C, and 15C. The 14C isotope is used for carbon dating. The carbon atom has 6 electrons, and 4 of them are in the outer (2S and 2P) electron shells. These 4 valence electrons are available to form stable covalent bonds (shared pairs of electrons between 2 neighboring atoms) with other carbon atoms or atoms of other elements. Carbon atoms can thus form chains or rings of high complexity. If all 4 electrons are used in covalent bonds, the resulting materials are generally transparent in the visible part of the electromagnetic spectrum. On the other hand, if not all valence electrons are used for covalent bonds then the unused electrons can form a cloud of non-localized electrons, as in graphite, and the material will start showing a definite degree of absorption and anisotropic electric conductivity. There is an enormous variety of organic compounds of carbon. They are compounds primarily of carbon with oxygen, hydrogen, and nitrogen, although compounds with a number of other elements, including sulfur, phosphorus, and halogens, are also formed. When heated, all organic substances have one thing in common: they always produce, in addition to steam and carbon dioxide, a black material commonly called the char, soot, black, or elemental carbon. This is due to the fact that almost all of the combustion processes taking place are incomplete (oxygen-deficient): they do not provide sufficient oxygen for the full oxidation of the fuel, and generally some of the carbon will end up in a condensed phase rather than in gaseous oxide form. periodic potential formed by a hexagonal lattice of carbon atoms in graphite. Nonlocalized electrons are responsible for good electric conductivity of graphite in the basal plane of the hexagonal structure and for its absorption properties in the visible part of the electromagnetic spectrum. Separation between lattice planes is about 2.4 times the nearest neighbor distance (about 0.142 nm) within the plane. Graphite is often called a semimetal, indicating that it has some properties similar to that of metals. However, its effective density of charge carriers is of order 1018 cm  3; several orders of magnitude below that of typical good metals (1022 cm  3). In the diamond lattice, the 4 nearest neighbors form the vertices of a regular tetrahedron; all 4 valence electrons of each carbon atom are used to form strong covalent bonds with the four nearest neighbors. There is no planar anisotropy and no free electrons. A diamond is an extremely hard, high-density, transparent nonconductor. The basic characteristic of graphite and the basis of its high absorption in the visible part of the spectrum is its planar hexagonal structure. Whenever a sufficiently large number of carbon atoms get arranged in the form of a planar hexagonal lattice, some electrons will be only weakly bound to their respective atoms; they will form almost a free electron cloud and the material will manifest an increased conductivity and light absorption. This happens even if there are other atoms involved with carbon, as long as the number of other atoms is relatively small (usually below 20% by mass). Such a material, which is not a pure graphite, but at least partially manifests the basic graphite characteristics (elevated conductivity and increased absorption at visible wavelengths), is generally referred to as black carbon. There are two ways in which carbon atoms can be induced into the planar hexagonal lattice: either through a process of soot formation (high-temperature combustion) or by charring (lower-temperature burning). Graphite and Black Carbon Particulate Emission by Fossil Fuel and Biomass Burning The two basic forms of solid elemental carbon are diamond and graphite. They differ from each other in the form of lattice structure into which carbon atoms are arranged. This difference leads to vast dissimilarities in physiochemical and optical properties between the two carbon forms. Graphite has a structure of a planar hexagonal lattice with 4 carbon atoms per primitive cell. Within the lattice plane each carbon atom is bound to 3 neighboring atoms by strong covalent bonds. The 4 valence electron of each atom contributes to a relatively weak bond between planes of hexagonal lattices. These electrons are not bound to any particular carbon atom (nonlocalized electrons) and they can move relatively freely within a Particulate material ejected into the atmosphere during combustion processes contains soot, charcoal and ash. Ash originates from an inorganic component of fuel. Its mass is usually small (around 1%) compared with the mass of other forms of particulate matter ejected to the atmosphere. Unburned hydrocarbons react with atmospheric oxides of nitrogen and solar radiation to form smog. Soot Soot production generally proceeds through condensation of vaporized organic matter, usually through a SOOT 2095 number of polycyclic aromatic hydrocarbons (PAH). This is a complex process involving, first, the production of benzene and acetylene from the original biomass of fossil fuel. It is believed that most fuels break down into the same species at the beginning of the sooting process. In the second step of soot formation, the acetylene and benzene are transformed into the phenyl, a simple aromatic hydrocarbon with just one ring. The chain of aromatic rings then grows through a fast polymerization process (replacement of hydrogen atoms by C2H2 groups). With an increasing number of aromatic rings, a nucleus of a soot particle is formed. Some models consider four rings to be sufficient for soot nucleation. Thus the soot is produced by gas to particle conversion. A typical size of a soot nucleus is a few nanometers. The nucleus grows by additional condensation and coagulation. Freshly produced soot particles are almost spherical and they have tendency to form, through coagulation, chainlike structures with fractal geometry (Figure 1), with a fractal dimension of about 1.8. Soot production takes place at high temperatures, above 10001C, during the fossil fuel combustion or during the flaming stage of biomass burning. A mature soot particle is typically composed of a stack of layers, each of them having a graphite-like hexagonal structure. Not all layers are arranged in a parallel fashion. In addition to carbon, soot contains remnants of other elements present in the original fuel. A typical carbon content of soot is between 90 and 98%. Soot X-ray analyses indicates the presence of a regular graphite structure throughout the soot volume. Generally, fuels with higher C/H (carbon-to-hydrogen) ratio produce more soot. For a given amount of fuel, the variable flow conditions produce more soot than a steady-flow regime. Soot can also be produced by the oxidation of almost pure elemental carbon. At high temperatures, a carbon vapor is formed, which in colder regions, away from the flame, condenses to form solid carbon structures. In this way graphite-like soot as well as the famous fullerenes C60 and C70 are formed. Charcoal Charring of organic materials starts at temperatures considerably lower than that of soot formation. Burning of food during cooking (i.e., the production of nicely black toast) is an example of low-temperature charring. At temperatures above about 3001C, most of the organic materials undergo a slight thermal decomposition; hydrogen and other noncarbon elements are stripped from carbon chains and rings and the carbon condenses into a graphite-like structure. The density of black porous residuum depends on the mass ratio of carbon to other elements in the original material. X-ray analysis confirms that at temperatures above 3001C the hexagonal, graphite-like structure begins to form. This structure becomes more evident and more regular with an increasing temperature of oxidation. As hydrogen and other elements (e.g., nitrogen and sulfur) are released to the atmosphere and carbon atoms start forming the basic planar hexagonal structures, the optical properties of the material undergo a drastic transformation. With an increasing number of planar hexagonal rings, there is an increasing number of nonlocalized, almost free, electrons, and the material starts showing some of the graphite characteristics, especially an increased absorption of visible electromagnetic radiation. Original organic material becomes dark brown or black. In the case of biomass burning the charring process takes place during the smoldering phase. Fossil fuel combustion often produces carbonaceous particles that are in a form of hollow spheres (cenospheres) or of spongy spherical structures (Figures 2 and 3). The sizes of particles produced by charring are from submicrometer to several hundred micrometers. Smaller sizes are uplifted during the turbulent conditions produced by localized heating. They reside in the atmosphere for an extended time and are transported over long distances. Organic and Black Carbon Figure 1 Morphology of freshly produced soot, showing a characteristic chain-like structure of nanometer-size soot particles. Regarding aerosol radiative effects, the total carbon in atmospheric aerosols (excluding inorganic carbon in the form of carbonates as a part of soil and mineral dust particles) is usually divided into so-called organic and black carbon. This division is based not on aerosol chemistry, but rather on the aerosol optical properties. Carbon of atmospheric carbonaceous aerosols that absorbs visible radiation strongly is called black carbon; the remaining carbon (carbon of nonabsorbing carbonaceous particulate matter) is organic carbon. 2096 SOOT wavelengths. Should they be a part of organic or of black carbon? If we are interested in radiative effects of carbonaceous aerosols then all absorbing material should be kept in a category of black carbon. On the other hand, if we are interested in chemical reactions of organic aerosols then we may keep even absorbing organic compounds in the inventory of organic (rather than black) carbon. From the point of view of absorption of solar radiation in the atmosphere, it is reasonable to divide the total carbon into organic and black carbon, even if this division is not chemically well defined. Figure 2 A typical black carbon (charcoal) particle structure from a coal-fired power plant. (Photograph by R. Cheng; reproduced with permission from Chylek P, Ramaswami V, Pinnick R, and Cheng R (1981). Optical properties and mass concentration of carbonaceous smokes. Applied Optics 20: 2980–2985. The black carbon defined in this way contains pure graphite (elemental carbon), soot, and charcoal as well as their internal mixtures and their mixtures with organic carbon. Black carbon generally resists oxidation at temperatures below about 4001C, while organic carbon is oxidized easily at lower temperatures. The separation of total carbon into organic and black carbon is not unambiguously defined chemically. Some of more complex organic compounds may show a substantial absorption in the range of visible Black Carbon Measurements Black carbon (soot) properties, such as density, absorption coefficient, size, and morphology are highly variable. They depend on conditions of generation, source strength, atmospheric transport, transformations due to mechanisms such as catalytic surface reactions, and their degree of mixing of black carbon with other atmospheric particles as well as of their removal due to wet and dry deposition processes. Measurements of mass concentration, absorption, and size distribution of black carbon are relatively sparse up to the late 1970s, owing mainly to lack of suitable instrumentation. An increased interest in the role of soot in the atmosphere brought about the development and evaluation of new analytical methods and measuring techniques. Mass Concentration and Size Distribution Figure 3 Black carbon (charcoal) particle structures from an oilfired power plant (Photograph by R. Cheng; reproduced with permission from Chylek P, Ramaswami V, Pinnick R, and Cheng R (1981). Optical properties and mass concentration of carbonaceous smokes. Applied Optics 20: 2980–2985. Most soot size distribution measurements have been obtained from filter samples using multistage impactors combined with either conventional or transmission and scanning electron microscopy. More recent techniques include the use of an optical scattering aerosol sizing probe equipped with a heated intake. Soot particle size resides predominantly in the submicrometer accumulation mode regime, with a geometric mean diameter in the range 0.05–0.2 mm and with a geometric standard deviation 1.33 to 2.0. The average particle size increases with time during long-range atmospheric transport. Typical soot mass concentration values (Table 1) range from about 1 ng m3 for remote Antarctic locations to more than 1 mg m3 for polluted urban air. The number concentration varies from about 0.1 to 4100 cm  3. The mass extinction coefficient in polluted urban environments has typical values in the range 10  3 to 10  4 m  1, while representative values for a more remote atmosphere are r10  5 m  1. SOOT 2097 8 Remote (Antarctic/Arctic) Mid troposphere Marine Rural/continental Urban Extinction coefficient Mass concentration (m  1) (mg m  3) 1  10  8 1–3  10  8 1–5  10  7 1–5  10  6 1–4  10  5 0.001 0.001–0.003 0.01–0.05 0.1–0.5 41.0 _ Region Soot concentration (ng g 1) Table 1 Summary of black carbon (soot) measurements Ice core Recent snow 7 6 5 4 3 2 1 0 320 Black Carbon in Precipitation 322 324 326 328 330 Ice date (Year AD) The removal of black carbon from the atmosphere is believed to be primarily by wet deposition. However, there are only a few measurements of black carbon concentration in rain and snow. The method used consists of the filtering of collected precipitation through quartz fiber filters, followed by a thermooptical method of determination of the amount of black carbon on the filter. The range of black carbon concentration measured in rainwater and in snow is summarized in Table 2. Black Carbon in Ice Cores Ice cores preserve the information concerning the state of the atmosphere at the time of snow deposition (analysis of ice and aerosols) and at the time of enclosure of air bubbles (analysis of gases trapped in bubbles). Black carbon concentration in ice cores can be used to deduce the information concerning the past climate and the effect of man’s activities on the atmosphere. The black carbon concentration changes in Alpine glaciers indicate the increase of atmospheric black carbon concentration due to an increase in the regional industrial activities. On the other hand, no increase in black carbon concentration has been found in several analyses of Greenland ice cores. A comparison of black carbon (soot) concentrations found in the Greenland Summit GISP2 (Greenland Ice Sheet Project 2) ice core dated to around the years Table 2 Black carbon concentration in cloud water and in precipitation Type of cloud or precipitation Black carbon (mg kg  1) Marine Stratus, North Atlantic Stratocumulus, eastern Pacific Rain Water, eastern Canada Rain Water, Seattle Snow, eastern Canada Snow, New Mexico and western Texas Snow, Cascade Mountains Snow, Camp Century, Greenland Snow, Antarctica 8–60 20–80 1–11 3–400 1–32 5–16 22–59 2–3 0.2 Figure 4 Comparison between soot concentrations (ng g  1) in the Greenland Summit GISP2 ice core dated about 320–330 AD with that in recent snow (1989–1990) from the same location. There is no change of an average soot concentration in remote Greenland location between the current snow and the ice core more than 1700 years old. Large, ancient forest fires somewhere in the Northern Hemisphere are represented by peaks in soot concentrations around the years 324 and 326 AD. 320–330 AD and recent (1989–1990) snow from the same location suggests the same average concentration of about 2 ng g  1 (Figure 4). Optical constants of Black Carbon (Soot) Determination of the complex refractive index m ¼ n þ ik, where i is an imaginary unit, n and k are real and imaginary parts of refractive index, respectively, is a difficult task for soot or atmospheric black carbon. A number of different approaches have been made to determine the refractive indices of soot carbon. One of the principal methods used has involved the measurement of reflectivity of electromagnetic radiation from polished soot-like materials. Reflectivity methods have been applied to soot material, which has been compressed into pellets with nearly specular surfaces. The compression does not result in a uniform carbon material, but contains voids, which have to be considered in the determination of the optical constants. A combination of transmission and reflection has been used on an amorphous thin film of carbon. Another approach has involved extinction measurements for a suspension of carbon particles (of mean diameter 75 nm), which overcomes uncertainties associated with purity, crystal microstructure variations and void fraction of the sample. Indirect determination of the refractive indices of flame soot has been carried out in situ using light scattering combined with extinction measurements. However, the soot particle size and number concentrations were not measured directly. 2098 SOOT Table 3 Black carbon optical constants (real and imaginary part of refractive index) in the 0.35–1.5 mm wavelength range Material Real part Imaginary part Method of measurement Amorphous carbon 1.85–2.8 1.2–0.9 Carbon black Polycrystalline graphite Coal samples Soot 1.92 2.24 0.95 1.04 Transmission, reflection Extinction Fresnel reflection 1.6–2.1 1.5–1.9 0.3–0.5 0.4–0.8 Reflectance Reflectance A summary of measured optical constants of soot is presented in Table 3. The variability in the data can be attributed largely to factors such as degree of sample homogeneity, compositional change such as C/H ratio, density, sample preparation, etc. Recommended values for the refractive index of black carbon within the wavelength range from 0.3 to 1.5 mm (measurements indicate that the parameters do not greatly change with wavelength in the solar spectrum region) are: m ¼ ð1:9 to 2:0Þ þ ið0:7 to 1:0Þ. Effect of Soot on Radiative Properties of Aerosols and Clouds When soot gets incorporated inside cloud droplets or within a composite aerosol particle (to form an internally mixed aerosol) it modifies their radiative properties. The main effect of soot is to increase the absorption by droplets and aerosol particles. Since soot exhibits a strong absorption at all wavelengths from UV to far infrared, while liquid water has a strong absorption only in the infrared region, it is mainly the absorption of the visible and UV radiation that is enhanced by the presence of soot. Consequently, the presence of soot will decrease the single scattering albedo, o, at visible wavelengths of cloud droplets and aerosol particles. The intensity of the electromagnetic field within a water droplet or sulfate aerosol is higher than the intensity in free space, owing to the focusing effect of the droplet. On average, a soot particle within a droplet, or as an internally mixed aerosol, will absorb more than twice as much of the incoming radiation than (externally mixed) soot in the free atmosphere. Effective Medium Approximation The single scattering albedo (the ratio of the scattering to the sum of the scattering and absorption crosssections) of a spherical aerosol particle or water droplet can be calculated using the standard Mie scattering formalism. Mie scattering calculations require as input the size of a spherical particle and its refractive index. A refractive index of a composite particle (a droplet of given material with soot inclusions inside) can be calculated using an effective medium approximation. For soot inclusions considerably smaller than the wavelength of a considered radiation, the Maxwell–Garnett effective medium approximation with soot inclusions surrounded by water matrix (or other material of the original particle) is an appropriate form of an effective medium approximation. The effective refractive index, meff , of a composite droplet is given by  m2eff ¼ m20 1 þ 4f ðm2c  m20 Þ m2c þ 2m20  2f ðm2c  m20 Þ  ½1 where m0 and mc are refractive indices of a matrix material (water or sulfate) and soot inclusion, and f the soot volume fraction. The single scattering albedo of a composite water–soot or aerosol–soot particle is then obtained by applying the Mie scattering formalism to a homogeneous particle whose optical properties are described by an effective refractive index. Soot and Direct Radiative Effect of Aerosols Soot incorporated within an aerosol particle will increase the particle’s absorption in the visible part of solar spectrum and thus it will decrease the particle’s single scattering albedo. The direct top of the atmosphere radiative forcing, DF, of an optically thin aerosol layer is given by DF ¼  h i S0 2 Tatm ð1  NÞ ð1  aÞ2 2btsc  4atabs ½2 4 where S0 is the solar constant, N the fraction of sky covered by clouds, Tatm the transmittance of the atmosphere above the aerosol layer, a the surface albedo, b the fraction of the scattered radiation that is scattered into the upper hemisphere, and tsc and tabs the scattering and the absorption optical thickness of an aerosol layer. The negative value of radiative forcing implies cooling of the system, while a positive value implies heating. For nonabsorbing aerosol tabs ¼ 0, and eqn [2] implies always a cooling effect. When soot is present within an aerosol, aerosol absorption increases and the direct aerosol effect will be either cooling or heating, depending on the relative magnitudes of the terms inside the bracket on the right-hand side of eqn [2]. For an optically thin aerosol layer, o ¼ tsc =ðtsc þ tabs Þ. The critical single scattering albedo, ocr , which determines whether an aerosol will heat or cool the system, is derived from eqn [2] in the form ocr ¼ 2a bð1  aÞ2 þ 2a ½3 SPECTRAL MODELS 2099 1.0 Model Cl 0.9 0.8 Cloud reflectivity For given surface albedo, a, and backscattering fraction, b, an aerosol with single scattering albedo o > ocr will cool the system, while aerosols with ooocr will cause heating. Thus the sign of a direct top-of-the-atmosphere aerosol forcing depends – in addition to the fraction of radiation scattered into the upward hemisphere and the albedo of an underlying surface – on the amount of soot within an aerosol particle (which determines the single scattering albedo o). Most aerosols will cause cooling over the ocean and heating over fresh snow. Thus, the soot heating effect will be especially significant over clouds, ice, and snow. 0.7 0.6 0.5 0.4 0.3 0.2 Pure water Soot volume _ fraction V = 10_ 7 V = 10_ 6 V = 10_ 5 V = 10 4 0.1 0 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8 Wavelength (Pm) Soot and Absorption of Solar Radiation by Clouds Soot within cloud droplets will again increase the droplets’ absorption of electromagnetic radiation and decrease their single scattering albedo. This leads to an increased absorption of solar radiation within a cloud layer, to heating, and to a possible increased rate of evaporation of cloud droplets. A small amount of soot, of the order of 10  9 to 10  7 by volume, in cloud droplets has little effect on cloud optical properties. However, soot in highly polluted regions, produced by industrial activities or biomass burning, can affect cloud absorption. Soot in cloud water concentration of the order of 10  6 and above will increase cloud absorption significantly. The effect of soot volume fraction, varying from 10  7 to 10  4, on the reflectivity of cloud is quite pronounced at visible wavelengths, as shown in Figure 5. Most accumulation-size soot particles can propagate up to several thousands miles away from their sources without a significant decrease in soot concentration. Thus, for example, an extensive biomass burning can affect cloud absorption and regional climate in regions several hundred miles away from source. Figure 5 Cloud reflectivity as a function of radiation wavelength for an optically thick (semi-infinite) cumulus cloud. The cases of pure water cloud droplets and for varying soot volume fractions in cloud droplets are shown. (Adapted with permission from Chylek P, Ramaswamy V, and Cheng RJ (1984). Effect of graphitic carbon on the albedo of clouds. Journal of the Atmospheric Sciences 41: 3076–3084. A significant reduction of cloud reflectivity at visible and near-infrared wavelengths is obtained for soot volume fractions at and above 10  6. See also Aerosols: Role in Cloud Physics; Role in Radiative Transfer. Aircraft Emissions. Biogeochemical Cycles: Carbon Cycle. Boundary Layers: Overview. Cloud Chemistry. Cloud Microphysics. Further Reading Cachier H (1998) Carbonaceous combustion aerosols. In: Harrison RM and van Grieken R (eds.) Atmospheric Particles, pp. 295–348. New York: Wiley. Goldberg ED (1985) Black Carbon in the Environment. New York: Wiley. Horvath H (1993) Atmospheric light absorption – a review. Atmospheric Environment 27A: 293–317. SPECTRAL MODELS F Baer, University of Maryland, College Park, MD, USA Copyright 2003 Elsevier Science Ltd. All Rights Reserved. Introduction With the advent of digital computers, weather forecasting was cast as a computational problem based on the fundamental prediction equations of fluids. Since analytic solutions are unavailable, approximations evolved to convert the differential equations to numerical equations suitable for computation on large computing machines. From this perspective, the concept of modeling was conceived. Thus weather forecasting – and more recently climate prediction – is approached by defining a numerical ‘model’, and solutions to this model are sought. A variety of models have been developed over time to meet this goal, and the ‘spectral model’ is one of these. 2100 SPECTRAL MODELS The atmosphere is represented by variables describing molecular composites of its gases; the primary variables are velocity, temperature, density, water content in all phases, and aerosols. These variables are considered to be distributed continuously in three-dimensional space and to vary with time. The evolution of these variables in time may be determined at each point in space (the Eulerian method) or by following the particles through time (the Lagrangian method) and both methods are in use. The differential equations defining the future state of the variables are based on physical and dynamical principles, some well known and others under study. These principles include the equations of motion (the Navier – Stokes equations), conservation of mass, an equation for change in entropy, equations for changes in water substance in its various phases, and chemical equations for changes of aerosols. To define the notation of this article, these equations are presented below (see Dynamic Meteorology: Primitive Equations). Using the Eulerian reference, the time derivative is taken locally at each point in the fluid. The motion of the fluid is determined by an equation for the vector velocity V relative to the rotating Earth in all three space dimensions (eqn [1]). qV 1 ¼ ðV . =ÞV  2X 3 V  rp  gk þ F qt r ½1 Here X is the angular velocity of the earth; r and p are density and pressure, respectively, at each atmospheric point; g is the gravitational acceleration in the k (unit vertical vector) direction; and F comprises all frictional forces per unit mass. Conservation of mass is represented by the equation of continuity and the system thermodynamics are described by changes in entropy as in eqns [2], [3] and [4]. qr ¼ = . rV qt  qs 1  ¼ V . =s þ q qv ; ql ; qi ; aj ; . . . ; 1 qt T qqk ¼ Qk qt ½2 j J ½3 ½4 In eqn [3], s represents specific entropy, q is the rate of heating per unit mass, and T is the temperature. Additionally, q depends on the heating rates associated with water vapor (qv ), ice (qi ), liquid water (ql ), aerosols (aj for j ¼ j1 ), and other factors such as radiation. Each of the variables qj and aj has its own prediction equation [4], where the Qk represent complex parametric formulas relating some or all of the dependent variables. This entire system of equa- tions constitutes the basis for selecting the ‘model’ that is integrated in time to predict the future state of the fluid. Additional features needed to complete the model are boundary conditions, initial conditions, scale truncation, external forces, and computational resources. The final step in constructing a model is to select a technique to convert the basic nonlinear differential equations that describe the forecast system [1]–[4] into a numerical form suitable for computation and integration on a digital computer. Finite differencing in both the time and space dimensions was the first method attempted. Since the vertical and horizontal dimensions in the atmosphere have unique properties, they may be and often are considered separately. Given that at any given height in the atmosphere a closed spherical surface exists on which the dependent variables describing the fluid are prescribed and predicted, the spectral method, which assigns a set of known continuous orthogonal functions over the domain to represent these variables, may be applied. When all the variables are described in this way, the resulting equations are integrated over the global domain, leading to a set of ordinary nonlinear differential equations in time and on each vertical level. Concurrently, differentiation in the vertical space coordinate and time is normally, but not universally, transformed to finite differences. The spectral method is most appropriate for the larger space scales since the functions usually used are global. However, regional models can be cast in the spectral framework if the boundary conditions are suitable. Alternative methods that have been applied include finite elements and spherical geodesic grids. In comparison with other modeling techniques, the spectral method has no pole problem; its resolution is essentially homogeneous and isotropic; it allows for a simple solution of the Helmholtz equation in various settings; and, with an appropriate choice of the transform grid, it produces nearly alias-free solutions. In addition to these advantages, it is also very computationally efficient. On the basis of these virtues, it has had a long run of success and has been the method of choice at many modeling centers. Computational Methods Models represented by finite differences are often denoted as gridpoint models and the grids for these models have been selected in a variety of ways. Despite their popularity, these models have many problems leading to significant computational errors, and the spectral method with its simple lateral boundary conditions over the globe is a natural alternative. Both methods are applied in the horizontal space SPECTRAL MODELS 2101 domain, and are combined with an alternate discretization in both the vertical and time domains. The techniques were developed with the prediction equations represented in the Eulerian framework; that is, all calculations are made locally in the space domain, including time extrapolation of the dependent variables. Although the structural characters of the two methods are substantially different, they can be cast in a similar representational form allowing for more systematic comparison. To elucidate this similarity, consider the dependent variables presented in the prediction equations [1]–[4] represented by the vector B ¼ fBb g ¼ ðVr s qv ql ; qi ; aj . . .ÞT where T represents transpose. The dimensions of B are determined by the number of variables in the system; let that be N. As the equations stand, the left-hand side of the set is simply qB=qt and the right-hand side can be summarized by a vector F with the same dimension as B to yield the following system [5]. qB ¼ FðB; r; tÞ qt ½5 F depends both differentially and nonlinearly on B, the space coordinates r, and time. Altering these equations by a transformation with the linear matrix operator L leads to the more general form [6] for the prediction system. L qB ~ ¼ FðB; r; tÞ qt ½6 Consider first the finite difference process applied to this system. Selecting a three-dimensional grid with M points to approximate the continuum in space with suitably prescribed boundary conditions, and a difference operator to describe derivatives, B is represented at each of the points and has dimensions (NM); if the values of B are available at some initial time, a numerical integration can proceed. The matrix L becomes by virtue of the difference operator an ðNMÞðNMÞ matrix, which can in principle be inverted, and F also becomes a numerical vector with NM elements after utilization of the difference operator at each grid point. Using a circumflex to represent numerical vectors and matrices at gridpoints, the finite difference system is written as eqn [7]. ^ qB ^ 1 F ^ ðB ^ ; ^r; tÞ ¼L qt ½7 The solution is thus reduced to a matrix computation provided a numerical scheme is introduced to step the variables forward in time, and the resulting computational errors and stability issues are dependent on the numerical and physical approximations made. The spectral method uses a different approach. Given a continuous domain over which the model variables are to be evaluated, a set of linearly independent global functions that are continuous over the domain with at least continuous first and second derivatives are selected. The model variables Bb are expanded in these functions with time dependent coefficients. Thus instead of a set of values for the Bb at each grid point (iDx1 , jDx2 , kDx3 ) one has eqn [8], where Za are the global expansion functions (with their requisite properties). Bb ðr; tÞ ¼ Me X a¼1 Bb;a ðtÞZa ðrÞ ½8 The choice of these functions is arbitrary, but some guidelines may optimize their selection. It would be ideal to select functions that fit the observation points of the expanded variables exactly, but the distribution of observations is not sufficiently uniform to make this feasible. The expansion functions might be chosen to fit statistics of observations interpolated to a more uniform grid such that the least number of functions (N) was required to describe most of the variance of the variables at those points. Additionally, functions could be chosen that fit boundary conditions most efficiently and/or with convenient orthogonalization properties. For application to the prediction system, eqn [8] is introduced into eqn [6]. To maintain the exact form of eqn [6], the series given by eqn [8] must be infinite. Using a truncated form creates the spectral model, and also generates errors analogous to those from reduction to a grid (eqn [7]). Selection of an optimum truncation is therefore a significant issue. The operator L, often used with the spectral method, is a diagonal matrix with Lb elements because the system is always linearly decoupled. The scalar spectral representation of eqn [6] is thus eqn [9] and the variables remain nonlinearly coupled in the ~b . functions F Lb qBb ~ ¼ Fb ðB; r; tÞ qt ½9 Substitution of eqn [8] into eqn [9] leads to the error equation [10]. Me  X qBb;a a¼1 qt Lb Za  ~ b ¼ eb F ½10 To solve this system for the unknown expansion coefficients Bb;a , multiply eqn [10] by suitable test ^ k ðrÞ and require the integral over the space functions Z domain to vanish, a least-squares error minimization procedure. The test functions must be continuous over 2102 SPECTRAL MODELS the domain, and can be arbitrary. In practice they are frequently chosen to be the expansion functions, but this is not required. With this approximation, the prediction equations for the expansion coefficients become eqn [11], yielding NMe equations for the unknown quantities, qBb:a =qt, which can be solved for Bb:a at future times using a suitable time extrapolation procedure. Z Me  X qBb;a a¼1 qt  Z ^ k dS  F ^ k dS ¼ 0 ~b Z Lb Za Z ½11 To cast eqn [11] in a form more comparable to the finite difference equations [7], let Bb ¼ ðBb;a Þ and Z ¼ ðZa Þ, both vectors with Me elements. Additionally, assume that the test functions can be similarly ^ ¼ ðZ ^ k Þ. Since the functions Fb are represented, i.e., Z implicitly functions of (r; t) (see eqn [5]), their projection onto the expansion functions yields eqn [12]. X Fb ¼ Fb;a ðtÞZa ¼ ZT Fb ½12 a Generating the coefficients Fb;a is nontrivial, resulting from nonlinear combinations of the expansion coefficients Bb;a , and efficient procedures will be discussed subsequently. Using the defined vectors, eqn [11] becomes eqn [13], representing Me equations for the expansion coefficients of each dependent variable. Z Z qBb T ^ ZT dS . Fb ^ . ¼ Z ½13 ZLb Z dS qt To combine the N equations of eqn [13] Rinto one ^ Lb ZT Z expression, define Me Me matrices Ab R the T ^ ZZ dS, and then create ðNMe Þ dS and A ðNMe Þ matrices AL ¼ diagðAb Þ and AR ¼ diagðAÞ. Extended vectors for the expansion coefficients to include all the variables can be constructed such that Bs ¼ ðBb Þ and Fs ¼ ðFb Þ, leading to an equation (eqn [14]) formally identical to the finite difference equation [7]. qBs ¼ A1 L AR F s qt ½14 The corresponding grid point values from this spectral representation may be calculated at each point (iDx1 , jDx2 , kDx3 ) for each dependent variable Bb by use of eqn [8]. Spectral Modeling Since most significant prediction models represent their dependent variables on a grid of points in the vertical and use nonspectral methods on that grid, the subsequent discussion of the spectral method will focus on the horizontal domain of the model representation. This requires that the variables Bb be represented on K surfaces in the vertical, with the surfaces separated by the grid intervals, and the variables described in those surfaces by eqn [8]. When selecting appropriate spectral functions for the expansion (8), in addition to fitting observations well, the functions should also be chosen with the properties of the system in mind. Several conditions have been accepted as suitable requirements. First, require the functions Za to satisfy the eigenvalue problem (eqn [15]). Lb Za ¼ cb;a Za ½15 In practice the selection of Lb almost always represents a conversion of wind components to vorticity and divergence, which is given by a linear differential operator. Application of this operator in eqn [15] leads to a variety of useful and simple functions. The second condition is to require the expansion functions to be orthogonal and normal over the domain in a Hermitian sense (eqn [16]). Z Zi Znj dS ¼ di;j ½16 This condition is reasonably simple to satisfy, since most function sets can be orthogonalized. Finally, the test functions when selected as the expansion functions do not lead to a significant loss of generality, thus this ^ ¼ Z. Utilizing condition is uniformly imposed as Z these three conditions greatly simplifies the calculations required to perform each prediction time step since both integrals in eqn [13] become diagonal matrices. A variety of functions have been used for the expansion [8], most satisfying the conditions just enumerated, with the selection depending on the degree of generality desired to approximate the general system [14]. When the atmosphere is represented on a channel with rigid boundaries at fixed northern and southern latitudes short of the poles, double Fourier series in latitude and longitude are found to be convenient expansion functions. They satisfy the boundary conditions easily and, because of the very simple addition rules for these functions, nonlinear products are rapidly calculated. For the full global domain approximated by spherical surfaces over the Earth, the obvious expansion functions that satisfy the boundary conditions are surface spherical harmonics (often denoted as solid harmonics), and they have become the functions of choice for spectral modeling. Surface spherical harmonics are constructed as the product of associated Legendre polynomials and complex exponential functions. Selecting coordinates in spherical surfaces such that m ¼ sin j, where SPECTRAL MODELS 2103 j is latitude, and l is longitude, normalized Legendre polynomials represent the latitudinal structures with the form of eqn [17]. Pm n ðmÞ  ð2n þ 1Þ ðn  mÞ! ðn þ mÞ! 1=2 ð1  m2 Þm=2 2n n!  d nþm 2 ðm  1Þn  dm  ½17 These are polynomials of degree n with n  m roots in the domain p=2ojop=2 and m roots at the poles. Together with Fourier series in longitude the solid harmonics are given by [18]. Yn;m ðl; mÞ ¼ Pn;m ðmÞe iml ½18 These are the complex expansion functions Za used in eqn [8] for the horizontal structures. All functions vanish at the poles except the zonal ones (m ¼ 0), and these remain finite there. The indices (n; m) define the roots of the functions and thus may be considered scaling elements; that is, the larger the indices, the smaller the scales represented by the functions. An example is given in Figure 1, which shows the cellular structure of the function for fixed n and various values of m. The total number of cells over the domain remains the same because some of the roots appear at the poles, but the cell structures differ. It is convenient to represent the indices as a single complex index, say a ¼ ðn þ imÞ. The functions are orthogonal over their respective domains and normalized; this is expressed (in a Hermitian sense) as eqn [19] with integration taken over the surface of the unit sphere. Z 1 Ya Yan0 dS ¼ da;a 0 ½19 4p The asterisk signifies complex conjugation, and d is the Kroneker delta. If Lb r2 (the Laplacian operator), substitution of Ya for Za in eqn [15] yields the eigenvalues [20]. ca ¼ nðn þ 1Þ ½20 Thus solid harmonics satisfy the conditions desired for suitable expansion functions. Most atmospheric variables (Bb ) are sufficiently smooth that, when expanded in these functions, the series converges rapidly. That expansion takes the form [21], where zk is any selected vertical level and the series truncates at Me . X Bb ðl; m; zk ; tÞ ¼ Bb;a;k ðtÞYa ðl; mÞ ½21 a The range of (a) is n  0 and, because of the complex nature of Fourier series, m takes both positive and negative values. When eqn [21] is introduced into m=0 m=1 m=2 m=3 m=4 m=5 Figure 1 Cellular structure of solid harmonic functions for n ¼ 5 and all allowed values of m. (From Baer (2000).) 2104 SPECTRAL MODELS eqn [14] and suitably integrated over the space domain, the resulting equations become a set of ordinary nonlinear differential equations in time for the expansion coefficients. Spectral Vorticity Model To better understand the details of this methodology, it is advantageous to simplify eqn [14] by approximations but still maintain a system that can describe the elements of the technique. The simplest appropriate system is represented by the barotropic vorticity equation. Consider a barotropic fluid, which exists if the thermodynamic variables are uniquely related to one another and are independent of position in the fluid. In this setting, fluid motions need consideration in only one horizontal surface and are independent of height. Assuming further that the fluid is incompressible, it is then also three-dimensionally nondivergent. If no divergence is introduced at the upper and lower boundaries, no divergence exists in any horizontal surface. Finally, under the condition of hydrostatic equilibrium, the vertical component of velocity can be ignored. The horizontal velocity is then represented by two scalar variables, which themselves may be transformed to any other two scalar functions; because rotation plays such a major role in atmospheric motions, vorticity and divergence are universally chosen. For the approximations stated, the divergence vanishes and hence the velocity is represented uniquely by the vorticity and the prediction equation for vorticity derived from eqn [1] is denoted the barotropic vorticity equation (BVE). Applying these approximations to eqn [1] and ignoring friction, the simplified equation of motion is eqn [22], where the subscript 2 denotes two-dimensionality. qV2 1 =2 p ¼ ðV2 . =ÞV2  2X 3 V2  rðpÞ qt ½22 The Earth’s vorticity is expressed here as 2X ¼ f k with f ¼ 2O sin j, the Coriolis parameter, and j is latitude. Transform the velocity to rotation and divergence by the definitions [23]. V2 ¼ k 3 =c þ =w = . V2 ¼ =2 w k . = 3 V 2 ¼ r2 c d z ½23a divergence ½23b relative vorticity ½23c The equation for predicting the vorticity (BVE) is established by applying the operator k . = 3 to eqn [22] and substituting [23], as in eqn [24]. qz ¼ V2 . =Z ¼ k 3 =c . =Z ¼ Jðc; ZÞ qt Z z þ f absolute vorticity ½24 This equation represents a very simplified atmosphere but contains prominent features of the full atmospheric system and is a useful tool for evaluating prediction techniques. Nondimensionalizing eqn [24] using the Earth’s radius (a) for space and its rotation rate (O) for time, and noting that the Coriolis parameter becomes f ¼ 2m, eqn [24] in terms of the stream function (c) is then eqn [25]. qr2 c qc ¼ 2  FðcÞ qt ql FðcÞ qc qr2 c qc qr2 c  ql qm qm ql ½25a ½25b Indeed, c ¼ B, the only variable remaining of the set Bn in eqn [21] and for only one k level. Equation [25] contains a linear term and two quadratic nonlinear terms; these latter terms constitute F, the remains of Fb in eqn [12]. A representation in terms of expansion coefficients ca ðtÞ is attained using eqn [15] for the Laplacian operator, eqn [21] for the expansion of c, and eqn [12] for expansion of F, yielding eqn [26]. X a ca Ya ðl; mÞ X qca ðtÞ ¼ 2i ma ca Ya ðl; mÞ qt a X þ Fa Ya ðl; mÞ a ½26 .As a final step, eqn [26] is multiplied by the test functions (in this case the conjugates of solid harmonics) and integrated over the unit sphere, noting orthogonality. This results in the prediction equation for each of the expansion coefficients (eqns [27]). qca ðtÞ ¼ 2ima c1 a ca ðtÞ þ Fa ðtÞ qt Fa ðtÞ ¼ Z FðcÞYan ðl; mÞdS ½27a ½27b It is evident how eqns [27] can be extended to involve more dependent variables and any number of levels in the vertical. However, if more variables exist in the system, these variables will be coupled nonlinearly through the coefficients Fa . Suppose that the series for a is truncated at Me as suggested. This implies that all values of ca for a > Me vanish. However, on calculating the nonlinear product SPECTRAL MODELS 2105 FðcÞ, a 2Me coefficients Fa are generated; thus at each time step the number of nonvanishing coefficients could double. This complication is resolved in the spectral method by always ignoring all computations for a > Me. The truncation of a at Me is somewhat intricate since, from eqn [17], n  0 and n  jmj, whereas mmax m mmax . The set of all allowed indices is best described by the intersections of integers in a grid on an n; m plane as depicted on Figure 2. The allowed points fall on an infinite triangle bounded by the lines n ¼
m, but it is sufficient to present only the triangle for m  0. All sequential values of n and m beginning at the origin are generally selected to satisfy convergence requirements for the dependent variables that they represent, but a relationship between maximum values must be chosen. Two options are preferred. The first, denoted as rhomboidal truncation, has a maximum value of mmax M (specified) and allows for all values of n jmj þ M for each jmj M. The corresponding figure (this configuration describes a parallelogram) is represented on Figure 2 and the notation is written as, for example, R30 if M ¼ 30. The advantage of this truncation is that each planetary wave m is represented by the same number of expansion coefficients, thereby allowing equal resolution for all waves. However, since the energy of atmospheric flow decreases rapidly with increasing wave number (m), resolution of the shorter waves may be less important than for the longer waves. This observation leads to triangular truncation, in which n N for each jmj M, with N  M, a predetermined integer. Usually N is selected equal to M and this option is described as a triangle on Figure 2 with the notation n 2M n = m + Mo n=N M = n + im n = m o  M N 2M m Figure 2 The domain and allowable range of indices m and n for triangular and rhomboidal truncations. (From Baer (2000).) T30 if N ¼ 30, for example. In terms of scaling, this truncation has some advantages. The ultimate choice for truncation is to optimize the resolution of the model in terms of the number of scales included and to minimize the computing requirements by selecting the fewest degrees of freedom compatible with resolution. Interaction Coefficient Method Since all prediction models are computationally intensive, the spectral method must compete in the efficient utilization of available computing resources. It is apparent from eqns [27] that most of the computing time required involves the calculation of the coefficients Fa and much effort has gone into optimizing this calculation. Early attempts followed the procedure of substituting the expansion series [21] for c into eqn [25] to represent FðcÞ and calculating Fa from eqns [27]. This results in eqns [28]. Fa ðtÞ ¼ Ia;b;g i XX c ðtÞcg ðtÞIa;b;g 2 b g b ðcb  cg Þ  Z  qYg qYb  mg Yg Yan dS  mb Yb qm qm ½28a ½28b The indices b and g go over the same range as a, which is determined by the selected truncation, and the integration is over the unit sphere. The integrals Ia;b;g are denoted as interaction coefficients and have exact solutions. Applying eqns [28] in eqns [27] shows that the time change of any expansion coefficient of the set a depends on the coupling of all the coefficients allowed in the spectral domain (refer to Figure 2) and each couple is weighted by its own interaction coefficient. Since each index consists of two real numbers, the set of interaction coefficients can be as large as the largest allowed index to the sixth power. In actuality, because of the simple addition rules for trigonometric functions, the integration over longitude reduces this by one order. The vector of these coefficients can be stored and need be computed only once. However, the number of multiplications that must be performed at each time step is daunting as the truncation limit becomes large. The more complex system [14] can be represented identically to [27] by simply increasing the number of expansion coefficients to include additional variables. But a shortcoming of using interaction coefficients concerns the convergence rate for the series of several dependent variables included in the general set (Bb) when expanded in global functions, in particular liquid water and precipitation. Significant truncation errors may ensue with time integration utilizing such functions. 2106 SPECTRAL MODELS Transform Method A technique denoted as the transform method is an alternate procedure for calculating the coefficients Fa , yielding the same (or better) results than the interaction coefficient method. This technique involves the transformation of the integrand in [27] onto a special numerical grid and solving the integral by quadrature. If the grid is selected appropriately, the integral is evaluated exactly and at a great reduction in computing cost. In the longitudinal direction, the quadrature is most conveniently done by a trapezoidal formula since it is known that eqn [29] holds. 1 2p Z 2p 0 eiml dl ¼ J 1 X imlj e J j¼1 ½29 The summation is taken over an equally spaced grid of points lj , and uses twice the number of points as the maximum wavenumber. Since the functions in latitude are Legendre polynomials, a Gaussian quadrature is preferred. In this case the quadrature is such that eqn [30] holds and is exact if the polynomial H is of degree 2K  1 or less. Z 1 1 HðmÞ dm ¼ K X k¼1 Gk ðm; KÞHðmk Þ ½30 The Gk are Gaussian weights and the grid points mk are the roots of the Legendre polynomial PK ðmÞ. The appropriate grid for this calculation contains all allowed values (lj ; mk ) as specified. The range of the grid points is determined by the functions of the integrand in eqns [27]. The derivatives in FðcÞ (see eqn [25]), must be taken before evaluating the function on the grid. Based on eqns [18] and [17], the differentiation with l is straightforward, but the m-derivative requires more information. The Legendre polynomials satisfy the differential equation [31], where the coefficients ba are constants, and this defines the latitudinal derivatives. ð1  m2 Þ1=2 dPa ¼ ba Pa1  baþ1 Paþ1 dm ½31 Following this procedure, FðcÞ is reduced to a quadratic series over the indices (b; g) in terms of the complex exponential functions in longitude and the associated Legendre polynomials in latitude, both of which can be evaluated on the specified grid. The actual calculation proceeds as follows. First the quadrature over longitude is taken (eqn [32]), where the sum goes over the value J ¼ 3M  1 if triangular truncation is chosen. Z 1 Fðcðl; m; tÞÞeima l dl Fma ðm; tÞ ¼ 2p ¼ J  1X  F cðlj ; m; tÞ eima lj J j¼1 ½32 Gk ðmk ; KÞFma ðmk ; tÞPa ðmk Þ ½33 The calculation is made over those latitudes m specified from the quadrature (eqn [33]). Z 1 Fma ðm; tÞPa ðmÞdm Fa ðtÞ ¼ 2 ¼ K X 1 Since the polynomial under summation in eqn [33] is HðmÞ and is the product of three Legendre polynomials less one order, and each has a maximum order of N, it can be shown that K ¼ ð3N  1Þ=2. Analysis of the computing requirements for eqns [32] and [33] indicates that the maximum number of calculations is proportional to N 3 , significantly less than the N 5 needed by the interaction coefficient method. When using the transform method with those variables that have unacceptable convergence properties yet contribute to eqn [7], their series representation is not essential. Their input is included directly into the quadrature formula by their distribution on the transform grid. Since all the forcing functions are summed over the grid before quadrature is completed, any singularities from individual terms are smoothed out and their effects are minimized. History Since the 1960s, spectral models have become by far the most popular representation for describing the global atmospheric prediction equations in computational form. They overcome many of the limitations inherent in finite difference models. Most international prediction centers have adopted this modeling procedure. Canada and Australia implemented the model in 1976, the National Meteorological Center of National Oceanic and Atmospheric Administration (NOAA) did so in 1980, the French in 1982 and the European Center for Medium-range Weather Forecasts (ECMWF) in 1983. As an example of how the models have evolved, production spectral models at ECMWF have grown in resolution from T63 in 1983 to T213 in 1998 with experiments currently running at T319. See also Boundary Layers: Modeling and Parameterization. Climate Prediction (Empirical and Numerical). Convec- STANDARD ATMOSPHERE 2107 tion: Laboratory Models of. Convective Cloud Systems Modelling. Coupled Ocean–Atmosphere Models. Mesoscale Meteorology: Models. Numerical Models: Methods. Predictability and Chaos. Weather Prediction: Adaptive Observations; Data Assimilation; Ensemble Prediction; Regional Prediction Models; Seasonal and Interannual Weather Prediction. Further Reading Baer F (2000) Numerical weather prediction. In: Zelkowitz MV (ed.) Advances in Computers. vol. 52, pp. 91–157. London: Academic Press. Boyd JP (2000) Chebyshev and Fourier Spectral Methods, 2nd edn. New York: Dover. Krishnamurti TN, Bedi HS and Hardiker VM (1998) An Introduction to Global Spectral Modeling. Oxford: Oxford University Press. Machenhauer B (1991) Spectral methods. In: Numerical Methods in Atmospheric Models Volume 1, pp. 3–86. (Reading, UK: European Center for Medium-range Weather Forecasts). Washington WM and Parkinson CL (1986) An Introduction to Three-dimensional Climate Modeling. Mill Valley, CA: University Science Books. STANDARD ATMOSPHERE W W Vaughan, University of Alabama in Huntsville, Huntsville, AL, USA Copyright 2003 Elsevier Science Ltd. All Rights Reserved. Introduction A ‘standard atmosphere’ is a vertical description of atmospheric temperature, pressure, and density that is usually established by international agreement and taken to be representative of the Earth’s atmosphere. The first ‘standard atmospheres’ established by international agreement were developed in the 1920s primarily for the purposes of pressure altimeter calibrations and aircraft performance calculations. Later, some countries, notably the United States, also developed and published ‘standard atmospheres’. The term ‘reference atmosphere’ is used to identify vertical descriptions of the atmosphere for specific geograp500 (A) (B) (C) (D) 90 80 400 60 1% Extremes 50 40 Geometric altitude (km) Geometric altitude (km) 70 300 30 200 20 10 0 120 140 160 180 200 220 240 260 280 300 320 Temperature (K) Figure 1 Range of systematic variability of temperature around the US Standard Atmosphere, 1976. (From Sissenwine et al. (1976).) 100  500 0 +500 Temperature difference (K) +1000 Figure 2 Departures of the temperature–altitude profiles from that of the US Standard Atmosphere, 1976, for various degrees of solar activity. (From Sissenwine et al. (1976).) Model (page no.) Geographic region Altitude range (km) Parameters CIRA, 1972 (1) Northern latitude Global 25 to 120, 110 to 2000 T ; p; d ; composition winds CIRA, 1986 (3) Global 130 to 2000 T ; p; d ; composition New Middle Atmosphere, 1985 (5) Global 801 S–801 N 20 to 80 T ; p; d , zonal ISO Reference Atmosphere, 1982 (7) Annual–151 N Seasonal–301, 451, 601, 801 N Cold/warm middle atmosphere  601, 801 N 0 to 80 T ; p; d ISO Standard Atmosphere, 1975 (9) 451 N  2 to 80 Monthly Mean Global Climatology, 1988 (11) Global 0 to 120 Species included Temporal variation Output data present Principal application N2 , O2 , O, A, He, H Seasonal, diurnal, solar activity, magnetic activity Tables, figures Aerospace vehicle design and evaluation, atmospheric reference Tables, figures, Seasonal, solar activity, geomagnetic computer code activity Aerospace vehicle design and evaluation, atmospheric reference – Monthly, interannual, tidal, planetary wave Tables, figures Aerospace vehicle design and evaluation, atmospheric reference Data on water vapor Seasonal, diurnal, daily, Tables, figures Aerospace vehicle and aircraft design and performance studies, atmospheric reference T ; p; d , composition, – sound speed, coll. freq. mfp, viscosity, spec. wt, scale ht, therm. cond. – Tables only Aerospace vehicle design and performance studies, atmospheric reference T ; p, zonal winds – Tables only Reference Climatology, numerical model initialization, instrumental design, scientific studies – 2108 STANDARD ATMOSPHERE Table 1 Summary of reference and standard atmospheres Global coverage 0 to 2500 T, p, d, wind velocity, wind shear, composition H2O, N2O, CH4 , N2 , O, Random perturbation, Computer code He, O3 , CO, CO2 , monthly NASA–MSFC O2 , A, H and COSMIC Aerospace vehicle design and simulation studies, space vehicle reentry, atmosphere reference for scientific studies US Standard, 1962 (16) Mid-latitudes (451)  5 to 700 T ; p; d , composition, part. speed, coll. freq., mfp, mean mol wt, viscosity, therm. cond., sound speed – – Tables, figures Aerospace vehicle design, atmospheric reference US Standard, 1966 Supplement (18) Mid-latitudes with variation  5 to 1000 Same as USS 1962 O2 , N2 , O, He, H Seasonal, diurnal, solar activity, magnetic activity Tables, figures Illustrate atmospheric variability US Standard, 1976 (19) Mid-latitudes (451)  5 to 1000 Same as USS 1962 Some data on N2 , O2 , H, He, O Diurnal, seasonal, solar cycle Tables, figures Aerospace vehicle design, atmospheric reference  5 to 1000 T ; p; d , composition, part. speed, coll. freq., mean mol. wt, viscosity, therm. cond., sound speed N2 , O2 , O Ar, He None Tables, figures Aerospace vehicle design studies, atmospheric reference 0 to 80 T ; p; d ; – Monthly, annual Tables, figures Design of aerospace vehicles, science applications 20 to 80 T ; p; d zonal winds – Monthly, latitudinal Tables, figures Aerospace vehicle design, atmospheric reference International Tropics Tropical Reference Atmosphere 1987 (21) Reference Atmosphere for Indian Equatorial zone, 1985 (23) Tropics Reference Model South 0–701 S Middle Atmosphere Southern Hemisphere 1987 (24) STANDARD ATMOSPHERE 2109 GRAM-95 (13) (Current Edition: GRAM-99) Model (page no.) Geographic region Altitude range (km) Parameters Species included Temporal variation Output data present Principal application AFGL (Phillips Laboratory) Atmospheric Constitution Profiles, 1986 (26) Global coverage 0 to 120 Number density, aerosol properties H2O, CO2 , N2O, O3 , CH4 , CO, O2 , N2 , 20 others, aerosols None Tables, figures, computer code Design and performance evaluation, scientific studies Extreme Envelope 601 S–901 N of Climate Elements 1973 (28) 0 to 80 Climatic elements: – T , p, humidity, wind shear, etc. Monthly Tables, figures Systems design Profiles of Temperature and Density, 1984 (30) Global except Antarctic 0 to 80 T;d – Monthly Tables, figures Systems design Global Reference Atmosphere, 1985 (32) Global 18 to 80 T ; p; d ; number density, scale ht. Wind velocity – Monthly Tables, figures Reference model for scientific studies 4120 km solar fluxEarth’s Upper dependent Atmosphere Density Model (Russia), 1984 (33) 0 to 1500 d – Solar flux, geomagnetic activity, daily and semi-annual effects Tables, computer code Aerospace vehicle design and orbital lifetimes Jacchia J70 (34) Mean global 90 to 2500 T ; p; d , scale ht N2 , O2 , O, Ar, He, H Diurnal, seasonal, geomagnetic activity Tables Design and simulation, lifetime analysis Jacchia J71 (35) Mean global 90 to 2500 T ; p; d , scale ht N2 , O2 , O, Ar, He, H Diurnal, seasonal, geomagnetic activity Tables, some computer code Design and simulation, lifetime analysis Jacchia J77 (36) Mean global 90 to 2500 T ; p; d , scale ht N2 , O2 , O, Ar, He, H Diurnal, seasonal, geomagnetic activity Tables, some computer code Design and simulation, lifetime analysis Model of Atmospheric Structure, 1987 (38) Global 70 to 130 T ; p; d – Monthly latitudinal, solar activity, magnetic activity Tables Connect Phillips Lab (AFGL) profiles of T ; p to MSIS-86 2110 STANDARD ATMOSPHERE Table 1 Continued Global coverage 85 to 2000 T ; p; d , composition N2 , O2 , O, He, H, Ar, N Diurnal, semiannual, latitudinal longitudinal solar activity, magnetic activity Computer code (NSSDC), floppy disk General scientific and engineering studies NASA Marshall Engineering Thermospheric Model, 1988 (41) (Current Edition: Version 2.0) Global 90 to 2500 T ; p; d , mean mol. wt, scale ht, spec. heat N2 , O2 , O Ar, He, H Solar activity, magnetic activity, seasonal, diurnal Computer code (NSSDC), floppy disk Orbital vehicle design and simulation, lifetime analysis Range Reference Models of the Atmosphere, 1982 (43) 0 to 70 Specific locations (e.g., Cape Canaveral, FL; Kwajalain, MI, etc.) T ; p; d ; wind velocity Water vapor Monthly, seasonal, means, monthly, parameter variations Tables, figures Site-related engineering analyses Reference Atmosphere for Edwards AFB, CA, 1975 (46) Edwards/Dryden, only ’Same as Reference Atmosphere for Patrick AFB- Hot and Cold Atmosphere for Edwards AFB, CA, 1975 (47) Edwards/Dryden only ’Same as Hot and Cold Atmosphere for Kennedy Space Center- Hot and cold Atmosphere for Kennedy Space Center, FL, 1971 (48) Kennedy Space Center only 0 to 90 T ; p; d – Seasonal Tables, figures Engineering studies Reference Atmosphere for Patrick AFB, FL, 1963 (49) Cape Kennedy only 0 to 700 T ; p; d ; composition, mean mol. wt, sound speed, viscosity, etc. – – Tables, figures Engineering studies Reference Atmosphere for Vandenberg AFB, CA, 1971 (50) Point Arguello only ’Same as Reference Atmosphere for Patrick AFB- STANDARD ATMOSPHERE 2111 NASA MSIS-86 (39) (Current Edition: NRL-MSIS-00) Model (page no.) Geographic region Hot and Cold Atmosphere for Vandenberg AFB, 1973 (51) Arguello only Mars-GRAM, 1996 (52) Global Global Venus International Reference Atmosphere (VIRA), 1985 (53) Altitude range (km) Parameters Species included Temporal variation Output data present Principal application ’Same as Hot and Cold Atmosphere for Kennedy Space Flight Center- 0 to B1000 T ; p; d , winds – Seasonal, diurnal, latitudinal longitudinal Tables, computer code Spacecraft design, atmospheric entry, orbital drag 0 to 3500 T ; p; d , composition o100 km CO2 , N2 , Ar, Ne, Kr, O2 , H2 , H2O, SO2 , D, NH3 o100 km latitudinal solar zenith angle, diurnal Tables, figures Spacecraft design, atmospheric entry, orbital drag 4100 km CO2 , O, CO, He, N, N2 , H, O2 , D, C 4100 km solar zenith angle, decimal, latitudinal, solar activity Source: AIAA Guide to Reference and Standard Atmosphere Models, Vaughan et al. (1996). T 5 kinetic temperature; p 5 pressure; d 5 mass density; mfp 5 mean free path; part. speed 5 particle speed; coll. freq. 5 collision frequency; mean mol. wt 5 mean molecular weight; therm. cond. 5 thermal conductivity; scale ht 5 scale height; spec. wt 5 specific weight; spec. heat 5 specific heat. CIRA: COSPAR (Committee on Space Research) International Reference Atmosphere; ISO: International Organisation for Standardization; GRAM: Global Reference Atmosphere Model; AFGL: Air Force Geophysics Laboratory; NASA: National Aeronautics and Space Agency; MSIS: Mass Spectrometer and Incoherent Scatter; NRL: Naval Research Laboratory. 2112 STANDARD ATMOSPHERE Table 1 Continued STANDARD ATMOSPHERE 2113 hical locations or globally. These were developed by organizations for specific applications, especially as the aerospace industry began to mature after World War II. The term ‘standard atmosphere’ has in recent years also been used by national and international organizations to describe vertical descriptions of atmospheric trace constituents, the ionosphere, aerosols, ozone, atomic oxygen, winds, water vapor, planetary atmospheres, and so on. A standard unit of atmospheric pressure is defined as that pressure exerted by a 760 millimeter, (or 29.22 inch) column of mercury at standard gravity at 45.54251 N latitude and sea level (9.80665 m s 2) at a temperature of 01C (321F). The recommended unit for meteorological use is 1013.25 hectopascals (1 hPa 5 1 mb). Standard temperature is used in physics to indicate a temperature of 01C (321F), the ice point, and a pressure of one standard atmosphere (1013.25 hPa). In meteorology, the term standard temperature has no generally accepted meaning, except that it may refer to the temperature at zero pressure-altitude in the standard atmosphere (151C) with a density of 1.2250 g m 3. The standard sea-level values of temperature, pressure, and density that have been used for decades are temperature of 288.15 K, 151C, or 591F; pressure of 1013.25 mb, 760 mm Hg, or 29.22 inches Hg; and density of 1225.00 g m  3 or 0.076474 lb ft 3. In 1925 the US National Advisory Committee for Aeronautics (NACA) Standard Atmosphere (or US Standard Atmosphere) was published. In 1952 the International Civil Aeronautical Organization (ICAO) produced the ICAO Standard Atmosphere, and in 1964 an extension to 32 km. Subsequently there have been a succession of ‘Standard and Reference Atmospheres’, some extending to altitudes above 1000 km, produced by the US Committee on Extension to the Standard Atmosphere (COESA), Committee on Space Research (COSPAR), Comitet Standartov (USSR), International Standardization Organization (ISO), US Air Force Research and Development Command (ARDC), US Range Commanders Council (RCC), and US National Aeronautics and Space Administration (NASA), plus others. In 1975 the International Standards Organization published a Standard Atmosphere for altitudes from  2 to 50 km that is identical to the ICAO Standard Atmosphere from  2 to 32 km. Subsequently the ISO published in 1982 a family of five Reference Atmospheres for Aerospace Use for altitudes up to 80 km and latitudes of 151, 301, 451, 601, and 801 N. Figure 1 provides an illustration of the temperature– height profiles to 100 km of the COESA US Standard Atmosphere, 1976, and the lowest and highest mean monthly temperatures obtained for any location between the Equator and Pole. The portion of the US Standard Atmosphere up to 32 km is identical with the ICAO Standard Atmosphere, 1964, and below 50 km with the ISO Standard Atmosphere, 1973. For altitudes above approximately 100 km, significant variations in the temperature, and thus density, occur due to solar and geomagnetic activity over the period of a solar cycle. Variations in the temperature– height profiles for various degrees of solar and geomagnetic activity are presented in Figure 2. Profile (A) gives the lowest temperature expected at solar cycle minimum; profile (B) represents average conditions at solar cycle minimum; (C) represents average conditions at a typical solar cycle maximum; and (D) gives the highest temperatures to be expected during a period of exceptionally high solar and geomagnetic activity. Currently some of the most commonly used Standard and Reference Atmospheres include: ICAO Standard Atmosphere, 1952/1964 ISO Standard Atmosphere, 1973 US Standard Atmosphere, 1976 COSPAR International Reference Atmosphere (CIRA), 1986 NASA Global Reference Atmosphere Model (GRAM), 1999 In 1996 the American Institute of Aeronautics and Astronautics (AIAA) published a Guide to Reference and Standard Atmosphere Models. This document provides information on the principal features for a number of global, regional, middle atmosphere, thermosphere, test range, and planetary atmosphere models. Summary information on these reference and standard atmosphere models is given in the Table 1. See also Evolution of Earth’s Atmosphere. Static Stability. Further Reading Champion KSW (1995) Early Years of Air Force Geophysics Research Contributions to Internationally Recognized Standard and Reference Atmospheres, Technical Report PL-TR-95-2164. Hanscom AFB, MA: Air Force Phillips Laboratory. Sissenwine N, Dubin M and Teweles S (COESA CoChairmen) (1976) US Standard Atmosphere, 1976, Stock No. 003-017-00323-0. Washington, DC: US Government Printing Office. Vaughan WW, Johnson DL, Justus CG, et al. (1996) Guide to Reference and Standard Atmosphere Models, Document ANSI/AIAA G-003A-1996. Reston, VA: American Institute of Aeronautics and Astronautics. 2114 STATIC STABILITY STATIC STABILITY J A Young, University of Wisconsin, Madison, WI, USA Copyright 2003 Elsevier Science Ltd. All Rights Reserved. monly equal to the restoring force per displacement, or B=dz in this case. Thus, if pressure effects are ignored in eqn [1], the simple buoyancy frequency N is given by N2 ¼ ðg=yv0 Þqyv0 =qz Introduction Static stability measures the gravitational resistance of an atmosphere to vertical displacements. It results from fundamental buoyant adjustments, and so it is determined by the vertical stratification of density or potential temperature. It influences the dynamics of many kinds of atmospheric motions, which in turn are responsible for determining its variations. Static stability is represented commonly by the square of the buoyancy frequency N, which plays a role in theories for flow instabilities, wave propagation, and forced motions. As summarized below, these theories apply to a wide range of spatial scales, from small-scale turbulence to convection, mesoscale motions, and large-scale circulations for which the ratio of N to the Coriolis frequency f is paramount. Basic Buoyant Stability and Instability The role of density fluctuations in a gravity field is best in the vertical component of the equations of motion. In an absolute sense, the gravity and pressure gradient forces are usually in a state of hydrostatic balance to within 1%. However, the slight imbalances account for vertical accelerations dw=dt which are often driven by buoyancy: 0 dw=dt ¼ r1 0 ½dp =dz þ B ½1 Here, w is the vertical velocity dz=dt, t is the time, r0 ðzÞ is the density of a static ‘environmental’ reference state, and a prime indicates deviation from that reference state. B is the buoyancy force per unit mass, given by B ¼ r 0 =r0 g. For many buoyant motions, B is an upper bound on vertical accelerations dw=dt since the pressure gradient term tends to oppose B. The most useful approximate form for B is 0 B ¼ ðyv =yv0 Þg ½2 where yv is the potential temperature augmented by a small (at most, a few 1C) amount proportional to water vapor, reflecting the contribution of humidity fluctuations to buoyancy. For a dry adiabatic vertical displacement dz, a parcel 0 conserves yv so that yv ¼ ðqyv0 =qzÞdz. For a stable system, the squared frequency of oscillation is com- ½3 N, also known as the Brunt–Vaisalla frequency, is determined by the vertical gradient of yv0 or equivalently by the difference between virtual temperature lapse rate qTv =qz and the dry adiabatic rate Gd ¼ g=cp. Unless conditions are superadiabatic, yv0 increases upward, corresponding to static stability. In this case, N 2 is positive and eqns [1]–[3] imply d2 w=dt2 þ N 2 w ¼ 0 ½4 It follows that the solution is a simple oscillation wðtÞ ¼ W cosðNt þ eÞ, where W is the maximum vertical velocity amplitude and e is a phase constant. The period is 2p=N, typically about 10 min in the troposphere. Figure 1A shows the vertical oscillation, and its driving by buoyancy, which is a quarter cycle ahead of the parcel displacement dz. The buoyancy oscillation is analogous to that of a spring, so N 2 is equivalent to the ‘stiffness’ of the atmosphere when it is subjected to vertical displacements. The stiffness increases with the closeness of y surfaces. Figure 1B shows that a larger stability produces a faster oscillation and inhibits the maximum vertical displacements W=N. For smaller values of static stability, the restoring buoyancy forces are weaker and the oscillations are slower. Neutral stability occurs when qyv0 =qz is zero (dry adiabatic conditions); a displaced parcel with no initial buoyancy remains that way, so there is no vertical acceleration. ‘Absolute instability’ occurs when qyv0 =qz is further reduced to a negative value (superadiabatic lapse rate). In this case N 2 ¼  N 2 is negative, and the solutions to eqn [4] are exponential in time (Figure 1C). The growing mode ðexpðjN jtÞÞ corresponds to a cooperative relation between buoyancy and motion (e.g., warm air rising) and may be thought of as the initial stage of convection. (A decaying mode ðexpðjN jtÞÞ corresponds to a mismatch of B and w (e.g., cold air rising) and so it is of no long-term consequence.) For convective motions, the increase of vertical kinetic energy is equal to the R buoyancy work B dz, known as the convective available potential energy (CAPE) along the parcel’s vertical path. (In vertically confined convective systems, a growing mode requires that thermal and viscous dissipation must be overcome, so a critical STATIC STABILITY 4 Z 3 2 (A) Time N 2 = _1 N=0 Z N=1 N=f (B) N=2 Time Figure 1 Simple buoyancy motions and varying environmental static stability. (A) Stable oscillation for N ¼ 1. Isentropic surfaces are shown; increasing labels indicate warmer y. Impulsive force creates initial vertical motion W (thin arrow), adiabatic displacements of y surfaces, changes in air parcel volume (circles), and buoyancy force (vertical arrows). (B) Parcel motions for five stability conditions. Moderate stability: N ¼ 1, shown in (A). Stronger stability: N ¼ 2 stable oscillation has shorter period, smaller vertical displacements. Extreme stability: N 5 infinity has no vertical displacement. Neutral stability: N ¼ 0 has displacements growing linearly, with no restoring force. Unstable conditions: N 2 ¼ 1 has buoyancy forces creating amplifying vertical parcel displacements. 2115 displacements depends most strongly upon qye =qz, with negative values corresponding typically to instability. (This criterion is used to describe ‘potential instability’, an often-misused concept that describes the stability of an unsaturated layer which is lifted hypothetically until it becomes a cloud layer.) The most important example of moist processes affecting stability occurs when rising, saturated parcels in cumulus clouds penetrate a dry ‘environmental’ layer. In this case, ‘conditional instability’ may occur even when qyv0 =qz is positive and the ‘dry dynamics’ of the environment are stable. This instability criterion may be expressed as qyes =qzo0, where yes is the saturation equivalent potential temperature, a known function of T and pressure p. This criterion is met if the virtual temperature lapse rate qTv =qz exceeds the smaller moist adiabatic rate Gm . The result is that the unstable combination of positive buoyancy with a rising parcel occurs if a saturated parcel moves upward through a layer of air where N 2 is insufficiently positive. Figure 2 illustrates the three fundamental types of stability for an atmosphere. The growth of cumulus clouds is overestimated by this simple parcel reasoning, because updrafts require compensating subsidence of the environment. The resulting adiabatic warming decreases the relative buoyancy of the cloud. A simple ‘slice’ theory shows that the effective stability of the system is then increased for finite-sized clouds; it can be represented as a combination of the moist and dry static stabilities. Additional stabilizing influences are turbulent mixing of momentum and thermodynamic quantities between the cloud and the environment, and the effects of pressure adjustments. value of N 2 must be exceeded, as expressed in a critical ‘Rayleigh number’ necessary for convection.) For many applications the distinction between y and yv is of secondary importance, as is assumed in the remaining discussion. CU AU S D M Moist Instability In a humid atmosphere, phase changes in the water content may cause instability even when N 2 is positive. In this case, a parcel conserves its equivalent potential temperature ye , rather than y. ye exceeds y by a temperature-dependent amount depending on humidity. As an example, conservation of ye is consistent with upward motions leading to saturation, the release of latent heat of condensation, and the diabatic increase of y. These ‘moist’ diabatic processes reduce the effective static stability for cloud systems. For example, the stability of a cloud layer to internal Z T Figure 2 Vertical temperature profiles (solid) for three categories of static stability. Temperature changes for dry and moist adiabatic parcel displacements are dashed. AU: absolutely unstable; CU: conditionally unstable (for saturated parcels); S 5 absolutely stable. 2116 STATIC STABILITY Climatology of Static Stability In the simplest terms, the dry and moist static stability indices depend upon vertical profiles of potential temperature, and to a lesser extent on the profile of water vapor. Figure 3 shows some typical features in a vertical cross-section. Strong static stability (N 2 ) regions are associated with isentropic surfaces that are closely spaced in the vertical, a symptom of the vertical ‘stiffness’. Weak stability regions have greater spacing, and the limit of zero stability may correspond to a vertical orientation of the isentropic surface. Regions of moist unstable motions are possible where there is a conditionally unstable temperature profile and sufficient moisture supply (e.g., the tropical boundary layer). The distribution of static stability qy=qz can be explained first by considering the processes that change the spacing Dz of potential temperature surfaces. From the first law of thermodynamics, it is easily shown that local changes of stability are caused by (1) advection of stability from upwind, (2) (vertically) differential temperature advection, and (3) differential diabatic heating. The differential diabatic term (3) explains many basic stability features in the atmosphere. The term is proportional to dJ=dz, where J is the diabatic heating rate per unit mass; negative J connotes cooling. This term increases the stability where J increases with height, and decreases it where J decreases with height. Examples of diabatic influence on static stability are seen in Figure 3. The strongest stability is seen in the stratosphere, where stability is maintained by the radiative heating increase due to absorption of solar ultraviolet radiation by ozone. The tropospheric static stability is several times smaller, due especially to downward long-wave radiation. Near the Earth’s surface, strong stability at high latitudes is created by long-wave radiative cooling, while weaker stability at other latitudes is driven by sensible heat from the surface. The sensible heating is concentrated in the atmospheric boundary layer, which often resembles a 100 404 400 396 392 388 384 380 376 372 368 364 360 356 352 348 344 340 336 332 328 324 320 316 312 308 304 300 296 150 200 ST 250 300 CU 292 400 288 284 500 280 700 850 925 1000 0; _150 Equator 276 FZ TB BL ML PS 272 268 264 260 256 90; _150 North pole Figure 3 Vertical cross-section of y from Equator to pole. Static stability is indicated by vertical closeness of y surfaces. Left scale is pressure in hPa. Dark shading: Earth’s topography. Light shading: boundary layer air with moisture mixing ratio exceeding 12 g kg  1. Strong stability cases: ST – stratosphere, PS – polar surface, BL – boundary layer top, FZ – frontal zone, TB – topographic blocking by mountains. Weak stability: CU – conditionally unstable tropical troposphere, ML – convectively mixed boundary layer. STATIC STABILITY ‘convective mixed layer’ of low stability, especially over land. In the tropics, the troposphere is moist at low levels, conditionally unstable, and deep; heavy cumulus convection is prevalent and its latent heating is the essential driving of the tropical climate system. Some of the smaller-scale features in the figure are affected by adiabatic circulation processes. Term 2 includes the effect of vertical wind shear in a baroclinic region; near the Earth’s surface, warm (cold) advection situations are associated commonly with stabilization (destabilization) of the lower atmosphere by this process. This term also explains the development of strong static stability by subsidence at the top of the atmospheric boundary layer and in frontal zones. There are seasonal and diurnal variations in stability that cannot be represented in the snapshot (Figure 3). These variations are caused by those of solar radiative forcing of the Earth’s surface, which results in variations of sensible and latent heating. Broadly speaking, the static stability fields tend to shift poleward in the summer season, and Equatorward in the winter season. The destabilization of the lower atmosphere is a maximum over land on summer days, while it is a maximum over the midlatitude oceans in winter. Static Stability and Circulation Dynamics Static stability influences the motions of the atmosphere on a range of scales and may permit waves to connect distant regions. Simple vertical buoyancy concepts are not sufficient for understanding these effects. In reality, one must also consider the coupling to horizontal winds and the ways in which pressure links the motion of different air parcels. The spatial distributions of static stability and wind determine the outcomes, which range from flow instability to various kinds of wave propagation in the horizontal and vertical. Small-Scale Turbulence Turbulence in the atmosphere may be caused by convection or by wind shear, and static stability is influential in each case. Ignoring moist dynamics, convection requires qyv =qz to be negative, which occurs most commonly when the air is in contact with a warmer Earth’s surface, such as a sunny day over dry land. In such cases, N 2 is strongly negative in the surface layer (roughly the lowest 50 m), reflecting a superadiabatic lapse rate of virtual temperature. Static stability is then near-neutral (N2 ¼ 0) in a deeper ‘mixed layer’ up to the boundary layer top. Thus, neutral boundary layers are symptoms of surfaceinduced convection. 2117 Positive static stability inhibits turbulence induced by wind shear. The production of shear turbulence may be understood by imagining a layer of concentrated wind shear which, when perturbed by vertical displacements, creates a pressure feedback that amplifies the displacements of the layer. The result is mixing of fast and slow air parcels by a growing pattern of Kelvin–Helmholtz instability (KHI) motions. Obviously, the vertical restoring forces of a statically stable atmosphere will oppose the vertical components of such KHI displacements. The competition between shear instability and stable stratification is best measured by the Richardson number Ri ¼ N 2 =SH 2 ½5 where SH is most generally the magnitude of the vector wind shear qV=qz. Ri is the squared ratio of the stable buoyancy oscillation frequency N to the maximum shear-induced growth rate SH. Theory and observation show that when Ri > 14, shear growth is eliminated: static stability wins, and perturbations are stable oscillations as in Figure 1. On the other hand, when static stability is reduced so that Rio14, the shear instability is not suppressed totally, and the perturbations may grow into turbulence. In the free atmosphere, intense frontal zones are associated commonly with ‘clear air turbulence’, despite the zones having a maximum static stability. This is because they are sloping regions of strong gradients, and Ri is reduced more effectively by the strong shear as the vertical width of the zone becomes small. The mixing by this turbulence is thought to modify the mesoscale structure of the static stability and shear near jets. Very near the Earth’s surface, strong shear is created by frictional drag, but the turbulence is limited by the surface and by static stability. In such surface boundary layers, the intensity of shear turbulence is greatest beneath the height L, the Monin–Obukhov length. L varies inversely with the stable air–surface temperature difference and static stability near the ground. Higher in the boundary layer, the turbulent fluxes are often represented by eddy mixing coefficients which are a decreasing function of Ri (and hence static stability). Mesoscale Motions Static stability and its spatial variations may produce complex mesoscale motions. Since wind speeds and the frequencies of weather systems are strongly subsonic, it follows that the pressure fields are in a state of ‘anelastic’ balance with the temperature and velocity patterns. The simplest balance involving buoyancy B is described by the three-dimensional p.d.e. 2118 STATIC STABILITY r2 p ¼ qB=qz, where r2 is the elliptic Laplacian operator in three spatial dimensions. The buoyancy gradient term ‘forces’ a smooth pressure response which decreases inversely with distance. For a vertically oriented pattern of B, the pressure response is negligible, and simple buoyancy forces dominate the motion. However, a pattern of B tilted toward the horizontal produces a pressure gradient force that opposes B. Thus, static stability may be associated with motions that may or may not be in hydrostatic balance, depending on the distribution of buoyancy in the vertical plane. The simplest tool for understanding these motions is the theory of buoyancy waves (see ‘Atmospheric Waves’). For patterns of motion and temperature with phase fronts tilted at an angle a from the vertical, the free oscillation has a frequency o ¼ N cos a. We see that N is actually an upper limit on the frequency, corresponding to the vertical orientation for a simple buoyancy oscillation. Such motions are nonhydrostatic. Much slower oscillations occur when the wave patterns are tilted toward the horizontal, a result of the ‘braking’ effect of the pressure field on the buoyant parcel. These motions are nearly hydrostatic, and the waves may propagate with a nondispersive phase speed obeying c2G ¼ N 2 =m2 ½6 cies predicts that (1) a wide mountain may cause upwind ‘blocking’ of low-level air with high static stability, and (2) motions over the mountain are nearly in hydrostatic balance. The theory for higher frequencies suggests that very narrow mountains do not disturb the flow far above the mountain, but an intermediate mountain width yields a complex pattern of vertically propagating wave patterns extending upward and downwind of the mountain. In order for energy to propagate upward, the wave fronts must tilt upwind with increasing altitude and the waves transport wind momentum down into the mountain. An example is shown in Figure 4. Static stability and wind variations influence the vertical fluxes of mesoscale wave energy and momentum which may link the upper atmosphere with the surface. For example, the vertical structure of the steady response with horizontal wavenumber k is governed by a propagation coefficient PðzÞ ¼ ½N 2 =U2  k2  The wave profile ‘propagates’ vertically only when P is positive, or when static stability makes the Scorer parameter N 2 =U2 sufficiently large. The vertical wavenumber is then P1=2 . Variations in stability or wind will cause PðzÞ to vary, which corresponds to 8 where m is the vertical wavenumber. Strong static stability corresponds to fast horizontal wave speeds. There are dramatic consequences of the simple frequency dispersion relation. For example, the energy of the waves is transmitted along the sloping wave front at a group speed ½7 where K is the two-dimensional wavenumber (inverse scale) of the wave pattern. We see that the energy propagation rate increases with static stability, and with angle a from the vertical. It follows that the response to a confined impulse will rapidly spread lowfrequency energy horizontally, while higher frequencies will be found immediately above and below the region. Imposed frequencies greater than N are ‘evanescent’: such energy cannot be propagated away from the forcing. Interestingly, the orthogonal relation between phase and group velocity vectors implies that downward phase propagation is associated with upward energy propagation. These properties have implications for a variety of mesoscale responses of a stable atmosphere to surface heating or mountains. For example, steady airflow U over a mountain complex may be envisioned in terms of periodic forcing. The above theory for low frequen- 6 Height (km) cg ¼ N sin a=K ½8 4 2 0 _ 60 _ 30 0 x (km) 30 60 Figure 4 Streamlines and y surfaces for flow over an isolated ridge. Upwind conditions have high static stability below 3 km, so PðzÞ decreases upward. Wind speeds vary along streamlines in proportion to closeness of streamlines. Proceeding from the left, note the slowing of air on the upwind side, strong downslope wind, vertically tilted flow pattern, downwind jump, and lee waves trapped in the stable layer. Shading denotes possible clouds due to lifting of moist layers. (Reproduced with permission from Houze (1993, Figure 12.9). Courtesy of Dale Durran (1986).) STATIC STABILITY wave refraction in the vertical plane. Two categories of phenomena result, depending upon whether PðzÞ decreases or increases with height. If stability decreases with height, then PðzÞ may become negative, and the wave may be reflected downward. Since the rigid Earth is also a reflecting surface for the wave vertical motion, the mountaininduced wave energy may become trapped in this layer. In this case, intense downslope winds and resonant ‘lee’ waves are possible. Other wave mechanisms, such as wave absorption at a critical layer where U ¼ 0, depend more strongly on the wind profile. In the other extreme, weak static stability in the boundary layer causes PðzÞ to increase with height above the surface. A common idealization is a mixed layer (N 2 ¼ 0) capped at height H by a sharp inversion of strength Dyv . In this case, horizontal scales larger than H are hydrostatic and move with speeds of ‘shallow water’ gravity waves obeying c2G ¼ ½g 0 H  ½9 We see that g 0 ¼ gðDyv =yv ), the ‘reduced gravity’ parameter for the inversion, plays an analogous role to static stability for these hydrostatic motions. An example of this kind of motion is the propagation of a gust front, the leading edge of thunderstorm outflow in the boundary layer. Another example is where this kind of air layer is forced to flow over a mountain at speed U; the inversion stability appears inversely in the Froude number F ¼ U2 =ðg 0 HÞ. This number represents a competition between the flow inertia and the inversion stability, or equivalently between advection by U and gravity wave propagation cG . Values exceeding O(1) may be associated with blocking on the upwind side of mountains, and strong downslope winds and hydraulic jumps on the downwind side. Large-Scale Circulations Large-scale circulations are those of large horizontal dimension, associated with low frequencies and hydrostatic balance. For such motions, static stability and the rotation of the Earth are important. Coriolis effects limit horizontal parcel motions in a fashion somewhat analogous to the buoyancy oscillation. The natural frequency of this ‘inertia oscillation’ is simply the Coriolis parameter f, which is about 100 times smaller than N. Thus, large-scale dynamics is ruled by the two fundamental frequencies of geophysical fluid dynamics: N and f . The most important large-scale flow variable is the combination known as the potential vorticity q ¼ ðf þ BÞN 2 ½10 2119 which is proportional to both the absolute vorticity of the winds and the static stability. Two frequency classes of large-scale waves are possible. The higher frequency class is inertio-gravity waves that obey o2 ¼ f 2 þ N 2 ðk2 =m2 Þ ½11 Static stability is seen to increase the minimum frequency f. These motions are never in a state of geostrophic balance, so they play an important role in the transient adjustments to thermal and mechanical forcing of the atmosphere. Vertical propagation of wave energy occurs only when frequency o exceeds f . For example, diurnal atmospheric tides propagate vertically only Equatorward of 301 latitude. Horizontal energy propagation is highly dispersive as a result of the Coriolis term: the largest scales propagate energy very slowly, while the smallest scales do so at the fast gravity wave speed cG . The separation between large and small horizontal scales occurs at l ¼ cG =f ¼ ðN=f Þm ½12 known as the Rossby deformation radius. The deformation radius is the natural horizontal scale for large-scale atmospheric dynamics. From eqn [12], it is the distance traveled by a gravity wave in the time (f 1 ) required for Coriolis forces to deflect the velocity. It represents the spatial scale for adjustment of wind and pressure to geostrophic balance. This scale of adjustment increases with the static stability parameter N, and it decreases with rotation f . The lowest-frequency class of large-scale dynamics is that of quasi-geostrophic (QG) dynamics for which ‘o  f ’ (see Quasi-geostrophic Theory). These motions are always near a state of geostrophic and hydrostatic balance, and are influenced strongly by static stability and the Earth’s rotation. The QG form of the potential vorticity corresponding to eqn [10] has a variable part proportional to  2 r2 p 0 þ f 2 q2 p 0 =qz2 qn ¼ N ½13 The response of p 0 to thermal or vorticity forcing is determined by eqn [13], which is a three-dimensional Laplacian in coordinates that are stretched vertically according to N=f . It follows that point forcing yields an elliptically shaped response, with the major axis lying in the direction of least resistance. For example, large static stability of the stratosphere yields responses that are stretched horizontally and compressed vertically. For a given vertical scale, this property implies a horizontal influence distance equal to the deformation radius. For a given horizontal scale L, it implies a vertical influence distance called the Rossby depth, given by HR ¼ ðf =NÞL, so that 2120 STATIC STABILITY increasing the stability decreases the vertical coupling distance HR . Similar considerations may be applied to the QG ‘omega equation’ to distinguish the total response to various patterns of thermal and vorticity forcing, illustrating the crucial importance of static stability on large-scale dynamics through the ratio N=f . There are obvious global implications, since f is small at low latitudes. Two major regimes of large-scale atmospheric circulation are the result. For example, QG instability theory indicates that baroclinic wave and cyclone growth are possible only at mid–high latitudes. Hence the tropics are less variable, except in concentrated areas of moist convection (such as tropical cyclones) where conditionally unstable air lowers the effective ratio N=f . Similar arguments account for the difference among the atmospheric circulations of other planets. often associated with static instability, dry convective motions, and sensible heating. Neutrally stable conditions are also very common, in which case turbulence transports latent energy away from the surface, enhancing the possibility of subsequent conditional instability. In summary, the three regimes of static stability account for much of the variety of weather and climate. Ultimately, the various kinds of circulations feed back on the static stability field itself, leading to increased complexity of its space–time variability. See also Buoyancy and Buoyancy Waves: Theory. Convective Storms: Overview. Dynamic Meteorology: Overview. Thermodynamics: Moist (Unsaturated) Air; Saturated Adiabatic Processes. Vorticity. Conclusions Static stability acts through gravitational buoyancy forces to suppress vertical motions, and helps to control the weather systems and climate of the Earth. In the Earth’s atmosphere, radiation and surface energy fluxes act to create three main categories of static stability. 1. Strong stability: The stratosphere is the most extensive example. Strong stability there encourages the vertical propagation of forced planetary waves through westerly winds regions, but it suppresses the growth of synoptic-scale circulations and convection. 2. Weak static stability: The troposphere is the atmosphere’s dominant region of lesser, more variable static stability. As a result, instabilities may produce weather systems on a range of scales. For example, baroclinic wave circulations create variable weather in middle and high latitudes, and conditional instability may be realized as moist convection. Moist convection may be organized on the global scale (e.g., Hadley and Walker circulations), the synoptic scale (e.g., tropical cyclones), or the mesoscale (deep cumulus convection and severe weather). The static stability for dry processes may be strong enough to allow mesoscale mountain influences on the upper atmospheric wind, or to suppress small-scale shear instability which would otherwise produce clear air turbulence. 3. Static instability: The energy balance of the Earth system requires that the Earth’s surface provides energy to the atmospheric boundary layer. This is Further Reading Andrews DG, Holton JR and Leovy CB (1987) Middle Atmosphere Dynamics. Orlando, FL: Academic Press. Chapman S and Lindzen RS (1970) Atmospheric Tides. Thermal and Gravitational. Dordrecht: Reidel. Durran DR (1986) Another look at dowslope windstorms, Part 1. Journal of Atmospheric Science 43: 2527–2543. Durran DR (1990) Mountain waves and downslope winds. In: Blumen W (ed.) Atmospheric Processes over Complex Terrain, pp. 59–82. Boston: American Meteorological Society. Emanuel KA (1994) Atmospheric Convection. New York: Oxford University Press. Gill AE (1982) Atmosphere–Ocean Dynamics. New York: Academic Press. Holton JR (1992) An Introduction to Dynamic Meteorology, 3rd edn. New York: Academic Press. Houze RA Jr (1993) Cloud Dynamics. San Diego: Academic Press. Irbane JV and Godson WL (1981) Atmospheric Thermodynamics, 2nd edn. Dordrecht: Reidel. Pedlosky J (1987) Geophysical Fluid Dynamics, 2nd edn. New York: Springer Verlag. Scorer RS (1978) Environmental Aerodynamics. Chichester, UK: Ellis Horwood. Sorbjan Z (1989) Structure of the Atmospheric Boundary Layer. Englewood Cliffs, NJ: Prentice-Hall. Stull RB (1988) An Introduction to Boundary Layer Meteorology. Boston: Kluwer. Tritton DJ (1996) Physical Fluid Dynamics, 2nd edn. New York: Oxford University Press. Turner JS (1973) Buoyancy Effects in Fluids. London: Cambridge University Press. Yih CS (1965) Dynamics of Non-Homogeneous Fluids. London: Macmillan. STATIONARY WAVES (OROGRAPHIC AND THERMALLY FORCED) 2121 STATIONARY WAVES (OROGRAPHIC AND THERMALLY FORCED) S Nigam, University of Maryland, College Park, MD, USA E DeWeaver, University of Wisconsin, Madison, WI, USA Copyright 2003 Elsevier Science Ltd. All Rights Reserved. Introduction The term stationary waves refers to the zonally asymmetric features of the time-averaged atmospheric circulation. They are also referred to as standing eddies, where standing refers to the time averaging over a month to season, and eddy is a generic term for zonally asymmetric patterns. The zonal asymmetries of the seasonal circulation are particularly interesting because they occur despite the longitudinally uniform incidence of solar radiation on our planet. Stationary waves must arise, ultimately, due to asymmetries at the Earth’s surface – mountains, continent–ocean contrasts, and sea surface temperature asymmetries. Understanding precisely how the stationary waves are generated and maintained is a fundamental problem in climate dynamics. Stationary waves have a strong effect on the climate through their persistent northerly and southerly surface winds, which blow cold and warm air. Advection of moisture by the stationary wave flow contributes to hydroclimate variations over the continents. Beyond their direct advective impact, stationary waves control the location of stormtracks – the preferred paths of synoptic weather systems in the midlatitudes, and the zone of tropical–extratropical interaction in the subtropics. Stationary waves are important also on longer time scales, since interannual climate variability projects substantially on the zonally asymmetric component of the flow. Finally, stationary waves contribute significantly to the maintenance of the complementary zonally symmetric circulation, in both climatological and anomalous states; the contribution is through quadratic fluxes of meridional momentum and heat. Stationary waves are thus a fundamental feature of the general circulation of the troposphere. Observed Structure Stationary waves are stronger in the Northern Hemisphere because of greater orography and continentality. Wave amplitudes in the Northern Hemisphere are largest during winter, modest during the transition seasons of spring and autumn, and weakest during summer. The Southern Hemisphere stationary waves and their seasonal variation are substantially smaller in comparison. Northern Hemisphere Winter Structure Because of the geostrophic balance condition, stationary waves in the upper-level flow can be conveniently displayed using the height of the 300 hPa pressure surface. The geostrophic wind blows along the height contours, with lower heights to the left in the Northern Hemisphere, and with a speed proportional to the gradient of the height field. The height of the 300 hPa surface varies considerably with latitude and longitude (Figure 1A), with the mean height being close to 9 km. The polar vortex is clearly recognizable in this projection. The vortex is due to insolation and planetary rotation, both zonally symmetric inputs, but the vortex has notable departures from symmetry: troughs over northern Canada and western Siberia, and ridges over the eastern Atlantic and Pacific. (The zonally asymmetric component of the field which highlights the troughs and ridges is shown later, in Figure 2A.) The regions where the height contours are close together correspond to strong westerly (coming from the west) jets: the Asian–Pacific and North American jets. Stationary waves at the Earth’s surface can be identified using the sea-level pressure field, which in elevated areas is the surface pressure reduced to sea level. The lightly shaded regions in Figure 1B are surface lows, and the dark regions are highs. Lows are found over both ocean basins, the Aleutian Low in the Pacific and the Icelandic Low in the Atlantic. The Aleutian Low is centered off the tip of the Aleutian Islands chain, and the counterclockwise flow around the low brings southerly marine air to coastal Canada and Alaska, lessening the severity of the winter season. To the south of the Icelandic Low is a high-pressure center known as the Azores High. Strong onshore surface flow occurs between the Icelandic Low and the Azores High, again lessening the severity of coastal winters in Europe. Much higher surface pressure can be found over central Asia in a center called the Siberian High. Between the Siberian High and the Aleutian Low is a region of strong northerly flow, which brings down colder air and lowers near-surface temperature along the east coast of Asia. The winter 2122 STATIONARY WAVES (OROGRAPHIC AND THERMALLY FORCED) 9 8.5 9 9.5 (A) H L L H (B) Figure 1 (A) Average height of the 300 hPa pressure surface in northern winter months (December, January, February, and March: DJFM). The average is over 20 winter seasons (December 1979 through March 1999), and is computed from the reanalysis fields produced by the US National Center for Environmental Prediction (NCEP). The contour interval is 100 m. (B) Average sealevel pressure (SLP) for the same months and years, with a contour interval of 5.0 hPa. Sea-level pressure data come from Trenberth’s analysis, which is archived at NCAR. Dark (light) shading represents values above 1015 hPa (below 1010 hPa). The letters ‘L’ and ‘H’ designate the prominent centers of action: the Aleutian Low, Siberian High, Icelandic Low, and the Azores High. Map domain begins at 201 N. sea-level pressure field can be broadly characterized as being high over the continents and low over the comparatively warmer northern oceans. Since sea- level pressure is related to column temperature, vertical coherence of the continent–ocean temperature contrast in the lower troposphere is key, as discussed later. The stationary wave pattern changes considerably between the surface and 300 hPa, and these changes are highlighted in Figure 2. The top panels show eddy heights at the 300 and 850 hPa levels, revealing the troughs and ridges. These features are displaced westward with increasing height, i.e., westward tilted, assuming that the same features are being tracked at the two levels. The low-level trough over the Pacific is positioned 15–201 westward of the Aleutian Low, and gives way to a trough centered on the east Asian coast at 300 hPa, which is associated with the Asian–Pacific jet. The 850 hPa trough over the North Atlantic is likewise shifted relative to the Icelandic Low, and migrates further westward towards Hudson Bay at upper levels; it brings cold Arctic air into the central and eastern United States and Canada. The low-level feature over Eurasia (Figure 2B) is more definitely linked to the surface Siberian High, but there is no corresponding feature of significance present at the upper level – in contrast with the vertically coherent structure of the Azores High. The vertical structure of stationary waves is plotted in Figures 2C and D, which are cross-sections of the eddy height field at 401 N and 601 N. The shading in these panels depicts the eddy temperature field. (In hydrostatic balance, this is the vertical derivative of the height field in log(p) coordinates.) These plots allow for the tracking of features. The northern section shows the pronounced westward tilt of the east Asian trough, the Rocky Mountain ridge, and the Azores High. The tilt is a consequence of meridional temperature advection by the associated geostrophic wind, which induces cooling to the west (east) of the low (high). Interestingly, connection with the prominent surface features is not strong, except in case of the highs. The Aleutian and Icelandic Lows, in particular, are quite shallow (p 0 800 hPa), exhibiting little connectivity to the westward displaced upper-level troughs. The southern section (401 N) nicely reveals the limited vertical extent of the Siberian High, in contrast with the deep structure of the Azores High. Comparison of the two cross-sections indicates a striking difference in vertical variation of the eddy heights, particularly in the Eastern Hemisphere. In the northern section, the wave amplitude keeps growing with height up until the tropopause, and even beyond. The structure is indicative of upward propagation of stationary wave energy into the polar lower stratosphere. (Note that the wave’s phase is stationary, so the phase velocity is zero, but its group velocity – the STATIONARY WAVES (OROGRAPHIC AND THERMALLY FORCED) 2123 (A) (B) 0 60° N 0 200 400 0 0 600 800 1000 (C) SH AL IL 40° N 0 200 0 400 600 800 1000 (D) 0° 0 60° E 120° E 180° 120° W 60° W AH 0° Figure 2 (A) Eddy height at 300 hPa during northern winter (DJFM), or height of the 300 hPa pressure surface after subtracting the zonal average. The contour interval is 50 m, with dark (light) shading for positive (negative) values in excess of 50 m. (B) Eddy height at 850 hPa for the same period, with a contour interval of 25 m and dark (light) shading for positive (negative) values in excess of 25 m. The thick contours in B enclose regions where the surface pressure is less than 850 hPa. In these regions, the pressure surface is interpolated below ground. (C and D) 1000–100 hPa zonal–vertical cross-sections of eddy height and temperature at (C) 601 N and (D) 401 N. Contour interval for eddy height is 50 m, and dashed contours represent negative values. Eddy temperature is plotted in 3 K contours with dark (light) shading for positive (negative) values in excess of 3 K, and zero contours suppressed. ‘SH’, ‘AL’, ‘IL’, and ‘AH’ are the surface lows and highs of Figure 1B. velocity of energy propagation – is not zero.) In contrast, the 401 N structure is indicative of trapping of wave energy within the troposphere. The eddy height at the 10 hPa level, displayed using shaded contours in Figure 3, reveals the presence of a large-amplitude stationary wave at an altitude of nearly 30 km. The zonal wavelength of this pattern is evidently close to the circumference of the latitude circle, i.e., the largest possible. Both observations and theory (see Rossby Waves) suggest that disturbances of such large wavelengths can propagate into the stratosphere. The wave pattern in Figure 3 moves the center of the polar vortex away from the geographical pole and reduces the strength of the vortex. Equatorial westerly duct An important circulation feature in the deep tropics during northern winter is the presence of strong upper-level westerlies (B10 m s  1) over the Pacific and Atlantic longitudes. This is notable because the equatorial belt is otherwise occupied by easterly winds. Zonal winds at 200 hPa are shown in Figure 4A, with the easterly region shaded. A vertical section at the equator (Figure 4B) shows westerly zones to be confined to the near-tropopause region (100– 300 hPa), with maximum values (B15 m s  1) at 200 hPa. The origin of equatorial westerly zones is not well understood, but their absence in northern summers and El Niño winters suggests that their occurrence is linked to the absence of 2124 STATIONARY WAVES (OROGRAPHIC AND THERMALLY FORCED) dynamical barrier, shielding the equatorial zone from the influence of midlatitude perturbations. Openings in this barrier, or westerly ducts, thus provide a conduit for equatorward penetration of midlatitude waves during northern winter – a timely opening, since the midlatitude stationary and transient wave activity is most vigorous in winter. Interaction between midlatitudes and the equatorial zone can impact convection and water vapor distribution in the tropics and subtropics. Lateral mixing from extratropical intrusions can also influence tracer transports. 31 30 Northern Hemisphere Summer Structure Figure 3 Average height of the 10 hPa pressure surface in northern winter (DJFM). Thick contours show the height field in 500 m increments. The eddy height is plotted at 100 m intervals, with dark (light) shading for positive (negative) values in excess of 100 m. The zero contour for eddy height is suppressed. strong convection in the central equatorial Pacific and Atlantic longitudes. Rossby wave propagation theory (see Rossby Waves) suggests that tropical easterlies are an effective The northern polar vortex is much weaker in summer than in winter. The summer vortex is shown at the 150 hPa level in Figure 5A, and is evidently quite symmetric. It also lacks the tight meridional gradients that characterized the winter vortex. A somewhat higher level was chosen for displaying the summer pattern in order to capture fully the divergent monsoonal flow and accompanying rotational circulations over the warmer landmasses. The upper-level asymmetries include the very prominent anticyclone over Tibet, and troughs over the subtropical ocean basins which are easier to appreciate in the eddy height plots, shown later. 45° N 10 0° −10 0 10 45° S (A) 0 5 200 400 600 800 1000 0° (B) −5 90° E 180° 90° W 0° Figure 4 (A) The 200 hPa zonal wind in northern winter (DJFM), contoured in 10 m s  1 intervals. Regions where the zonal wind blows from the east are shaded. (B) Zonal–vertical cross-section of zonal wind at the Equator, contoured in 5 m s  1 increments, with shading for easterly regions. Top level in (B) is 50 hPA. STATIONARY WAVES (OROGRAPHIC AND THERMALLY FORCED) 2125 L 14.3 14 H 13.7 H 14 (A) (B) TA 30° N 200 0 400 600 800 1000 (C) 0° 0 0 0 0 60° E 120° E 180° PH 120° W 60° W BH 0° Figure 5 (A) Average height of the 150 hPa pressure surface in northern summer months (June, July, and August: JJA). The average is over 20 summer seasons (June 1980 through August 1999), and is computed from NCEP reanalysis. The contour interval is 100 m. (B) Average sea-level pressure for the same months, plotted as in Figure 1. The letters ‘L’ and ‘H’ designate the prominent centers of action: the Bermuda High, the Pacific High, and the broad region of low pressure associated with the Asian monsoon. Map domain begins at 151 N. (C) The 1000–100 hPa zonal–vertical cross-section of eddy height and temperature at 301 N, with contours and shading as in Figure 2D. ‘TA’ gives the location of the Tibetan anticyclone, which is enclosed by the 14.3 km contour near the top of panel (A). The summertime sea-level pressure (Figure 5B) has almost a reversed winter structure. Two subtropical anticyclones of comparable strength are present in the ocean basins, underneath the upper-level troughs. They are referred to as the Pacific High and the Bermuda High. The Bermuda High is the summer equivalent of the Azores High, which expands while the Icelandic Low retreats northward during the transition from winter to summer. The Pacific sector undergoes a similar winter to summer transition. The subtropical anticyclones constitute the descending branch of the regional Hadley cells which are driven by deep convection in the tropics. Descending motions induced to the northwest of subtropical monsoonal heating may also contribute to anticyclone development. Over the continents, sea-level pressure is low during summer. A large region of low sea-level pressure is present over Asia beneath the Tibetan anticyclone (which is actually centered over northern India). The continental-scale anticyclone is an integral element of the Asian monsoon circulation, being the rotational response to deep heating. The cross-section of eddy height at 301 N (Figure 5C) shows the internal baroclinic structure that is typically produced by deep heating in the tropics. The Tibetan anticyclone reaches maximum amplitude at 150 hPa, the level displayed in Figure 5A. Over the oceans, the structure is also baroclinic, but the Pacific and Bermuda Highs are evidently shallow features – although not as shallow as their winter counterparts in Figure 2C. The strong positive temperature centered over the Tibetan plateau is caused by latent heat release in the monsoon rains. On the other hand, negative temperatures over the Pacific and Bermuda Highs are produced, in part, from the long-wave radiative cooling to space. The eddy height fields during summer are displayed in Figure 6. The Tibetan anticyclone is the prominent feature at upper levels. Baroclinic structure is evident in the Northern Hemisphere, with upper-level troughs positioned over the subtropical highs. Also evident at the upper level is a weak ridge over North America that is associated with the local monsoon system, which includes the Mexican monsoon. The western edge of the Bermuda High produces low-level southerly flow, which brings in significant amounts of moisture from the Gulf of Mexico into the US Great Plains. A notable low-level feature in the Southern Hemisphere is the Mascarene High centered south of Madagascar, which generates strong easterlies along its northern flank (recall that the flow around a 2126 STATIONARY WAVES (OROGRAPHIC AND THERMALLY FORCED) 90° N 45° N 0° 45° S (A) 90° N 45° N 0° 45° S (B) 90° N 45° N 0° 45° S (C) 0° 90° E 180° 90° W 0° Figure 6 Eddy height at (A) 150 hPa and (B) 850 hPa during northern summer (JJA), with a contour interval of 25 m and dark (light) shading for positive (negative) values in excess of 25 m. (C) Eddy wind vectors at 850 hPa. Regions where the eddy wind speed is in excess of 5 m s  1 are shaded, and the longest arrow represents a wind speed of 18 m s  1. Eddy winds with speeds below 2 m s  1 are suppressed. As in (B), the thick closed contours in (C) surround mountainous regions where the surface pressure is less than 850 hPa. Southern Hemisphere High is counterclockwise). After turning northward along the African coast and crossing the Equator, this flow evolves into the southwesterly monsoon flow over the Arabian Sea. In the Asian monsoon circulation, equatorial and cross-equatorial flows play an important role, and these cannot be appreciated in the height field, since the geostrophic relationship breaks down at the Equator. The summer circulation figures are thus complemented with a vector-wind plot at 850 hPa (Figure 6C); only the zonally asymmetric components of winds are plotted. Strong cross-equatorial flow occurs along the east coast of Africa, bringing moisture to the Asian continent. Easterly flow is found all along the Equator, particularly along the southern flank of the Pacific and Bermuda Highs. Southern Hemisphere Stationary Waves The Southern Hemisphere has much less land than the Northern Hemisphere, resulting in weaker asymmetries at its lower boundary. A more zonally symmetric circulation, with smaller-amplitude stationary waves, is thus expected. Due to the larger fraction of ocean, the seasonal cycle will also be muted. The seasonal change in surface temperature, for example, will be smaller than in the Northern Hemisphere. The southern vortex is shown during the December– March (southern summer) and June–August (southern winter) periods in Figure 7. As before, the winter vortex is shown at 300 hPa and the summer one at the higher 150 hPa level. Thick lines mark the height STATIONARY WAVES (OROGRAPHIC AND THERMALLY FORCED) 2127 14 9 13 8.2 (A) (B) (C) (D) Figure 7 (A) Height of the 150 hPa pressure surface in the Southern Hemisphere during DJFM months (southern summer). (B) Height of the 300 hPa surface during JJA months (southern winter). Thick solid contours show the total height field in 200 m increments, while thin contours represent the eddy height. The contour interval for eddy height is 25 m, with dark (light) shading for positive (negative) values in excess of 25 m. The zero contour for eddy height is suppressed. (C, D) Sea-level pressure for (C) DJFM and (D) JJA months, in 2.5 hPa increments, with dark (light) shading for values above 1015 hPa (below 1012.5 hPa). Map domain is from the Equator to the South Pole. Sea-level pressure values over Antarctica are unreliable and hence suppressed. contours while the shaded region shows the corresponding eddy height patterns. Note that in the Southern Hemisphere the flow around a low is clockwise rather than counterclockwise. The southern vortex is considerably more symmetric than the northern one. Eddy heights are thus smaller, and contoured at 25 m in both summer and winter (Figure 7). The summer and winter patterns are both dominated by the wave number 1 component in the high latitudes so that opposite points along a latitude circle have opposite polarities. The wave component exhibits similar phase and amplitude structure in the two seasons, indicating a significant role of Antarctic orography in its forcing. The subtropics shows greater seasonality, with a ridge over northern Australia in summer; this upper-level feature is linked to the Australian monsoon outflow. The extent of zonal asymmetries at the surface is examined using sea-level pressure which is contoured with a 2.5 hPa interval as opposed to 5.0 hPa in the Northern Hemisphere. The summer distribution (Figure 7C) is much like the one in the Northern Hemisphere (Figure 5B), with high-pressure cells occupying the midlatitude ocean basins. In summer, the subtropical highs are interrupted by continental heat lows, caused by the warmer land temperatures. The winter sea-level pressure (Figure 7D) is more zonally symmetric, unlike the Northern Hemisphere where asymmetries are most pronounced during winter (cf. Figures 1 and 2). A prominent feature of the southern winter pattern is the Mascarene High extending from Africa to Australia, which generates strong south-easterly flow along its northern flank. Its linkage with south-westerly flow over the northern Indian Ocean and Asian summer monsoon can be seen in Figure 6C. The vertical structure of Southern Hemisphere stationary waves is shown in Figure 8 at 301 S in summer and 601 S in winter – the latitude of the subtropical highs and the polar wavenumber 1 pattern, respectively. Contour intervals in Figure 8A are 10 m for height and 1.5 K for temperature, as opposed to 50 m and 3.0 K in northern summer (Figure 5C). As in the Northern Hemisphere, the subtropical highs have a baroclinic structure with upper-level troughs superimposed on surface highs. The heat lows over 2128 STATIONARY WAVES (OROGRAPHIC AND THERMALLY FORCED) DJFM 30° S 200 0 0 0 400 0 600 800 1000 (A) 0 JJA 60° S 200 0 400 0 600 800 1000 (B) 0° 60° E 120° E 180° 120° W 60° W 0° Figure 8 Zonal–vertical cross-sections of eddy height and temperature in the Southern Hemisphere at (A) 301 S in DJFM and (B) 601 S in JJA months. In (A), the contour interval for eddy height is 10 m, with dashed contours for negative values. The contour interval for eddy temperature in (A) is 1.5 K, with dark (light) shading for positive (negative) values in excess of 1.5 K, and zero contours suppressed. In (B), contour intervals for eddy height and temperature are 25 m and 1.5 K, respectively, with plotting conventions as in panel (A). Australia and southern Africa are quite shallow and intense: this is typical of arid regions where rainfall and mid-tropospheric latent heating do not occur in response to the surface heat low. The winter height and temperature structures (Figure 8B) are plotted using 25 m and 1.5 K intervals, as opposed to 50 m and 3.0 K in northern winter (Figures 2C, D), due to their relative weakness. The southern winter pattern evidently changes little with height. There is much less westward tilt in comparison with the northern winter structure (Figure 2C), indicating less upward propagation of wave energy. Although westerlies are necessary for upward propagation, theoretical considerations suggest that propagation is hindered by the presence of excessive westerlies (westerlies exceeding the Rossby critical velocity), the southern winter vortex is substantially stronger than its northern counterpart (cf. Figures 1A and 7B; note the larger contouring interval in the latter figure). Transience in the Atmosphere The above review of stationary wave structure does not convey the extent to which these waves are representative of the instantaneous circulation. For example, how stationary (or transient) is the upperlevel circulation during northern winter? Can the stationary waves be ‘seen’ on synoptic weather charts? The degree to which these charts depart from the climatological pattern is a measure of the strength of the transient flow. Transient activity is estimated in northern winter in Figure 9 because it is expected to be strongest in this season. The greater vertical shear of the thermally balanced Asian–Pacific and Atlantic jets in winter makes them prone to hydrodynamic instability, which in the context of geostrophic flows is called baroclinic instability. Baroclinic instability produces transient disturbances on subweekly time scales. The extent to which the stationary wave structure is representative of instantaneous flow is depicted in Figure 9 by projecting the daily, instantaneous (00UTC), 300 hPa circulation on the climatological wave pattern (Figure 2A) during the winters of 1980/ 81 and 1989/90. Correlation – a measure of the structural similarity of the two maps (without regard to amplitude) – is plotted on the y-axis. The correlation ranges from 0.2 to 0.8 in these winters, indicating that the climatological pattern accounts for up to 65% of the spatial variance. High correlation is however achieved only on a few days in each winter. More typically, the correlation is between 0.5 and 0.6. Interestingly, the correlation drops and recovers over a 2–3 week period, 1–2 times each winter, revealing the establishment time scale of the climatological pattern. Dynamical analysis of such episodes, especially of the recovery phase, can shed light on the establishment mechanisms of stationary waves. The question of whether the climatological wave pattern can be ‘seen’ on synoptic charts is addressed in Figures 9B and C, which show the instantaneous STATIONARY WAVES (OROGRAPHIC AND THERMALLY FORCED) 2129 1980/81 1989/90 0.8 0.6 0.4 0.2 1 Dec (A) 16 Dec 1 Jan 16 Jan (B) 1 Feb 16 Feb 1 Mar 16 Mar (C) Figure 9 (A) Spatial correlation between the instantaneous (00UTC) and climatological 300 hPa eddy heights during 1980/81 (solid curve) and 1989/90 (dashed curve) winters. Correlations are area weighted and include height data from 201 N to the north pole. (B) 300 hPa eddy height on 1 February 1981, when the spatial correlation was 0.74. (C) Eddy height on 16 February 1990, when the correlation was 0.29. In (B) and (C), the contour interval is 100 m (twice the interval in Figure 2A), with dark (light) shading for positive (negative) values in excess of 100 m. (00UTC) wave pattern on two days: 1 February 1981, when the spatial correlation is high (0.74; Figure 9B), and 16 February 1990, when the correlation is low (0.29; Figure 9C). The climatological pattern (Figure 2A) can be clearly recognized in the former plot, but not in the latter. Even when structurally similar, the patterns can evidently have very different wave amplitudes; the contour interval is 50 m in Figure 2A but 100 m in Figure 9B. Forcing of Stationary Waves Stationary waves are generated, ultimately, by the zonal asymmetries at the Earth’s surface: orography, continent–ocean contrasts, and sea surface temperature gradients. Through dynamic and thermodynamic interactions with the zonal-mean flow, and subsequent mutual interactions, surface inhomogeneities produce zonally asymmetric circulation and precipitation features at upper levels. Comprehensive numerical models of the atmosphere, which include coupling between physical and dynamical processes, are able to realistically model the observed stationary waves. In a sense, the often posed question – on relative contribution of orography and other processes in forcing of stationary waves – has been addressed by such prognostic general circulation models (GCMs). Comparison of GCM simulations obtained with and without orography provide insight. In these assessments, the change in the heating distribution is attributed to orographic forcing, whose circulation impact is found to be comparable to that of all other processes put together. Historically, answers to the above question were sought in a framework where ‘orographic forcing’ was used more restrictively – to refer to the dynamical forcing of flow from mechanical diversion. In such analysis, the entire heating distribution, regardless of its origin (e.g., from condensation in adiabatically cooled upslope flow), was regarded as an independent forcing. This framework was adopted, perhaps, because mechanical diversion of flow by an orographic barrier is conceptually easier to model. Such studies lead to the rapid advancement of stationary wave theory, including construction of potential vorticity conserving models for the responses to orography, meridional and vertical wave propagation analysis, and understanding of troposphere–stratosphere interaction. Theoretical Considerations Large-scale atmospheric motions in the extratropics are approximately hydrostatic and quasi-geostrophic (QG) in character. The hydrostatic approximation recognizes the operative balance between the 2130 STATIONARY WAVES (OROGRAPHIC AND THERMALLY FORCED) horizontally varying pressure and density perturbations, while the QG approximation acknowledges the near-balance between the Coriolis force and the horizontal pressure gradient. QG flow is thus dominated by the rotational component. Its evolution, however, is determined, in part, by the comparatively weaker divergent flow component, as described by the vorticity equation qz þ Vh =z þ bv ¼ ð f þ zÞð= Vh Þ ½1 qt ^ ð=V Þ ¼ Here, z is the QG relative vorticity ð¼ k h 2 = cÞ, c is the geostrophic streamfunction, f is the Coriolis parameter with qf =qy ¼ b, and Vh is the horizontal QG flow. The right-hand term is the product of absolute vorticity ðf þ zÞ and horizontal convergence, and is often called the ‘stretching term’ because convergent flow leads to stretching of vortex tubes. Due to compressibility of air, evolution of the thermodynamic state is conveniently described using potential temperature, y ¼ Tðp0 =pÞR=Cp , which is conserved in adiabatic motion; p0 is the reference pressure (1000 hPa). Potential temperature . thus changes only in response to diabatic heating Q, as follows: . qy qy Q þ Vh =y þ w ¼ ðy0 =T0 Þ ½2 qt qz Cp where y0 ¼ T0 ðp0 =pÞR=Cp , with T0 ¼ Tðp0 Þ. Since stationary waves refer to the zonally varying component of the flow (here onward denoted by prime), their dynamics can be described, to first order, by linearizing eqns [1] and [2] about the zonal-mean  ðy; zÞ and  circulation, U yðy; zÞ. The linearized equations are valid for small-amplitude perturbations: 0 0 0  yy Þ  f ð= V Þ  z þ v 0 ðb  U zt þ U x h ½3 . 0  y0 þ v0 yt þ U yy þ w 0  yz  ðQ 0 y0 =T0 cp Þ x ½4 For convenience, subscripts are used to denote the partial derivatives. Orographic Forcing and Response The forcing and propagation of stationary waves can be discussed using eqns [3] and [4]. In contrast with diabatic heating, which is explicitly present as righthand forcing in the thermodynamic equation, the mechanical forcing by orography ðh 0 Þ is implicitly present through its kinematic impact on vertical velocity at the lower boundary ðws Þ. In the presence of the zonal-mean circulation, the linearized vertical 0  h0 . velocity, ws , equals U x In simplified treatments of orographic interaction, the geophysical fluid is additionally considered to be homogeneous and incompressible ð= Vh ¼ qw=qzÞ, so that response is determinable using the vorticity equation alone – this ‘shallow water’ approximation is indeed reasonable for the interaction of oceanic flows with underwater topography, but somewhat limited in capturing aspects of the atmospheric interaction. In shallow water theory, density (or temperature) is constant, and the horizontal flow, including horizontal divergence, is height independent. Assuming that a rigid lid is placed at the top of the fluid ðz ¼ HÞ, so that vertical velocity vanishes there,  h 0 Þ=H. The forced waves one obtains qw 0 =qz  ðU x  has been are then modeled by eqn [5], where U additionally assumed to be latitude independent, and perturbation vorticity is dissipated (e.g., by Ekman spin-down) on an e1 time scale:  q=qx þ eðc 0 þ c 0 Þ ½q=qt þ U yy xx 0  h0 þ bcx  ð f =HÞU x ½5 To understand the forced response, consider an arbitrary Fourier component of the geostrophic ^ eiðkxþlyorÞ g, streamfunction: c 0 ðx; y; tÞ ¼ Realfc k; l where the hat denotes the complex amplitude corresponding to zonal and meridional wavenumbers, k and l, and associated frequency o. For such a perturbation, eqn [5] yields the solution ^¼ c ^ fh   cÞ=U   b=U   ieðk2 þ l2 Þ=ðU  kÞ H½ðk2 þ l2 ÞðU ½6 For stationary waves ðo ¼ 0Þ, the zonal phase speed, cð¼ o=kÞ, vanishes. This simplifies the first term in the denominator to ðk2 þ l2 Þ. In the presence of dissipation, the orography and streamfunction are not in phase, since the denominator in eqn [6] is complex. The possibility of resonance is also indicated in the  . Dissipainviscid case ðe ¼ 0Þ, when ðk2 þ l2 Þ ¼ b=U tion however limits the wave amplitude at resonance, ^. The streamfunction and orography are ^ / ih with c 901 phase-shifted (or in quadrature) in this case, with the trough in the flow being a quarter wavelength downstream of the mountain ridge. When forcing is on  , planetary vorticity adlarger scales, ðk2 þ l2 Þob=U vection dominates zonal advection of relative vorticity in balancing the orographically induced vorticity on upslopes and downslopes. Both the real and imaginary parts of the denominator in eqn [6] are negative in this case, which puts the trough within a quarter wavelength downstream of the ridge. STATIONARY WAVES (OROGRAPHIC AND THERMALLY FORCED) 2131 It is interesting that troughs in the observed 300 hPa stationary wave pattern (cf. Figure 2A) are also downstream of the orographic features, but the extent to which these are forced by local orography remains somewhat uncertain, as discussed later. Also, the assumed Fourier representation of the streamfunction implies the presence of meridional boundaries which confine wave energy to a midlatitude channel – a set-up conducive for resonance. The sinusoidal zonal structure is also unrealistic, since orographic features are generally localized. In nature, the wave energy propagates away zonally, meridionally, and vertically from the localized forcing region, thus calling into question the validity of the solution [6]. The pedagogically useful shallow-water model of orographic interaction is limited for other reasons as well. The tropospheric flow cannot be assumed to be homogeneous and incompressible, since cooling (heating) from adiabatic expansion (compression) during ascent (descent) is important in the thermodynamic budget. Moreover, a flow configuration which satisfies the vorticity eqn [3] can be unbalanced from the viewpoint of this budget. For example, when planetary vorticity advection dominates the lefthand side in balancing the upslope divergent flow in eqn [3], the thermodynamic budget [4] is unbalanced in the absence of condensation heating (precipitation), since both adiabatic ascent and equatorward flow lead to colder temperatures. The generation of orographic response must thus be understood by considering the vorticity and thermodynamic equations together, so that any implications that one may have for the other are fully accounted for. Such considerations lead to the development of the QG potential vorticity equation. QG Potential Vorticity Equation The prediction equation for QG flow that does not explicitly reference the divergent flow component is called the QG potential vorticity equation. Although it can be derived quite generally, the focus here is on its simplified linearized  is version, when the zonal-mean flow U independent of latitude. The equation is derived by eliminating the divergent flow from eqns [3] and [4]. In the zn ¼ ðRT0 =gÞ lnðp0 =pÞ coordinate, which reduces to geometric height in an isothermal atmosphere, the QG potential vorticity equation is q 0t  q0 þ v0q y ¼ þU x  .  R q r0 Q 0 Hcp r0 qzn N 2 ½7 where 0 q ðx; y; zÞ ¼ c 0xx þ c 0yy   f2 q qc 0 þ 2 r0 n N r0 qzn qz and   f2 q qU y ðzÞ ¼ b  2 q r0 n qz N r0 qzn ½8 Since divergent flow is not referenced by this equation, it is of some interest to examine the manifestation of orographic forcing in this analysis framework. Not surprisingly, this forcing enters as a lower boundary condition, but in the thermodynamic equation [4]. This is because of the direct reference to vertical velocity in eqn [4], in contrast with the vorticity equation [3] which refers only to its vertical gradient. 0  h 0 , the boundary condition conveying With ws ¼ U x orographic forcing is 0  y 0 þ v 0 yy yt þ U x .  h 0 yz þ ðQ 0 y0 =T0 cp Þ at the surface ¼ U x ½9 Assume for purposes of this discussion that diabatic heating vanishes at the surface, so that only adiabatic cooling (warming) is occurring on the upslope (downslope). In steady flows, the heating can be balanced by zonal eddy advection and/or meridional advection of mean temperature. If upslope cooling is compensated by the latter, the upslope flow will be poleward, and a high-pressure center will be positioned over the mountain ridge near the surface. The response at upper levels depends upon the zonal scale of mountains: large wavelengths will propagate into the lower stratosphere, and phase lines will tilt westward with increasing height, all as depicted in the Figure 10A schematic. It is interesting that although thermal advection and vertical wave propagation are absent in the shallow water model, the horizontal structure of the long-wavelength solution (in the presence of damping, eqn [6]) is not too different from that indicated at upper levels in Figure 10A. Heating Response The stationary wave response to heating can be qualitatively understood from the thermodynamic equation [4]. In the deep tropics, horizontal variations of geopotential (and temperature) are much smaller since it is difficult to maintain them in the presence of the weak Coriolis force. Consequently, horizontal temperature advection is ineffective in balancing diabatic heating in eqn [4]. Away from the surface, heating is thus balanced, almost entirely, by adiabatic cooling, with the vertical 2132 STATIONARY WAVES (OROGRAPHIC AND THERMALLY FORCED) C H W L (A) L (B) In the midlatitudes, heating does not extend as deeply into the troposphere as in the tropics. Heating in the Pacific and Atlantic stormtracks, for example, is confined mostly to the lower troposphere, as shown later in Figure 11B. Midlatitude heating is offset to a large extent by horizontal temperature advection; larger temperature gradients are sustainable in midlatitudes due to the greater Coriolis force. Large-scale heating in midlatitudes is balanced, mostly, by cold advection from the north; the near-surface low is thus positioned eastward of the heating. Interestingly, vertical motion in the vicinity of midlatitude heating is determined by vorticity balance considerations – a complete reversal of the tropical situation: cold advection from the north brings with it higher vorticity air as well, and this induced vorticity advection must be offset if a steady state is to be maintained. The compensation is accomplished by vortex compression, which has implications for the temperature field. Tropics Wave Propagation L (C) Mid-lat Figure 10 Schematic depiction of the longitude–height response forced by (A) westerly flow over midlatitude orography, (B) tropical heating, and (C) midlatitude heating, all taken from Hoskins and Karoly (1981). The orographic response is shown for the longwavelength case, and is determined from both dynamic and thermodynamic (i.e., quasi-geostrophic potential vorticity) considerations. The arrows depict vertical motion, and circled crosses and dots denote poleward and equatorward flow, respectively. H and L denote the pressure ridge and trough, with the lines showing the vertical tilt of the pressure wave. W and C are the warmest and coldest air, respectively. profile of w0 closely following that of heating. A substantial portion of heating in the tropics results from deep convection, which produces strongest heating in the mid-to-upper troposphere, as shown later in Figure 11C. Such heating distribution leads to convergence (divergence) in the lower (upper) troposphere, which results in vortex stretching (squashing). The rotational response to the induced vorticity depends on the horizontal forcing scale: if the scale is large, the stretching is offset by poleward advection of planetary vorticity, which is tantamount to the surface low being positioned westward of the heat source, as schematically illustrated in Figure 10B. The qualitative arguments discussed above are helpful in understanding the nature of response in the forcing region. The stationary wave response is however not confined to the forcing region alone, since Rossby waves propagate zonally, meridionally, and vertically, carrying the disturbance (energy) into the far field (unforced region). The energy propagation, or group velocity, characteristics depend both on the perturbation scale and structure of the basic state. Some zonal-mean zonal wind configurations encourage Rossby wave propagation, while others impede it. Basic state flow can thus profoundly impact wave propagation into the tropics and the stratosphere. Theoretical analysis helps to focus on the basic state attributes that are influential, e.g., the direction and curvature of the zonal-mean zonal wind. A useful quantity in wave propagation analysis is the refractive index which seizes on these and other relevant attributes. A display of refractive index variations is often helpful, since it conveys, to first order, the wave propagation pathways, as waves are generally refracted towards higher refractive index regions. Such analysis suggests that midlatitude stationary waves are refracted towards the Equator, drawn there by the large index values resulting from diminishing westerly winds. The tropical easterlies, in contrast, present an effective dynamical barrier to equatorward propagation of midlatitude stationary waves. In the vertical, waves with large horizontal scales alone can propagate upward, but only when the upper-level westerlies are not too strong. STATIONARY WAVES (OROGRAPHIC AND THERMALLY FORCED) 2133 60° N b 30° N c 0° 30° S 60° S (A) 37.5° N 200 400 600 800 1000 (B) 5° N 200 400 600 800 1000 (C) 0° 60° E 120° E 180° 120° W 60° W 0° Figure 11 (A) Mass-weighted vertical average of diabatic heating, calculated as a residual from the thermodynamic equation. The winter season (DJFM) diagnosis is obtained from NCEP reanalysis fields for 20 winter seasons (1979/80–1998/99). The contour interval is 0.5 K day  1, with dark (light) shading for positive (negative) values in excess of 0.5 K day  1, and zero contours suppressed. (B, C) Zonal– vertical (1000–100 hPa) cross-section of diabatic heating at (B) 37.51 N and (C) 51 N, with contours and shading as in panel (A). The latitudes of the cross-sections in (B) and (C) are marked with thick lines at the edges of panel (A). A 9-point smoother is applied to the heating field before plotting. Diabatic Heating in Northern Winter Diabatic heating plays a prominent role in the forcing of stationary waves. In stationary wave theory, it is an explicit forcing in the QG potential vorticity equation [7], and even orographic forcing in this theoretical framework manifests as surface heating [9]. In nature, heating resulting from the change of phase of water substance, turbulent eddy diffusion, and short-wave and long-wave radiative fluxes is referred to as diabatic heating. (Note that the temperature of air parcels can change even without any diabatic heating, from adiabatic compression or expansion.) In contrast with Earth’s orography, whose highly accurate measurements are widely known, the three-dimensional structure of diabatic heating is only beginning to be described. The main reason why the heating distribution has remained uncertain is that, unlike other quantities, heating is not directly measured. It is, instead, estimated, usually as a residual in the thermodynamic budget. Since the heating estimate is only as good as the quality of atmospheric data from which it is diagnosed, the quality of atmospheric analysis is critical for the diagnosis. Fortunately, data coverage and quality and analysis methods have all improved in the last two decades, and are reflected in the modern reanalysis data sets. Heating diagnosis from one such data set, the US National Centers for Environmental Prediction (NCEP) reanalysis, is shown in Figure 11. The mass-weighted vertical average of diabatic heating, 1 ðps  100Þ Z ps 100 . Qc1 p dp is shown during northern winter in units of K day  1; here, ps is the surface pressure, cp is the specific heat of air at constant pressure, and the integration is from the surface to 100 hPa. Key features in Figure 10A include the heating centers in the extratropical Pacific and Atlantic basins, which effectively define the two 2134 STATIONARY WAVES (OROGRAPHIC AND THERMALLY FORCED) midlatitude stormtracks. The northern continents, in contrast, constitute the cooling regions. In the tropics, heating is strong over the South Pacific convergence zone and the Amazon basin. A narrow zone of heating is also present in the Pacific just northward of the Equator; this intertropical convergence zone (ITCZ) is much stronger during northern summer when it is positioned a few degrees farther northward and fully extended across the Pacific basin. Diabatic heating has a complicated vertical structure which changes with latitude and season. The changes with latitude are shown in Figures 10B and C, which depict height–longitude cross-sections through the midlatitude stormtracks (37.51 N) and the ITCZ (51 N). The stormtrack heating is evidently strongest near the surface, with peak values close to 6 K day  1, and diminishes rapidly with height. Latent heating due to precipitation in baroclinically unstable synopticscale disturbances is the primary contributor to stormtrack heating. Diabatic cooling, on the other hand, is comparatively weaker, and focused more near the tropopause in this estimation, for reasons that are not clear. The vertical structure of ITCZ heating is strikingly different. Although the entire column is being heated, heating is generally strongest in the mid-to-upper troposphere. For example, over the tropical Pacific warm pool – the site of persistent deep convection – heating is strongest (B3 K day  1) at 400 hPa. In contrast, heating over land (e.g., equatorial Africa) is strongest near the surface due to sensible heating. The heating structure over Central America is also similar, except that elevated surface heating there has produced some deep convection as well. Interaction with Transients The climatological stationary waves coexist with vigorous atmospheric motions occurring on a variety of time scales (Figure 9), and there are strong interactions between these transient motions and the stationary waves. Transient motions are the instantaneous departures of the flow from its climatological state, and the time mean of transient motion thus vanishes, by definition. However, fluxes of heat and vorticity by transients do not vanish in general. For example, the contribution of transients to the advection terms in eqn [2] can be written as 00 Vh =y 0 0 þ w 0 0 qy 0 0 =qz 00 00 00 00  = ðVh y Þ þ qðo y Þ=qp right-hand side of eqn [10] is the heat-flux divergence from transient motions. In synoptic systems, northward (southward) transient motions are typically accompanied by positive (negative) temperature fluctuations, so that heat flux diverges to the south of a stormtrack and converges to the north. The heat-flux divergence acts as heat sources and sinks for the timemean flow, and the stationary waves respond to this thermal forcing just as they respond to diabatic heating. Likewise, the convergence of transient vor00 ticity flux ð= ðVh z 0 0 Þ  qðo 0 0 z 0 0 Þ=qpÞ provides sources and sinks of vorticity for the stationary waves. The net effect of transient thermal and vorticity fluxes on stationary waves is not easy to characterize. However, it is clear that transient forcing is strong enough in northern winter to exert a powerful influence on stationary waves. The 700 hPa heat-flux convergence by perturbations lasting less than 1 month is superimposed in Figure 12A on the local winter eddy temperature pattern. The two fields evidently oppose each other. For example, transient heat fluxes diverge from the warmer regions over the Atlantic and the west coast of North America and converge in the colder regions above north-eastern Canada. Thus, throughout most of the northern midlatitudes, transient thermal fluxes have a damping effect on the lower tropospheric stationary eddy temperature pattern. Furthermore, the forcing by transient thermal fluxes is on the order of 0.5 K day  1, , while the eddy temperatures are about 4 K. In the absence of other processes, it would take little more than a week for the thermal fluxes to reduce dramatically the 700 hPa eddy temperature field. Such a reduction also implies a substantial weakening of the upper-level geopotential pattern, since temperature is the vertical derivative of geopotential in hydrostatic balance. While transients are an important influence on stationary waves, the stationary waves can be equally influential for the transients. One way in which stationary waves organize transients is by creating localized regions of strong cyclogenesis. Synoptic systems tend to develop in regions of strong lower tropospheric temperature gradients, and such regions are present off the coasts of Asia and North America in Figure 12A. The ability of the local temperature gradient to enhance the growth of synoptic systems can be measured by the Eady growth parameter ðeÞ: e ¼ 0:31 ½10 where the double prime denotes the transient component and o is the vertical velocity in p-coordinates. The j=Tj jTðT0 q ln y= g qzÞj1=2 Its plot in Figure 12B shows large values in the same regions where stormtrack heating occurs in Figure 11A. Comparison of these panels suggests that STATIONARY WAVES (OROGRAPHIC AND THERMALLY FORCED) 2135 (A) steering winds that determine storm paths and by exchanging mechanical energy with the storms. The Eady growth rate is applicable to synoptic transients, which grow by extracting energy from the thermal gradients (or vertical shear) of the climatological state. Synoptic systems, in fact, account for less than half of the transient forcing in the vorticity and thermodynamic equations for the climatological stationary waves. Furthermore, the slower transients (those with time scales between, say, 10 days and 1 month) are quite distinct from synoptic transients. They do not generally travel along concentrated stormtracks, nor do they typically grow by extracting energy from the climatological temperature gradients. Although these transients are certainly influenced by the climatological stationary waves, the nature of this influence is rather complex and cannot be easily summarized. Modeling of Northern Winter Stationary Waves (B) Figure 12 (A) The 700 hPa eddy temperature (thick contours) and heat-flux convergence by transient motions (shading and white contours) in northern winter months (DJFM). The contour interval for temperature is 2 K, and dashed lines represent negative values. The contour interval for heat-flux convergence is 0.5 K day  1, with dark (light) shading for positive (negative) values in excess of 0.5 K day  1. Zero contours for temperature and heat-flux convergence are suppressed, and regions where the surface pressure is less than 700 hPa are masked out. (B) Eady growth parameter in northern winter (DJFM), calculated from the 700 hPa temperature gradients. The contour interval is 0.1 day  1, with dark shading for values in excess of 0.6 day  1. stationary waves play an important role in determining the locations of stormtracks. In addition to this effect on growth rate, stationary waves can also have a mechanical effect on stormtracks, by changing the Modeling of orographically forced stationary waves dates back to the seminal paper of Charney and Eliassen in 1949, in which linear shallow water theory was applied to the longitudinal distribution of orography at 451 N. The earlier discussion here, including the development of eqn [5] and its solution [6], largely follows the analysis reported in that paper. Charney and Eliassen found the midlatitude mountains to be rather influential, accounting for almost all of the observed signal in their analysis. Since that time, diabatic heating due to continent– ocean contrasts, and transient fluxes of heat and momentum have also been advocated as important mechanisms for the generation of stationary waves. In the intervening period, the atmosphere has been more closely observed, both spatially and temporally, and there has been a tremendous increase in computational power for modeling studies. A reassessment of the relative roles of various forcing mechanisms is thus in order. In effect, more complete versions of the dynamical and thermodynamical equations can now be solved globally at high resolution, and verified against the extensive record of upper-air observations that have been compiled since. It is still advantageous of linearize the system of equations, at least initially, since this allows the influence of individual forcing terms to be examined separately. Of course, the forcing terms can have strong mutual interactions. For example, heating can cause eddy flow which then impinges on mountains, and the subsequent orographically induced uplift can generate convection and lead to further heating. 2136 STATIONARY WAVES (OROGRAPHIC AND THERMALLY FORCED) However, linear diagnostic models serve a valuable purpose in assessing the relative importance of the different forcing terms in different regions. They also indicate the degree to which linear perturbation theory applies to stationary waves. A linear simulation of northern winter stationary waves is presented in Figure 13. The linear model uses s ¼ p=ps as the vertical coordinate (here ps is the surface pressure) so that mountains do not intersect the model levels. There are 15 levels in the vertical, ranging from 1000 to 25 hPa, and the horizontal resolution is 7.51 in the zonal direction and about 2.51 in the meridional direction. As in eqns [3] and [4], the model is linearized about the zonal-mean climatological state. The momentum and thermal dissipation in the model is roughly equivalent to a 5-day damping time scale in eqns [3] and [4]. The forcing consists of three-dimensional diabatic heating (Figure 11), orographic height, and transient fluxes of heat and momentum. The 300 hPa response obtained with all forcings is shown in Figure 13A, and plotted using the convention used in Figure 2A, with which it should be compared. The linear model can simulate the ridge over the Atlantic, the low off the east Asian coast, and the trough over Canada. A notable flaw in the simulation is the weakness of the ridge over the Rockies. The solution shows that linear perturbation theory can explain many, though not all, aspects of the observed stationary wave pattern. The model response when forced separately by diabatic heating, mountains, and transients is shown in panels (B)–(D), respectively; note that contour interval in these three panels is half that in panel (A). All three forcings contribute significantly to the total pattern. The heating response includes jets in the Asian–Pacific and Atlantic sectors. Heating is evidently important in establishing the ridge over the Atlantic and northern Europe, and contributes significantly to the trough over Canada as well. The response to mountains in panel (C) shows troughs downstream of the Himalayan–Tibetan complex and the Rockies, as suggested by the shallow water solution [6] and also QG potential vorticity considerations (cf. Figure 10A), in each case for long wavelengths. Thus, orography contributes to the (A) (B) (C) (D) Figure 13 (A) The 300-hPa height response of a linear stationary wave model forced by heating, mountains, and transient fluxes of heat and momentum. The contour interval is 50 m, with dark (light) shading for positive (negative) values in excess of 50 m. (B–D) Response of the model when forced separately by (B) heating, (C) mountains, and (D) transient fluxes. The contour interval in (B–D) is 25 m. STRATOSPHERE–TROPOSPHERE EXCHANGE / Global Aspects 2137 forcing of the jets as well. The high amplitudes directly over Greenland and Tibet are a consequence of the linearization of the hydrostatic equation in s-coordinates. Examination of geopotential heights gives a somewhat misleading impression that waves generated by mountains propagate primarily in the zonal direction. Examination of the modeled streamfunction (a more suitable variable for describing the rotational response in the tropics; not shown), however, reveals considerable equatorward propagation of the forced waves. The forcing by submonthly transients (panel D) produces a somewhat intricate pattern with no clear relationship to the synoptic stormtracks. Transients are apparently responsible for a large part of the response over eastern Atlantic and northern Europe. Studies of stationary wave dynamics have traditionally focused on the question of the relative importance of the various forcing terms in generating the observed pattern. Yet recent simulations such as the one in Figure 13 show clearly that the northern winter stationary waves do not constitute a simple linear response to a single form of forcing. Furthermore, linearized equations, such as eqns [3] and [4], neglect the advection of eddy heat and vorticity by the stationary waves themselves, and also the effect of eddy winds impinging on the mountains. These terms play an important role in generating some features of the stationary wave pattern, such as the ridge over the Rockies. Future examinations of stationary wave dynamics will have to assess not only the relative importance of various forcing terms but their mutual interactions, and the nonlinear interactions of the stationary waves themselves. See also Climate Variability: Seasonal to Interannual Variability. Coriolis Force. Cyclogenesis. Dynamic Meteorology: Overview; Waves. Stratosphere–Troposphere Exchange: Global Aspects. Further Reading Gill AE (1982) Atmosphere–Ocean Dynamics. Orlando: Academic Press. Grotjahn R (1993) Global Atmospheric Circulations: Observations and Theories. New York: Oxford University Press. Holton JR (1992) An Introduction to Dynamic Meteorology. New York: Academic Press. Hoskins BJ and Karoly DJ (1981) The steady linear response of a spherical atmosphere to thermal and orographic forcing. Journal of the Atmospheric Sciences 38: 1179–1196. Hoskins BJ and Pearce R (1983) Large-Scale Dynamical Processes in the Atmosphere. London: Academic Press. James IN (1994) Introduction to Circulating Atmospheres. Cambridge: Cambridge University Press. Saltzman B and Manabe S (eds) (1985) Advances in Geophysics, vol. 28, Issues in Atmospheric and Oceanic Modeling. Orlando: Academic Press. STRATOSPHERE–TROPOSPHERE EXCHANGE Contents Global Aspects Local Processes Global Aspects J R Holton, University of Washington, Seattle, WA, USA Copyright 2003 Elsevier Science Ltd. All Rights Reserved. Introduction The troposphere and the stratosphere are separated by a boundary called the tropopause, whose altitude varies from about 16 km in the tropics to about 8 km near the poles. The troposphere is characterized by rapid vertical transport and mixing caused by weather disturbances; the stratosphere is characterized by very weak vertical transport and mixing. The tropopause thus represents a boundary between the troposphere, where chemical constituents tend to be well mixed, and the stratosphere, where chemical constituents tend to have strong vertical gradients. The two-way exchange of material that occurs across the tropopause is important for determining the climate and chemical composition of the upper troposphere and the lower stratosphere. This cross-tropopause 2138 STRATOSPHERE–TROPOSPHERE EXCHANGE / Global Aspects transport is referred to as stratosphere–troposphere exchange. The upward transport of tropospheric constituents into the stratosphere occurs primarily in the tropics, and initiates much of the chemistry that is responsible for global ozone depletion. The downward transport of stratospheric constituents into the troposphere occurs mostly in the extratropics and not only serves as the major sink for some of the constituents involved in stratospheric ozone depletion, but also provides a source of upper tropospheric ozone. This pattern of upward cross-tropopause transport in the tropics and downward cross-tropopause transport in the extratropics is part of a global mass circulation in the stratosphere that occurs as an indirect response to zonal (westward) forcing in the stratosphere, which is caused by the breaking of largescale waves propagating from the troposphere. The magnitude and variability of this stratospheric mass circulation, and its consequences for atmospheric chemistry, are primary considerations in the study of stratosphere–troposphere exchange. The Dynamics of Mean Mass Exchange The Dynamical Definition of the Tropopause The tropopause is traditionally defined by meteorologists as the lowest level at which the rate of decrease of temperature with respect to height (normally about 6 K km  1 in the troposphere) decreases to 2 K km  1, and the average from this level to any level within the next 2 km does not exceed 2 K km  1. This definition, however, does not always clearly mark the boundary between stratospheric and tropospheric air. The physical tropopause is better defined in terms of a specified critical value for a long-lived tracer such as ozone, which has distinctly different stratospheric and tropospheric values. Because global observations of ozone in the vicinity of the tropopause are very limited, it has become common to use as an alternative marker for the tropopause a dynamical field called the potential vorticity. Potential vorticity is somewhat analogous to spin angular momentum. For large-scale atmospheric motions potential vorticity is approximately given by P ¼ r1 ðz þ f Þðqy=qzÞ where r is the air density, z is the vertical component of relative vorticity, f is the Coriolis parameter (twice the local vertical component of the Earth’s angular velocity), y is the potential temperature (a measure of entropy, which increases rapidly with height in the stratosphere), and z is the height above sea level. Since qz=qy may be regarded as a local measure of the depth of the layer between two potential temperature surfaces, an increase in qz=qy implies stretching of vortex tubes and an increase in the absolute vorticity, while a decrease in qz=qy implies shrinking of vortex tubes and a decrease in the absolute vorticity; this is somewhat like the spin angular momentum of a ballerina or figure skater. Outside the tropics, potential vorticity is positively correlated with ozone in the extratropical lower stratosphere, and is a particularly suitable tracer for defining the tropopause. Potential vorticity increases dramatically from troposphere to stratosphere, and it can be readily calculated from conventional wind and temperature data. As shown in Figure 1, the tropopause defined in terms of a critical value of potential vorticity does not coincide with an isentropic surface, 20 15 100 10 300 5 500 1000 90 60 South Pole Altitude (km) Pressure (hPa) 30 30 0 Equator 30 60 90 North Pole Latitude Figure 1 Latitude–altitude cross-section for January 1993 showing longitudinally averaged potential temperature (solid contours) and temperature (dashed contours). The heavy solid contour (cut off at the 380 K potential temperature surface) denotes a constant potential vorticity contour, which approximates the tropopause outside the tropics. Shaded areas denote the ‘lowermost stratosphere’ region whose potential temperature surfaces span the tropopause. (Reproduced with permission from Holton JR, Haynes PH, McIntyre ME et al. (1995) Stratosphere–Troposphere exchange. Reviews of Geophysics 33: 403–439.) STRATOSPHERE–TROPOSPHERE EXCHANGE / Global Aspects 2139 but rather cuts across the isentropes as it slopes downward toward the poles in midlatitudes. The region of the stratosphere where the isentropes intersect the tropopause is called the ‘lowermost stratosphere’, and must be clearly distinguished from the region above where the isentropes lie entirely in the stratosphere. This latter region is often referred to as the ‘overworld’. The Diabatic Circulation Potential temperature is a function of specific entropy alone and is thus conserved by fluid parcels when the motion is adiabatic. Since diabatic processes operate on the time scale of weeks in the lower stratosphere, on shorter time scales parcels move approximately along constant potential temperature surfaces. Transfer of mass and chemical constituents from the troposphere into the stratospheric overworld, however, clearly requires motion across isentropic surfaces. This transport is accomplished by a mean meridional crossisentropic mass circulation. Such a circulation was first deduced by Brewer and Dobson, who showed that observations of the stratospheric distributions of water vapor and ozone were consistent with the notion that upward transport in the stratosphere is limited to the tropics, while downward transport occurs in the extratropics. Thus, the stratosphere is dehydrated by the freeze drying of air passing upward through the extremely cold tropical tropopause, and ozone accumulates at high latitudes in the lower stratosphere through poleward and downward transport from its source region in the upper tropical stratosphere. This transport circulation is now commonly referred to as the Brewer–Dobson circulation. It is often called the ‘diabatic circulation’, since it is associated with the diabatic processes of radiative heating and upward motion across isentropes in the tropics, and with radiative cooling and downward motion across the isentropes in the extratropics. For long-lived trace constituents, such as methane and nitrous oxide, this pattern of meridional overturning, moving up in the tropics and down in the extratropics, tends to produce surfaces of constant mixing ratio that are elevated in the tropics and slope downward toward the poles, while mixing along the isentropes by planetary waves tends to flatten the slopes of surfaces of constant tracer mixing ratio. Although it is associated with diabatic heating and cooling, the Brewer–Dobson circulation is not forced by radiative heating, nor is it forced directly from below by penetration of convection into the stratosphere. Rather, it is a nonlocal response to an extratropical wave-driven pumping action. This pumping is caused by the wave-induced westward force in the extratropical stratosphere. Because the Earth is rotating rapidly, pushing air westward produces a gyroscopic effect in which the air drifts poleward. By mass continuity a poleward drift in midlatitudes is compensated by upward motion accompanied by expansion and adiabatic cooling in the tropics and downward motion accompanied by compression and adiabatic warming in the extratropics (Figure 2). This distribution of adiabatic cooling and heating maintains the temperature below radiative equilibrium in the tropical upwelling region, and above radiative equilibrium in the extratropics. Thus, the distribution of radiative heating and cooling in the stratosphere does not drive the mean meridional mass flow, rather it is a response to the dynamically driven mass flow. Rossby Waves Wave driving in the extratropical stratosphere is caused primarily by Rossby wave breaking. Rossby waves owe their existence to the latitudinal gradient of potential vorticity along isentropic surfaces. Because of this gradient, a fluid parcel displaced poleward or equatorward (and materially conserving its potential vorticity) will have potential vorticity different from that in the local environment and will induce a perturbation velocity disturbance. This will cause parcel displacements of the same sign to the west of the original displaced parcel, and of the opposite sign to the east. The result is a wave pattern in the potential vorticity field that propagates westward relative to the mean flow. When such a wave breaks in the stratosphere it produces a westward directed zonal force, or wave drag (see Rossby Waves). Global Exchange: The Lowermost Stratosphere and the Overworld Global mass exchange into and out of the chemically important region of the stratosphere is to a large degree controlled by the extratropical wave-driven pump discussed above. Air in the overworld, where isentropic surfaces lie entirely in the stratosphere, cannot reach the troposphere without first slowly descending across isentropic surfaces, a process that must be accompanied by diabatic cooling. The isentropic surface bounding the overworld and the lowermost stratosphere generally has a potential temperature around 380 K, depending on cloud top heights (see Figure 2). The distinction between the overworld and the lowermost stratosphere implies that it is not always essential to measure stratosphere–troposphere exchange by the transport across the tropopause. 2140 STRATOSPHERE–TROPOSPHERE EXCHANGE / Global Aspects 10 Pressure (hPa) 30 100 300 1000 Pole Equator Latitude Figure 2 Dynamical aspects of stratosphere–troposphere exchange. The tropopause, defined by a specified constant potential vorticity in the extratropics and the 380 K potential temperature surface in the tropics, is shown by the thick line. Thin lines are potential temperature surfaces labeled in Kelvin. The shaded region is the ‘lowermost stratosphere’ where potential temperature surfaces intersect the tropopause, and isentropic exchange by tropopause folding occurs. Light shading in the stratosphere denotes the wave-induced westward zonal force. Wavy arrows indicate quasi-isentropic transport and mixing by large-scale waves. The two-way exchange results in blocking of anti cyclones, cyclone cut off, and tropopause folds in the troposphere. The broad horizontal arrows are the meridional drift that balances the wave-induced zonal force, and the broad vertical arrows show the nonlocally driven equatorial upwelling and extratropical downwelling referred to here as the diabatic circulation. In the tropics some cumulonimbus clouds penetrate the stratosphere. (Reproduced with permission from Holton JR, Haynes PH, McIntyre ME et al. (1995) Stratosphere–Troposphere exchange. Reviews of Geophysics 33: 403–439.) For many purposes transport into and out of the stratospheric overworld may be more relevant, and more effectively evaluated. For example, consider the case of a chemical species such as methane (CH4) that has a tropospheric source and a stratospheric sink, with the sink being about 18 km or so, in the overworld. The transport across the 380 K potential temperature surface, which can largely be understood as part of the global-scale circulation of the overworld, is then an acceptable measure of exchange; indeed it is often more relevant because of the higher location of the photochemical sink. The same applies to a species that has a stratospheric source and a largely tropospheric sink. In this context, details of the transport across the tropopause are largely irrelevant. For the understanding of mass and tracer transport into and out of the overworld the replacement of the tropopause by a more convenient isobaric or isentropic control surface located in the lower stratosphere is useful. However, for some purposes the mass transport across the actual tropopause is required. The net downward mass fluxes across the extratropical tropopause and across an isobaric or isentropic control surface that nearly coincides with the tropopause in the tropics should be equal for a sufficiently long time average. Such equality will not hold on seasonal or shorter time scales since the amount of mass in the layer between the extratropical tropopause and an isobaric or isentropic surface coinciding with the tropical tropopause may vary with time. To focus on the important aspects of global scale exchange, it is useful to distinguish between the transport along isentropic surfaces, which can occur in rapid adiabatic motions (wavy arrows in Figure 2), and transport across isentropic surfaces, which requires diabatic processes. Since the tropopause intersects the isentropes, transport can occur in either way, and is likely to occur in both ways. In the region of the atmosphere called the lowermost stratosphere, where isentropic surface intersect the tropopause, air and chemical constituents can be irreversibly transported across the tropopause as adiabatic eddy motions lead to large latitudinal displacements of the tropopause, followed by irreversible mixing on small scales. The dark shading in Figure 2 shows the region within the lower stratosphere most directly affected by these eddy transport effects. The lowermost stratosphere must be distinguished from the rest of the stratosphere, STRATOSPHERE–TROPOSPHERE EXCHANGE / Global Aspects 2141 being the only part of the stratosphere accessible from the troposphere via transport along isentropic surfaces. Transport in the overworld must be clearly distinguished from transport in the lowermost stratosphere. Transport in the lowermost stratosphere requires consideration of the details of synoptic-scale and small-scale processes. Horizontal mixing can be especially significant in the lowermost stratosphere, especially during ‘blocking’ events, when meridional motions are enhanced. Thus exchange between the troposphere and the lowermost stratosphere can be significantly faster than exchange between the overworld and the lowermost stratosphere. annual temperature cycle near the tropical tropopause, which is characterized by temperatures that are several degrees colder in January than in July throughout the tropics, and several degrees warmer in January than in July in the extratropics, as would be expected from the influence of the annual cycle in adiabatic cooling associated with the vertical motions. The annual cycle in mass transport is also consistent with the observed cycle in tropical total ozone, which is a minimum in February and a maximum in August, as would be expected from the enhanced vertical advection of ozone-poor tropospheric air into the tropical lower stratosphere during the Northern Hemisphere winter. The Annual Cycle in Global Stratosphere–Troposphere Exchange Isentropic Exchange in the Extratropics Diagnostics of the wave-driven zonal force in the extratropical stratosphere from analysis of conventional global meteorological data can be used to produce estimates of the vertical mass flux across a convenient control surface, such as the 100 hPa isobaric surface (B380 K potential temperature surface in the tropics), which can be regarded as approximating the lower boundary of the overworld. This technique works best for the solstice seasons, when the time change of zonal momentum is small compared to the wave-induced force. Results for the downward mass fluxes across the 100 hPa surface for the Northern and Southern Hemispheres (and by continuity the upward flux in the tropics) are shown in Table 1. The deduced upward mass flux in the tropics is sufficient to completely replace the mass above the 100 hPa surface in about 2–2.5 years. The observations suggest that the mass transport across the tropical tropopause is twice as large in Northern Hemisphere winter as in Southern Hemisphere winter. This deduced annual cycle in mass transport across the 100 hPa surface is consistent with the observed As mentioned above, in midlatitudes the tropopause cuts across isentropic surfaces so that two-way stratosphere–troposphere exchange can occur through isentropic transport and mixing processes. Nevertheless, the boundary between stratospheric and tropospheric air remains very distinct in this region. This suggests that there must be a rather strong dynamical resistance to exchange along the isentropes. This resistance is supplied by the mechanism of Rossby wave propagation. Because of the strong isentropic gradient of potential vorticity that marks the tropopause in the lowermost stratosphere, there is a very strong Rossby wave restoring force in that region, which limits the extent of parcel displacements across the potential vorticity gradient. Hence, wave breaking only occurs for large-amplitude disturbances. Because there is a large store of available potential energy associated with the strong meridional temperature gradient in this region, large-amplitude weather disturbances quite frequently develop in this region, especially in wintertime. The vertical circulations associated with such disturbances create deep intrusions of stratospheric air into the troposphere, which may then become mixed with tropospheric air to produce irreversible transport into the troposphere. Much of the ozone transport from the lowermost stratosphere into the troposphere occurs in connection with such ‘tropopause fold’ events. The quasi-isentropic exchange initiated by tropopause folding could in theory occur in the absence of the diabatic circulation. But in that case there would necessarily be an equal quasi-isentropic reverse transport of air from the troposphere into the stratosphere in order to maintain mass balance. The extremely low water vapor mixing ratios observed throughout the stratosphere indicate, Table 1 Solstice season mass flux across the 100 hPa surface Mass fluxa (108 Kg s 1 ) Location NH extratropics Tropics SH extratropics a DJF JJA Annual mean  81 114  33  26 56  30  53 85  32 Negative sign indicates downward flux. DJF, December, January, February; JJA, June, July, August; NH, Northern Hemisphere; SH, Southern Hemisphere. Data from Rosenlof KH and Holton JR (1993) Journal of Geophysical Research 98: 10465–10479. 2142 STRATOSPHERE–TROPOSPHERE EXCHANGE / Global Aspects however, that the quasi-isentropic exchange in midlatitudes is mostly a one-way transport into the troposphere. Furthermore, the large-scale diabatic circulation is required to transport stratospheric constituents such as ozone downward from the overworld to the lowermost stratosphere. The average rate at which such a species can be transported into the troposphere is thus ultimately determined by the rate at which the dynamically controlled large-scale circulation transports mass into the lowermost stratosphere. For this reason the details of mesoscale tropopause fold events may not be important for determining the global flux of ozone from the stratosphere, although they will certainly strongly influence the time and space distribution of such transport. intrusions of stratospheric air can penetrate into the troposphere. These intrusions may also to some extent be regarded as the result of the systematic effect of the large-scale ageostrophic circulations associated with the development of frontal structures near the tropopause. The stretching deformation that occurs during frontal development stretches stratospheric intrusions to ever finer scales and leads to irreversible transport, often speeded up by turbulence resulting from shear instabilities. Much of the ozone transport from the lowermost stratosphere into the troposphere is believed to occur in connection with such tropopause fold events. Many studies have confirmed that large episodic stratosphere– troposphere exchange can occur in association with tropopause folding. Tracer Exchange in the Lowermost Stratosphere The Role of Tropical Convection in Stratosphere–Troposphere Exchange Exchange of trace constituents cannot be treated in the simple manner used above for the net mass flux because net tracer exchange can occur in the absence of mass exchange. For example, if one unit of air containing a high ozone mixing ratio flows into the troposphere and an equal unit with low ozone mixing ratio flows into the stratosphere, there will be a net downward ozone flux, but zero net mass flux. This sort of process could lead to tracer exchange at the lower edge of the lowermost stratosphere, where the tropopause cuts across isentropic surfaces. Such exchange does not, however, occur on a continuous basis. As noted above, the boundary between stratospheric and tropospheric air along isentropes that span the tropopause is normally marked by strong isentropic potential vorticity gradients. The existence of this band of strong potential vorticity gradients, and indeed similarly strong gradients in mixing ratios of species such as ozone and water vapor, itself suggests that there must be rather strong dynamical resistance to cross-tropopause transport along the isentropes, since otherwise vigorous mixing of stratospheric and tropospheric air would destroy the band of strong gradients. Nevertheless, stratosphere–troposphere exchange of tracers can occur by isentropic transport in the extratropical region. Development of strong upperlevel weather disturbances can lead to displacement of the tropopause from its equilibrium position, followed by nonconservative processes such as diabatic heating or cooling or small-scale turbulent mixing. It is only in the presence of such vigorous eddy motions near the tropopause that the dynamical resistance to crosstropopause exchange can be overcome, and deep Although convection does not control the rate at which the diabatic circulation moves mass into or out of the overworld, penetrative convection may influence a layer a few kilometers in depth in the region of the tropical tropopause. This tropopause layer plays an essential role in establishing the stratospheric water vapor budget. Aircraft and satellite observations of the water vapor distribution in the tropical lower stratosphere reveal the existence of a water vapor mixing ratio minimum (referred to as the ‘hygropause’). The hygropause often occurs well above the tropopause. This feature is now believed to result from dehydration to the saturation mixing ratio at the tropopause (the freeze drying process), followed by vertical advection of the resulting minimum in water vapor mixing ratio into the overworld by the diabatic circulation. Since, as pointed out above, the tropical tropopause temperature is colder in Northern winter than in Northern summer, the driest air enters the stratosphere in Northern winter (when the hygropause is observed to be just above the tropopause), and is advected upward to produce the observed elevated hygropause near 19 km in Northern summer. See also Dynamic Meteorology: Overview; Primitive Equations; Waves. El Niño and the Southern Oscillation: Observation. Instability: Symmetric Stability. Middle Atmosphere: Transport Circulation. Rossby Waves. Stratosphere–Troposphere Exchange: Local Processes. STRATOSPHERE–TROPOSPHERE EXCHANGE / Local Processes 2143 Further Reading Andrews DG, Holton JR and Leovy CB (1987) Middle Atmospheric Dynamics. New York: Academic Press. Brewer AM (1949) Evidence for a world circulation provided by the measurements of helium and water vapor distribution in the stratosphere. Quarterly Journal of the Royal Meteorological Society 75: 351–363. Dobson GMB (1956) Origin and distribution of polyatomic molecules in the atmosphere. Proceedings of the Royal Society of London A236: 187–193. Holton JR, Haynes PH, McIntyre ME, et al. (1995) Stratosphere–troposphere exchange. Reviews of Geophysics 33: 403–439. Salby ML (1996) Fundamentals of Atmospheric Physics. New York: Academic Press. Local Processes J F Lamarque and P Hess, National Center for Atmospheric Research, Boulder, CO, USA Copyright 2003 Elsevier Science Ltd. All Rights Reserved. The tropopause separates the stratosphere from the troposphere. It is located at the interface between two air masses with distinctly different characteristics in water vapor, ozone, potential vorticity, and other chemical or physical quantities. Stratospheric–tropospheric exchange (STE) refers to the processes whereby mass and chemical species are transported between these two atmospheric regions across the tropopause. This exchange is important to the chemistry of both regions as it regulates the transport of species with tropospheric sources into the stratosphere (e.g., chlorofluorocarbons (CFCs), water vapor, and hydrocarbons) and species with stratospheric sources into the troposphere (e.g., ozone and nitric acid). In this article we identify and describe the small-scale processes occurring in the vicinity of the tropopause that govern this exchange. We distinguish between these processes and the large-scale regulation of the exchange by the zonal mean-meridional Brewer–Dobson circulation (Figure 1). The Brewer–Dobson circulation, a circulation largely forced by wave breaking remote from the tropopause, acts to drive air parcels up through isentropic surfaces in the tropics (corresponding to a mean heating of the air parcels) and down through the isentropic surfaces in the extratropics (corresponding to a mean cooling of the parcels). The dynamics of this circulation determines the net STE on an annual time scale. The upward branch of this circulation forces a net exchange from the troposphere to the stratosphere in the tropics and from the stratosphere to the troposphere in the extratropics. Small-scale processes influence precisely where and when STE of mass and chemical species occur. Modeling studies suggest the timing of the exchange, in particular, is important to tropospheric chemistry. Small-scale processes also influence the composition of the stratosphere in the vicinity of the tropopause. In particular, the compo- sition of the lowermost extratropical stratosphere (the part of the stratosphere that shares isentropic surfaces with the troposphere) is strongly affected by smallscale processes. The operational definition of the tropopause by the World Meteorological Organization is given in terms of the temperature lapse rate. However, in the extratropics a dynamically based definition of the tropopause is in terms of a potential vorticity (usually taken at a potential vorticity equal to 2  10  6 m2 s  1 K kg  1). This definition cannot be extended to the tropics, where it is convenient to simply define the tropopause as the 380 K potential temperature surface. Regardless of the definition, the interface between stratospheric and tropospheric air masses forms a wavy surface with substantial geographic variations in height, latitude, and longitude. Significant displacements of the tropopause can occur without STE. The tropopause is a dynamic surface so that transport across it cannot be considered in the same manner as transport across a surface unaffected by transport (e.g., a constant altitude surface). For example, while the mean height of the Northern Hemisphere tropopause lowers during winter, the STE peaks in the spring months. The advantage of using potential vorticity or potential temperature to mark the tropopause is that these quantities act as tracers of air mass motion, making them ideal to mark the interface between stratospheric and tropospheric air masses. Potential vorticity and potential temperature are conserved along trajectories except for the processes of diabatic heating (the vertical gradient of diabatic heating in the case of potential vorticity) and mixing. The extent to which these quantities are not conserved can be taken as a measure of the STE. Therefore, STE can be defined as the amount of mass or constituents transported across potential temperature surfaces in the tropics and potential vorticity surfaces in the extratropics. Diabatic heating and its vertical gradient are generally small in the upper troposphere and lower stratosphere. Mixing is also expected to be slow due to the high static stability of the stratosphere (which resists 2144 STRATOSPHERE–TROPOSPHERE EXCHANGE / Local Processes 10 Wave-driven 30 extratropical Pressure (hPa) Large-scale ascent Large-scale subsidence ‘pump’ 100 400 380 Some cumulonimbus clouds penetrate stratosphere 350 330 300 300 Two-way exchange blocking anticyclones cutoff cyclones tropopause folds 1000 Pole Equator Latitude Figure 1 Dynamical aspects of stratosphere–troposphere exchange. The tropopause is shown by the thick line. Thin lines are isentropic surfaces labeled in Kelvin. The heavily shaded region is the ‘lowermost stratosphere’, where isentropic surfaces span the tropopause and isentropic exchange by tropopause folding occurs. The region above the 380 K surface is the ‘overworld’, in which isentropes lie entirely in the stratosphere. Light shading in the overworld denotes wave-induced forcing (the extratropical ‘pump’). The broad arrows show transport by the global-scale circulation, which is driven by the extratropical pump. This global-scale circulation is the primary contribution to exchange across isentropic surfaces (e.g., the 400 K surface) that are entirely in the overworld. (Reproduced with permission from Holton et al. (1995).) vertical displacements) and the high potential vorticity gradients (which resist horizontal displacements) of the lower extratropical stratosphere. However, as discussed later, under specific circumstances, these nonconservative processes are large enough to allow for significant STE. Because of the different processes involved, the description of STE by small-scale processes is split between the tropics and extratropics. In each section we show how small-scale mixing and diabatic heating at the tropopause result in exchange between the troposphere and the stratosphere. Although intensive research has taken place in the last 40 years, there are still a large number of uncertainties and unknowns in the small-scale processes involved in STE. In particular, the precise mechanisms for the exchange across the tropical tropopause are still not completely understood. Tropical Regions The tropical tropopause (located at approximately 380 K) is located in the upwards branch of the Brewer– Dobson circulation (Figure 1) at a pressure of approximately 100 hPa and a temperature of approximately 70 to 801C. Constituents lofted across the isentropic surface 400 K (approximately 90 hPa) subsequent to crossing the tropical tropopause are likely to be transported into the middle and upper stratosphere by the large-scale Brewer–Dobson circulation. There they can affect the composition of the stratosphere for years. Between the tropical tropopause and 400 K STRATOSPHERE–TROPOSPHERE EXCHANGE / Local Processes 2145 theoretical calculations and measurements of both water vapor and atomic bomb debris (from the 1950s and 1960s explosions) indicate considerable poleward transport of trace constituents. This suggests that a fraction of the constituents which cross the tropical tropopause are not transported much above 400 K, but are rapidly transported into the lowermost extratropical stratosphere, through mostly isentropic transport. STE in the tropics is governed by a complex and poorly understood interplay between convection and the large-scale Brewer–Dobson circulation. Parcels that cross the tropopause are initially transported upwards in deep convective clouds. However, above some height, the Brewer–Dobson circulation will govern the subsequent uplift of the parcel. The transition height between convection and the largescale circulation is not firmly fixed. At least the tropical tropopause is often not clearly demarcated. Instead it may be more accurate to regard the tropical tropopause as a rather deep transition region between the troposphere and the stratosphere. It is still an open question whether the transition between convection and the large-scale circulation typically occurs above or below the defined tropical tropopause. Convective turrets do penetrate the tropopause on occasion, as observed in the Indonesian region, for example. However, there is some doubt as to whether these very deep convective events occur frequently enough to supply the requisite upward mass flux. In this case the upward motion across the tropical tropopause could be of large scale, in which case frequent high cloudiness near the tropopause would be expected. Subvisible cirrus clouds are observed over the warm pool of the western Pacific over 90% of the time during Northern Hemisphere winter, but the cause of this cloudiness is yet undetermined. On the other hand, if convection supplies more than the requisite mass flux above the tropopause, only the highest and coldest convective events may end up impacting the stratosphere. In this case, outside the convective updrafts the equatorial tropopause is in a subsident region. The dryness of the air entering the equatorial stratosphere (approximately 3 ppm by volume during the Northern Hemisphere winter and 4.2 ppm by volume during Northern Hemisphere summer) tightly constrains the possible pathways through which tropical air can enter the stratosphere. As this is much drier than tropospheric air on average and typically drier than the saturation water vapor mixing ratio at the tropical tropopause, any theory of tropical STE must account for the dehydration of air parcels entering the stratosphere. A possible mechanism for such low water vapor mixing ratio is that air that enters the stratosphere has been processed through a cloud. Indeed, as a parcel travels upward and cools, water in excess of the saturation vapor pressure condenses out. Efficient dehydration requires that the parcel remain at cold enough temperatures for ice crystals to grow to sufficient size for rapid sedimentation. Otherwise, as the parcel continues to rise into the stratosphere, the ice crystals may reevaporate. Air with low stratospheric mixing ratios of water vapor has sometimes been measured in association with deep convective clouds. However, processes other than convection may also play a role in dehydrating air. For example, gravity waves propagating near the tropopause may provide sufficient uplift to allow for additional condensation and loss of water vapor. Cloud processing will also affect the STE of chemical species through the attendant loss of soluble species. Zonally averaged tropical tropopause temperatures are not consistent with the extreme dryness of the stratosphere. This suggests the hypothesis that there are preferred regions in which air enters the stratosphere; air passes locally upwards through the tropical tropopause only where the saturation vapor pressure is low enough (from the very cold temperatures) to allow for the sufficient dehydration of air parcels as described above. One such region occurs in the western Pacific (mostly in the vicinity of Indonesia) during Northern Hemisphere winter, in accord with the idea of a localized stratospheric ‘fountain’ through which air enters the stratosphere. However, during the Northern Hemisphere summer the temperature distribution from the large-scale meteorological analyses indicates no region with temperatures persistently cold enough to explain the water vapor record. At this time of year the cold temperatures and dehydration events must occur only sporadically in association with spatially and temporally restricted events not captured in the large-scale meteorological analyses. Another hypothesis, introduced recently and still being developed, is based on the existence of a deep tropopause transition layer. The dehydration of air occurs in convective systems but the transport of the dehydrated air into the stratosphere occurs in a slow ascent due to the overall net radiative heating in this part of the atmosphere. In this view, the dehydration and transport into the stratosphere occur at different times and locations. This view of tropical STE is more dynamic than the stratospheric ‘fountain’ and involves vertical and horizontal processes at very different scales. None of the hypotheses described above have yet been able to fully and consistently explain the observed distribution of water vapor in the tropical stratosphere. Longitudinal variations in tropopause height and temperature, and therefore the preferred locations of equatorial STE, can be ascribed to an array of poorly 2146 STRATOSPHERE–TROPOSPHERE EXCHANGE / Local Processes understood local processes. The coldest tropopause heights are associated with the western Pacific warm pool and the Northern Hemisphere monsoon. This is consistent with convection playing an active role in shaping the morphology of the tropopause. However, the relationship between convection and the tropopause height is not straightforward. In particular, there is indication that minimum temperatures at the tropopause in January are centered on the Equator, while convection maximized slightly south. The radiative effects of convective clouds and the wave motions forced by their diabatic heating obscure any straightforward relationship between convection, the height and temperature of the tropopause, and the location of STE. Extratropics Moving poleward from the Equator, the tropopause is conveniently defined in terms of a potential vorticity surface. STE occurs between the lowermost stratosphere and the troposphere through transport across this surface. While the transport can occur in either direction, it is predominantly from the stratosphere to the troposphere. The effect of transport in the opposite sense is short-lived due to the downwards large-scale mean meridional circulation, which acts to flush out the lowermost stratosphere within a relatively short period. In distinction to the tropical tropopause, the extratropical tropopause is usually clearly demarcated by strong gradients in potential vorticity and trace constituents. The Subtropics Stratosphere–troposphere exchange in this region occurs between the upper and mid-equatorial troposphere and the lowermost stratosphere. The subtropical tropopause drops rapidly near 301 from tropical heights to the level of the extratropical tropopause (approximately from 100 hPa to 300 hPa) (Figure 1). Trajectories from analyzed winds suggest very little STE occurs across this portion of the tropopause during the winter months but that considerable STE occurs during the summer months. The subtropical tropopause cuts through the subtropical jet stream. This jet undergoes a substantial annual cycle in amplitude with the strongest winds occurring during the winter season. When the jet is strong, mixing across it between the troposphere and the stratosphere is inhibited; inhibited both through the large potential vorticity gradients associated with the jet, and the fact that breaking of tropospheric waves and the resulting mixing is unlikely to penetrate the jet core. Indeed, in the case of a strong jet the wind speeds are substantially larger than those associated with most tropospheric waves, implying that the critical layers (where the phase speed of the wave is equal to that of the large-scale flow field and therefore where the wave is unstable and breaks) will occur away from the jet core. During the summer months the subtropical jet weakens considerably, allowing mixing across the jet to be enhanced. Not only do the critical layers occur closer to the jet core during the summer months, but the smaller gradients of potential vorticity associated with the summer jet make for wider critical layers and weaker barriers to mixing. The transport across the summertime subtropical jet is primarily associated with the Asian monsoon (Figure 2), and to a lesser extent the Mexican monsoon. While the monsoons of South America, Africa, and Australia probably play a similar role during the austral summer, their comparatively weak circulations are much less effective in transporting air across the tropopause. As indicated by the arrows in Figure 2, monsoon circulations are able to tap a particularly rich source of water vapor in the midlatitudes. The resulting STE is believed to be of primary importance to the seasonal cycle of water vapor in the extratropical lowermost stratosphere and does not involve the pronounced dehydration that occurs in the tropics. The tropopause is elevated over monsoon regions with the associated anticyclonic circulation penetrating into the lowermost stratosphere. A steady state monsoon circulation will not in itself result in STE. However, due to the proximity of the monsoon circulation to the jet core, perturbations in the circulation are likely to be important, resulting in isentropic mixing between the troposphere and the stratosphere (Figure 2) and associated STE. Moreover, the interaction between monsoon and midlatitude synoptic disturbances or large-scale low-frequency transients will act to transport species across the tropopause. It has been demonstrated in the case of the Asian monsoon that the interaction can act to pull filaments (see the following section on extratropical STE) of moist tropospheric air into the stratosphere, and filaments of dry stratospheric air into the troposphere. The Extratropics In the extratropics a number of local processes result in STE. These include: stratospheric intrusions in the troposphere and their subsequent fragmentation; tropopause folds; cutoff lows; gravity waves; deep convection; radiative processes in the vicinity of the tropopause; and local dynamical instabilities. All the processes listed above are examined in more detail below. The process of fragmentation (i.e., breaking into smaller and smaller structures such as filaments) of STRATOSPHERE–TROPOSPHERE EXCHANGE / Local Processes 2147 _1 10 m s 10 16q E_121q E July U 20 30 0 0 0 70 T 100 T 0 150 200 0 Pressure (hPa) 50 T 300 500 _60q 0 0 1000 _90q S 0 700 _30q 0q Latitude 30q 60q 90qN Figure 2 Schematic diagram of meridional transport and mixing adjacent to monsoon regions in northern summer, superposed on contours of zonal wind (interval 10 m s  1). Heavy contours (interval 2 m s  1) and one-way bold arrows indicate climatalogical meridional transport; two-way arrows illustrate mixing along isentropic surfaces. The large bold arrow at 301 N represents the western side of the Asian monsoon. Its direction should be reversed for the eastern side, where v (meridional velocity) is opposite and slightly smaller. The tropopause is shown (heavy dotted line, T) and the zero-wind line is labeled 0. (Reproduced with permission from Dunkerton (1995).) stratospheric intrusions is strongly related to isentropic mixing. Parcel advection calculations suggest that this mixing occurs vigorously throughout the year on isentropic surfaces below 330 K and is therefore responsible for most of the STE. The fragmentation of stratospheric intrusions occurs as the large-scale velocity field causes tongues of stratospheric air to undergo large latitudinal excursions (Figure 3). Subsequently, these tongues can stretch and thin until they become mere filaments of stratospheric air embedded in the troposphere. This process can be viewed as the fragmentation of the tropopause itself. Once the filaments reach small enough scales they are rapidly and irreversibly mixed into the troposphere. This final mixing may occur due to dynamical instabilities growing at the interface of the filaments (e.g., Kelvin–Helmholtz shearing instabilities) (Figure 4) or through radiative decay. The associated potential vorticity anomalies become increasingly susceptible to radiative decay as they are stretched to small scales. Satellite measurements of ozone and water vapor suggest that fragmentation occurs continually in the vicinity of the tropopause. The fragmentation of intrusions across the tropopause is similar to the fragmentation of the polar stratospheric vortex, creating the so-called stratospheric ‘surf’ zone. In both processes the associated irreversible mixing can be traced to the large meridional parcel displacements that occur in the vicinity of a Rossby wave’s critical layer. As only large-scale waves can propagate into the stratosphere (due to the vertical structure of the zonal wind), the waves which break near the tropopause are of much smaller scale (generally wave number 4–7) than those that break higher up. Consequently the mixing regions are of smaller scale. The waves that break at the tropopause can often be linked to baroclinic instability. Depending on the horizontal shear of the flow, the mixing can occur on the equatorward side of the jet stream, in which case stratospheric air extrudes anticyclonically into the troposphere. In cases of enhanced horizontal shear, mixing can occur on the poleward side of the jet, in which case tropospheric air is entrained into the stratosphere. Owing to the large potential vorticity jump across the tropopause, strong ageostrophic circulations are often created in association with baroclinic wavebreaking events. The ageostrophic circulations enhance the deformation fields due to the large-scale winds and drive stratospheric air deep into the troposphere along isentropic surfaces. During these tropopause folding events, sheets of stratospheric air with a small vertical to horizontal aspect ratio become embedded deep within the troposphere. The ageostrophic circulation associated with tropopause folding is transverse to the jet, with the strongest downward motion generally occurring in the northerly flow to the west of the upper level trough, near the jet 2148 STRATOSPHERE–TROPOSPHERE EXCHANGE / Local Processes Figure 3 (A) Isentropic contours of potential vorticity on the 320 K surface for 14 May 1992, at 1200 UT, calculated from European Centre for Medium-Range Weather Forecasts (ECMWF) operational analyses. The instantaneous tropopause appears as the first solid contour (2 PVU); contours for 1, 1.5, 3, and 4 PVU are also shown. (B) Meteosat water vapor image for the same time as (A). The black structures in the upper-left and right are indicative of dry stratospheric air. entrance region (point A Figure 5). The exchange is associated with both mixing and diabatic processes. The mixing occurs mostly in areas of strong upward and downward motion, as shown by several high-resolution modeling studies. Diabatic effects (latent heat release in clouds and radiative heating/ cooling in the vicinity of clouds) seem to occur mostly in the center of the curvature of the jet stream. STRATOSPHERE–TROPOSPHERE EXCHANGE / Local Processes 2149 2152 2155 2158 2201 2204 2207 2210 2230 2233 3318.0 11435.2 3331.5 11436.4 3344.9 11437.9 3358.1 11439.4 3411.2 11440.9 3423.9 11442.1 3452.1 3440.6 3504.2 3516.8 3526.6 11444.8 11445.5 11447.0 11450.0 11448.5 Electra location 2236 2239 2242 2245 2300 2303 2306 2309 3537.3 11449.3 3550.8 11450.3 3602.3 11451.7 3614.2 11442.4 3626.8 11432.5 2312 Altitude (km) 6.0 4.5 3.0 1.5 0 q N 3305.0 qW 11434.0 30 (A) 2152 Relative atmospheric backscattering 2155 2158 2201 2204 2207 2210 2229 3318.0 11435.2 3331.5 11436.4 3344.9 11437.9 3358.1 11439.4 3411.2 11440.9 3423.9 11442.1 3437.2 3448.7 3500.2 11444.7 11445.6 11446.5 Electra location 2232 2235 280 2243 2238 2246 2301 2304 2307 2310 3606.4 11448.8 3618.2 11439.0 3631.0 11429.3 Altitude (km) 6.0 4.5 (B) 3530.0 3540.8 11448.8 11449.8 3554.5 11451.0 LAS VEGAS, NEVADA 300 3512.6 11448.9 240 180 0 YUMA, ARIZONA 120 q N 3305.8 qW 11434.0 60 3.0 Ozone mixing ratio (ppbv) Figure 4 Cross-section of tropopause fold event on 20 April 1984. Color-scale displays of (A) relative aerosol distributions and (B) ozone mixing ratios. In each case, the higher values of the parameter are indicated by yellow and orange. The mixing of the fold by shearing instabilities can be seen at its leftmost edge. Significant STE occurs during this process. In fact, tropopause folding is considered the most evident form of STE. Under some circumstances tropospheric or stratospheric filaments wind up so as to consist of interwoven regions of stratospheric and tropospheric air. This can create medium-scale potential vorticity anomalies; positive when stratospheric filaments wind up in the troposphere, and negative when tropospheric filaments wind up in the stratosphere. These anomalies will typically be associated with closed circulations – circulations that are temporarily resilient to deformation by the large-scale flow field. The corresponding cutoff cyclones (when a high potential vorticity region becomes trapped in the troposphere) and cutoff anticyclones (when a low potential vorticity region North East West B A Upslope Downslope Figure 5 Three-dimensional view of the tropopause (PV 5 2 PVU) during a tropopause folding. The jet entrance (exit) is indicated by A (B). Due to the ageostrophic circulation, the tropopause undergoes a strong downward motion at the western edge of the fold, followed by a strong upward motion at the eastern edge of the fold. (Reproduced with permission form Lamarque and Hess (1994).) 2150 STRATOSPHERIC CHEMISTRY AND COMPOSITION / Overview becomes trapped in the troposphere), are often rather long-lived, subject only to slow decay through mixing and turbulent processes, radiative processes, and convective mixing (in the cutoff cyclones). All the above processes will result in STE. Other processes may also contribute to extratropical STE. Because potential vorticity is not conserved in the presence of a heating rate gradient, radiative heating in the vicinity of the tropopause is likely to be important, for example, the local heating induced by the high cirrus cloud shield associated with synoptic storms. Lidar (light detection and ranging) ozone measurements and high-resolution modeling suggest that gravity waves excited at the surface by strong winds over steep terrain may at times be responsible for mixing at the tropopause level. Strong thunderstorms in the extratropics occasionally penetrate the tropopause, presumably resulting in STE, although the effect of these storms has not been adequately documented. Both extratropical convection and topographic gravity waves will only result in STE under specific conditions: topographically forced gravity waves only occur under specific wind conditions and intense convection is most likely to during the summer months and over land. While it is difficult to extrapolate from local events to their global effects, the global importance of these processes is likely to be small, although their local effects may be significant. See also Baroclinic Instability. Critical Layers. Monsoon: Overview. Ozone: Ozone Depletion. Tropopause. Further Reading Danielsen EF (1968) Stratospheric–tropospheric exchange based upon radioactivity, ozone, and potential vorticity. Journal of the Atmospheric Sciences 35: 502–518. Dunkerton TJ (1995) Evidence of meridional motion in the summer lower stratosphere adjacent to monsoon regions. Journal of Geophysical Research 100: 16675–16688. Holton JR, Haynes PH, McIntyre ME, et al. (1995) Stratosphere–troposphere exchange. Review of Geophysics 33: 403–439. Hoskins BJ, McIntyre ME and Robertson AW (1985) On the use and significance of isentropic potential vorticity maps. Quarterly Journal of the Royal Meteorological Society 111: 877–946. Pierrehumbert RTand Yang H (1993) Global chaotic mixing on isentropic surfaces. Journal of the Atmospheric Sciences 50: 2464. Randel WJ, Wu F, Russell JM III, Zawodny JM and Oltmans SJ (2001) The seasonal variation of water vapor in the lower stratosphere observed in HALOE data. Journal of Geophysical Research 106: 14313–14325. Shapiro MA (1980) Turbulent mixing within tropopause folds as a mechanism for the exchange of chemical constituents between the stratosphere and the troposphere. Journal of the Atmospheric Sciences 37: 994–1004. STRATOSPHERIC CHEMISTRY AND COMPOSITION Contents Overview Halogen Sources, Anthropogenic Halogen Sources, Natural Halogens HOx Hydrogen Budget Hydroxyl Radical Reactive Nitrogen (NOx and NOy) Overview J A Pyle, University of Cambridge, Cambridge, UK Copyright 2003 Elsevier Science Ltd. All Rights Reserved. Introduction and Background Ozone is perhaps the most important stratospheric constituent. It absorbs solar ultraviolet radiation, particularly strongly at wavelengths below about STRATOSPHERIC CHEMISTRY AND COMPOSITION / Overview 310 nm where stratospheric ozone acts as a filter to protect life at the surface from these potentially harmful wavelengths. Absorption of solar radiation by ozone also results in heating of the stratosphere and leads to the observed stable temperature structure, where temperature increases with height throughout the stratosphere. Ozone is also infrared active and is an important gas for the climate system. For these reasons, the chemistry of the stratosphere is essentially the chemistry of ozone and the minor constituents involved in ozone chemistry. In the troposphere, ozone is present in mixing ratios (the ratio of the concentration of ozone to that of air) of a few tens of parts per billion by volume ( few109 , or a few ppbv) but its peak mixing ratio is much greater in the stratosphere, reaching almost 10 parts per million (10106 , or 10 ppmv) at just above 30 km (10 hPa) in low latitudes. Figure 1 shows seasonally averaged mixing ratios of ozone in the stratosphere and mesosphere based on satellite and ozone sonde data. In contrast, the largest absolute concentrations of ozone are found in high latitudes between 20 and 25 km and reach a few times 1012 molecules cm3. Another useful measure for ozone is its column abundance, the vertically integrated density of ozone above the surface (also often called 2151 ‘total ozone’). For most species column densities are measured in molecules per square centimeter. For ozone, the traditional unit is the Dobson unit (DU), named after the Oxford scientist who in the 1920s pioneered the routine measurement of column densities using spectrophotometers to measure the absorption by atmospheric ozone of the solar spectrum. A Dobson unit is a thickness of 1 millicentimeter at standard temperature and pressure. Typical column densities are 250 DU in the tropics, with little seasonal variation, and 400 DU in high latitudes in winter and spring. Figure 2 shows the average variation of the ozone column, as a function of latitude and month, obtained from satellite measurements by the total ozone mapping spectrometer (TOMS) satellite instrument. As we will see below, the ozone distribution is in part controlled by radical species which themselves are present in even lower concentrations. Typical mixing ratios of the oxides of nitrogen are in the part per billion range while for active chlorine species peak values are usually around or below a part per billion. Mixing ratios of odd hydrogen species are even lower. The radicals themselves are produced from source gases, of both natural and anthropogenic origin and emitted in the troposphere. Thus, nitrous oxide is FUB Ozone (ppmv) 0.1 0.1 0.3 0.5 Pressure (hPa) 1 1 0.3 0.5 2 2 3 3 1 4 76 8 3 5 10 7 6 30 50 100 4 5 6 8 7 10 9 2 3 1 45 9 7 65 34 2 1 30 50 100 DJF MAM 300 300 90° S 60° S 30° S EQ 30° N 60° N 90° N 90° S 60° S 30° S EQ 30° N 60° N 90° N 0.1 0.1 1 Pressure (hPa) 0.3 0.5 1 6 7 3 5 10 30 50 100 3 5 10 32 1 5 JJA 2 3 1 8 9 7 6 1 0.3 0.5 2 3 4 5 4 30 50 100 6 7 4 5 8 3 2 1 9 76 5 4 SON 300 300 90° S 60° S 30° S EQ 30° N 60° N 90° N 90° S 60° S 30° S EQ 30° N 60° N 90° N Latitude (°) Latitude (°) Figure 1 Seasonally averaged zonal mean cross-section of ozone mixing ratio (ppmv) constructed from a combination of satellite and ozone sonde data by the Free University of Berlin. (Courtesy of Dr Ulrike Langematz (FUB).) 2152 STRATOSPHERIC CHEMISTRY AND COMPOSITION / Overview TOMS LTMM Total Ozone (DU) 90° N 420 60° N 400 380 360 420 340 30° N Latitude (°) 400 380 360 300 280 320 300 340 260 320 300 280 EQ 260 260 260 260 280 300 30° S 320 340 280 300 300 360 60° S 320 340 280 360 320 90° S J F M A M J J A S O N D J F M A M J J A S O N D Time (month) Figure 2 Long-term average latitude–time variation of the monthly and zonally averaged ozone column density (Dobson units) based on the available TOMS satellite data from November 1978 to December 1999. The values are repeated for 2 years to emphasize the annual cycle. The instrument measures reflected solar radiation, and the data gaps are in regions of darkness or twilight. (Figure produced by Dr Peter Braesicke (University of Cambridge) from the original TOMS data (http://toms.gsfc.NASA.gov/ozone/ozone.html).) emitted from the Earth’s surface and is relatively inert, and hence well mixed, in the troposphere with a present concentration of about 310 ppbv. It is oxidized in the stratosphere to produce NO (and hence NO2). Similarly, water vapor (2–6 ppmv in the stratosphere) and methane (about 1.5 ppmv at the tropical tropopause) are oxidized to yield the odd-hydrogen species H, OH, and HO2. The halogen species, which have played an important role in ozone depletion during the last two decades, are mainly of recent anthropogenic origin. Their major source gases include CH3Cl (with predominantly natural sources), CF2Cl2 , and CFCl3. These latter species are the so-called freons, which were widely used in aerosol spray cans, refrigeration, and foam blowing and are now regulated under the Montreal Protocol. Along with a number of other chlorinated species, these led to the present-day abundance of chlorine in the stratosphere of about 3.5 ppbv. Similarly, there is about 20 pptv (parts per trillion) of bromine in the stratosphere, arising from the degradation of methyl bromide (which has both natural and anthropogenic sources) and other industrially produced bromocarbons, used, for example, as fire retardants. Odd Oxygen and the Chapman Reactions In 1930 Sidney Chapman proposed a series of reactions to explain the distribution of stratospheric ozone of which the most important are: J1 O2 þ hn ! O þ O k2 O þ O2 þ M ! O3 þ M J3 O3 þ hn ! O þ O2 k4 O þ O3 ! O2 þ O2 ½1 ½2 ½3 ½4 (M represents any third body, usually N2 or O2 , required to conserve energy and momentum in a termolecular reaction.) Note that, in the troposphere and stratosphere, the photolysis of oxygen is much slower than the photolysis of ozone. Reactions [2] and [3] are rapid and have very short time constants for the conversion of O to O3 and vice versa, and they establish a steady state much more rapidly than reactions [1] and [4]. J3 ½O3  ¼ k2 ½O½O2 ½M ½5 ([ ] represents concentration). Note also that reactions [2] and [3] only interconvert the ‘odd oxygen’ species O and O3; i.e., they conserve odd oxygen ð½O þ ½O3 Þ. In contrast, odd oxygen is formed by reaction [1] and removed by reaction [4]. STRATOSPHERIC CHEMISTRY AND COMPOSITION / Overview Thus, we can write for the rate of change of odd oxygen dð½O þ ½O3 Þ=dt ¼ 2J1 ½O2   2k4 ½O½O3  ½6 The time scale for steady state between reactions [1] and [4] varies strongly with altitude, being on the order of hours at 40 km but on the order of many years at 20 km. Invoking steady-state in the upper stratosphere (i.e., setting dð½O þ ½O3 Þ=dt ¼ 0Þ is thus a good approximation. In the low stratosphere, it would clearly be a poor approximation since many external factors (the intensity of solar radiation, temperature, atmospheric transport, etc.) will all vary much more rapidly. In the upper stratosphere, setting dð½O þ ½O3 Þ=dt ¼ 0, we can calculate the steady-state distribution of ozone from eqns [5] and [6]: ½O3  ¼ ðk2 J1 ½O2 2 ½M=k4 J3 Þ1=2 ½7 Equation [7] describes the steady-state ozone concentration in an oxygen-only atmosphere. The vertical profile derived from eqn [7] is consistent with the shape (but not the magnitude) of the observed profile, especially in low latitudes. Thus, eqn [7] predicts a peak in the ozone mixing ratio at a little above 30 km. However, this equation also predicts that the ozone concentration should be very low in high latitudes, when, for example, the photolysis rate of molecular oxygen, J1 , becomes very low. However, observations show large column amounts of ozone in high latitudes in winter and spring (see Figure 2), when photolysis will be at its slowest. The reason for the discrepancy lies in the long photochemical time constant for ozone at low altitudes. When the time constant is long, the transport of ozone must also be considered so that the continuity equation for odd oxygen (eqn [6]) must also include terms to describe the transport. In reality, ozone is produced in a source region in the low latitude middle stratosphere and moved to higher latitudes, where ozone is slowly destroyed, by the action of the stratospheric general circulation. For many years, it was thought that Chapman’s model could adequately explain the distribution of stratospheric ozone, at least in the middle and upper stratosphere. However, with improved measurements – both in the laboratory and in the atmosphere – it became apparent that reaction [4] only removes about 25% of the odd oxygen produced by oxygen photolysis. Calculations based on just the Chapman reactions will seriously overestimate stratospheric ozone concentrations, even when the photochemical time constant is short. 2153 Catalytic Cycles Reaction [4] has an unexpectedly high activation energy for such an exothermic reaction. It was realized that, at stratospheric temperatures (200–290 K), odd oxygen could be removed efficiently in catalytic cycles which achieve the same result as reaction [4] without loss of the catalytic species X or XO: X þ O3 ! XO þ O2 Net : XO þ O ! X þ O2 O þ O3 ! O2 þ O2 (i.e., the two reactions effectively catalyze reaction [4]). Cycles of this kind were discussed for mesospheric chemistry by David Bates and Marcel Nicolet in the 1950s. In the late 1960s and early 1970s attention switched to their role in stratospheric chemistry, pioneered by, for example, Harold Johnston, Paul Crutzen, Mario Molina, and Sherry Rowland, who all highlighted an important potential role in ozone depletion for these cycles. They showed that if the concentration of X increases, the ozone concentration will fall: ozone would be depleted. There are a number of candidates for X present in the stratosphere. These include NO, H, OH, Cl, and Br, all discussed in detail in separate articles. Here, we will take the cycle involving the oxides of nitrogen (NO and NO2 , members of the odd-nitrogen family) as a single example of odd-oxygen destruction by these catalytic cycles. So, substituting X ¼ NO in the catalytic cycle, k5 NO þ O3 ! NO2 þ O2 k6 NO þ NO2 ! NO þ O2 ½8 ½9 This cycle is responsible for about 50% of oddoxygen removal from the stratosphere, despite a number of competing reactions of which the most important is NO2 photolysis, very rapid even at low altitudes. This produces a ‘null cycle’: NO2 þ hn ! NO þ O ðlo400 nmÞ O þ O2 þ M ! O3 þ M NO þ O3 ! NO2 þ O2 ½10 ½2 ½8 Assuming steady state between NO and NO2 based on reactions [8]–[10] (a very good approximation) then, after some simple algebra, the rate of oddoxygen change by the nitrogen oxides can be written, dð½O þ ½O3 Þ=dt ¼ 2k6 ½NO2 ½O ½11 2154 STRATOSPHERIC CHEMISTRY AND COMPOSITION / Halogen Sources, Anthropogenic and the total rate of odd-oxygen change for the combined Chapman and odd-nitrogen cycles would be given by adding eqn [6] to eqn [11]. Similarly, other cycles, where X 5 Cl, OH, etc., have loss rates of the form given by eqn [11]; the rate-limiting step usually involves the reaction of XO with atomic oxygen, O. The concentration of O is low in the low stratosphere (since the rate at which O recombines to form O3 , reaction [2], increases with increasing pressure) and thus the odd-oxygen loss rates are lower in the low stratosphere, leading to the longer photochemical time scales there. These cycles dominate the middle atmosphere away from polar latitudes. In polar latitudes severe ozone depletion has been observed in recent years, forced by halogen chemistry and with the halogens turned into active form by reactions on polar stratospheric clouds, at the low temperatures found there. The cycles are again catalytic, and involve both ClO and BrO. For further details, (see Ozone: Ozone Depletion). One final general point is worth making, again to be discussed in more detail in the articles discussing the individual chemical families. This is that in addition to the radical species involved in the catalytic cycles, other family members exist and can play important roles. For example, HNO3 is an important reservoir species for odd nitrogen; i.e., a species which is a ‘holding-tank’ for NO and NO2 (and indeed OH and HO2) but does not take part in ozone-destruction cycles. Similarly, HCl and ClONO2 , the reservoirs for odd chlorine, are usually the dominant form of chlorine in the lower stratosphere, a fact which limits chlorine-catalyzed ozone destruction, away from polar latitudes, mainly to the upper stratosphere. See also Chemistry of the Atmosphere: Chemical Kinetics. Middle Atmosphere: Transport Circulation. Observations for Chemistry (In Situ): Ozone Sondes. Ozone: Photochemistry of Ozone. Stratospheric Chemistry and Composition: HOx; Halogen Sources, Anthropogenic; Halogen Sources, Natural; Halogens; Hydrogen Budget; Hydroxyl Radical; Reactive Nitrogen (NOx and NOy). Stratospheric Water Vapor. Further Reading An excellent series of reviews of stratospheric ozone have been published by the World Meteorological Organization as part of their ‘Global Ozone Research and Monitoring Project’. The most recent report is No. 44, Scientific Assessment of Ozone Depletion: 1988. Brasseur G and Solomon S (1986) Aeronomy of the Middle Atmosphere. Dordrecht: Reidel. Finlayson-Pitts BJ and Pitts JN Jr (2000) Chemistry of the Upper and Lower Atmosphere. New York: Academic Press. Wayne RP (2000) Chemistry of Atmospheres. Oxford: Oxford University Press. Halogen Sources, Anthropogenic A McCulloch, University of Bristol, Bristol, UK P M Midgley, M & D Consulting, Leinfelden Musberg, Germany Copyright 2003 Elsevier Science Ltd. All Rights Reserved. Introduction None of the anthropogenic carriers of halogens in the stratosphere is actually released there. They are emitted close to ground level and have to survive transport through the troposphere, requiring a lifetime in the atmosphere of at least a year. Thus soluble halogen-containing materials, such as hydrogen chloride (HCl), which are rained out of the atmosphere in a matter of days, do not provide a significant halogen input into the stratosphere; neither do the more reactive materials, such as trichloroethene or tetra- fluoroethene, which are oxidized in the lower atmosphere within a similar time scale. The bulk of the halogen input into the stratosphere is from anthropogenic gases that have atmospheric lifetimes significantly longer than 2 years. These are released from industrialized regions, principally in the Northern Hemisphere. Chlorofluorocarbons (CFCs), with atmospheric lifetimes of 45–1700 years, were introduced in the 1930s as refrigerants that were safer than the toxic and flammable materials then used. Despite the fact that small amounts of CFCs have been measured in volcanic vents, the natural contribution is negligible compared with man-made sources. Carbon tetrachloride, which has an atmospheric lifetime of 35 years, had been used as a solvent until the middle of the twentieth century; subsequently it was mainly used as a raw material for CFC manufacture, and emissions STRATOSPHERIC CHEMISTRY AND COMPOSITION / Halogen Sources, Anthropogenic 2155 into the atmosphere grew with CFC production. Hydrochlorofluorocarbons (HCFCs), with lifetimes between 1.4 and 19 years, were introduced in the 1940s for deep freezing applications otherwise served by ammonia. More recently, the HCFCs have become partial replacements for CFCs. Halons, fire-extinguishing chemicals with lifetimes of 11–65 years and containing bromine, were introduced in the 1960s. At the same time the use of methylchloroform (1,1,1trichloroethane, atmospheric lifetime 4.8 years) as a precision cleaning solvent was expanding rapidly. Together with methyl chloride and methyl bromide, which have significant natural fluxes, these carry potentially reactive halogens (chlorine and bromine) into the stratosphere and, with the exception of methyl chloride, all are ozone-depleting substances controlled by the Montreal Protocol. The history of anthropogenic emissions and the resulting atmospheric concentration is described here, together with the consequential rise of chlorine and bromine in the troposphere. CFCs, carbon tetrachloride, and most of the halons are removed from the atmosphere only by photolysis in the stratosphere, hence their relatively long atmospheric lifetimes. HCFCs, methylchloroform and methyl halides are oxidized in the troposphere and generally have shorter atmospheric lifetimes, but for all compounds the average time delay between release in the lower atmosphere and decomposition in the ozone layer to generate stratospheric halogen is 3 years. Furthermore, the relative effectiveness in ozone depletion of each compound varies with the altitude at which its halogen is released, and this, together with the time delay, is taken into account when calculating the total Equivalent Effective Stratospheric Chlorine (EESC), which is a measure of the combined effect of all ozonedepleting substances. Fluorine does not react with stratospheric ozone. Consequently, the hydrofluorocarbons (HFCs) that are designed to replace CFCs are not controlled under the Montreal Protocol. There is already a significant stratospheric fluorine concentration arising from decomposition of CFCs and this is starting to be augmented by HFCs, which have lifetimes between 1.5 and 240 years. The extent of this and the concentrations of the much less reactive perfluorocarbons (PFCs) are also discussed. The Chlorine Flux Chlorofluorocarbons Most of the anthropogenic chlorine content of the atmosphere is a consequence of CFC emissions. Historically, the largest single source was aerosol spray cans from which the CFC propellant was released immediately the can was used. Currently, the principal release into the atmosphere is from a declining stock of CFC contained in refrigeration and air-conditioning systems and foam insulation; in these applications, release of the substance occurs some time after it was manufactured. The delay is variable. Automobile air conditioners can release all of their contents in a matter of a few years; on the other hand, a domestic refrigerator has a typical service life of 20 years and the CFCs it contains leak only when it is dismantled. Insulating foam can be in use for much longer, with only slow release until (and if) the foam is crushed. Uncertainties in estimates of the delay between manufacture and emission into the atmosphere, characteristic of such uses, contribute significantly to the uncertainty of the estimated emission. For almost 30 years the manufacturers of CFCs have organized an annual collection of audited industrial production and sales data for CFC-11 and CFC-12. Historical production and sales records were also extracted by the manufacturers for the period back to first production in 1931 and the combined data provide the basis for calculation of emissions of these compounds. Annual emissions are estimated for each major category of application based on the quantities used, coupled with emission functions that take account of the rates of release of the materials during actual use and disposal (which are specific to the application). The survey procedure and emissions estimation have been extended to most of the industrial halocarbons: CFC-11 (trichlorofluoromethane), principally used in aerosols and foam insulation CFC-12 (dichlorodifluoromethane), principally used in aerosols and refrigeration CFC-113 (1,1,2-trichlorotrifluoroethane), a solvent CFC-114 (1,1,2,2-tetrafluorodichloroethane), principally used in aerosols and refrigeration CFC-115 (chloropentafluoroethane), a refrigerant. In much of the world, CFC production was carried out by subsidiaries of companies that reported their production and sales into the database and up to 15 years ago the only producing country not included was the USSR. Since then India, China, and Korea have become significant producers, although they too do not report into the industrial database. However, national aggregate CFC production now has to be reported to the Secretariat of the Montreal Protocol by all parties. The estimated historical quantities released, shown in Table 1, are based on a composite global estimate of annual production from the industrial and legislative databases. 2156 STRATOSPHERIC CHEMISTRY AND COMPOSITION / Halogen Sources, Anthropogenic Table 1 Emissions of CFCs (Gg y  1) Year CFC-11 (CFCl3; 45 y)a CFC-12 (CF2Cl2; 100 y) CFC-113 (CF2ClCFCl2; 85 y) CFC-114 (CF2ClCF2Cl; 300 y) CFC-115 (C2F5Cl; 1700 y) 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 209.0 229.8 259.2 296.6 327.1 318.4 325.9 314.0 294.9 276.1 264.6 263.8 257.2 273.8 295.6 308.6 326.8 345.7 353.5 304.7 211.4 213.2 168.9 146.8 101.1 100.7 101.7 99.2 97.2 313.8 338.0 368.9 408.7 444.3 435.5 425.2 406.1 376.7 375.7 373.3 385.3 385.3 394.4 413.6 426.1 437.5 451.0 462.7 436.9 378.8 335.7 319.9 302.8 239.9 239.3 220.4 185.0 155.5 28.0 32.1 36.9 42.2 48.4 55.5 63.5 72.8 83.4 95.5 109.4 119.4 124.6 138.3 171.1 201.7 216.6 236.4 260.3 271.6 233.8 181.5 147.5 80.5 52.0 43.2 27.0 9.5 5.4 9.7 10.1 10.5 10.9 11.3 11.7 12.2 12.7 13.2 13.7 14.2 14.2 13.7 14.1 15.1 16.2 18.0 18.2 16.2 14.5 10.3 6.3 5.2 4.6 4.0 3.1 2.4 2.3 2.7 1.3 1.6 1.9 2.2 2.5 3.0 3.5 4.0 4.7 5.4 6.3 7.2 8.1 8.9 9.6 10.1 10.6 11.0 11.4 11.9 12.2 12.6 12.6 12.6 11.8 10.9 9.5 7.8 6.0 Sources: AFEAS (Alternative Fluorocarbons Environmental Acceptability Study) (2000), Production, Sales and Atmospheric Release of Fluorocarbons through 1998. Washington, DC: AFEAS. Midgley PM and McCulloch A (1999) Production, sales and emissions of halocarbons from industrial sources. In: Fabian P and Singh ON (Eds) The Handbook of Environmental Chemistry, vol 4. Part E, Reactive Halogen Compounds in the Atmosphere, pp. 157–190. Heidelberg: Springer-Verlag. a The formula and atmospheric lifetime in years are given in parentheses. For each sales category a characteristic pattern of emissions in time was established by market surveys carried out by the producers. This enabled estimates of annual emissions as outlined above. These have been the subject of a sensitivity study that confirmed that the largest contributions to the uncertainties came from the fraction of production that was not reported in the industrial data and the rate of release of materials from closed-cell foams. The first of these has been addressed using the database from the Montreal Protocol, which now matches the industrial data reliably (to within 1% over the same set of countries). The second is a particular problem for CFC-11, where the range of emissions resulting from the lowest credible estimate of the release from closedcell foams to the highest is 13.1%. For the period up to 1992, a mid-range estimate was used in Table 1. In recent years, as the degree of containment of CFCs in systems has improved, the historical emissions functions have tended to overestimate releases. This was allowed for in the estimates developed for recent Scientific Assessments and, from 1992 onwards, it is those values that are shown in Table 1. In all cases the release estimates show substantial falls during the 1990s. The fall in consumption has actually been faster than that required under the Montreal Protocol; nevertheless, large quantities of CFCs remain in systems and may be released in the future: for example, over 700 Gg of CFC-11 and 250 Gg of CFC-12 are currently unreleased. Figure 1 shows how the atmospheric concentrations of CFCs have grown. These were calculated using a simple single-box model of the atmosphere and current estimates of atmospheric lifetimes, according to eqn [1], where C0 and Cy are the atmospheric concentrations in the starting year and in year y, S is the annual rate of release of the substance, and T is its atmospheric lifetime. Cy ¼ ST þ ðC0  STÞ expðy=TÞ ½1 Tropospheric chlorine loading (ppt) STRATOSPHERIC CHEMISTRY AND COMPOSITION / Halogen Sources, Anthropogenic 2157 Table 2 Emissions of chlorocarbons (Gg y  1) 2500 (a) (b) 2000 1500 (d) 1000 500 0 1970 (e) 1974 Year Carbon tetrachloride (CCl4 35 y)a Methyl chloroform (CH3CCl3, 4.8 y) 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 69 74 84 94 99 85 92 86 83 97 95 78 78 92 74 82 100 91 89 72 64 36 44 45 36 32 10 8 8 133 170 214 266 305 309 382 462 513 511 538 549 523 536 585 593 602 623 666 691 718 635 593 380 283 234 84 30 16 (c) 1978 1982 1986 1990 1994 1998 Year Figure 1 Contributions to tropospheric chlorine loading from chlorofluorocarbons: (a) chloropentafluoroethane (CFC-115); (b) dichlorotetrafluoroethane (CFC-114); (c) trichlorotrifluoroethane (CFC-113); (d) dichlorodifluoromethane (CFC-12); (e) trichlorofluoromethane (CFC-11). The units are parts per trillion (ppt, 1 in 1012) of tropospheric chlorine loading, which is the calculated concentration of each CFC multiplied by the number of atoms of chlorine in its molecule. Thus, for CFC-11 (fluorotrichloromethane), the CFC concentration is multiplied by 3. While the growth in chlorine loading arising from CFC-12 emissions has slowed in recent years, it is still the largest of the CFC contributors and its absolute concentration is still growing. The concentrations of CFC-11 and CFC-113 have fallen discernibly, and those from CFC-114 and CFC-115 are not large enough to matter. Overall, the CFC contribution to chlorine loading is now level in time. Chlorocarbons The next largest contribution to the chlorine loading of the atmosphere comes from carbon tetrachloride (CCl4 , tetrachloromethane) and methylchloroform (CH3CCl3 , 1,1,1-trichloroethane). Estimates of their emissions are shown in Table 2. Carbon tetrachloride is hepatotoxic at relatively low concentrations and so has not been used as a solvent in developed countries for many years. Its principal use is as raw material for the manufacture of CFC-11 and CFC-12 and it is thought that the accumulation in the atmosphere now results solely from process losses. It has not been possible to quantify these losses in the same way as for the CFC releases, consequently emissions into the atmosphere have been calculated from the change in atmospheric concentration over the period 1979 to 1996, using an inverted form of eqn [1]. For methyl chloroform, an audited production and sales database has been maintained from information Sources: Midgley PM and McCulloch A (1999) Production, sales and emissions of halocarbons from industrial sources. In: Fabian P and Singh ON (eds) The Handbook of Environmental Chemistry, vol 4. Part E, Reactive Halogen Compounds in the Atmosphere, pp. 157–190. Heidelberg: Springer-Verlag. Simmonds PG et al. (1998) Global trends and emission estimates of CCl4 from in situ background observations from July 1978 to June 1996. Journal of Geophysical Research, 103: 16017–16027. a The formula and atmospheric lifetime in years are given in parentheses. collected by the producers in much the same way as for CFCs. With the exception of use as a chemical intermediate (in which it is totally converted and not released), methyl chloroform was used as an industrial solvent, with total emission into the atmosphere. Hence the emission function is relatively simple; 34 of annual sales are estimated to be emitted in that year and 14 in the following year. Long-term storage, over one or two years, was accommodated by a linear displacement of emissions in time. Prompt emissions, coupled with a relatively short atmospheric lifetime, have meant that the concentration of methyl chloroform shows the sharpest fall as a consequence of the Montreal Protocol. Figure 2 shows the contributions to chlorine loading from the individual chlorocarbons superimposed on that from the CFCs. Tropospheric chlorine loading (ppt) 2158 STRATOSPHERIC CHEMISTRY AND COMPOSITION / Halogen Sources, Anthropogenic 3500 3000 (a) 2500 (b) 2000 1500 1000 (c) 500 0 1970 1974 1978 1982 1986 Year 1990 1994 1998 Figure 2 Contributions to tropospheric chlorine loading from CFCs and chlorohydrocarbons: (a) methyl chloroform (1,1,1trichloroethane); (b) carbon tetrachloride (tetrachloromethane); (c) all CFCs combined, as shown in Figure 1. Hydrochlorofluorocarbons Despite their potential to replace CFCs, hydrochlorofluorocarbons (HCFCs) have relatively little impact on atmospheric chlorine loading. The principal member of this group of substances, chlorodifluoromethane (HCFC-22), has been used as a refrigerant fluid since 1946; its low boiling point makes it suitable for lowtemperature duties and some airconditioning. As shown in Table 3, HCFC-22 emissions have grown to about 250 Gg y  1 and are now stable. Emissions of the other HCFCs are one or two orders of magnitude lower. HCFC-124 (1,1,1,2-tetrafluorochloroethane), introduced comparatively recently, is an aerosol propellant and refrigerant fluid that is produced in modest amounts. HCFC-141b (1,1-dichloro-1-fluoroethane), again a relative newcomer, is produced in much larger quantities for use either as a blowing agent for rigid plastic foams, such as those used for insulation, or as a solvent. HCFC-142b (1-chloro-1,1-difluoroethane) is also used to blow plastic foam. Over 94% of all HCFC production is in the developed world. HCFCs were considered to be suitable temporary replacements for CFCs because of their low intrinsic potential to impact the ozone layer. In general, they Table 3 Emissions of HCFCs (Gg y  1) Year HCFC-22 (CHF2Cl, 11.8 y)a 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 45.5 50.4 55.1 63.2 71.5 72.3 81.4 90.6 99.1 106.4 114.7 121.3 123.1 135.1 145.0 150.9 160.2 169.1 188.5 207.6 214.7 224.7 236.3 234.0 240.8 247.2 266.2 245.9 255.8 HCFC-124 (CF3CHFCl, 6.1 y) 0.1 0.3 0.4 1.8 3.5 3.7 3.2 HCFC-141b (CH3CFCl2, 9.2 y) HCFC-142b (CH3CF2Cl, 18.5 y) 0.4 3.9 13.1 24.8 36.6 39.5 42.7 49.8 0.6 0.5 0.5 0.6 0.4 1.6 1.7 2.0 2.9 5.8 8.4 10.8 10.2 10.7 12.0 11.7 11.6 10.6 Sources: AFEAS (Alternative Fluorocarbons Environmental Acceptability Study) (2000) Production, Sales and Atmospheric Release of Fluorocarbons through 1998. Washington, DC: AFEAS. WMO (World Meteorological Organization) (1999) Scientific Assessment of Ozone Depletion: 1998, WMO Global Ozone Research and Monitoring Project Report No. 44. Geneva: WMO. a The formula and atmospheric lifetime in years are given in parentheses. STRATOSPHERIC CHEMISTRY AND COMPOSITION / Halogen Sources, Anthropogenic 2159 contain less chlorine than CFCs, have shorter atmospheric lifetimes, so that they do not accumulate in the atmosphere to the same extent as CFCs, and are not photolyzed as effectively in the stratosphere, so that the chlorine they contain is not released directly into the ozone layer. Nevertheless, HCFCs are ozonedepleting substances and are to be phased out under the Montreal Protocol by 2020 in the developed world and 2040 elsewhere. The contribution of HCFCs to chlorine loading is shown in Figure 3; no allowance has been made for the relative effectiveness of their chlorine content. Natural Source of Chlorine Although the quantities released by human activities are small, methyl chloride (CH3Cl, chloromethane) is produced by natural processes in sufficient amounts to contribute significantly to stratospheric chlorine. The lifetime of methyl chloride is only 1.3 years. However the flux of 4 Tg y  1, mainly from the oceans, biomass burning, and terrestrial fungi, is large enough to maintain an atmospheric concentration of 550 ppt. The Bromine Flux Halons Tropospheric chlorine loading (ppt) The natural contribution to bromine in the stratosphere is similar to that from anthropogenic sources; of the total bromine loading of about 17 ppt, 9 ppt is attributable to man’s activities and most of this comes from halon emissions. Halons were first produced as fire-extinguishing agents in 1963 and their use expanded to almost 20 Tg y  1 by the mid 1990s. Two 3500 (a) (b) 3000 (c) 2500 2000 1500 1000 (d) 500 0 1970 substances predominated; Halon-1211 (bromochlorodifluoromethane), used mainly in portable extinguishers, and Halon-1301 (bromotrifluoromethane), an agent used in fixed systems. In addition, Halon2402 (1,2-dibromotetrafluoroethane) was produced in somewhat smaller quantities and used in Eastern Europe. Halon-1202 (dibromodifluoromethane) has also been detected in small, but growing, amounts in the atmosphere. Bromine is 60 times more potent in ozone depletion than chlorine in the current background stratospheric composition. This was recognized in the Montreal Protocol and halon production was phased out earlier than CFCs in the developed world (in 1994). However, production of Halon-1211 and Halon-1301 will continue in China, India, and Korea for the next few years and Russia has dispensation to continue the manufacture of Halon-2402. In much the same way as for CFCs, audited production statistics are available from industry in the developed world and from the submissions required under the Montreal Protocol for the controlled halons, but the proportion of annual halon production that is unreleased is much higher than is the case for CFCs. Currently, halons should be released into the atmosphere only when they are used in earnest – on a fire or when the fire protection system is activated. Although historically they were also released during training and system testing, there remains a considerable time delay between production and release and a stock of halon (known as the ‘bank’) has accumulated in systems and equipment. The emissions of Halon1211 shown in Table 4 were based on a small part of the bank (currently 12%) being emitted each year. In the case of Halon-1301, calculated similarly, the emission factor is now 4% of the bank each year. Production data for Halon-2402 do not exist in the same form and so the values for emissions in Table 4 were calculated by inverse modeling from measured atmospheric concentrations, using eqn [1]. The source of Halon-1202 has yet to be identified, although it is a well known by-product of the manufacture of Halon1211. The emissions shown in Table 4 were calculated by inverse modeling. Methyl Bromide 1974 1978 1982 1986 Year 1990 1994 1998 Figure 3 All contributions to tropospheric chlorine loading from (a) the combined concentrations of HCFC-124 (1,1,1,2-tetrafluorochloroethane), HCFC-141b (1,1-dichloro-1-fluoroethane), and HCFC-142b (1-chloro-1,1-difluoroethane); (b) HCFC-22 (chlorodifluoromethane); (c) all chlorohydrocarbons (as in Figure 2); (d) all CFCs (as in Figure 1). Methyl bromide contributes a total of 10 ppt to bromine loading. Of this, only about 1.9 ppt arises from human activities that are controlled under the Montreal Protocol, principally use of manufactured material for pest control in growing and harvested agricultural produce. Minor other anthropogenic sources that are not controlled add a further 0.4 ppt into the atmosphere; these include the exhausts of 2160 STRATOSPHERIC CHEMISTRY AND COMPOSITION / Halogen Sources, Anthropogenic Table 4 Emissions of halons (Gg y  1) Year Halon-1211 (CF2ClBr,11 y) a Halon-1301 (CF3Br, 65 y) Halon-2402 (CF2BrCF2Br, 25 y) Halon-1202 (CF2Br2, 3y) 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 0.29 0.41 0.56 0.74 0.95 1.18 1.49 1.74 2.08 2.44 2.84 3.18 3.60 4.18 4.96 5.96 7.10 8.41 9.82 10.08 10.05 11.49 11.99 12.49 12.99 13.49 11.12 11.46 11.12 0.05 0.11 0.19 0.30 0.41 0.57 0.86 1.10 1.37 1.70 1.98 2.35 2.90 3.27 3.81 4.39 5.02 5.61 6.25 6.01 5.62 3.56 3.61 1.16 4.66 3.47 2.77 2.80 2.71 0.20 0.25 0.31 0.39 0.47 0.54 0.62 0.70 0.77 0.85 0.93 1.00 1.11 1.18 1.26 1.36 1.43 1.50 1.73 1.73 1.73 1.74 1.72 1.70 1.68 1.30 0.85 0.70 No data 0.04 0.04 0.06 0.07 0.08 0.10 0.11 0.13 0.15 0.16 0.19 0.21 0.23 0.25 0.27 0.29 0.31 0.33 0.35 0.37 0.39 0.41 0.43 0.51 0.59 0.67 0.73 0.79 No data motor vehicles running on leaded gasoline and also chemical process emissions. Although there is much uncertainty, the bulk of methyl bromide entering the atmosphere seems to come from natural processes. The role of the oceans is particularly difficult to untangle because they act as both sources and sinks. Methyl bromide is released into the atmosphere particularly from the polar oceans and is absorbed from the atmosphere into tropical waters where it is destroyed by bacteria. A number of other bromine compounds are produced naturally: dibromomethane, bromochloromethane and dibromochloromethane together can add up to 6 ppt to bromine loading at ground level, particularly in the Arctic. However, these are very short-lived species and are not considered normally to be transported into the stratosphere. Figure 4 shows the increase in anthropogenic bromine loading since 1970, subdivided into contributions from individual compounds. In the absence of Tropospheric bromine loading (ppt) Sources: Midgley PM and McCulloch A (1999) Production, sales and emissions of halocarbons from industrial sources. In: Fabian P and Singh ON (eds) The Handbook of Environmental Chemistry, vol 4. Part E, Reactive Halogen Compounds in the Atmosphere, pp. 157–190. Heidelberg: Springer-Verlag. Fraser PJ et al. (1999) Southern Hemispheric halon trends (1978–1998) and global halon emissions. Journal of Geophysical Research 104: 15985–15999. a The formula and atmospheric lifetime in years are given in parentheses. 9.00 8.00 (b) 7.00 6.00 (a) (c) 5.00 (d) 4.00 3.00 2.00 (e) 1.00 0.00 1970 1975 1980 1985 Year 1990 1995 Figure 4 Contributions to tropospheric bromine loading from (a) that part of methyl bromide emissions that is controlled under the Montreal Protocol; (b) Halon-1202 (dibromodifluoromethane); (c) Halon-2402 (1,2-dibromotetrafluoroethane); (d) Halon-1301 (bromotrifluoromethane); (e) Halon-1211 (bromochlorodifluoromethane). STRATOSPHERIC CHEMISTRY AND COMPOSITION / Halogen Sources, Anthropogenic 2161 better information, the contribution from methyl bromide has been shown as constant. The Fluorine Flux Tropospheric fluorine loading (ppt) Neither F nor CF3 radicals, nor their oxygenated derivatives, interact with stratospheric ozone; fluorine released into the stratosphere is converted into hydrogen fluoride (HF), which does not react further and is eventually removed when the stratospheric air circulates into the troposphere. However, it is a significant component of the stratospheric halogen budget. In much the same way as for chlorine and bromine, fluorine loading of the troposphere may be calculated from the atmospheric concentrations of CFCs, HCFCs, and halons, with the results shown in Figure 5. The contribution from hydrofluorocarbons (HFCs) is currently small but is increasing at a significant rate. This is largely a consequence of releases of trifluoromethane (HFC-23, fluoroform), which is a by-product of the manufacture of HCFC22, has a long atmospheric lifetime, and is decomposed in the stratosphere, so adding to the fluorine burden there. More recently this has also been augmented by releases of HFC-134a (1,1,1,2-tetrafluoroethane, CF3CH2F), which is manufactured for use as a refrigerant and now has a tropospheric concentration of 9 ppt. Other fluorine-containing substances do not contribute significantly to fluorine loading either because the quantities released are currently too small to matter (the case with hydrofluorocarbons other than HFC-23 and HFC-134a) or because they are so inert that they do not decompose to release fluorine in the stratosphere (the case with perfluorocarbons and sulfur hexafluoride). Perfluorocarbons, in particular those that are formed as by-products of primary aluminum production, are much more abundant than hydrofluorocar2500 (a) 2000 (b) (c) 1500 1000 (d) 500 0 1970 1974 1978 1982 1986 1990 1994 1998 Year Figure 5 Contributions to tropospheric fluorine loading from (a) all halons; (b) all HCFCs; (c) all CFCs. bons. Tetrafluoromethane (CF4 , PFC-14) has now reached a concentration of 80 ppt, half of which is due to aluminum production. The other 40 ppt is volcanic in origin and has accumulated in the atmosphere over many thousands of years. Hexafluoroethane (C2F6 , PFC-116), another aluminum by-product, has no natural source and its atmospheric concentration now stands at 3 ppt. These substances have atmospheric lifetimes over ten thousand years and are so inert that they do not contribute to the stratospheric loading of fluorine; indeed, the trend of their stratospheric concentrations with altitude is a good indicator of their historic tropospheric concentrations. As for perfluorocarbons, the atmospheric lifetime of sulfur hexafluoride (SF6) is long (3200 years) and it too does not contribute to the stratospheric loading of fluorine. Although there is a volcanic source, it is too small to be significant and most of the 4 ppt of sulfur hexafluoride that is now present in the atmosphere has been used in industrial applications such as electrical switchgear. Equivalent Effective Stratospheric Chlorine and the Future The concentrations so far discussed may be verified by direct measurement of the individual species in the troposphere, which is comparatively well mixed. Tropospheric loading describes the concentration of potentially active chlorine and bromine in the flux of air entering the stratosphere; it is not exactly equal to the loading of active halogen at the ozone layer. This is parameterized by equivalent effective stratospheric chlorine loading (EESC). To calculate the EESC, the tropospheric loadings of all compounds are adjusted by an overall factor to take account of the transport time between the troposphere and the stratospheric ozone layer and the contributions from individual chlorine- and bromine-containing compounds are adjusted by factors that accommodate their different effects on the ozone layer. The delay due to transport is set at 3 years. The effectiveness factor for the difference between chlorine and bromine is set at 60, as described above. The differences between individual chlorine compounds are much smaller than that; they range from 1.11 for HCFC-123 (CF3CHCl2) to 0.35 for HCFC-22 (CF2HCl). Figure 6 shows the way that EESC has developed in the past and the changes expected over the twenty-first century. Although it is clear that the peak loading is past, it will take the whole of the twenty-first century for the stratospheric loading to return to 1970 levels and, all other things being equal, the return to loadings that predate the Antarctic ozone hole is expected to occur only towards the middle of the century. Equivalent effective stratospheric chlorine loading (ppt) 2162 STRATOSPHERIC CHEMISTRY AND COMPOSITION / Halogen Sources, Natural 3500 3000 (a) 2500 (b) (c) 2000 1500 (d) 1000 500 0 1970 (e) 1990 2010 2030 2050 Year 2070 2090 Figure 6 Contributions to the equivalent effective stratospheric halogen loading from (a) halons and controlled sources of methyl bromide; (b) all HCFCs; (c) controlled chlorocarbons; (d) all CFCs; and (e) methyl chloride and the natural (and other uncontrolled) sources of methyl bromide. Hydrofluorocarbon (HFC) A chemical compound consisting only of carbon, fluorine, and hydrogen. Tropospheric chlorine loading The concentration, in the lower mixed layer of the atmosphere, of a compound that could transport chlorine to the stratosphere, expressed as the product of actual concentration and the molecular chlorine content. Units are parts per trillion (ppt). Parts per trillion (ppt) (1 in 1012) Used here to describe the atmospheric concentration of a substance in terms of its molar mixing ratio. It is equivalent to picomoles mole  1. See also Lightning: Production of Nitric Oxide. Ozone: Ozone Depletion Potentials; Photochemistry of Ozone. Tropospheric Chemistry and Composition: Oxidizing Capacity. Glossary Further Reading Atmospheric lifetime The ratio of the atmospheric concentration of a substance to its instantaneous loss rate. Chlorofluorocarbon (CFC) A chemical compound consisting of carbon, chlorine, and fluorine only. Equivalent effective stratospheric chlorine loading The calculated concentration of halogen that could be effective in ozone depletion, expressed as an average for the stratosphere and taking account of the relative effectiveness of chlorine and bromine in ozone depletion and the relative efficiency by which individual species releases halogen into the ozone layer. Units are chlorine equivalent parts per trillion (ppt). Halon A chemical compound consisting only of carbon, bromine, fluorine (and in some cases chlorine). Hydrochlorofluorocarbon (HCFC) A chemical compound consisting only of carbon, chlorine, fluorine, and hydrogen. AFEAS (Alternative Fluorocarbons Environmental Acceptability Study) (2001) Production, Sales and Atmospheric Release of Fluorocarbons through 2000. Washington, DC: AFEAS. Midgley PM and McCulloch A (1999) Production, sales and emissions of halocarbons from industrial sources. In: Fabian P and Singh ON (eds) The Handbook of Environmental Chemistry, vol 4, part E, Reactive Halogen Compounds in the Atmosphere, pp. 157–190. Heidelberg: Springer-Verlag. SORG (United Kingdom Stratospheric Ozone Review Group) (1999) Stratospheric Ozone: 1999, Seventh Report of the UK SORG. London: Department of the Environment, Transport and the Regions. UNEP (United Nations Environment Programme) (1998) Production and Consumption of Ozone Depleting Substances 1986–1996. Nairobi: The Ozone Secretariat to the Vienna Convention and Montreal Protocol. WMO (World Meteorological Organization) (1999) Scientific Assessment of Ozone Depletion: 1998, WMO Global Ozone Research and Monitoring Project Report No. 44, Geneva: WMO. Halogen Sources, Natural J H Butler, National Oceanic and Atmospheric Administration, Boulder, CO, USA Copyright 2003 Elsevier Science Ltd. All Rights Reserved. Introduction The depletion of stratospheric ozone (O3) has been driven by long-lived, anthropogenic halocarbons emitted into the atmosphere during the past few decades. When these gases, which in large part resist degradation in the troposphere, reach the stratosphere, their halogen atoms are released as free radicals. Here, the radicals accelerate the removal of ozone through a series of catalytic reactions. Because fluorine radicals are removed effectively as HF from the stratosphere and because iodinated compounds STRATOSPHERIC CHEMISTRY AND COMPOSITION / Halogen Sources, Natural 2163 react readily in the troposphere, persistent halocarbons containing chlorine and bromine are the main halogenated compounds implicated in the destruction of stratospheric ozone, and chlorine and bromine radicals are the primary halogens of concern. Not all halocarbons in the atmosphere are entirely anthropogenic, however. Although attention in atmospheric chemistry has centered on halocarbons resulting from human activities – the chlorofluorocarbons (CFCs), halons (CBrF3 , CBrClF2), chlorinated solvents (CH3CCl3 , CCl4 , CH2Cl2 , CHCl3), and their replacements, the hydrochlorofluorocarbons (HCFCs) – the methyl halides (methyl chloride (CH3Cl) and methyl bromide (CH3Br)) are present in significant amounts in the troposphere (Figure 1). Other halogenated methanes, such as CHBr3 , CHBr2 , and CH3I, can be locally high in atmospheric concentration, but their short tropospheric lifetimes significantly reduce their impact on stratospheric ozone. Nevertheless, halogen atoms from short-lived compounds do at times reach the stratosphere through deep convection of these compounds. The contribution of these gases of lower concentration to ozone depletion is unknown, although considered by most to be small. Of the naturally produced halocarbons, CH3Br and CH3Cl are the largest contributors to stratospheric ozone depletion, accounting for about one-quarter of the equivalent chlorine in the atmosphere (Figure 1). Methyl bromide is the single largest carrier of bromine to the stratosphere. Bromine, on a per-atom basis, is about 50 times more effective in depleting ozone than is chlorine. Although natural sources dominate the methyl bromide budget, there is a sizable anthro- pogenic flux to the atmosphere through its use as a fumigant. By multilateral international agreement, its industrial production is due to be phased out largely because of its high ozone-depletion potential. Methyl chloride, on the other hand, is the most abundant chlorine-containing compound in the atmosphere, contributing over 15% to the total tropospheric burden of organic chlorine. Its sources are believed to be largely natural and there is some evidence that it was present at over 90% of today’s levels during the early twentieth century. Both of these methyl halides have lifetimes of around a year, making them much shorter-lived than the CFCs, solvents, and halons currently banned by international agreement. Nevertheless, their large fluxes into the atmosphere mean that they reach the stratosphere, where they become involved in ozone depletion. Methyl Bromide Methyl bromide is present in the atmosphere at a mole fraction, or volume mixing ratio, of around 10 parts per trillion (1 ppt 5 10 12 moles of specific gas per mole of air) and its known sources include oceanic emission, biomass burning, agricultural application as a biocide, combustion of leaded gasoline, and disinfestation of buildings and structures. Until the 1990s, little attention had been paid to this gas in the atmosphere, in part because of its low mixing ratio and short atmospheric lifetime. In the early 1990s atmospheric methyl bromide was thought to emanate naturally from a large oceanic source and to be destroyed exclusively by reactions in Relative contributions to stratospheric ozone depletion CH3Cl(n) 12% CH3Br(a) 3% CH3Cl(a) 1% CFC-113 6% CFC-11 18% CH3Br(n) 11% Halons 9% HCFCs 3% CCl4 9% CH3CCl3 CFC-12 24% 4% Figure 1 Equivalent chlorine in ozone-depleting halogenated gases. Data are for 1999 concentrations in the troposphere. (n) and (a) signify estimates of natural and anthropogenic contributions. Halogens in the two methyl halides make up about one-quarter of the equivalent chlorine from persistent organic compounds in the atmosphere. Equivalent chlorine is the total number of chlorine atoms plus a weighting factor times the total number of bromine atoms in these compounds. 2164 STRATOSPHERIC CHEMISTRY AND COMPOSITION / Halogen Sources, Natural the atmosphere, predominantly with tropospheric hydroxyl (OH). Anthropogenic emissions, mainly from disinfestation of soils, commodities, and structures, were considered responsible for about 3 ppt of CH3Br in the atmosphere. Biomass burning and emissions from burning of leaded gasoline were thought to be possible contributors, but were not quantified at that time. Recognizing that there was a paucity of information on this important atmospheric gas, scientists began working to understand more completely its cycling and atmospheric budget. The results were surprising in a number of areas. The first of these surprises was that the ocean was not the large source it was thought to be, but rather a small net sink for atmospheric CH3Br. This net sink, however, results from rapid aquatic production and degradation working in opposition in the surface ocean, leaving it largely undersaturated. In some areas where production exceeds degradation, the ocean is supersaturated in methyl bromide, but in most of the surface ocean, most of the time, methyl bromide is undersaturated. Because the degradation rate of CH3Br is so high in most of the surface ocean, it had to be included as a significant component of the atmospheric lifetime computation. Subsequent calculations of the atmospheric lifetime of CH3Br yielded a rate that was almost equal to the loss due to reaction with OH in the troposphere. This alone lowered the atmospheric lifetime of CH3Br from around 2 years to 1 year. At about the same time, studies of the terrestrial environment revealed additional sinks and sources of atmospheric CH3Br. The discovery that CH3Br was degraded rapidly in a variety of soils, mainly by prokaryotic bacteria, lowered the estimates of atmospheric lifetime even further. Later, studies of isolated plant leaves and stems from over 100 species of plants demonstrated that the biosphere also was involved in the degradation of methyl bromide. Whether this loss to plants turns out to be a significant sink will depend upon further research. At present it appears to be small on a global basis. However, the few field studies of CH3Br fluxes between plants and plant ecosystems and the atmosphere reveal net emissions from the plants rather than net losses (Figure 2). These are each small but collectively significant in the global atmospheric budget of this gas (Figure 3). Our current understanding of atmospheric CH3Br is that of a gas with numerous, diverse sources and significant sinks on land, in the ocean, and in the atmosphere. Its lifetime, including atmospheric, oceanic, and soil sinks, is now computed at 0.7 years, but its calculated atmospheric budget is largely out of balance, with sinks outweighing sources by B40%. New findings continue to reveal previously unidenti- fied sources, which seem gradually to be closing the gap between calculated sources and sinks (Figure 2). Anthropogenic emissions of CH3Br are scheduled for phase-out by 2005 in developed countries and by 2015 in developing countries. However, the extent to which this will actually reduce the atmospheric burden of methyl bromide depends upon the actual atmospheric budget. Methyl Chloride Like methyl bromide, methyl chloride, at roughly 600 ppt in the atmosphere, received little attention until the past decade or so, as most research efforts were directed toward the rapidly increasing anthropogenic halocarbons. Methyl chloride also has a short atmospheric lifetime, B1 year, relative to the anthropogenic halocarbons and its anthropogenic sources are very small. Until recently, it was thought that most of the global emissions of CH3Cl came from the oceans. Although the oceans are still considered a major source of CH3Cl, new and more detailed studies show that the oceanic source is responsible for at most 15% of the methyl chloride in the atmosphere. Similarly, wood-rotting fungi contribute only a small amount and anthropogenic emissions of CH3Cl constitute only about of 1% of the total budget. Most of the known emissions of CH3Cl are accounted for by biomass burning, although there remains a sizable deficit in the overall budget. Yokouchi et al. (2000) recently noted that emissions from tropical plants could potentially make up this deficit. The identified sinks for methyl chloride, mainly loss via reaction with OH in the troposphere, suggest that, as for CH3Br, about half of the CH3Cl in the atmosphere is unaccounted for. Other Gases Most of the remaining naturally produced organic halogens are of low concentration and short lifetime. They are therefore thought to pose only a small threat to stratospheric ozone. However, they have been observed at the tropopause (Table 1) and they can be convected rapidly from the Earth’s surface into the upper troposphere and lower stratosphere. Of the purely chlorinated gases, chloroform (CHCl3) and perchloroethylene (C2Cl4) appear to have significant natural sources, although their budgets have been little studied. The naturally occurring brominated species (e.g., CHBr3 , CH2Br2 , CHBr2Cl), although low in concentration, are of some concern because of the efficiency of bromine in depleting stratospheric ozone. These gases are produced in the ocean and are STRATOSPHERIC CHEMISTRY AND COMPOSITION / Halogen Sources, Natural 2165 1.2 _ Annual CH3Cl flux (Tg year 1) 1 0.8 0.6 0.4 0.2 0 Oceans Bioburning Fungi Industrial (A) Salt ‘Missing’ marshes 70 _ Annual CH3CBr flux (Gg year 1) 60 50 40 30 20 10 (B) ‘Missing’ Fungi Rapeseed Salt marshes Bioburning Gasoline Fumigation Oceans 0 Figure 2 Identified sources of atmospheric methyl chloride (A) and methyl bromide (B). Budget deficits, calculated as the sum of all identified sinks minus the sum of all identified sources, are shown on the right. supersaturated throughout, by tens to hundreds of percent. Their fluxes from the ocean are large. Together, these lesser gases represent most of the total flux of organic bromine into the troposphere (Table 2). Closing the Budgets From recent research, it is clear that the missing sources of methyl bromide and methyl chloride are not oceanic. The saturations of these gases have now been mapped over most of the oceans. Although the surface concentrations of the two gases vary spatially and temporally to some degree, the ocean, for the most part, is undersaturated in methyl bromide, making it a net sink, not a net source of this gas. The ocean also is insufficiently supersaturated in methyl chloride to explain more than a small percentage of its total flux to the atmosphere. Further, although the atmospheric mixing ratios of both gases show marked seasonal 2166 STRATOSPHERIC CHEMISTRY AND COMPOSITION / Halogen Sources, Natural CH3Cl concentration (ppt) CH3Cl inferred history 1995 firn air 600 500 400 300 200 (A) 1880 1900 1920 1940 1960 1980 2000 1960 1980 2000 CH3Br concentration (ppt) CH3Br inferred history 1995 firn air (B) 10 8 6 4 1880 1900 1920 1940 Figure 3 Inferred atmospheric histories of (A) methyl chloride and (B) methyl bromide from measurements made in 1995 of air trapped in Antarctic snowpack. Modern-day measurements in the atmosphere are also shown on the plots. It is clear from these plots that both of these gases have significant natural sources, as both were present before the onset of large-scale agriculture and widespread use of industrial solvents and agricultural chemicals. Concentrations of both gases, however, have increased during the twentieth century, although the increase in methyl bromide is much larger. (Adapted from Butler et al. (1999).) cycles, the cycles in the Northern Hemisphere are amplified over those in the Southern Hemisphere, particularly for methyl bromide. Although tropospheric OH is responsible in part for the seasonal cycling, the uneven match between hemispheres, especially with smaller amplitude in the Southern Hemisphere, speaks for a more complicated involvement of sources and sinks. Because fluxes from the ocean to the atmosphere are retarded significantly at the ocean surface, it is not possible for cycles in the oceanic flux to drive seasonal variations in the atmosphere. This has led to several studies to determine whether methyl halides are released in significant amounts elsewhere, and it appears that they are. A number of investigations have shown that natural and cultivated terrestrial plants emit both of these gases, and others as well. The emissions seem to be related to the amount of Table 2 Potential global bromine fluxes Compound Table 1 Organic bromine in the troposphere Compound CH3Br CBrF3 CBrClF2 C2Br2F4 CH2Br2 CHBr2Cl CHBr3 CH2BrCl CHBrCl2 Totals Compound mole fraction  109 Bromine mole fraction  109 10 2.3 3.5 0.45 0.75–1.5 0–0.5 0.5–4 0–0.5 0–0.5 10 2.3 3.5 0.9 1.5–3 0–1 1.5–12 0–0.5 0–0.5 417.5 419.7 Purely anthropogenic compounds appear in bold type. Compounds that are natural or have significant natural components to their budgets (e.g., CH3Br) are shown in normal type. CH3Br CH3Br CH2Br2 CHClBr2 CHBr3 CH2BrCl CBrClF2 CBrF3 C2Br2F4 Total Source Anthropogenic Natural Ocean Ocean Ocean Ocean Anthropogenic Anthropogenic Anthropogenic Flux (Gmol Br year 1 ) 0.5 1.0 2.0 1.5 2.0 0.5 0.05 0.012 0.005 7.5 (6.0) Fluxes of naturally produced compounds are shown in bold type. Although these gases contribute only a small part of the bromine measured at the tropopause, their fluxes, mostly from the ocean, make up about half of the flux of organic bromine into the troposphere. STRATOSPHERIC CHEMISTRY AND COMPOSITION / Halogens halide in the soil. Coastal plants, such as those in salt marshes and in tropical environments, emit large quantities of methyl halides and, although their global area of coverage is small, they seem to contribute significantly to the global budget. To date, only a few plants and a few ecotomes have been studied for emission of methyl halides. Additional investigations are likely to locate more sources from the terrestrial biosphere, and with this a possible closing of the atmospheric budgets of these gases. Although the natural fluxes of these gases existed long before there were problems with stratospheric ozone depletion, this does not mean that they are not important. Contributions of bromine and chlorine from the anthropogenic gases are now declining and should continue to do so into the future; this should provide some relief, albeit slowly. If the fluxes of natural compounds remain the same and all countries abide by the Montreal Protocol and its amendments, then ‘pre-ozone-hole conditions’ could be reached by the mid-21st century. However, everything may not remain the same. A question that will become more pressing with global change is: How will the fluxes of methyl bromide, methyl chloride, and other halogenated gases between the Earth’s surface and atmosphere will change in the future? A change in sea surface temperature or soil temperature will certainly affect the fluxes, as will changes in precipitation or land use patterns. It is possible that such alterations of natural fluxes could offset or delay the timing of recovery, but we cannot know until we more fully understand the natural cycles of these ozonedepleting gases. 2167 See also Observations for Chemistry (In Situ): Gas Chromatography. Ozone: Ozone Depletion Potentials. Stratospheric Chemistry and Composition: Halogens. Further Reading Butler J and Rodrigues J (1996) Methyl bromide in the atmosphere. In: Bell C, Price N and Chakrabarti B (eds) The Methyl Bromide Issue, Vol. 1, pp. 27–90. John Wiley and Sons, Ltd. Butler JH, Battle M, Bender M, et al. (1999) A record of atmospheric halocarbon concentrations during the twentieth century from polar firn air. Nature 399: 749–755. Lobert JM, Butler JH, Montzka SA, et al. (1995) A net sink for atmospheric CH3Br in the East Pacific Ocean. Science 267: 1002–1005. Rhew RC, Miller BR, Vollmer MK and Weiss RF (2001) Shrubland fluxes of methyl bromide and methyl chloride. Journal of Geophysical Research – Atmospheres 106: 20875–20882. Schauffler SM, Atlas EL, Blake DR, et al. (1999) Distributions of brominated organic compounds in the troposphere and lower stratosphere. Journal of Geophysical Research 104(D17): 212, 513–521, 535. Shorter JH, Kolb CE, Crill PM, et al. (1995) Rapid degradation of atmospheric methyl bromide in soils. Nature 377: 717–719. Yokouchi Y, Machida T, Barrie LA, et al. (2000) Latitudinal distribution of atmospheric methyl bromide: measurements and modeling. Geophysical Research Letters 27: 697–700. Yvon-Lewis SA and Butler JH (1997) The potential effect of oceanic biological degradation on the lifetime of atmospheric CH3Br. Geophysical Research Letters 24(10): 1227–1230. Halogens D Toohey, University of Colorado, Boulder, Colorado, USA Copyright 2003 Elsevier Science Ltd. All Rights Reserved. Introduction The potential impact of halogen atoms (fluorine, chlorine, bromine, and iodine) on the chemistry of stratospheric ozone (O3) was first recognized in the early 1970s, not long after researchers proposed that nitrogen oxides (NOx ) and hydrogen oxides (HOx ) could destroy ozone. These halogen atoms are produced by compounds that are relatively unreactive in the troposphere but that decompose photochemically in the presence of short-wave ultraviolet radiation in the stratosphere. Among such compounds are those known as halocarbons, which are predominantly industrial in origin. For much of the second half of the twentieth century, a number of halocarbons were used for a variety of purposes, including refrigeration, manufacturing of foam products, extinguishing of fires, fumigation of crops, and production of polymers. Organisms in the upper ocean produce small, but significant, amounts of several halocarbons. There are only a few ways to destroy most halocarbons once they are released to the atmosphere, including reaction with hydroxyl (OH) (if the halocarbon contains a hydrogen atom), ultraviolet photolysis, and reaction with 2168 STRATOSPHERIC CHEMISTRY AND COMPOSITION / Halogens electronically excited oxygen atoms, O(1D). However, these processes also initiate the cycle of ozone destruction in the stratosphere. Halogen atoms are examples of free radicals, species that typically (although not exclusively) possess an odd number of electrons and require an additional electron to fill a molecular orbital to become more stable. Upon collision, a free radical can acquire this additional electron by stripping it from another molecule (called electron transfer), by pulling an atom off the collision partner (called extraction), or by attaching to another free radical (called addition or recombination). With a variety of collision partners available, there are literally hundreds of possible reactions and dozens of inorganic halogen compounds that must be considered for an accurate description of halogen chemistry in the stratosphere. However, only halogen atoms react rapidly with ozone. Thus, atmospheric chemists refer to two types of inorganic halogen species in the stratosphere, free radicals and reservoirs. Whereas the free radicals are directly involved in ozone destruction, the reservoirs are more stable compounds that do not react directly with ozone. However, the revervoirs can react with other free radicals or break down in sunlight to form free radicals, hence the origin of their name. Laboratory studies show that the reactivities of the families of halogen compounds follow the general trend I4Br4Cl4F; however, stratospheric abundances follow the trend [Cl]4[F]4[Br]4[I]. Consequently, most of the ozone destruction by halogens in the stratosphere is due to chlorine and bromine species. Destruction of ozone has been quantitatively linked to chlorine and bromine free radicals, whereas inorganic fluorine species have little impact on ozone. The role of iodine remains to be determined. Gas-Phase Halogen Photochemistry Organic source gases released at Earth’s surface are mixed throughout the lower atmosphere, a process that takes about a year. Once they reach the upper troposphere, these gases are slowly transported across the tropopause, primarily in the tropics. As air in the lower tropical stratosphere ascends, these compounds are broken down into their constituent atoms by shortwave ultraviolet radiation. The atoms react with ozone or with other species present to form inorganic compounds. Thus, organic halogens that contain halogen atoms and at least one carbon atom are called the source gases, whereas the inorganic halogens are those that contain only halogen, hydrogen, nitrogen, and oxygen atoms. If there is no selective separation (e.g., precipitation), the number-weighted sum of the mixing ratios of all forms of a particular halogen will be conserved. Thus, as the organic compounds break down in sunlight, the abundances of inorganic compounds increase concomitantly. Ultimately, the inorganic halogen compounds are removed from the stratosphere by slow downward transport into the upper troposphere at high latitudes. Because these compounds generally are acidic and water-soluble (unlike the organic source gases), they are readily scavenged in the relatively wet troposphere, returning to Earth’s surface with precipitation. Halogen atoms released in the stratosphere destroy ozone by a series of catalytic reactions, so called because the halogen free radicals are cycling between various forms with no net change in abundance while the ozone is converted into diatomic oxygen, O2. The main catalytic cycles for ozone destruction can be written symbolically (where X and Y represent F, Cl, Br, or I, hn represents a photon of wavelength c=n, and M is shorthand for a N2 or O2 molecule) as follows. Cycle one X þ O3 ! XO þ O2 ½I
O3 þ hn ! O þ O2 ½II
XO þ O ! X þ O2 ½III
X þ O3 ! XO þ O2 ½I
Cycle two Y þ O3 ! YO þ O2 XO þ YO ! X þ Y þ O2 ½IVa
for : XO þ YO ! XY þ O2 XY þ hn ! X þ Yg ½IVb
Cycle three X þ O3 ! XO þ O2 ½I
OH þ O3 ! HO2 þ O2 ½V
XO þ HO2 þ M ! HOX þ O2 þ M ½VI
HOX þ hn ! X þ OH ½VII
The net result of each of these cycles is loss of two ozone molecules with no change in radical abundance as in eqn [VIII]. O3 þ O3 ! 3O2 ½VIII
STRATOSPHERIC CHEMISTRY AND COMPOSITION / Halogens From the law of mass action, the rate of ozone destruction by a catalytic cycle can be written as the product of the concentrations of the reactants and the rate constant for the rate-determining step. Additional reactions (shown below) shift the steady-state balance between the two forms of the halogen radicals in favor of XO, such that the rate-determining steps are the final reactions in each of the three cycles above. Consequently, we can write the change of ozone versus time as d½O3
¼ 2ðkIII ½XO
½O
þ kIV ½XO
½YO
dt þ kVII ½XO
½HO2
þ minor cyclesÞ lished as the various chemicals cycle from one form to another. This cycling is shown symbolically in Figure 1. Because the rates of the analogous reactions vary among the different chemical families, the partitioning between the different chemical forms also varies. HF is the main form of inorganic fluorine; HCl and ClNO3 account for more than 90% of inorganic chlorine, except in the polar regions in winter; BrO, BrNO3 , and HOBr are the primary inorganic bromine species; and it is believed that IO, I, and HOI are the primary iodine species. ½1
The factor of 2 appears because two ozone molecules are destroyed for each pass through the cycle. If there were no other reactions to consider, a considerable amount of ozone would be destroyed before stratospheric air mixed back into the troposphere. However, the halogen radicals are deactivated by reactions with other species in the stratosphere. The main reactions of importance are shown in eqns [IX] to [XI]. X þ CH4 ! HX þ CH3 ½IX
X þ HO2 ! HX þ O2 ½X
XO þ NO2 þ M ! XNO3 þ M Heterogeneous Halogen Chemistry Early studies of stratospheric halogens focused primarily on reactions between gaseous species, or socalled gas-phase chemistry; however, a new class of reactions was necessary to explain the rapid appearance of the Antarctic ozone hole in the 1980s. These reactions, called heterogeneous because they involve the interactions of species in different phases (e.g., XNO3 h
NO ½XI
OH þ HX ! X þ H2 O ½XII
XNO3 þ hn ! X þ NO3 ½XIIIa
HX X XO H2O, CH4 HO2 O3 h
½IVc
OXO þ hn ! XO þ O ½XIV
Nitric oxide produced by photolysis of NO2 and the reaction O1NO2 reacts rapidly with halogen oxides to release halogen atoms, thereby strongly influencing the partitioning between the atomic and diatomic radical forms (eqn [XV]). ½XV
Except at very high solar zenith angles and at very low altitudes in the stratosphere, most of these reactions occur rapidly, and a steady state is estab- h
OH N2 O2 h
h
½XIIIb
XO þ YO ! OXO þ Y XO þ NO ! X þ NO2 NO2 h
O, YO OH There are also important reactions that re-release the radicals or that produce short-lived reservoirs, including those shown in eqns [XII] to [XIV]. ! XO þ NO2 2169 HOX HO2 YO XY XYO2 Tropopause Halocarbon ‘RAINOUT’ Figure 1 Schematic diagram of gas-phase halogen cycling in the Earth’s atmosphere. Open arrows are used for fast exchange between the radical forms of inorganic chlorine. Large dashed arrows represent transport across the tropopause. X and Y are halogen atoms, Cl, Br, I, or F. Processes that are underlined result in catalytic ozone loss. See text for further discussion. 2170 STRATOSPHERIC CHEMISTRY AND COMPOSITION / Halogens between gases and species dissolved in liquids or solids), are typically less rapid than gas-phase reactions because they require the additional processes of adsorption and dissolution. However, under dim sunlight, such as in the winter polar regions or near the bottom of the stratosphere, where most ultraviolet light has been removed by the column of ozone overhead, the rates of heterogeneous reactions can become competitive with those of gas-phase reactions. It is also in these regions that ozone production is slow, such that ozone-destruction cycles can have a large local impact. Extensive laboratory and modeling studies have shown that the following heterogeneous reactions of halogen species have the greatest impact on stratospheric chemistry (n denotes a reactant dissolved in liquid or solid phase). ClNO3 þ HCln ! Cl2 þ HNOn3 ½XVI
ClNO3 þ H2 On ! HOCl þ HNOn3 ½XVII
HOCl þ HCln ! Cl2 þ H2 On ½XVIII
BrNO3 þ H2 On ! HOBr þ HNOn3 HOBr þ HCln ! BrCl þ H2 On ½XIX
½XX
These reactions all serve to convert relatively longlived reservoirs of chlorine and bromine into species that photolyze readily in weak sunlight to release ozone-destroying radicals, and simultaneously convert short-lived reservoirs of NOx radicals into longerlived species. Because NOx limits the reactivities of the halogen compounds to ozone through reaction [XI] or reaction [XV] followed by reactions [IX] or [X], its removal results in further enhancements of the halogen oxides, and consequently more severe ozone loss. The reaction N2 O5 þ H2 On ! HNOn3 þ H2 On ½XXI
also indirectly enhances the abundances of halogen oxides by converting nitrogen oxides into the longlived reservoir nitric acid. Several of these heterogeneous reactions also influence the budget of HOx by either producing (e.g. [XVII] and [XIX]) or removing (e.g. [XVIII] and [XX]) its short-lived reservoirs. Many of these heterogeneous reactions depend strongly on temperature, and become important in the lower stratosphere only when temperatures drop below about 210 K. Because of this, and the interactions of the halogen radicals with NOx and HOx , the response of ozone to changes in temperature or changes in abundances of the halogen source gases can be quite complicated and sometimes counterintuitive. Therefore, detailed computer models are required for accurate assessments of the impact of halogen species on stratospheric ozone. Observations of Halogen Species Direct observations of inorganic halogen species form the basis for descriptions of present, and prediction of future, decreases of ozone in the stratosphere. Based on the rate-determining step, it is sufficient to measure the species that control ozone loss (that is, the halogen oxides) in order to compute the consequent rate of ozone destruction. However, to develop a more definitive understanding of the mechanisms controlling the abundances of the free radicals it is necessary to measure as many of the inorganic halogen species as is possible. Since the mid-1970s there have been many observations of a large number of inorganic halogen species in the stratosphere. By the year 2000, the following species had been quantified: fluorine family, HF, CF2O, CFClO; chlorine family, Cl, ClO, HCl, ClNO3 , HOCl, OClO, and Cl2O2; bromine family, BrO, HBr, and possibly HOBr; iodine family, IO. The remaining species are predicted to exist at abundances that represent significant challenges for current measurement techniques. Each of the families is examined separately below. Chlorine Emissions of industrially produced halocarbons such as CFC-11 (CFCl3) and CFC-12 (CF2Cl2) have delivered about 3–4 parts per billion (ppb) of chlorine to the stratosphere, more than all the other halogen families combined. Emissions by volcanoes and solid rocket motors can significantly enhance the local abundances of inorganic chlorine, but otherwise these sources have a small global impact following diffusion and mixing. Reactions [IX] and [X] proceed rapidly for chlorine, so that in the tropics and middle latitudes ClO rarely exceeds 20% of the inorganic chlorine budget. The remainder of the budget consists primarily of HCl and ClNO3 in roughly equal proportions, except at very high altitudes where HCl dominates. This partitioning is illustrated schematically in Figure 2A. At high latitudes in winter, where photolysis rates are slower and particles are larger and more abundant, heterogeneous reactions can activate all of the inorganic chlorine into short-lived reservoirs that rapidly produce radicals. Under these low-illumination conditions, reactions such as [IVa] and [IVb] proceed rapidly for weeks and months, destroying ozone at rates of a few percent per day. In a region where ozone STRATOSPHERIC CHEMISTRY AND COMPOSITION / Halogens 40 ClO 35 Altitude (km) HOCl HCl 30 ClNO3 25 20 15 10 Clorganic 5 0 0 (A) 1000 2000 3000 4000 5000 Cumulative abundance (ppt) 40 Br BrO HOBr Altitude (km) 30 BrNO3 20 Brorganic 10 0 0 (B) 5 10 15 Cumulative abundance (ppt) 20 2171 production is very slow owing to the lack of shortwavelength ultraviolet light necessary to break the O2 bond, significant losses of ozone occur. When solar illumination increases in springtime, ozone loss will cease if nitric acid is present to produce NO2 , which rapidly ties up ClO into ClNO3. However, if nitric acid is largely removed (as occurs over the Antarctic by sedimentation of cloud particles that contain nitric acid and water, called polar stratospheric clouds, or PSCs), ClNO3 and HCl form at rates that are far too slow to avoid complete destruction of ozone. In such regions measurements have identified a clear correlation of ozone loss with enhanced abundances of ClO. More typically, ozone production and loss are in closer balance, and only a gradual year-by-year erosion of ozone has been detected as abundances of the halogen species increase. Consequently, the impact of halogens on stratospheric ozone at mid-latitudes and in the Tropics is assessed by long-term monitoring of as many chemical species as is feasible; these observations are incorporated into detailed photochemical models of the stratosphere for interpretation. Such studies indicate that trends in industrial chlorine and bromine compounds can account for at least half of the downward trends in column ozone abundances that have been observed over the past several decades. A great deal of international cooperation was necessary to formulate regulations (e.g., the Montreal Protocol) that have only recently begun to impact on the abundances of ozone-destroying forms of chlorine in the stratosphere. 40 CF2O 35 HF Altitude (km) 30 25 20 15 CFClO Florganic 10 5 0 0 (C) 500 1000 1500 2000 2500 Cumulative abundance (ppt) Figure 2 Mid-latitude vertical profiles of the cumulative partitioning of (A) chlorine, (B) bromine, and (C) fluorine in the Earth’s atmosphere from the ground to 40 km. Mixing ratios are typical of values in the 1990s. The slight fall-off of total abundance with altitude in the middle to upper stratosphere reflects the lag time for air to reach these altitudes and the upward trends in source gas abundances in the 1990s. The partitioning of inorganic iodine is ignored because it is too uncertain at the present time. Bromine There are several sources of stratospheric inorganic bromine, including the halons and methyl bromide, the latter a compound that originates from both natural and industrial processes. Only two inorganic bromine species, BrO and HBr, have been accurately measured in the stratosphere and they represent 40– 60% and o10%, respectively, of the inorganic bromine budget. The abundances of the remaining species have been deduced by photochemical models and through the response of BrO to changing abundances of compounds with which it reacts. The partitioning of inorganic bromine is illustrated schematically in Figure 2B. It is believed that the sum of the abundances of all bromine-containing species in the stratosphere approaches 20 parts per trillion (ppt), which is about a factor of 20 smaller than the sum of all the chlorinecontaining species. However, the catalytic cycles involving bromine free radicals proceed much faster than their chlorine counterparts, and the percentage of bromine in free radical form is larger than that of 2172 STRATOSPHERIC CHEMISTRY AND COMPOSITION / Halogens chlorine. Therefore, ozone destruction due to bromine is almost as significant as that due to chlorine, especially in the winter polar regions and in the lowermost stratosphere. However, because the natural sources of methyl bromide are not well characterized and because there are still some important uncertainties in kinetic parameters, it is more difficult to attribute the anthropogenic contribution of brominecatalyzed ozone losses. Consequently, international regulations for the industrial bromine compounds have taken longer to formulate than those for the chlorine compounds. There have been some important developments in the study of atmospheric bromine based on new observations in the last decade of the twentieth century that call into question some assumptions about the stratospheric bromine budget. First, by deploying grab-samplers in the upper tropical troposphere, investigators have detected small, but significant, abundances of bromine-containing organic species that have fairly short lifetimes (days) in the troposphere. Presumably, these compounds are lofted to the upper troposphere by strong convective systems. Because abundances of these compounds vary widely at the surface, their contribution to the atmospheric bromine budget is hard to ascertain. These compounds have both natural and industrial origins, which further complicates assessments of anthropogenic ozone losses due to bromine. Second, observations have shown that large enhancements of bromine radicals occur sporadically near the Earth’s surface in the polar regions by an uncertain, but probably heterogeneous, mechanism. It is possible that a similar mechanism operating in the lowermost stratosphere could alter the present understanding of bromine photochemistry. Third, elemental bromine has been detected in particles near the tropopause, raising questions about the sources and sinks of atmospheric bromine. It is likely that the present understanding of bromine chemistry will change over the next decade with new and improved measurements. Iodine Of all the halogens in the stratosphere, least is known about the iodine family. Laboratory measurements indicate that any organic iodine that is transported to the lower stratosphere will very quickly oxidize and release constituent atoms, and that these atoms will be even more destructive to ozone than chlorine and bromine. However, tropospheric measurements of potential source compounds suggest that the abundance of iodine in the stratosphere is on the order of 1 ppt or less, such that the iodine radicals will be at least an order of magnitude or smaller in abundance than BrO. Initial attempts to observe IO in the stratosphere have had mixed results, but generally indicate that there is no more than about 1 ppt of IO. However, little is known about the reactions that partition iodine into its various forms. It is likely that abundances of this species are highly variable. There have been no attempts to measure other gasphase inorganic iodine species whose concentrations are well below the detection limits of most modern instruments. However, the recent detection of elemental iodine in particles near the tropopause, an observation that is similar to the detection of elemental bromine, raises additional questions about the processes that control abundances of iodine in the atmosphere. Even at low abundances, iodine could play an important role in ozone destruction in the lowermost stratosphere through its synergistic interactions with the bromine and chlorine free radicals. It has also recently been proposed that iodine could destroy ozone by reactions that involve OID. Clearly, there is much to be learned about the role of iodine in the lower stratosphere, and this halogen is likely to be the focus of vigorous scientific study in the early part of the twenty-first century. Fluorine Inorganic fluorine is produced by the photodecomposition of fluorocarbons, predominantly the chlorofluorocarbons (CFCs) CFCl3 and CF2Cl2. Measurements show that HF and the photodecomposition intermediates CF2O and CFClO can account for the entire inorganic fluorine budget, in agreement with models that incorporate laboratory measurements of fluorine reactions. Fluorine atoms react rapidly with hydrogen-containing species, especially H2O and CH4 that are present at parts-per-million (ppm) abundances in the stratosphere. In addition, there are no known ways to release fluorine atoms from HF, a very thermodynamically stable species. Consequently, immeasureably small abundances of fluorine radicals are present as extremely short-lived intermediates in the photodecomposition of fluorocarbons, and their contribution to ozone loss is negligibly small. Therefore, the primary role of fluorine in stratospheric chemistry is as a marker or tracer for other halogen species, in particular the CFCs. Ground-based and satellite measurements have shown that the rate of increase of HF in the stratosphere can be explained entirely by the buildup of chlorofluorocarbons in the troposphere followed by gradual transport into the stratosphere where they photodecompose in the presence of UVand chemical oxidants (primarily OH and O(1D)). The STRATOSPHERIC CHEMISTRY AND COMPOSITION / Halogens partitioning of inorganic fluorine is illustrated schematically in Figure 2C. 1500 2173 3000 O3 1250 2500 1000 2000 750 1500 500 1000 250 O3 (ppbv) 500 ClO 0 0 61° S 63° S 65° S 67° S 69° S 71° S 73° S Latitude 2500 2000 50 Ozone 40 30 1500 20 1000 10 500 ClO mixing ratio (ppb) (A) Ozone mixing ratio (ppb) There are three distinct regimes in which ozone losses attributable to halogens have been detected. These are the middle stratosphere year-round, in the Antarctic and Arctic polar vortices in springtime, and in rocket plumes within hours following launches. In all cases, two conditions are met that establish the link between halogens and ozone loss. First, the regions where ozone losses are detected are correlated with larger abundances of halogen radicals than in adjacent regions where there is less or no ozone loss. Second, the rates at which ozone losses occur are equivalent (within measurement uncertainties) to the corresponding rates predicted with models that include laboratory measurements of the rate constants for the rate-determining reactions. Examples of the anticorrelation between abundances of chlorine oxide and ozone are shown in Figure 3. The sources of enhancements in the chlorine radicals differ (direct local injection in the case of the rocket and heterogeneous reactions of HCl and ClNO3 in the case of the Antarctic ozone hole). Consequently, the time scales for ozone loss in these two cases are dramatically different, less than one hour for the rocket plume and a month for the Antarctic ozone hole; however, in both cases the amounts of ozone destroyed over these periods are consistent with the known kinetics of the halogen radicals to within the uncertainties of the measurements. In the two cases presented in Figure 3, the ozone destruction rates vastly exceeded the rates at which ozone could be produced. Thus, regions of low ozone formed adjacent to regions of higher ozone where the abundances of the halogen free radicals were significantly lower. In the middle stratosphere, the situation is quite different, and ozone production and loss rates nearly match (that is, ozone is in a photochemical steady state). In addition, the spatial variability of the halogen radicals is small. Consequently, it is difficult to attribute an instantaneous ozone value to a particular abundance of halogen radicals. Rather, it is by correlating the long-term downward trend of ozone abundances with concomitant increases in halogen radical abundances that the link is deduced. Observations for the last twenty years of the twentieth century showed a decrease in ozone of approximately 10–15% between 35 and 50 km, an amount that agrees well with the decrease predicted as a result of the steady rise in abundances of chlorine, the primary agent of halogen-induced ozone loss in the middle stratosphere. ClO (pptv) Halogens and Ozone Loss ClO 0 0 77 380 (B) 77 390 77 400 77 410 77 420 UT (s) Figure 3 Examples of the measured correlation of ozone loss with ClO radical abundances in (A) the Antarctic ozone hole in 1987 (16 September) (excerpted with permission from Anderson JG et al. (1991) Science 251: 39–46. Copyright 1991 American Association for the Advancement of Science) and (B) in a plume of a Delta II rocket (adapted from Ross MN, et al. (2000)). Long-term reductions in ozone have been reported for other regions of the stratosphere, in particular the Arctic and the lowermost stratosphere at middle latitudes. In the first case, springtime ozone losses are consistent with calculations based on observed abundances of the radicals ClO and BrO. However, losses in the middle of winter appear to be significantly greater than expected, for reasons that are not yet clear. It is possible that transport between regions of differing ozone concentrations confounds efforts to attribute ozone loss to specific halogen radical abundances. In the mid-latitude lower stratosphere, the ozone losses themselves are highly uncertain, because they occur in a region where there is a strong vertical 2174 STRATOSPHERIC CHEMISTRY AND COMPOSITION / HOx gradient in the ozone mixing ratio and because measurement techniques are not optimized for these altitudes. Nevertheless, there have been several attempts to attribute these trends to increases in the abundances of anthropogenic chlorine and bromine, accelerated by naturally occurring iodine radicals. This is a region of the stratosphere where temperatures are very low (B200 K) and heterogeneous reactions on sulfuric acid aerosols or thin cirrus clouds could enhance halogen radical abundances. New measurements at the turn of the century suggest that chlorine radicals are not significantly enhanced in this region, but bromine and iodine radicals may be. Whether or not these enhancements are sufficient to explain the unexpectedly large trends deduced in the lower stratosphere will be a focus of much attention in the coming years. See also Chemistry of the Atmosphere: Chemical Kinetics. Numerical Models: Chemistry Models. Observations for Chemistry (In Situ): Particles. Observations for Chemistry (Remote Sensing): Lidar; Microwave. Ozone: Ozone Depletion Potentials; Photochemistry of Ozone. Stratospheric Chemistry and Composition: HOx; Halogen Sources, Anthropogenic; Halogen Sources, Natural. Further Reading Anderson JG, Toohey DW and Brune WH (1991) Free radicals within the Antarctic vortex: the role of CFCs in Antarctic ozone loss. Science 251: 39–46. Brune WH (1998) Stratospheric chemistry – perspectives in environmental chemistry. In: Macalady DL (ed.) Perspectives in Environmental Chemistry, pp. 292–324. Oxford: Oxford University Press. Finlayson-Pitts BJ and Pitts JN (2000) Chemistry of t he Upper and Lower Atmosphere. London: Academic Press. Roan SL (1989) Ozone Crisis: The 15-Year Evolution of a Sudden Global Emergency. New York: Wiley. Ross MN, et al. (2000) Observation of stratospheric ozone depletion associated with Delta II rocket emissions. Geophysical Research Letters 27: 2209–2212. Wayne RP, et al. (1995) Halogen oxides – radicals, sources and reservoirs in the laboratory and in the atmosphere. Atmospheric Environment 29: 2677–2881. Wayne RP (2000) Chemistry of Atmospheres, 3rd ed. Oxford: Oxford University Press. HOx T F Hanisco, Harvard University, Cambridge, MA, USA HOx Sources Copyright 2003 Elsevier Science Ltd. All Rights Reserved. HOx is produced from the oxidation of the stable hydrogen-containing species, water, methane, and molecular hydrogen. The relative strength of these sources is largely determined by their concentrations at the tropopause: H2O B4 parts per million (ppm), CH4 B1.5 ppm, and H2 B0.5 ppm. The numerous pathways that participate in this oxidation are diagrammed in Figure 1. The oxidation occurs mostly through gas phase reactions with highly reactive species, i.e. excited oxygen atoms, chlorine atoms, and OH. The oxidation of water also occurs via hydrolysis reactions catalyzed by acid aerosols found in the lower stratosphere. In most cases, the production mechanisms include a large number of reactions, most of which have little direct effect on HOx . These mechanisms are presented in terms of reaction sequences that can be simplified in terms of ratedetermining reactions and net yields. Introduction The hydrogen radical family (HOx ) consists of the hydrogen (H), hydroxyl (OH), and hydroperoxyl (HO2) radicals. Concentrations of these highly reactive radicals are small, between 1 part per trillion (ppt) in the lower stratosphere and 400 ppt in the upper stratosphere. Despite this, HOx is important because it participates in many reactions that control the photochemistry of stratospheric ozone. The hydrogen radicals are important in the removal of O3 through direct reaction with O3 and indirectly through reaction with the halogen oxides. HOx also removes ozone through reactions with the nitrogen and halogen chemical families. The extensive coupling of HOx to these chemical families leads to a particularly complex set of reactions that control HOx photochemistry. Understanding these mechanisms is important in understanding both HOx and, more broadly, the mechanisms that control ozone photochemistry. Gas Phase Processes The largest single source of HOx throughout the stratosphere is the oxidation of H2O by the electronically STRATOSPHERIC CHEMISTRY AND COMPOSITION / HOx 2175 HNO3 40 OH NO2 HNO4 OH CH4 35 H2 O(1D) CH4 O(1D) OH Cl H2O O(1D) O3, ClO, O OH HO2 OH H2O O3, NO, ClO, BrO, O Altitude (km) NO2 30 Hydrolysis 25 H2O 20 ClONO2 HOCl BrONO2 HOBr N2O5 HNO3 h
Total h
15 h
10−3 10−2 10−1 100 OH −1 HOx production rate (ppt s ) Figure 1 Primary sources and sinks of HOx are shown. Gas phase reactions are denoted with solid lines and heterogeneous reactions with dashed lines. excited oxygen atom Oð1 DÞ which is generated from the photolysis of O3 at wavelengths less than B330 nm. The production from this mechanism is represented by a sequence of reactions that leads to a net conversion of H2O to HOx : Figure 2 Typical mid-latitude daytime production rates of HOx calculated from concentration profiles of the source gases are shown versus altitude. The CH4 oxidation sequences initiated by Oð1 DÞ, OH, and Cl are grouped together. Likewise, the hydrolysis reactions involving N2O5 , ClONO2 , and BrONO2 are combined. The production from H2 is less than 4% of the total at all altitudes. NO2 þ hn ! NO þ O ½6
O3 þ hn ! Oð1 DÞ þ O2 ½1
O þ O2 ! O3 ½7
Oð1 DÞ þ H2 O ! 2 OH ½2
CH3 O þ O2 ! CH2 O þ HO2 ½8
CH2 O þ hn ! H þ CHO ½9
CHO þ O2 ! HO2 þ CO ½10
Net: O3 þ H2 O ! 2OH þ O2 Only a small fraction of the Oð1 DÞ produced from reaction [1] subsequently reacts via reaction [2]; most Oð1 DÞ relaxes to ground state Oð3 PÞ after collisions with N2 and O2. The rate of this sequence is determined by the slowest or the ‘rate-determining step’, in this case reaction [2]. The rate of the production of HOx from H2O, which is equal to twice the rate of reaction [2], is shown in Figure 2. This and the following sequences are identified in the figures by the rate-determining steps. The oxidation of methane requires a greater number of reactions to liberate all four hydrogen atoms. The sequence initiated by Oð1 DÞ can produce as many as four HOx : O3 þ hn ! Oð1 DÞ þ O2 ½1
Oð1 DÞ þ CH4 ! CH3 þ OH ½3
CH3 þ O2 ! CH3 O2 ½4
CH3 O2 þ NO ! CH3 O þ NO2 ½5
Net: CH4 þ 3O2 ! H þ OH þ 2HO2 þ CO An alternate pathway for the photolysis of CH2O in reaction [9] is the production of H2 and CO: CH2 O þ hn ! H2 þ CO ½11
When the oxidation of methane follows this path, the net yield is only 2HOx for every CH4 consumed. Since reactions [9] and [11] compete, the yield of HOx from CH4 oxidation initiated by reaction [3] is somewhere between 2 and 4. The utility of the rate-determining step and net yields is particularly evident in this sequence, where nine reactions can be thought of in terms of one. In this case, the rate of production of HOx from CH4 is 2–4 times the rate of the ratedetermining step, reaction [3]. The rate of this and the following CH4 oxidation sequences decrease relative to the H2O source at higher altitudes. This results from the net conversion of CH4 to H2O as the air in the lower stratosphere ages during the slow ascent into the mid to upper stratosphere. 2176 STRATOSPHERIC CHEMISTRY AND COMPOSITION / HOx The methane oxidation sequence that is initiated by OH is autocatalytic. OH is consumed at the initiation step, OH þ CH4 ! CH3 þ H2 O ½12
but the subsequent reactions [4]–[10] can produce as many as 2HOx . In the lower stratosphere where the oxidation of CH4 is most important, the rate of reaction [12] is faster than that of reaction [3], so that the sequence initiated by OH is often more important than that initiated by Oð1 DÞ. Oxidation of CH4 is also initiated by reactions of CH4 with Cl. lower stratosphere: M NO2 þ NO3 ! N2 O5 Aerosol ½17
N2 O5 þ H2 O ! 2 HNO3 ½18
2ðHNO3 þ hn ! OH þ NO2 Þ ½19
NO2 þ O3 ! NO3 þ O2 ½20
Net: H2 O þ O3 ! 2OH þ O2 Outside the winter polar vortex this sequence is not a large source of HOx , because Cl concentrations are lower and the reaction The first step in this sequence is significant only at night because NO3 is easily photolyzed during the day. Thus, this sequence is not important in the highlatitude summer when continuous sunlit conditions occur. An alternate pathway for the removal of HNO3 in reaction [19] is reaction with OH, discussed in the next section. This pathway removes HOx so that the yield from the hydrolysis of N2O5 is less than 2HOx . In the winter and early spring polar vortex when ClO and aerosol concentrations are high, heterogeneous reactions of ClONO2 are the dominant source of HNO3. The hydrolysis of ClONO2 can be a particularly strong source of HOx : OH þ HCl ! H2 O þ Cl ClO þ NO2 ! ClONO2 Cl þ CH4 ! CH3 þ HCl ½13
The methane oxidation sequence that is initiated by Cl can be the most important source of HOx in the winter polar vortex when active chlorine levels are high and HCl concentrations are low. In these conditions, ClO controls the conversion of CH3 O2 ! CH3 O: CH3 O2 þ ClO ! CH3 O þ Cl þ O2 ½14
½15
is important, resulting in a lower yield for HOx production. In these conditions [13] and [15] are the dominant production and loss terms of HCl. The production of HOx from the oxidation of H2 is analogous to that of H2O: Aerosol ClONO2 þ H2 O ! HOCl þ HNO3 ½21
½22
HOCl þ hn ! OH þ Cl ½23
O3 þ hn ! Oð1 DÞ þ O2 ½1
Cl þ O3 ! ClO þ O2 ½24
Oð1 DÞ þ H2 ! H þ OH ½16
HNO3 þ hn ! OH þ NO2 ½19
Net: O3 þ H2 ! H þ OH þ O2 This source is much smaller than the H2O and CH4 sources, owing to the much lower concentration of H2 in the stratosphere and to a smaller rate constant. In the lower stratosphere where this sequence is most significant, H2 accounts for only 4% of the total production of HOx . Heterogeneous Processes Heterogeneous reactions are important in the partitioning of the nitrogen and halogen families. These same reactions are important sources of HOx in the lower stratosphere where aerosol concentrations are significant. The hydrolysis of N2O5 on sulfuric acid aerosols is a major source of HNO3 and HOx in the Net: H2 O þ O3 ! 2OH þ O2 The heterogeneous reaction [22] is strongly temperature-dependent, proceeding fastest at low temperatures (190–200 K). As in the case of N2O5 hydrolysis, the reaction of OH with HNO3 competes with [19] to reduce the yield of this sequence. In addition, the heterogeneous reaction: Aerosol HOCl þ HCl ! H2 O þ Cl2 ½25
competes with reaction [23] and to further reduce the yield of this sequence. When ClONO2 is absent, as in the polar winter lower stratosphere, reaction [25] can be a net sink of HOx . The analogous hydrolysis of BrONO2 is only weakly temperature-dependent, and it is only a small source throughout the lower STRATOSPHERIC CHEMISTRY AND COMPOSITION / HOx 2177 stratosphere. This reaction produces a distinct increase in HOx when HOBr produced is photolyzed at daybreak. 40 35 Sinks The removal of HOx in the stratosphere consists entirely of reactions that convert HOx to H2O. The numerous pathways involving NOy and HOx intermediates are shown in Figure 1. Unlike the nitrogen and halogen radical families, HOx radicals are not sequestered in reservoirs. This is because these potential reservoirs (HNO3 and HNO4) react rapidly with OH, serving as sinks instead. The primary removal mechanism for HOx in the lower and middle stratosphere is the reaction of OH with NO2 , followed by reaction with HNO3: M OH þ NO2 ! HNO3 ½26
OH þ HNO3 ! H2 O þ NO3 ½27
NO3 þ hn ! NO2 þ O ½28
Altitude (km) HNO3 30 25 HNO4 HO2 Hydrolysis Total 20 15 10−9 10−8 10−7 First-order HOx loss rate (s−1) Figure 3 Typical mid-latitude first-order loss rates determined from profiles of the sinks of HOx are shown versus altitude. The first-order loss rates, e.g. kOHþHNO3 ½HNO3
, are the independent variables that control the removal of HOx . Concentrations of HOx are determined nearly entirely by production from H2O and removal via the selfreaction. The most significant heterogeneous removal mechanism of HOx is the hydrolysis of N2O5: M Net: 2OH ! H2 O þ O NO2 þ NO3 ! N2 O5 The similar sequence that produces and removes HNO4 also removes HOx : N2 O5 þ H2 O ! 2HNO3 ½18
2ðOH þ HNO3 ! H2 O þ NO3 Þ ½27
NO3 þ hn ! NO2 þ O ½28
M HO2 þ NO2 ! HNO4 ½29
OH þ HNO4 ! H2 O þ NO2 þ O2 ½30
Net: OH þ HO2 ! H2 O þ O2 Net: 2OH ! H2 O þ O The HNO3 and HNO4 sequences are most important in the lower stratosphere where the concentrations of HNO3 and HNO4 are high. The first-order loss rates of the removal sequences are shown in Figure 3. When concentrations of ClO are elevated, the reaction OH þ ClO ! O2 þ HCl ½31
is a significant sink of HOx . Because concentrations of HCl are usually low when ClO concentrations are high, the reaction of OH with HCl [15] that completes the conversion of HOx to H2O is not significant. When the concentrations of HOx are large the selfreaction of HOx becomes significant: OH þ HO2 ! H2 O þ O2 Aerosol ½17
½32
In the mid to upper stratosphere reaction [32] is the dominant sink of HOx . In this region of the stratosphere HOx photochemistry is greatly simplified. This reaction is important only in the lower stratosphere where aerosol concentrations are high. As mentioned in the previous section, an alternate pathway to reaction [27] is the photolysis of HNO3 [19], so that the occurrence of [18] leads to the removal of less than 2HOx . The analogous removal mechanisms involving ClONO2 and BrONO2 are less significant because the reactions of OH with HOCl and HOBr are too slow to compete with the photolysis of these species. Secondary Sources and Sinks The production and loss mechanisms shown in Figure 1 are portrayed as part of a one-way flux from the primary sources (H2 , H2O, and CH4) through HOx and back into H2O. On average this flux is balanced, but there are certain situations when 2178 STRATOSPHERIC CHEMISTRY AND COMPOSITION / HOx other pathways are significant. These processes are part of null sequences. For example, the formation followed by photolytic destruction of HNO4 is ½29
HNO4 þ hn ! OH þ NO3 ½33
Net: H2 O þ NO2 ! OH þ NO3 This is a net null for HOx averaged over a 24-hour diurnal cycle. However at any given instant within the diurnal cycle these reactions are not required to balance. In particular, at sunrise and sunset HNO4 is not in steady state, i.e. the rate of [29] does not balance [30] and [33]. Under these twilight conditions, this sequence can be a source (sunrise) or a sink (sunset). Other null cycles that can be instantaneous sources or sinks of HOx include the formation and photolysis of HONO, HOCl, HOBr, and H2O2. Conceptually, these terms should be considered secondary processes. That is, they do not influence the interpretation of the 24-hour average abundance of HOx versus altitude or latitude, but they are significant in model calculations that attempt to reproduce HOx at twilight conditions. Diurnal Change The production of HOx is tied to the flux of ultraviolet (UV) radiation that drives the photolysis reactions and initiates the production sequences. This UV flux, which is strongly attenuated by O3 , depends on the O3 slant column (the amount of O3 between an air parcel and the Sun). During a diurnal cycle, this slant column changes with the angle of the Sun, resulting in large changes in UV flux and photolysis rates. The resulting change in the production rate of HOx for conditions typical of the mid-latitude lower stratosphere is shown in Figure 4. The sharp increase in the production rate indicates the strong dependence on UV flux, hence solar zenith angle. The relative strengths of the sources and sinks of HOx also depend on solar flux. For example, the oxidation of CH4 is faster than that of H2O at the highest solar zenith angles. This is because the oxidation of CH4 can be initiated by OH and Cl atoms that are produced more easily than Oð1 DÞ at twilight conditions. HOx Cycling The relative concentration of H, OH, and HO2 is controlled by fast cycling reactions that do not produce or remove HOx . H is converted to HO2 via 50 20 50 95 0.02 HOx production rate (ppt s−1) M HO2 þ NO2 ! HNO4 Solar zenith angle (degrees) 95 Total 0.01 H2O CH4 Hydrolysis 0.00 4 8 12 16 20 Local time (hours) Figure 4 The production rate of HOx is shown for a typical day in the mid-latitude lower stratosphere. The solar zenith angle is the angle between the Sun and the zenith (directly overhead). The sequences that oxidize CH4 and the hydrolysis reactions are grouped together. the extremely fast reaction with O2. H þ O2 ! HO2 ½34
This reaction is significantly faster than the reactions that produce H (i.e. [9] and [16]), so that concentrations of H are negligible compared to the OH and HO2. Because the conversion of H ! HO2 is so fast, H is often neglected and the production of H is considered equivalent to the production of HO2. The relative concentration of OH and HO2 is controlled by reactions that interconvert OH and HO2. The primary conversion mechanism in the lower and middle stratosphere is OH þ O3 ! HO2 þ O2 ½35
HO2 þ O3 ! OH þ 2O2 ½36
Net: 2O3 ! 3O2 This reaction sequence is a net loss for ozone, and accounts for a large fraction of the total ozone removal rate in the lower stratosphere. In this sequence, the rate constant for reaction [35] is much larger than that of reaction [36], so that concentrations of HO2 are almost always greater than OH in the lower to mid stratosphere. An important pathway for the conversion of HO2 ! OH is part of a null cycle: OH þ O3 ! HO2 þ O2 ½35
HO2 þ NO ! OH þ NO2 ½37
STRATOSPHERIC CHEMISTRY AND COMPOSITION / HOx 2179 NO2 þ hn ! NO þ O ½38
Cl þ O3 ! ClO þ O2 ½24
O þ O2 ! O3 ½39
HO2 þ ClO ! HOCl þ O2 ½41
HOCl þ hn ! OH þ Cl ½23
Cl þ O3 ! ClO þ O2 ½24
Net: null This and the prior sequence illustrate the interaction between HOx and NOx on ozone loss rates. Reaction [36] removes O3 and reaction [37] leads to the production of O3. The relative rate of [36] compared with [37] determines the net O3 removal rate following reaction [35]. The relative abundance of OH and HO2 is proportional to the concentrations of the species and rate constants that interconvert OH and HO2. In the midlatitude lower to mid stratosphere, where reactions [35]–[37] dominate the interconversion rate, the ratio HO2 is controlled by O3 , NO, and the rate constants for these reactions. Figure 5 shows how the ratio HO2/ OH responds to the changes in O3 and NO. At a fixed amount of NO, an increase in O3 leads to greater HO2/ OH because the rate constant for reaction [35] is roughly 15 times faster than that of reaction [36]. At the limit of very high concentrations of O3 (or low NO) the ratio of HO2/OH is determined by the ratio of the rate constants for reactions [35] and [36], i.e. HO2 =OH  k35 =k36 . For a fixed amount of O3 , increases in NO lead to decreases in HO2 because reaction [37] converts HO2 ! OH. At some limit of very high NO (or low O3), HO2/OH would approach zero. When the concentrations of ClO are high, such as the wintertime polar vortex, the reaction with ClO controls the balance between OH and HO2: Above B40 km this sequence dominates the conversion of OH and HO2. The rate constant of reaction [43] is roughly twice as large as that for reaction [42], so that concentrations of OH are greater than [HO2] in the highest part of the stratosphere. The OH-initiated oxidation of CH4 , reactions [12] followed by [4]–[10], converts OH to HO2. In addition, the subsequent oxidation of CO converts OH ! H, via OH þ ClO ! HO2 þ Cl OH þ CO ! H þ CO2 OH þ O ! HO2 ½42
HO2 þ O ! OH þ O2 ½43
Net: 2O ! O2 ½44
These reactions are important only in the lowest part of the stratosphere near the tropopause region, where they might account for a few percent of the total OH ! HO2 conversion rate. 15 HO2 /OH The analogous sequence involving BrO also contributes to the interconversion of OH and HO2 and to the removal of O3. These sequences are almost always rate limited by reaction [41] and the analogous reaction of HO2 with BrO. In the upper stratosphere, the reactions with O atoms become important: ½40
20 Nomenclature 10 O3 = 2.5 ppm 5 O3 = 0.5 ppm 0 Net: 2O3 ! 3O2 0.0 0.5 1.0 1.5 2.0 NO mixing ratio (ppb) Figure 5 The ratio of HO2/OH is controlled by NO and O3 in the mid-latitude lower stratosphere. The predicted ratio is shown versus NO for O3 ¼ 0:5 and 2.5 ppm. Concentration: in parts per million (ppm 5 10  6) and parts per trillion (ppt 5 10  12) Production rate: in ppt s  1 Nanometer: 1 nm 5 10  9 m Collision partner: M 5 O2 and N2 Photon: hv See also Aerosols: Physics and Chemistry of Aerosols. Chemistry of the Atmosphere: Chemical Kinetics; Principles of 2180 STRATOSPHERIC CHEMISTRY AND COMPOSITION / Hydrogen Budget Chemical Change. Ozone: Photochemistry of Ozone. Stratospheric Chemistry and Composition: Halogens; Hydrogen Budget; Reactive Nitrogen (NOx and NOy). Further Reading Dessler AE (2000) The Chemistry and Physics of Stratospheric Ozone. San Diego: Academic Press. Jacob DJ (1999) Introduction to Atmospheric Chemistry. Princeton: Princeton University Press. Johnston HS and Podolske JR (1978) Interpretations of Stratospheric Photochemistry. Review of Geophysics and Space Physics 16: 491–519. McElroy MB, Salawitch RJ and Minschwaner K (1992) The changing stratosphere. Planetary and Space Science 40: 373–401. Okabe H (1978) Photochemistry of Small Molecules. New York: Wiley. Wayne RP (2000) Chemistry of Atmospheres: An Introduction to the Chemistry of the Atmospheres of Earth, the Planets, and Their Satellites. 3rd edn. Oxford: Oxford University Press. Hydrogen Budget J E Harries, Imperial College of Science, Technology and Medicine, London, UK Copyright 2003 Elsevier Science Ltd. All Rights Reserved. Introduction The principal components of the hydrogen budget of the stratosphere are water vapor, H2O (mixing ratio between 3 and 4 parts per million by volume, ppmv, in the lower stratosphere), methane, CH4 (about 1.7 ppmv), and molecular hydrogen, H2 (about 0.5 ppmv). The first two of these are greenhouse gases, with strong absorption bands in the infrared, making their concentration and evolution of interest in studies of the balance of climate, and how this might change with time. The hydrogen free radicals such as the hydroxyl radical OH, HO2 , and atomic hydrogen, H, are present in much lower concentrations (mixing ratios of order 10  11 at 30 km), but are highly reactive, and are consequently of particular importance in atmospheric chemistry. Since this article is concerned with the budget of hydrogen in the stratosphere, we shall be concerned primarily with the three species that dominate the mass: the trace radicals will only be considered in so far as they enter into reactions which determine the concentrations of the three major constituents. Because of the importance to radiative energy exchange and chemistry, a wide range of studies of both the major and minor hydrogen-bearing species have been undertaken over the years. In what follows, we shall review the state of knowledge on the concentrations, budget, and variability of the three principal hydrogen species H2O, CH4 , and H2. Supply of Hydrogen Species to the Stratosphere Water vapor is continuously supplied to the stratosphere from the ocean surface, via the troposphere. While concentrations in the troposphere are high, sometimes approaching saturation near the surface, the Brewer–Dobson circulation transports air upward over the tropics, and through the tropical tropopause, which is characteristically very cold and high. This causes the ‘freeze-out’ of water by the tropopause cold trap, with the consequence that the air moving into the stratosphere is extremely dry (mixing ratios of order a few parts in 10  6 (see Stratospheric Water Vapor). In the stratosphere, oxidation of both methane and H2 takes place, adding to the concentration of water vapor, but simultaneously H2 can be produced by oxidation of formaldehyde, CH2O, which itself is derived from CH4. The overall net effect is to cause water vapor to increase with height, mainly at the expense of methane, which decreases in mixing ratio, while molecular hydrogen stays roughly constant in mixing ratio with height (since reactions which both add and delete H2 occur with similar rates). Later we will examine some data for H2O and CH4 from both a satellite and a balloon experiment, which indicate that the sum of total hydrogen, which may be expressed as c ¼ 2CH4 þ H2 O þ H2 , is, as far as can be measured, constant with height through most of the stratosphere. Loss of Hydrogen Species from the Stratosphere There are three principal mechanisms by which hydrogen species are lost from the stratosphere: 1. by being part of the overall global circulation of descending air at mid to high latitudes; 2. by loss, particularly of the lightest component, molecular hydrogen, from the top of the the atmosphere to space; 3. by removal of ice where condensation occurs and ice particles may be removed by movement into the troposphere and subsequent ice sublimation or ice STRATOSPHERIC CHEMISTRY AND COMPOSITION / Hydrogen Budget melting and liquid evaporation: this can happen, for example, at the top of cumulonimbus clouds once they begin to decay, or in the Antarctic polar vortex, where extremely low temperatures and sinking motion prevail. Photochemistry of Stratospheric Hydrogen Since original work in 1950 by two major figures in atmospheric chemistry and aeronomy, David Bates and Marcel Nicolet, the photochemistry of the hydrogen family has been treated many times. A valuable treatment is to be found in the book by Brasseur and Solomon (see under Further Reading). The three principal components, H2O, CH4 , and H2 all enter the stratosphere from the troposphere below, as part of the general circulation of the atmosphere. In the troposphere, water vapor derives, of course, ultimately from the large reservoir in the oceans; methane comes from anaerobic processes, and molecular hydrogen is thought to be produced from cars and biomass burning, as well as from natural sources at the surface. The approximate mean values of the mixing ratios of the three as they enter the stratosphere are:  H2O: 3–4.2 ppmv  CH4: 1.7 ppmv  H2: 0.5 ppmv 2181 further below. CH3 þ O2 ! CH3 O2 ½IV
CH3 O2 þ NO ! CH3 O þ NO2 ½V
CH3 O þ O2 ! CH2 O þ HO2 ½VI
The formaldehyde so formed may be photolysed, or may react with OH, as in reactions [VII], [VIII], and [IX] below. CH2 O þ hn ! H2 þ HCO ½VII
CH2 O þ hn ! HCO þ H ½VIII
CH2 O þ OH ! H2 O þ HCO ½IX
The formaldehyde produced eventually forms H2 and H2O, and so the sequence from methane to water vapor (and to molecular hydrogen) is established. Also, if HOx , HCl or HCO is produced from CH4 , it is quickly converted to H2O, much faster than it is produced from the methane, so that these particular processes are also essentially a means of producing H2 and H2O from CH4. Observations of Hydrogen-Containing Constituents in the Stratosphere There has been some controversy over the so-called ‘mean entry level’ mixing ratio for water vapor, which will be discussed further below (under Issues). Once the source molecules are in the stratosphere, they are carried by the circulation to higher altitudes and latitudes. In the mid-stratosphere, the methyl radical, CH3 , is formed by oxidation processes [I], [II], and [III] below, and then may be converted to formaldehyde, CH2O, by schemes [IV], [V], and [VI], in the presence of chlorine and nitrogen oxides. CH4 þ OH ! CH3 þ H2 O ½I
CH4 þ Oð1 DÞ ! CH3 þ OH ½II
CH4 þ Cl ! CH3 þ HCl ½III
Note that because of the distributions of OH, O, and Cl with height, these reactions are more effective in the middle and upper stratosphere than they are nearer the tropopause. This means that little methane is converted to water vapor or molecular hydrogen in the lower stratosphere. We will consider this Satellite Observations of H2O and CH4 The near-global perspective offered by satellites has been used in the study of the total hydrogen budget of the stratosphere. Not only are satellite instruments capable of near-global observations, but they are also, in principle, capable of making measurements of many atmospheric constituents simultaneously, and of the way they move around. In fact, water vapor and methane, being polyatomic molecules, have active infrared spectra that provide a mechanism for remote measurement of concentrations from space, either by measurement of thermal emission or by solar absorption at the frequencies of the relevant vibration– rotation bands. Molecular hydrogen does not, however exhibit an allowed infrared spectrum, and has not yet been measured from space. To date, three satellite projects have provided sufficient data to make a test of the hydrogen budget possible. These are the Nimbus 7 and the Upper Atmosphere Research Satellite projects, and the ATMOS Space Shuttle instrument, all from NASA in the USA (a useful web page listed under Further Reading will allow the viewer to investigate the details of past, present, and future NASA satellite missions). 2182 STRATOSPHERIC CHEMISTRY AND COMPOSITION / Hydrogen Budget On Nimbus 7, launched in 1978, the Limb Infrared Monitor of the Stratosphere, LIMS, provided first infrared measurements of water vapor, while the Stratospheric and Mesospheric Sounder, SAMS, made measurements of methane. On UARS, the Halogen Occultation Experiment, HALOE, produced a decade of measurements of both water vapor and methane, simultaneously and precisely bore-sighted (i.e., colocated), which have provided copious quantities of data with which to test theory. Also, the ATMOS experiment, a high-resolution solar tracking Fourier transform spectrometer, has been flown on a number of Space Shuttle missions, and has provided complete spectral coverage in the infrared from about 3 to 16 microns, at high spectral resolution. Each of these experiments has been used to study the distribution throughout the stratosphere of H2O and CH4 , and the ratio of the changes of mixing ratio of these two constituents with height, R ¼ DH2 O=DCH4 . R is a parameter which can provide a useful test of photochemical and dynamical processes that might determine the hydrogen budget of the stratosphere, as we will discuss further below. Values in the range R ¼ 1:5 to 2.0 have been found. The sum of ‘total hydrogen’ mixing ratio, c ¼ 2CH4 þ H2 O þ H2 , has also been examined using these satellite data, or at least that part of c that can be measured, i.e., cn ¼ 2CH4 þ H2 O (so that c ¼ cn þ H2 ). It has usually been assumed in such studies that the mixing ratio of H2 is constant at 0.5 ppmv. Figure 1 shows a result for cn obtained from HALOE data, for the stratosphere and mesosphere above 10 hPa. This and other work (see later) indicate that this parameter is, indeed rather constant in the stratosphere, and took values in the 1990s in the range 6.0 to 7.5 ppmv, though there does seem to be a significant trend with time, according to the SPARC 0.01 report on upper-tropospheric and stratospheric water vapor (see Further Reading). The value of cn starts to fall significantly from a constant value above about 0.1 hPa, where rapid production of H2 in preference to H2O takes over. Work by the author of this article and his colleagues has shown that, on the assumptions that the total hydrogen budget is constant, and that water vapor, methane, and molecular hydrogen are the only significant components, such measurements of cn may be used to derive the distribution of H2 , particularly in the mesosphere, where it varies significantly (see Further Reading). Aircraft and Balloon measurements of H2O, CH4 and H2 Very many observations of water vapor have been made from balloon and aircraft platforms, far too many to review here (see the SPARC assessment listed in Further Reading). Fewer, though still many, measurements have been reported of CH4 , and still fewer of H2. Those measurements in which all three species have been measured together are very few! However, because of the near-constancy of the H2 mixing ratio in the stratosphere, measurements of just water vapor and methane have proven valuable. The advantage of local measurements over satellite measurements is, of course, that they are sensitive to smaller spatial and time scales than are satellites, so that more detailed processes may be studied. Also, many of the sensors used on aircraft and balloons, especially some of the in-situ sampling sensors, are capable of higher relative and absolute accuracy than are satellite sensors. One example of the use of aircraft measurements of H2O and CH4 is shown in Figure 2. This shows scatter plots for four different flights of an aircraft at between 17 and 20 km, in the latitude range 15–401 N, in 1993. Water vapor was measured using a photofragment fluorescence sensor, methane by a tunable diode laser HALOE 2 × DCH4 + H2O ppmv Sunset 6 March 1993 _ 11 April 1993 V17 8.00 0.10 Mixing ratio Pressure (hPa) 6.67 1.00 5.33 4.00 2.67 1.33 10.00 _ 90 _ 60 _ 30 0.00 0 Latitude 30 60 90 Figure 1 Height–latitude cross section of HALOE measurements of cn ¼ 2CH4 þ H2 O, a proxy measure of total hydrogen. Units are ppmv. STRATOSPHERIC CHEMISTRY AND COMPOSITION / Hydrogen Budget H2O (ppmv) 7 6 6 5 5 4 4 0.9 1.0 1.1 1.2 1.3 1.4 Flight 930503 7 0.9 6 5 5 4 4 1.0 1.1 1.2 1.0 1.1 1.3 1.4 1.2 1.3 1.4 1.3 1.4 Flight 930426 7 6 0.9 Flight 930506 7 Flight 930507 2183 0.9 1.0 1.1 1.2 CH4 (ppmv) Figure 2 Airborne measurements (17–20 km altitude, 15–401 N) of H2O and CH4 made in 1993. The results of four different flights are shown, and the solid line is the result of a straight-line fit to all the data. instrument. Both measurements are local. The data are compared with a line obtained by linearly fitting the data from all four flights, which has a gradient of m ¼ DH2 O=DCH4 ¼ 1:94  0:27. This is close to the value of R ¼ 2:0 expected if oxidation of methane goes completely to water vapor, and if the production and loss of molecular hydrogen is in balance. This will be discussed further below. Other balloon and aircraft instruments have included infrared sensors, cryogenic trapping followed by laboratory analysis, gas chromatography, mass spectrometry, resonance fluorescence techniques, and frostpoint hygrometry. Some Issues about the Hydrogen Budget of the Stratosphere Entry-Level Mixing Ratio of Water Vapor There has been some controversy over the so-called mean ‘entry level’ of water vapor. Values based on Nimbus 7 data of 2.7 and 3.25 ppmv have been reported, while a value of 4.270.5 ppmv has been reported from aircraft data. However, the picture is confused by the fact that the value may actually have changed with time (reports of increases of about 1% year  1 have been published, i.e. about 10% decade  1, a very significant change), and because only rather limited measurements have been reported near the tropopause. What is beyond disagreement now is that such a global mean value for the entry-level water vapor is merely an average over quite a range, probably from as low as 2 ppmv to as high as 7 ppmv in different regions: moreover, there are probably a number of different processes that control the transport of water vapor in particular from the troposphere to the stratosphere (see Stratospheric Water Vapor), so that the globally averaged view cannot resolve these important individual processes. The Conversion of Methane to Water Vapor An important question is that of the ratio of R ¼ DH2 O=DCH4 , i.e. the increase in water vapor with respect to the decrease in methane with height. It is important in part because it is a ratio we may hope to test from observations. Is this ratio R ¼ 2:0, which would imply that all the hydrogen in the methane, including that converted into intermediate compounds like CH3 and CH2O, has been converted to water vapor, and also that the H2 mixing ratio has not changed significantly, at different heights? In other words, a value of R ¼ 2:0 reflects a balance between large production and loss rates for molecular hydrogen. Reaction [VII] above is the principal source of additional H2 in the stratosphere. However, H2 can also be destroyed in the stratosphere by a number of reactions with OH, O(1D), and Cl. The reaction rates 2184 STRATOSPHERIC CHEMISTRY AND COMPOSITION / Hydrogen Budget for these are similar to those for reactions [I]–[III]. It is thought that the balance of these formation and destruction reactions for H2 is such that the H2 mixing ratio does not change much with height, at least from the tropopause to 40 or 50 km, which is the domain of our concerns here. Thus, the ratio R ¼ 2:0 in those regions where the oxidation of methane is rapid: it is less than R ¼ 2:0 (as low as R ¼ 1:6 has been suggested) in the lower stratosphere, but this is where methane oxidation is much slower, so that the precise value of the ratio is not so important. If, in addition, as seems likely, the H2 profile is constant with height because production and loss mechanisms are roughly in balance, then in practice we would expect R ¼ 2:0 to hold virtually everywhere where significant conversion from CH4 to H2O occurs. There appears to have been some disagreement in the literature about the value of R, but this has arisen largely over a misunderstanding of the relative insignificance of methane oxidation near the tropopause, where the ratio is formally less than 2.0, but where the effect on the measured profiles is small. The Total Hydrogen Content of the Stratosphere Finally, in this section, we ask what is the total amount of hydrogen in the stratosphere? Ignoring the minor species such as CH2O, OH, and so on, we address the parameter c ¼ 2CH4 þ H2 O þ H2, which on the basis of current theory should be constant, unless some unknown significant sources or sinks exist. Estimates of this quantity based on observations vary from 6:5 to 7:0 ppmv from satellite observations from the Nimbus 7 (1979) and the ATMOS experiments (early 1980s), to 8.170.6 ppmv from aircraft measurements between 151 and 401 N, made in 1993. These observations were made at different epochs, and since there is known to have been a long-term upward trend in tropospheric methane, and an upward trend in stratospheric water vapor (of about 1% year  1) between about 1980 and the present, it is possible that these differences may be due to real changes. However, the uncertainties due both to experimental error and to variability and different sampling are probably also large enough to account for these differences. For the present we must adopt a value of c ¼ 2CH4 þ H2 O þ H2 in the range 6.5 to 8.0 ppmv. More accurate measurements are needed to distinguish a real change of the hydrogen budget. See also Climate: Overview. Global Change: Upper Atmospheric Change. Methane. Middle Atmosphere: Planetary Waves; Quasi-Biennial Oscillation. Observations for Chemistry (In Situ): Gas Chromatography; Resonance Fluorescence; Water Vapor Sondes. Observations for Chemistry (Remote Sensing): IR/FIR; Microwave. Satellite Remote Sensing: Water Vapor. Stratospheric Water Vapor. Further Reading Bates DR and Nicolet M (1950) The photochemistry of atmospheric water vapor. Journal of Geophysical Research 55: 301–327. Brewer A (1949) Evidence for a world circulation provided by the measurements of helium and water vapor distributions in the stratosphere. Quarterly Journal of the Royal Meteorological Society 75: 351. Brasseur G and Solomon S (1984) Aeronomy of the Middle Atmosphere. Dordrecht: Reidel. Dessler AE, Weinstock EM, Hintsa EJ, et al. (1994) An examination of the total hydrogen budget of the lower stratosphere. Geophysical Research Letters 21: 2563–2566. Gunson MR, Farmer CB, Norton RH, et al. (1990) Measurements of CH4 , N2O, CO, H2O and O3 in the middle atmosphere by the ATMOS experiment of Spacelab 3. Journal of Geophysical Research 95: 13867–13882. Harries JE, Ruth S and Russell JM (1996) On the distribution of mesospheric molecular hydrogen inferred from HALOE measurements of H2O and CH4. Geophysical Research Letters 23: 297–300. Holton J (1992) An Introduction to Dynamic Meteorology. San Diego: Academic Press. Jones RL, Pyle JA, Harries JE, et al. (1986) The water vapour budget of the stratosphere studied using LIMS and SAMS satellite data. Quarterly Journal of the Royal Meteorological Society 112: 1127–1143. Le Texier H, Solomon S and Garcia RR (1988) The role of molecular hydrogen and methane oxidation in the water vapour budget of the stratosphere. Quarterly Journal of the Royal Meteorological Society 114: 281–295. NASA: Upper Atmosphere research Satellite web site: http:// uarsfot08.gsfc.nasa.gov/ Earth Science Enterprise Programme: http://www.earth.nasa.gov/ Salby ML (1996) Fundamentals of Atmospheric Physics. San Diego: Academic Press. SPARC Assessment of Upper Tropospheric and Stratospheric Water Vapor (2000) World Climate Research Programme Report No. 113. (The author is particularly indebted to the authors of this report, which has provided very valuable background in writing this article.) STRATOSPHERIC CHEMISTRY AND COMPOSITION / Hydroxyl Radical 2185 Hydroxyl Radical D E Heard, University of Leeds, Leeds, UK Copyright 2003 Elsevier Science Ltd. All Rights Reserved. Introduction There has been a considerable desire to measure concentrations of the hydroxyl radical (OH) in the stratosphere over the last 30 years. The measurement is extremely challenging, owing to the short lifetime of OH and its minuscule concentration (typically less than 1 part per trillion, ppt). A range of instrumentation has been developed, but despite considerable expenditure of effort, knowledge of the distribution and variability of OH is still limited. This article will review the methods used to measure OH, and the closely related hydroperoxy radical (HO2), the measurement database of OH in the stratosphere, and some of the chemical insights gained from comparison with model predictions. A number of experimental methods (both in situ and remote sensing) have been used to determine stratospheric OH and HO2 for a range of temporal and spatial scales using ground-based, balloon, high-flying-aircraft, and satellite platforms. As OH is involved in a very large number of photochemical processes, one would expect the concentration of OH to be highly variable, depending on its local environment. However, in the lower stratosphere, the converse is true: OH is remarkably independent of all photochemical parameters except the ozone slant column. A quantitative understanding of the HO2/OH ratio is important, since many of the reactions that control this ratio are involved directly in catalytic removal of O3 in the lower stratosphere. The Importance of OH in the Stratosphere No other species is more intimately involved in the chemistry of the stratosphere than OH, and hence OH is the ideal target molecule for calculation by models. The degree of agreement between observations and model predictions is a powerful indicator of the completeness of our understanding of the chemical behavior of the stratosphere. Clearly, then, the most useful OH data are those collected simultaneously with concentrations of other closely coupled species within the same air mass that can be used to constrain the models. The chemistry of HOx , a collective term for the hydrogen radical family of H, OH, and HO2 , is considered in detail in another article (see Stratospheric Chemistry and Composition: HOx), and so only a very brief treatment is given here. Production of HOx (HOx OH þ HO2 for this article) in the stratosphere results from reaction of O(1D) with the hydrogen source gases H2O, H2 , and CH4 , with O(1D) generated predominantly from O3 photolysis. As H2O mixing ratios increase and CH4 mixing ratios decrease with altitude in the stratosphere, the relative importance of the O(1D)1H2O source increases with increasing altitude. Loss processes for stratospheric HOx involve the recombination of OH and HO2 radicals, either directly or through interactions with NOx . The interconversion of OH and HO2 occurs on a fast time scale compared with their rate of formation and loss. OH is extremely reactive; it acts as a scavenger by reacting with virtually all trace species (in particular those containing bonds to H atoms), and largely defines the oxidizing capacity of the lower stratosphere. The reaction of OH with CO, CH4 , SO2 , and the hydrogen-containing chlorofluorocarbons (HCFCs) initiates the oxidation and hence removal of these species in the stratosphere. OH and HO2 radicals react directly with ozone in a catalytic cycle that dominates O3 destruction in the lower stratosphere (o20 km) and in the upper stratosphere and mesosphere (445 km). OH is responsible also for the coupling between the NOx and ClOx families. For example, reactions of OH and HO2 with NO2 convert NOx into the less reactive reservoir compounds HONO2 and HO2NO2 , respectively, hence reducing the effectiveness of O3 destruction. Reaction of HO2 with NO alters the ratio of NO to NO2. Reaction between OH and HCl to form Cl atoms is the major reaction involved in the conversion of an inactive form of inorganic chlorine to more reactive forms, enhancing the degree to which chlorine can destroy O3. Techniques and Platforms for Measuring OH and HO2 in the Stratosphere Of all measurements in the stratosphere, detection of the OH radical has been the most elusive, and yet one of the most important. A number of groups have attempted to make estimates of OH globally using satellite measurements of other molecules. Satellite measurements of HNO3 and NO2 using limb IR monitoring have been used to derive OH in the 2186 STRATOSPHERIC CHEMISTRY AND COMPOSITION / Hydroxyl Radical stratosphere by assuming photochemical equilibrium in the production and loss of HNO3. Alternatively, photochemical equilibrium can be assumed for the sources and sinks of HOx and satellite measurements of O3 , H2O vapor, and other relevant species then used to infer OH. Techniques for detection of stratospheric OH (and the closely related species HO2) can be divided into two main categories, remote sensing methods and in situ methods, and each is now considered in turn. Table 1 summarizes the techniques for OH and HO2 detection, listing the platform used, the approximate detection limit and accuracy or precision, the temporal and vertical resolution, the working altitude range, and the locations and/or names of field campaigns in which the instruments have been deployed. Remote Sensing Techniques for OH and HO2 Detection Extensive vertical column measurements of OH from the ground have been made using instruments based on the fractional absorption of sunlight near 308 nm. The instruments use spectrometers with extremely high resolving power to measure high-resolution atmospheric spectra. The data set extends over 20 years, and provides a good history covering a range of seasons and solar zenith angles (SZA), but for a very limited number of locations. The column is the integral of OH number density along the vertical path and hence gives no vertically resolved information. However, the tropospheric contribution to the column is small (o5%), and the contribution above 60 km is B45%, and so the slant column is weighted towards stratospheric OH. The monthly average OH column abundance above Fritz Peak, Colorado (401 N), for the period 1980–1990 was stable to 71%, and significant diurnal and seasonal variations were established. Between 1990 and 1995 the OH column during the late summer and early fall decreased significantly (15%), and was connected with the appearance of Pinatubo aerosols at northern midlatitudes following the eruption in 1991. Average OH column abundances at 401 S, measured in New Zealand, were B20% higher than those at 401 N. Another ground-based instrument, this time in the millimeter wave region, has measured a profile of HO2 in the middle and upper stratosphere (435 km), detecting molecular rotational emission lines. The lines are pressure-broadened, and as the pressure depends upon altitude, the spectral line shape can be deconvoluted to give a profile with limited vertical resolution. Submillimeter and far-infrared detection of OH and HO2 from balloon-borne platforms have been used to measure vertical profiles of OH and HO2. These instruments measure thermal emission spectra while observing various angles above the earth’s limb (i.e., at different slant columns) and are able to measure vertical profiles and diurnal profiles for a given altitude. These methods have reduced sensitivity in the lower stratosphere because of water vapor absorption. The altitude of the balloon is usually fixed, and a complex retrieval procedure is required to generate altitude profiles. The average sampling latitude for a given telescope pointing angle can be quite different from that of the balloon itself. Laser-based fluorescence measurements of OH have been performed by the laser imaging, detection, and ranging (lidar) technique, using a balloon-based instrument. The results obtained are very limited, as the method cannot calibrate the signal, and suffers from a large solar scattered background, as the fluorescence is collected by a telescope from the open atmosphere up to several hundred meters from the balloon. Groundbased lidars have been used to detect OH in the mesosphere at altitudes from 75 to 85 km. Stratospheric OH has also been observed from space. Because of the large electronic cross-section of OH in the near UV, the OH resonant fluorescence emission can be detected by spacecraft-borne remote sensing spectrometers. The Middle Atmosphere High Resolution Spectrograph Investigation (MAHRSI) instrument, deployed on the CRISTA-SPAS satellite orbiting at 304 km, has been deployed and retrieved by the space shuttle for missions lasting 8 days. Limb scans of solar resonance fluorescence near 309 nm are inverted to yield OH from B80 km down to 38 km altitude. For several reasons, a complex data analysis is required to generate an OH vertical profile. The signal for a given field of view originates from different altitudes, the excited OH molecules are de-excited by collisions with O2 and N2 (the rate of which is therefore altitude dependent), and the OH fluorescence is absorbed by O3. In-Situ Methods Remote sensing methods suffer from poor spatial and vertical resolution, and require complex fitting routines to retrieve OH and HO2 concentrations. The first attempt to measure OH in the atmosphere in situ involved detection by solar-induced fluorescence using a rocket-borne instrument, and yielded OH concentrations in the upper stratosphere and mesosphere (45–70 km) under evening twilight conditions. The first OH measurements in situ in the mid-stratosphere (30–40 km) used the technique of resonance fluorescence (see Observations for Chemistry (In Situ): Resonance Fluorescence), in which a radio-frequency excited lamp generated light in the 308 nm region and was used to excite fluorescence from OH at the same Table 1 Remote sensing and in-situ measurement techniques and platforms used to measure OH and HO2 in the stratosphere Detection limit h Vertical resolution Solar-induced fluorescence 45–70 km Integrated column abundance only, weighted towards stratospheric OH Integrated column abundance (most sensitive to 38–65 km) B2 km 30–40 km In situ 23–43 km In situ OH Resonance fluorescence using RF-excited lampa Off-resonance laserinduced fluorescencea Fluorescence LIDAR 28–38.5 km Balloon OH, HO2 Far IR emission spectroscopy 23–50 km Remote sensing 4150 m from balloon Remote sensing, limb scanning 1 km Balloon HO2 Matrix isolation electron 3  107 spin resonance (MIESR) Solar induced fluorescence (MAHRSIb) Off-resonance laser5  104 induced fluorescencea Single data In situ point at 32 km 38–80 km Remote sensing, limb sounding 0.3 km at 68 km o 21 km In situ Species Method Ground-based OH UV absorption spectroscopy using sun as light source Ground-based HO2 mm wave absorption spectroscopy Rocket OH Balloon OH, HO2 Balloon OH, HO2 Balloon Satellite (spaceOH shuttle-launched) ER-2 OH, HO2 a OH 5  105 HO2 1  106 Uncertainty Above Fritz Peak, Colorado, 401 N; Lauder, New Zealand, 451 S; Table Mountain, California, 341 N, 1977 Above Mauna Kea, Hawaii, 19.51 N, 1982 7120% Above Wallops Island, Virginia, 1969; Above White Sands, New Mexico, 1971 Above Palestine, Texas, 321 N, 1976 OH735% HO2738% Accuracy 750% Above Palestine, Texas, 321 N, 1987–1989 Above Palestine, Texas, 321 N, 1982 5% OH, 3% HO2 Above Palestine, Texas, 321 N, 1983; Above Fort Sumner, New Mexico, 341 N, 1989; Above Fairbanks, Alaska, 691 N, 1997. Note: sampling latitude varies considerably depending where telescope points 531 N, 1976 7factor of 3 Orbit covers 521 S to 621 N, 1994 and 1997 OH 25% / 1% SPADE,c 15–601 N, 1992, 1993 HO2 30% / 0.5% ASHOE/MAESA,d 44–701 S, 1994 (accuracy / STRAT,e 16–261 N, 32–421 N, 1996 POLARIS,f 65–901 N, 1997 precision) SOLVE,g Arctic, 1999/2000 HO2 measurements also by conversion to OH via addition of NO and subsequent detection of OH. MAHRSI: Middle Atmosphere High Resolution Spectrograph Investigation. c SPADE: Stratospheric Photochemistry, Aerosols and Dynamics Expedition. d ASHOE/MAESA: Airborne Southern Hemisphere Ozone Experiment/Measurements for Assessing the Effects of Stratospheric Aircraft. e STRAT: Stratospheric Tracers of Atmospheric Transport. f POLARIS: Photochemistry of Ozone Loss in the Arctic Region in Summer. g SOLVE: SAGE III Ozone Loss Validation Experiment. h In units of molecule cm  3. At sea level, 2.5  107 molecule cm  3 5 1 ppt (part per trillion). b Location/date of measurements/campaigns STRATOSPHERIC CHEMISTRY AND COMPOSITION / Hydroxyl Radical Altitude range Platform 2187 2188 STRATOSPHERIC CHEMISTRY AND COMPOSITION / Hydroxyl Radical wavelength. The source and detector were mounted in a pod that was suspended from a balloon on a parachute generating a flow of clean air through the excitation region. In order to remove ambient OH for a background measurement chemical seeding was used. The method suffered from a large background because of Rayleigh scattering, and the altitude range and time of day during which measurements were made were limited. HO2 was also measured by this technique following titration to OH by the addition of NO. Previously only a single HO2 data point at 32 km was available using matrix isolation/electron spin resonance. Measurements of OH and HO2 (after conversion to OH) down to 23 km were made in situ by laser-induced fluorescence (LIF) from a balloonborne gondola during three summer flights from one location. The LIF method is able to make exquisitely sensitive and selective quantitative measurements of OH. A high-pulse-repetition-rate laser system (copper-vapor-pumped tunable dye laser) was used to generate radiation at 282 nm, and used to excite OH to the first vibrationally excited level (v 0 ¼ 1) of the electronically excited A2 Sþ state. At stratospheric pressures most of the v 0 ¼ 1 levels undergo vibrational energy transfer to v 0 ¼ 0, and fluorescence from this level at 308 nm is collected by a photomultipler using photon counting. As the wavelengths differ for laser excitation and off-resonance fluorescence, scattered light is enormously reduced. In contrast to the troposphere, laser-generated interference from O3 photolysis at 282 nm to give O(1D) atoms, followed by reaction with water vapor to generate OH, is not a limiting factor, as the H2O mixing ratios are very low. OH measurements in the troposphere using LIF have lagged behind those made in the stratosphere, and an alternative LIF method (known as FAGE) with 308 nm excitation has been developed. The above methods provide useful snapshots that allow comparisons with photochemical models, but are necessarily limited in frequency and spatial coverage, and suffer from a lack of supporting measurements of species that control the formation and loss rates of HOx . The greatest single advance in the understanding of stratospheric chemistry came with the advent of instruments deployed aboard the NASA ER-2 aircraft to measure OH and HO2 in combination with their sources and sinks. Off-resonance fluorescence, excited using a solid-state, laser-pumped dye laser, was able to detect OH down to 5  104 molecule cm  3 (o0.01 ppt in the lower stratosphere) with signal-to-noise ratios 430 achieved in o30 s averaging time, up to altitudes of 21 km. The observed signal was converted to absolute OH number densities using laboratory and in-flight calibrations employing Raman scattering and known concentrations of OH. Once again, HO2 was detected following its conversion to OH, with a similar sensitiviy. The instrument has flown on a large number of missions (see Table 1) covering a wide range of latitudes from the Arctic to the Antarctic, and has provided comprehensive data sets of OH and HO2 for comparison with the calculations of a number of atmospheric models. The models are constrained by supporting data from the ER-2, providing a stringent test of our understanding of lower-stratospheric chemistry. Although primarily sampling air from the upper troposphere, another LIF instrument aboard the NASA DC-8 aircraft has made occasional measurements of both OH and HO2 in air of lower-stratospheric origin. Distribution of OH in the Stratosphere There is still not a good picture of the diurnal, seasonal, latitudinal, and altitudinal profile of OH or HO2 , because the measurements are challenging and the instruments are deployed only for short periods. In this section some of the measurements are highlighted together with a comparison with model calculations. Vertical Distributions of OH and HO2 in the Middle and Upper Stratosphere Volume mixing ratios obtained for OH from balloonborne instruments range from about 3 ppt near 25 km to 400 ppt at 45 km, in reasonable agreement with satellite measurements. For HO2 , balloon-borne measurements indicate mixing ratios of about 10 ppt near 25 km, gradually increasing to 200 ppt near 45 km, again similar to the satellite measurements. Figure 1 shows measurements of OH and HO2 obtained by a balloon-borne far-infrared Fourier transform spectrometer, together with the altitude profiles predicted by various models in which O3 , H2O, and CH4 were constrained by simultaneous measurements, and adjustments made to the rate coefficients of key reactions that control the budget of HOx . The increase in HOx mixing ratios with altitude is explained by an increase in O(1D) concentrations and hence the formation rate of OH via reactions of O(1D) with H2O and CH4. In the upper stratosphere above 38 km (where HOx partitioning is no longer dependent upon [NO]), OH and HO2 are modeled best if the rate coefficient for the reaction O1HO2-OH1O2 is reduced by 25% from the currently recommended value. Above B40 km, catalytic cycles involving OH and HO2 dominate photochemical loss of O3. Although Figure 1 indicates generally good observation– model agreement for balloon-borne measurements (to within 10% for OH with the adjustment of certain STRATOSPHERIC CHEMISTRY AND COMPOSITION / Hydroxyl Radical 2189 Figure 1 Vertical profile of measured (points with error bars) and modeled concentrations of OH and HO2 , HOx , and the ratio HO2/OH at 691 N, 1491 W on 30 April 1997, 09.15 local solar time. The measurements were made by a balloon-borne thermal emission far-infrared Fourier transform spectrometer (FIRS-2). The model curves are constrained by simultaneous FIRS-2 measurements of temperature, O3 and H2O, using selected kinetic parameters. Model A: Photochemical and chemical data from the 1997 JPL Panel Evaluation. Model B: kOþHO2 decreased by 50%. Model C: kOþOH decreased by 20% and kOHþHO2 increased by 30%. Model D: kOþHO2 and kOHþHO2 both reduced by 25%. (Reproduced with permission from Jucks KW, Johnson DG, Chance KV et al. (1998) Observations of OH, HO2 , H2O and O3 in the upper stratosphere: implications for HOx photochemistry. Geophysical Research Letters 25: 3935–3938.) kinetic parameters), the vertical distribution of OH from 80 km down to 38 km measured from satellites cannot be fitted adequately by a single model. Typically observations are 30 to 40% lower than model predictions, while observations of total column concentrations are up to 30% higher than modeled values. During 80 orbits of OH measurements, the satellitebased MAHRSI instrument on the space shuttle obtained B1200 daytime limb scans of OH emission, and after retrieval of the data yielded OH as a function of altitude (50 to 80 km) and latitude (501 N to 551 S). As discussed below, aircraft observations suggest that models of O3 chemistry in the lower stratosphere are accurate, but models of the upper stratosphere and lower mesosphere underpredict the O3 abundance (the ‘O3 deficit’). Analysis of the MAHRSI OH measurements and coincident O3 observations from another instrument suggest that the dominant portion of the deficit is a consequence of the overestimation of OH. Full diurnal measurements of OH for different altitudes have also been recorded in the upper stratosphere for latitudes near 341 N using a balloon-borne far-infrared limb-observing spectrometer (FILOS). The measurements, taken on five flights over 2 years, were compared with a simple model that uses water and ozone fields obtained from instruments on the upper atmosphere research satellite (UARS). At 40 km or above (pressures below 3.2 hPa) the average ratio of the observed to modeled OH concentration was 0.9670.08. The agreement over the OH diurnal cycles for pressures between 3.2 and 21.5 hPa was generally very good, becoming worse at lower altitudes (e.g. 0.8770.24 at 10.0 hPa). The photochemical model assumed production of OH through the reactions of O(1D) with CH4 and water vapor, and HOx destruction via the reactions of OH with HO2 and nitric acid (the latter converting HOx to NOx ). The OH measurements retrieved as a function of 2190 STRATOSPHERIC CHEMISTRY AND COMPOSITION / Hydroxyl Radical altitude were used to calculate the diurnal variations of the OH column density above 25 km, and were compared with model-calculated column densities. The agreement was good, except for just after sunset when above 50–60 km OH is not likely to be in photochemical equilibrium (the key assumption of the model). A critical parameter used in the model is the rate coefficient for the reaction OH 1 HO2 that has an uncertainty at 250 K of B50%, leading to a 22% uncertainty in the modeled OH concentration. Measurements in the Lower Stratosphere The balloon- and satellite-borne instruments provide a general indication of how OH and HO2 change with altitude in the stratosphere, but the datasets are limited to altitudes above B25 km, and in general do not enable diurnal cycles to be measured with a high temporal resolution. In addition, the balloon measurements are above a single launch site (mostly Palestine, Texas). Mixing ratios of OH and HO2 in the lower stratosphere have been made with excellent temporal resolution using the LIF instrument on the NASA ER-2 instrument. Extensive measurement campaigns (see Table 1) under a wide dynamic range of atmospheric conditions have obtained a nearly pole-to-pole database (701 S to 901 N) of HOx and the species that control its chemistry in the lower stratosphere. OH is intimately involved in a large number of chemical processes (for example, in the partitioning of the nitrogen and halogen chemical families) and hence it is expected that OH concentrations will show considerable variability, depending upon the state of the atmosphere (e.g. high vs. low latitudes, levels of NOy ). The suite of ER-2 measurements, however, indicated the remarkable finding that OH is nearly independent of all dynamical and photochemical parameters except the O3 slant column, which is a function of the SZA and O3 column above the aircraft. OH concentrations are the most predictable of the free radicals in the lower stratosphere, enabling a parameterization with SZA to be made that can be used extensively in modeling the lower stratosphere. HO2 displays considerably more variability. During the 1993 SPADE (Stratospheric Photochemistry, Aerosols and Dynamics Expedition) ER-2 aircraft campaign, the first simultaneous measurements in situ were made of the species OH, HO2 , NO, NO2 , ClO, and BrO that are responsible for catalytic destruction of O3 at altitudes from 15 to 21 km and midlatitudes from 15 to 601 N. Throughout this region of the atmosphere the measurements showed that HOx catalysis, with the rate-limiting step being the HO21O3-OH12O2 reaction, constituted 30–50% of the total odd-oxygen loss. The measurements demonstrated quantitatively the coupling that exists between the radical families, with the coupled HO2/ ClO and HO2/BrO catalytic cycles (see Stratospheric Chemistry and Composition: HOx) responsible for 50% of the halogen-controlled O3 removal and 15% of the total odd oxygen loss rates. The vertical profile of [OH] measured during the SPADE campaign between 201 N and 601 N is shown in Figure 2. Except for the dependence upon SZA and altitude, the OH concentration is remarkably invariant in the lower stratosphere, despite a significant change in the concentration of O3 and NOy (such as HNO3) at different latitudes. The OH mixing ratio is nearly independent of the concentration of NO, NO2 , total NOy , O3 , or H2O, and is determined almost solely by solar flux. In contrast, HO2 is more variable, and is driven directly by atmospheric concentrations of O3 , NO, and ClO. The response of HOx to photochemical changes is observed via the [HO2]/[OH] ratio. The dependence of [HO2]/[OH] with [O3], [CO], [NO], [ClO], and [BrO] is reproduced within 710% by a steady-state model Figure 2 Vertical profiles of [OH] measured in situ between 201 and 601 N latitude during the SPADE campaign. To account for differences in solar illumination the data have been normalized to 301 solar zenith angle using the measured diurnal behavior. Despite the large change in NOy levels observed for different latitudes, there is little variation in [OH] for a given altitude. The measurements were made by laser-induced fluorescence spectroscopy aboard the NASA ER-2 high-flying aircraft. (Reproduced with permission from Wennberg PO, Cohen RC, Stimpfle RM et al. (1994) Removal of stratospheric O3 by radicals: in situ measurements of OH, HO2 , NO, NO2 , ClO and BrO. Science 266: 398–404.) STRATOSPHERIC CHEMISTRY AND COMPOSITION / Hydroxyl Radical constrained by the measured mixing ratios of these species. Hence the understanding of the chemistry that partitions OH and HO2 is complete and accurate. The steady-state model assumes that the rate of interconversion of OH into HO2 and vice versa is much faster than the production and loss of HOx . The precision of the measurements showed that the uncertainties of the rate coefficients for the HO21O3-OH12O2 , OH1O3-HO21O2 , and HO21BrO-HOBr1O2 reactions was considerably less (by factor of 2 or so) than described in recommendations. Despite the excellent agreement for the [HO2]/[OH] ratio, SPADE measurements of OH and HO2 at midday were B30% larger than predicted by models; and at high SZA HOx concentrations were B3 times larger. The higher than expected HOx implied the existence of unknown sources throughout the day, and there was a striking and rapid onset of OH and HO2 in the early morning (Figure 3). Initially it was thought that hydrolysis of peroxynitric acid (HO2NO2) on sulfate aerosols produced HONO that photolyzed to generate OH, but this was later shown not to occur. The hydrolysis in sulfuric acid of BrONO2 to form HOBr, followed by photolysis of HOBr via a low-lying state near 500 nm, was then postulated to generate the significant HOx at sunrise. As shown in Figure 3, the model can reproduce 2191 the OH and HO2 measurements, including the unusual behavior when the sun is near the horizon, but only with the inclusion of heterogeneous chemistry. High latitudes were sampled during the ASHOE/ MAESA (Airborne Southern Hemisphere Ozone Experiment/Measurements for Assessing the Effects of Stratospheric Aircraft) and POLARIS (Polar Ozone Loss in the Arctic Region in Summer) campaigns, where the ER-2 was deployed from Christchurch, New Zealand, and Fairbanks, Alaska, respectively. The aircraft payload included OH, HO2 , NO, NO2 , NOy, ClO, H2O, O3 , CH4 , CO, HCl, pressure, temperature, and spectrally resolved radiation fields. Measurement of the SZA dependence of OH and HO2 provides insight into their sources. Figure 4 shows the diurnal variations of OH from these campaigns for 17 to 21 km. The measurement accuracy for OH is 725% (2s) with an instrument precision of 71  104 molecule cm  3 (1%) for 1 min averaging. The diurnal change of OH is controlled primarily by the production of HOx from ozone photolysis followed by reaction O(1D) with H2O or CH4 , from HNO3 photolysis and from CH2O photolysis to generate HCO 1 H, which reacts with O2 to give HO2. Much of the variability can be reproduced using parameterized photolysis rates of O3 and CH2O. The interpretation of the Figure 3 Diurnal variations of [OH] and [HO2] during the SPADE campaign at B19 km altitude measured in situ from the ER-2 aircraft on 11/12 May 1993 at 371 N. The lines are various model simulations using photochemical and kinetic data from the 1992 JPL Panel Evaluation, neglecting all heterogeneous processes (blue dotted line), including hydrolysis of N2O5 and ClNO3 (blue solid line), and also including updated photochemical data for HNO3 , O3 and heterogeneous production of HONO (red solid line); ppt 5 parts per trillion. (Reproduced with permission from Salawitch RJ, Wofsy SC, Wennberg PO et al. (1994) The diurnal variation of hydrogen, nitrogen, and chlorine radicals: implication for the heterogeneous production of HNO2. Geophysical Research Letters 21: 2551–2554.) 2192 STRATOSPHERIC CHEMISTRY AND COMPOSITION / Hydroxyl Radical Figure 4 Measurements in situ of [OH] obtained from the ER-2 LIF instrument during the ASHOE/MAESA (1994) and POLARIS (1997) campaigns as a function of solar zenith angle. The data are shown at 1 min time intervals for the altitude range 17 to 21 km, and the line is a fit to the data using a parameterized form of the photolysis rates of O3 and CH2O. (Reproduced with permission from Hanisco TF, Lanzendorf EJ, Wennberg PO et al. (2001) Sources, sinks, and the distribution of OH in the lower stratosphere. Journal of Physical Chemistry A 105: 1543–1553.) measurements of OH and HO2 at higher SZA (4701) is more difficult, and measurements of OH and HO2 at sunrise and sunset at high latitudes (an ideal location, as the SZA is slowly varying) during POLARIS imply the existence of unknown photolytic sources of HOx . High SZA observations of HOx have demonstrated that a source of HOx  3103 molecule cm3 is missing from the photochemical description of the stratosphere. The wavelengths responsible for producing this HOx must be longer than 650 nm, because the flux at shorter wavelengths is significantly attenuated at high SZA by scattering and absorption. Modeling and recent laboratory experiments suggest that HO2NO2 can dissociate via excitation of overtone transitions to yield HOx . Figure 5 shows measurements and model calculations for HO2 close to sunrise and sunset during POLARIS, and shows the importance of including heterogeneous reactions, and also longer-wavelength photolysis of O3 , HNO3 , and HO2NO2 in the calculation of HOx . Only the very high precision and accuracy of the HO2 measurements has allowed such a detailed comparison with various models to be made. These findings provide a graphic illustration of the power of high-quality field measurements to improve our understanding of the detailed photochemistry of the stratosphere. Figure 5 Dawn and dusk ER-2 measurements in situ and model calculations for [HO2] on 30 April 1997 and 9 May 1997 during the POLARIS campaign. HO2 was measured by conversion to OH via addition of NO with LIF detection of the OH formed. The calculations use rate coefficients and cross-sections from the 1997 JPL Panel Evaluation, and recently reported rate coefficients for the reactions of OH with NO2 and HNO3. The various lines show the significant effect of including additional sources of HOx at high SZA to the predictions of a basic calculation (blue dashed line), including hydrolysis of BrONO2 (blue solid line), addition of spin-forbidden production of O(1D) from O3 photolysis in the near UV (red dashed line), excitation of overtones of the OH stretch in HNO3 and HO2NO2 (red dotted line), and additional photolysis of HO2NO2 at 800 nm (blue solid line, top). Only when all of these additional sources are included is there good agreement with the measurements; pptv 5 parts per trillion, by volume. (Reproduced with permission from Wennberg PO, Salawitch RJ, Donaldson DJ et al. (1999) Twilight observations suggest unknown sources of HOx . Geophysical Research Letters 26: 1373–1376.) STRATOSPHERIC CHEMISTRY AND COMPOSITION / Reactive Nitrogen (NOx and NOy) Further observations are necessary to illustrate the response of HOx to changes in halogen concentrations. The effect of chlorine or bromine partitioning on HOx has not been studied in detail, but measurements of OH in the Antarctic winter polar vortex show that OH concentrations are highly variable and strongly dependent upon chlorine partitioning. Other perturbations within the vortex, such as denitrification and decreased CH4 , will affect the OH concentration. The recent SOLVE (SAGE III Ozone Loss Validation Experiment) campaign (winter 1999/2000) included extensive OH and HO2 measurements at high latitudes in the Arctic, including measurements in the polar vortex, that should help to elucidate these mechanisms. The measurements were made from LIF instruments aboard the NASA ER-2 and DC-8 aircraft, with flights from Kiruna, Sweden. OH was found to be sensitive to the albedo of low clouds and distant high clouds. See also Observations for Chemistry (In Situ): Resonance Fluorescence. Stratospheric Chemistry and Composition: HOx; Hydrogen Budget. Tropospheric Chemistry and Composition: Hydroxyl Radical. 2193 Further Reading Brasseur GP, Orlando JJ and Tyndall GS (eds) (1999) Atmospheric Chemistry and Global Change. New York: Oxford University Press. Finlayson-Pitts BJ and Pitts JN (2000) Chemistry of the Upper and Lower Atmosphere. Theory, Experiments and Applications. San Diego, CA: Academic Press. Jacob DJ (1999) Introduction to Atmospheric Chemistry. Princeton, NJ: Princeton University Press. Jucks KW, Johnson DG, Chance KV, et al. (1998) Observations of OH, HO2 , H2O and O3 in the upper stratosphere: implications for HOx photochemistry. Geophysical Research Letters 25: 3935–3938. NASA Facts: ER-2 high altitude airborne science program web site: http://www.dfrc.nasa.gov/PAO/PAIS/HTML/ FS-046-DFRC.html Wayne RP (2000) Chemistry of Atmospheres: An Introduction to the Chemistry of the Atmospheres of Earth, the Planets, and their Satellites, 3rd edn. Oxford: Oxford University Press. Web pages for details of missions involving ER-2 measurements of OH and HO2: SOLVE: http://cloud1.arc.nasa.gov/solve/ POLARIS: http://cloud1.arc.nasa.gov/polaris/ ASHOE/MAESA: http://cloud1.arc.nasa.gov/ashoe_ maesa/ Wennberg PO, Cohen RC, Stimpfle RM, et al. (1994) Removal of stratospheric O3 by radicals: in situ measurements of OH, HO2 , NO, NO2 , ClO and BrO. Science 266: 398–404. Reactive Nitrogen (NOx and NOy ) Y Kondo, The University of Tokyo, Tokyo, Japan Copyright 2003 Elsevier Science Ltd. All Rights Reserved. Introduction Reactive nitrogen (NOy) plays important roles in controlling the abundance of stratospheric ozone. In this article, sources and sinks of NOy are first described, together with the resulting NOy distributions. Then, the role of NOx , which is the most reactive form of NOy, is explained. NOx destroys catalytically stratospheric ozone and couples with other radical mechanisms. Photochemical processes controlling the relative abundance of component species of NOy determine the NOx abundance. The importance of heterogeneous reactions on sulfate aerosol is described in comparison with gas phase chemistry, typical for midlatitudes. In polar regions, sunlit conditions are very different from those at midlatitudes both in winter and summer. Behaviors of NOx and NOy under these extreme conditions are also explained. NOy Sources and Sinks Reactive nitrogen in the stratosphere is comprised of several component species: NO (nitric oxide), NO2 (nitrogen dioxide), NO3 (nitrogen trioxide), N2O5 (dinitrogen pentoxide), HNO3 (nitric acid), HO2NO2 (peroxynitric acid), ClONO2 (chlorine nitrate), and BrONO2 (bromine nitrate). The sum of these species is defined as total reactive nitrogen NOy. Namely, NOy 5 NO1NO21NO312N2O51HNO31HO2NO21 ClONO21BrONO2. The reactions among NOy component species do not lead to a net change in NOy abundance. N2O, which is produced by bacteria in soil and released into the atmosphere, is the primary source of stratospheric NOy. Since N2O is very stable in the troposphere, it is transported into the stratosphere, mainly through the tropical tropopause. In the 1990s, the tropospheric concentration of N2O was about 310 ppbv and increased at about 0.25% y 1 over the 1978–96 period, due to imbalance between the global 2194 STRATOSPHERIC CHEMISTRY AND COMPOSITION / Reactive Nitrogen (NOx and NOy) sources and sinks of N2O. The stratosphere is a net photochemical source of NOy and a net photochemical sink of N2O. About 90 percent of the total loss of N2O in the stratosphere occurs via its photolysis by ultraviolet radiation yielding N2 and excited atomic oxygen (O(1D)), which does not lead to NO production. N2 O þ hn ! N2 þ Oð1 DÞ ½R1
ðlo230 nm : l is the wavelengthÞ Reactions with O(1D) are responsible for 10% loss of N2O. 1 N2 O þ Oð DÞ ! N2 þ O2 N2 O þ Oð1 DÞ ! 2NO ð42%Þ ½R2a
ð58%Þ ½R2b
where O(1D) is produced primarily by the photolysis of ozone at lo325 nm. Reactions [R1] and [R2] occur in the middle and upper stratosphere (at 35–45 km), mainly in the tropics where the intensity of the solar ultraviolet radiation is greatest, as shown in Figure 1. NOy is produced primarily via reaction [R2b]. Downward transport from the mesosphere and thermosphere, where NO is produced by solar radiation and auroral ionization, can provide an additional NOy input to the upper stratosphere. Production of NO by lightning associated with upward transport through the tropical tropopause may be an additional NOy source in the lower stratosphere at low latitudes, although its magnitude 70 January 2 Summer NOy (production − loss) Winter −1200 −800 Altitude (km) 60 −400 50 40 −320 −160 −80 0 0 160 120 80 40 30 20 −90 −60 −30 0 30 60 90 Latitude Figure 1 Contours of the local instantaneous value at noon of the net production rate of NOy (NOy (production  loss)), in units of 10 8 ppbv s 1. The rates depend on the abundance of N2O, O3 , and other species, rate coefficients of reactions [R1]–[R3], and solar radiation field. (Reproduced with permission from Fahey et al. (1990).) is poorly understood. Supersonic aircraft flying at much higher altitudes than subsonic aircraft injects NO molecules directly into the stratosphere. However, NO emissions from the currently operational Concordes are much smaller than the natural sources. There is also some emission, directly into the stratosphere, by long haul subsonic passenger aircraft, especially on flights routed over high latitudes. The net loss of NOy occurs in the upper stratosphere and lower mesosphere, where NO is reduced into N2 via the following reaction: N þ NO ! N2 þ O ½R3
N is produced by the photolysis of NO by ultraviolet radiation. NOy is also lost through transport of NOy down to the troposphere where HNO3 dissolves in water droplets and is removed from the atmosphere by precipitation. Distribution of NOy Remote spectroscopic measurements by the Atmospheric Trace Molecule Spectroscopy (ATMOS) and Mk IV instruments from the Jet Propulsion Laboratory on board the space shuttle and balloons, in situ measurements on board the NASA ER-2 aircraft, and balloon experiments, in addition to many other measurements, have provided extensive data on the distributions of NOy species, together with N2O. NOy mixing ratios obtained by these measurements increase with altitude from the tropopause up to about 35 km, where values peak at about 18 ppbv (parts per billion by volume) at midlatitudes as shown in Figure 2. This increase is due to an increase in the NOy production via reaction [R2b] and to NOy loss through the tropopause. In contrast, N2O mixing ratios decrease with altitude due to reactions [R1] and [R2] as shown in Figure 2. The NOy mixing ratio decreases with altitude above 35 km due to NOy loss by reaction [R3] and a decrease in the NOy production rate at low N2O mixing ratios. In the lower stratosphere, reaction [R3] is very slow and the lifetime of NOy defined by this loss process is longer than 10 years. Given the increase (decrease) of the NOy (N2O) mixing ratios with altitude up to 35 km, NOy is anticorrelated with N2O for N2O values larger than 120 ppbv, as shown in Figures 3 and 4. The relationship between NOy and N2O obtained by measurements on board the ER-2 in the lower stratosphere is expressed as ½NOy
 ½NOy
0 ¼ 0:07ð½N2 O
0  ½N2 O
Þ ½1
Here [NOy] and [N2O] are the mixing ratios of NOy and N2O in ppbv. [NOy]0 and [N2O]0 are values at the STRATOSPHERIC CHEMISTRY AND COMPOSITION / Reactive Nitrogen (NOx and NOy) 100 N2O (ppbv) 200 300 400 ATMOS/ATLAS-3 In situ balloon (941012) Mk IV (930925) 50 Altitude (km) NOy 40 N2O 30 20 10 0 5 10 15 20 NOy (ppbv) Figure 2 Profiles of NOy and N2O observed by ATMOS at 39–491 N in November 1994, Mk IV at 351 N in September 1993, and in situ balloon-borne measurements at 441 N in October 1994 at northern midlatitudes. 20 ATMOS/ATLAS-3 In situ balloon (941012) Mk IV (930925) ER-2 ASHOE/MAESA 12 January 1989 8 4 0 300 200 100 58 62 66 70 Latitude (°N) 74 78 10 12 January 1989 12 NOy (ppbv) NOy (ppbv) 15 12 N2O (ppbv) tropical tropopause, which are 0.25 and 310 ppbv, respectively. This relationship indicates that the increase in NOy is proportional to the decrease in N2O with a constant slope of  0.07. Namely, 3.5% of N2O molecules lost via reactions [R1]–[R3] are converted to NOy. In contrast, the NOy mixing ratio decreases along with the decrease in the N2O mixing ratio above 35 km where the N2O values are lower than 40 ppbv, leading to a positive correlation between NOy and N2O as shown in Figure 3. Due to its long lifetime, NOy produced mainly in the tropical upper stratosphere is transported to higher latitudes and lower altitudes. This large-scale transport process is known as the Brewer–Dobson circulation. All long-lived trace gases, including ozone, N2O, and CH4 , are subject to this transport. As a consequence, the NOy (N2O) mixing ratios in the lower stratosphere are higher (lower) at higher latitudes, as can be seen from Figure 4. In general, compact correlations have been observed between long-lived species whose local photochemical lifetimes exceed the time scales for atmospheric transport as predicted theoretically. The NOy–N2O correlation shown in Figures 3 and 4 is consistently compact, especially in the lower stratosphere, where the N2O mixing ratios are higher than 120 ppbv. This correlation has proved to be very useful in predicting NOy NOy (ppbv) 60 0 2195 5 8 4 0 0 0 50 100 150 200 N2O (ppbv) 250 300 350 Figure 3 Correlation between NOy and N2O. The data used are the same as shown in Figure 2. In addition, the data obtained by the ER-2 measurements at 30–401 N in February and November 1994 are also shown. 80 120 160 200 240 280 N2O (ppbv) Figure 4 NOy and N2O mixing ratios obtained around 20 km from the north-bound leg of the ER-2 flight on 12 January, 1989 in the lower Arctic stratosphere. (Reproduced with permission from Fahey et al. (1990).) 2196 STRATOSPHERIC CHEMISTRY AND COMPOSITION / Reactive Nitrogen (NOx and NOy) abundances from observed N2O mixing ratios as described below. 40 Gas Phase Chemistry 35 A scheme showing important reactions controlling the level of each reactive nitrogen species is shown in Figure 5. Altitude profiles of NOy species observed by Mk IV at 351 N in September 1993 are shown in Figure 6. The time constants for the photolysis of NOy species for local noon at 441 N in October are shown in Figure 7. NO and NO2 are the most reactive among NOy species. NO is oxidized to NO2 by ozone and NO2 is photolyzed by visible sunlight to reform NO and atomic oxygen (O). NO þ O3 ! NO2 þ O2 NO2 þ hn ! NO þ O ½R4
ðlo420 nmÞ ½R5
NO and NO2 are often treated as a sum, defined as NOx , because the time required for the exchange between NO and NO2 during daytime is about 1 min (Figure 7). NOx plays important roles in controlling stratospheric ozone. First, NOx destroys ozone catalytically via the following reactions: NO þ O3 ! NO2 þ O2 ½R4
NO2 þ O ! NO þ O2 ½R6
Net : Altitude (km) Role of NOx N2O5 (sr) NO NOx NOy O3, CIO, BrO h
CIO h
NO2 O3 BrONO2 BrO OH 20 10−11 N2O5 10−10 10−9 Figure 6 Observed (symbols) and calculated (lines) profiles of NOy species, as indicated, for sunset at 351 N on 25 September, 1993. Sunrise profiles for N2O5 are also shown. The NOy profile represents the sum of nitrogen oxides measured by Mk IV and was used to constrain the model. The model calculation used JPL 2000 kinetic data. (Reproduced with permission from Sen et al. (1998), modified for using updated model calculations by RJ Salawitch.) proceeds faster than reaction [R6] in the stratosphere. The ozone loss rate is therefore proportional to the product of the NO2 and O concentrations ([NO2][O]). Reactions analogous to [R4] and [R6] also represent catalytic ozone loss cycles by reactive hydrogen 1s 40 1 min SZA = 50° (noon) 1 day 1 week 1h HNO3 35 NO2 BrONO2 30 25 HNO4 HONO 20 HNO3 Sulfate aerosol N2O5 NO3 15 10 100 Figure 5 Schematic of the reaction pathways between the principal NOy component species in the lower stratosphere. Photolysis reactions are indicated by hn. ‘Sulfate aerosol’ denotes heterogeneous reactions on sulfuric acid aerosol particles. (Reproduced with permission from Gao et al. (1999).) 10−8 Volume mixing ratio OH, h
h
NOy N2O5 (ss) Attitude (km) h
HNO3 25 The cycle is catalytic since NOx is conserved: NO and NO2 are simply interchanged. Reaction [R6] is rate determining for the catalytic cycle since reaction [R4] Sulfate aerosol NO2 HNO4 O3 þ O ! 2O2 CIONO2 CINO3 30 NO CIONO2 101 102 103 104 105 106 107 Photolysis time constant (s) Figure 7 Time constant of the NOy species due to the photolysis between 15 and 30 km altitude, for noon at 441 N on 12 October. (Courtesy of MY Danilin.) STRATOSPHERIC CHEMISTRY AND COMPOSITION / Reactive Nitrogen (NOx and NOy) (HOx), chlorine (ClOx), and bromine (BrOx) species if NO is replaced with OH, Cl, and Br, respectively. Similar to the NOx catalytic cycle, the ratedetermining reaction for the HOx , ClOx , and BrOx cycles are ½R7
where X is OH, Cl, or Br, respectively. Ozone loss rates are therefore proportional to the product of the concentrations of XO and O ([XO][O]). Secondly, NOx buffers the ozone loss by HOx , ClOx , and BrOx by converting OH, ClO, and BrO into HNO3 , ClONO2 , and BrONO2 , which do not destroy the ozone directly, via the following reactions: NO2 þ OH þ M ! HNO3 þ M ½R8
NO2 þ ClO þ M ! ClONO2 þ M ½R9
NO2 þ BrO þ M ! BrONO2 þ M ½R10
Here, the third body M represents the major atmospheric molecules N2 and O2. Hence, the relative importance of catalytic loss cycles of ozone by HOx , ClOx , and BrOx is strongly dependent upon the NOx abundance. NOx buffers HOx and ClOx catalytic cycles also by the following interchange reactions: NO þ HO2 ! NO2 þ OH ½R11
NO þ ClO ! NO2 þ Cl ½R12
Reactions [R11] and [R12] decrease the HO2 and ClO levels, respectively, by shifting the HO2/OH and ClO/ Cl ratios. The reductions in HO2 and ClO lead to decreases in ozone loss rates, which are proportional to [HO2][O] and [ClO][O], as mentioned above. On the other hand, these reactions increase the NO2/NO ratio, enhancing the ozone loss rate by the NOx cycle. Profiles of the ozone loss rates by the NOx , HOx , ClOx , and BrOx cycles at 351 N in September are shown in Figure 8. Oxidation of NOx NOx levels are also controlled by chemical processes that lead to the production and loss of NOx as detailed below. NOx produced by reaction [R2b] is converted to higher oxides of nitrogen (N2O5 , HNO3 , HO2NO2 , ClONO2 , BrONO2). Since these NOy species do not react directly with ozone but produce NOx by photolysis and reactions with OH, they are called reservoir NOy species. N2O5 is produced 40 35 Altitude (km) XO þ O ! X þ O2 2197 NOx 30 ClOx O+O3 25 BrOx HOx 20 10−2 10−1 100 Fraction of total loss Figure 8 Odd oxygen sinks at 351 N on 25 September 1993. The fractional contribution of the dominant sinks and the diurnally averaged loss rate of odd nitrogen, computed using constraints imposed by the MK IV data. Losses due to each catalytic cycle are indicated as NOx , HOx , ClOx , and BrOx. O31O denotes loss by recombination reaction of odd oxygen. Heterogeneous reactions included in the model calculations increase the contributions from HOx , ClOx , and BrOx. (Courtesy of RJ Salawitch and B Sen.) through the following reactions: NO2 þ O3 ! NO3 þ O2 NO2 þ NO3 þ M ! N2 O5 þ M ½R13
½R14
Reaction [R13] is rate determining for the formation of N2O5 , which occurs only during nighttime since NO3 is photolyzed within a few seconds by visible radiation (lo670 nm) during daytime (Figure 7). Other reservoir NOy species are produced via reactions [R8]–[R10] and the following reaction: NO2 þ HO2 þ M ! HO2 NO2 þ M ½R15
Conversely, NOx is produced by decomposition of reservoir species, such as HNO3 þ OH ! NO3 þ H2 O ½R16
HNO3 þ hn ! OH þ NO2 ðlo310 nmÞ ½R17
N2 O5 þ hn ! NO2 þ NO3 ðlo360 nmÞ ½R18
ClONO2 þ hn ! Cl þ NO3 ðlo380 nmÞ ½R19a
ClONO2 þ hn ! ClO þ NO2 ½R19b
2198 STRATOSPHERIC CHEMISTRY AND COMPOSITION / Reactive Nitrogen (NOx and NOy) BrONO2 þ hn ! Br þ NO3 ðlo500 nmÞ ½R20
10−7 HNO3 HO2 NO2 þ hn ! OH þ NO3 HO2 NO2 þ OH ! NO2 þ O2 þ H2 O ½R21b
½R22
Here the photolysis wavelength thresholds are given for absorption cross-section limits of about 1  10–21 cm2. Typical lifetimes of N2O5 and HNO3 in the lower stratosphere as determined by the above decomposition processes are several hours and 1 week, respectively, for noontime midlatitude fall conditions, as shown in Figure 7. Below 25 km, HNO3 is the dominant NOy species since NOx is oxidized to form HNO3 via the three-body reaction [R8], which proceeds faster at lower altitude (higher pressure) as shown in Figure 6. At higher altitudes, NOx dominates among the NOy species due to enhanced NOx production by reactions [R17]–[R19]. The 2  N2O5 mixing ratio at sunrise becomes comparable to NOx in the lower stratosphere. The ClONO2 mixing ratio shows a broad peak of about 1 ppbv centered around 25 km. Diurnal and Seasonal Variations at Midlatitudes NOy species undergo temporal variations depending on their chemical lifetimes as shown in Figure 9. NO is oxidized to NO2 within a few minutes after sunset and NOx exists in the form of NO2 during nighttime. Part of the NO2 is photolyzed to produce NO soon after sunrise. Due to the formation of N2O5 during the nighttime and photolysis during the daytime, N2O5 mixing ratios reach maximum and minimum values at sunrise and sunset, respectively, as is partly shown by Mk IV observations (Figure 6). Corresponding to the diurnal variation of N2O5 , NOx shows a slow increase in the morning reaching a maximum value at sunset, when it starts to decrease again until sunrise. In contrast, HNO3 does not undergo significant diurnal variation since the lifetime of HNO3 is about a week as described above. Both N2O5 and HNO3 levels respond to the seasonal variations of solar elevation and sunlit hours. The HNO3/NOy and N2O5/NOy ratios at midlatitudes reach minimum values at summer solstice due to the largest rates of reactions [R17] and [R18]. Similarly, they reach maximum values near the winter solstice. The NOx mixing ratio and NOx/NOy ratio show almost sinusoidal seasonal variations reaching maximum and minimum values at Volume mixing ratio ½R21a
ðlo325 nmÞ 44° N, October, 20 km NOy 10−8 HO2 NO2 þ hn ! HO2 þ NO2 NO2 10−9 CIONO2 HNO4 10−10 N2O5 BrONO2 10−11 10−12 HONO 10−13 NO3 NO 10−14 0 3 6 9 12 15 18 21 24 Local time (h) Figure 9 Diurnal variation of odd-nitrogen species in the stratosphere at 20 km altitude calculated at 441 N for October 1994 conditions. (Courtesy of MY Danilin.) summer and winter solstices, respectively, as shown by ground-based and satellite remote sensing observations. Heterogeneous Chemistry Sulfate aerosol composed of liquid sulfuric acid (H2SO4) is ubiquitous in the lower stratosphere from the tropics to the polar regions. H2SO4 in the stratosphere is produced through oxidation of sulfurcontaining gases transported from the troposphere mostly in the form of carbonyl sulfide (OCS). In addition to the gas phase chemistry described above, the partitioning of NOy is also controlled by the heterogeneous reactions on sulfate aerosols listed below: N2 O5 ðgÞ þ H2 OðaÞ ! 2HNO3 ðgÞ ½R23
ClONO2 ðgÞ þ H2 OðaÞ ! HOClðgÞ þ HNO3 ðgÞ ½R24
ClONO2 ðgÞ þ HClðaÞ ! Cl2 ðgÞ þ HNO3 ðgÞ ½R25
BrONO2 ðgÞ þ H2 OðaÞ ! HOBrðgÞ þ HNO3 ðgÞ ½R26
Here (g) and (a) denote species in gas phase and in aerosol, respectively. Reaction [R23] occurs with a reaction probability of 0.1, weakly dependent on the composition STRATOSPHERIC CHEMISTRY AND COMPOSITION / Reactive Nitrogen (NOx and NOy) of aerosols, temperature, and particle size. On the other hand, the effects of reactions [R24] and [R25] are important only at very low temperatures in high-latitude winter (temperature o210 K). Reaction [R26] is fast, but its impact on nitrogen species partitioning in the stratosphere is smaller than that due to reaction [R23] because of the smaller bromine content (B10 pptv (parts per trillion by volume)). These reactions, especially reaction [R23], convert shorter-lived reservoir N2O5 , ClONO2 , and BrONO2 into longer-lived HNO3 , even at low and midlatitudes. They lengthen the time required to regenerate NOx via reactions [R18]–[R20], resulting in an effective decrease in the NOx levels (Figure 5). In addition, reaction [R23] oxidizes NOx without consuming OH, resulting in a higher OH level. The higher OH abundance accelerates reaction [R8] causing further reduction in NOx. The reduction of the NOx/NOy ratio in a stratospheric model due to the effect of heterogeneous reactions is shown in Figure 10. The reduction of the NOx level by heterogeneous reactions leads to increases in the HOx , ClOx , and BrOx levels because of the slower reaction rates of reactions [R8]–[R10] at lower NOx concentrations. The reduction of the NOx level by heterogeneous reactions leads to increases in the HOx , ClOx , and BrOx levels because of the slower reaction rates of 30 12 October 1994 2199 reactions [R8]–[R10] at lower NOx concentrations. The nonlinear dependence of ozone loss rates on HOx , halogen (ClOx and BrOx), and NOx abundances are shown in Figure 11. At the lowest NOx levels (left dotted lines in Figure 11), ozone loss rates increase due to increased abundances of HOx and halogen species that result from the lowering of NOx in an air parcel. Typical values at midlatitudes (right dotted lines in Figure 11) are near the mid-range NOx values where ozone loss rates have low sensitivity to the abundance of NOx. The effect of heterogeneous reactions on NOx and ozone is enhanced by volcanic eruptions. Associated with large volcanic eruptions, significant amounts of SO2 are injected into the stratosphere. The injected SO2 is oxidized to H2SO4 , which forms subsequently sulfuric acid aerosols within a short time. Mount Pinatubo in the Philippines erupted in June 1991 and the aerosol loading increased by up to a factor of 100 over background values. The aerosol loading remained high for a few years with a gradual decrease with time. Corresponding to this enhanced aerosol loading, significant reductions in NOx were observed as shown in Figure 10. It should be noted that the rate of reaction [R23] at large aerosol surface area is limited by the formation rate of N2O5 [R13] during the nighttime. Therefore the decrease in NOx by reaction [R23] saturates eventually at a certain surface area, depending on photochemical conditions. The reduction in NOx caused the decrease in the ozone levels as observed by a variety of in situ and remote sensing measurements, providing evidence of the effect of heterogeneous reactions on stratospheric chemistry. 20 Observed Model Gas Hetero 15 0 0.2 0.4 0.6 Increasing ozone loss Altitude (km) 25 Halogens HOx NOx NOx /NOy Increasing NOx Figure 10 Vertical profiles of the NOx/NOy ratios (small solid circles) derived from the balloon observations made at 441 N on 12 October 1994. Bars show the total uncertainties in the NOx and NOy measurements. The calculated NOx/NOy ratios incorporating heterogeneous chemistry and gas phase chemistry only are compared. (Reproduced with permission from Kondo et al. (2000), modified for using NOx/NOy ratio instead of NO/NOy ratio.) Figure 11 The O3 removal rate versus NOx levels. Because of the coupling that exists between the radical families, the response of the total O3 removal rate to changes in NOx abundance is highly nonlinear. At sufficiently low NOx levels, such as observed at midlatitudes in May 1993, the removal rates are inversely correlated with NOx abundance. 2200 STRATOSPHERIC CHEMISTRY AND COMPOSITION / Reactive Nitrogen (NOx and NOy) Polar NOx and NOy Winter Both NOx mixing ratios and NOx/NOy ratios have been observed to decrease sharply at latitudes higher than 50–601 in winter and early spring as shown in Figure 12. This large and sharp decrease in NOx is caused by the large reduction of sunlight in highlatitude winter, which reduces greatly the formation of NOx from NOy reservoir species. N2O5 produced in the dark stratosphere is converted effectively to HNO3 via reaction [R23] before being photolyzed. Therefore, HNO3 dominates among the NOy species in high-latitude winter. The temperature in the Antarctic and Arctic lower stratosphere decreases to as low as 190–185 K by midwinter in the absence of solar heating. Under these very low temperatures, HNO3 co-condenses with H2O and/or H2SO4 to form polar stratospheric cloud (PSC) particles. These particles provide sites for heterogeneous reactions, such as reactions [R24] and [R25], which convert unreactive inorganic chlorine species into reactive chlorine very efficiently. The NOx 30 Altitude (km) 26 22 Winter Ascent Descent 18 Summer Ascent Descent 14 0.1 1.0 Mixing ratio (ppbv) 10 Figure 12 Contrast in the vertical distribution of NOx in winter and summer. The balloon-borne in situ measurements were made at 511 N in August and December 1982. A strong reduction of NOx occurred at 20–28 km. (Reproduced with permission from Ridley et al. (1987).) reactive chlorine destroys the ozone rapidly when exposed to sunlight. In addition, a polar vortex with westerly winds forms as the high-latitude stratosphere cools each winter season. The polar vortex isolates partially high-latitude stratospheric air from midlatitude air. Extensive ozone depletion inside the vortex in early spring over the Antarctic is well known as the Ozone Hole. Similar processes occur during the cold Arctic winters, although the temperatures in the Arctic are much warmer and show larger year-toyear variations. In this way, HNO3 contributes to ozone destruction through the formation of PSC particles. With the reappearance of the Sun in early spring, NO2 is produced from HNO3 by reactions [R16] and [R17]. NO2 deactivates reactive chlorine and bromine by reactions [R9] and [R10]. This process decelerates effectively the ozone destruction by halogen radicals after the formation of PSCs ceases in spring when the temperature rises above the PSC formation threshold. However, spring HNO3 levels, and therefore NO2 levels, are often much lower than in late fall or early winter for the following reason. HNO3-containing PSC particles in the crystalline form sometimes grow to larger than 10 mm in radius under continued very low temperatures in midwinter. These particles fall out of the stratosphere in a few to several days, leading to permanent removal of NOy. This NOy loss process, called denitrification, lowers the level of HNO3 , resulting in a delay in the deactivation of chlorine and extending the period of ozone depletion throughout the winter and early spring. Denitrification has been detected as deviations of the NOy values from those anticipated from the reference NOy–N2O correlation observed prior to denitrification in late fall as shown in Figure 13. Equation [1] represents a good reference for N2O higher than 120 ppbv. Extensive denitrification occurs in the Antarctic winter, when temperatures fall persistently below even the ice saturation threshold. The temperature in the Arctic in winter is somewhat higher and much more variable than in the Antarctic, as described above, resulting in less extensive denitrification. Falling PSC particles evaporate if they experience temperatures higher than the HNO3–H2O condensation threshold temperature. This leads to local enhancement in NOy over the background value as has been observed at 12–15 km in the Arctic. Summer During the summer, large regions of the polar stratosphere receive uninterrupted sunlight for many weeks. H2O (ppmv) NOy (ppbv) N2O (ppbv) STRATOSPHERIC CHEMISTRY AND COMPOSITION / Reactive Nitrogen (NOx and NOy) Antarctic 23 August 1987 300 2201 Arctic 7 February 1989 200 100 12 8 4 0 5 4 3 2 1 −70 −66 −62 −58 Latitude (deg) −54 62 66 70 74 Latitude (deg) 78 Figure 13 N2O, NOy , and H2O mixing ratios observed by ER-2 at 16–19 km altitude from a portion of a flight in the Antarctic and Arctic missions. The average relation between NOy and N2O Equatorward of the vortex boundary (dashed line) is given by eqn [1]. The shaded areas, highlighting the difference between measured NOy and NOy calculated from eqn [1], represent denitrification in the sampled air masses. (Reproduced with permission from Fahey et al. (1990).) Under these conditions, daily N2O5 production via reactions [R13]–[R14] ceases abruptly with the onset of continuous photolysis in high-latitude air masses, because NO3 , the intermediate in its formation, is photolyzed rapidly, thereby preventing N2O5 formation. Depletion of N2O5 shuts off the hydrolysis of N2O5 in the heterogeneous reaction [R23]. In addition, the photolysis of HNO3 is augmented by continuous sunlight. The NOy family simplifies to a near ‘gas-phase-only’ system in summer air masses because the NOx/HNO3 ratios become primarily controlled by reactions [R8], [R16], and [R17]. Due to these conditions, the NOx/NOy ratios at 18– 20 km observed by the ER-2 aircraft in polar summer reach as high as 0.25, which are much higher than those at lower latitudes. It is noted that gas phase models predict NOx/NOy ratios close to those observed in polar summer even for midlatitude near equinox (Figure 10), thereby demonstrating the importance of reaction [R23] in determining the NOx levels in the lower stratosphere. High NOx abundances in polar summer have also been observed by satellite and ground-based spectroscopic measurements. The measurements by the ER-2 aircraft of related radicals have shown the predominance of the NOx catalytic ozone loss cycle over the HOx , ClOx , and BrOx cycles in polar summer under high NOx , as can be understood by the diagram of Figure 11. Total ozone loss rates calculated using aircraft data are as high as 10–20% per month at 18–20 km at 60–901 N in June. This ozone loss rate is consistent with that observed by satellites. See also Aerosols: Physics and Chemistry of Aerosols. Chemistry of the Atmosphere: Gas Phase Reactions. Middle Atmosphere: Polar Vortex; Transport Circulation. Ozone: Ozone Depletion; Photochemistry of Ozone. Stratospheric Chemistry and Composition: HOx; Halogens; Hydroxyl Radical; Overview. Further Reading Brasseur G and Solomon S (1986) Aeronomy of the Middle Atmosphere. Dordrecht: Reidel. Dessler A (2000) The Chemistry and Physics of Stratospheric Ozone. London: Academic Press. Finlayson-Pitts BJ and Pitt Jr JN (2000) Chemistry of the Upper and Lower Atmosphere. London: Academic Press. Kaye JA and Jackman CH (1994) Stratospheric ozone change. In: Hewitt CN and Sturges WT (eds) Global Atmospheric Chemical Change, pp. 123–168. London: Chapman & Hall. Kolb CE, Worsnop DR, Zahniser MS, et al. (1995) Laboratory studies of atmospheric heterogeneous chemistry. In: Barker JR (ed.) Progress and Problems in Atmospheric Chemistry, pp. 771–875. Singapore: World Scientific. Ridley B and Atlas E (1999) Nitrogen compounds. In: Brasseur GP, Orlando JJ and Tyndall GS (eds) Atmospheric Chemistry and Global Change, pp. 235–287. Oxford: Oxford University Press. Wayne RP (1991) Chemistry of Atmosphere. Oxford: Oxford University Press. World Meteorological Organization (WMO) (1992) Scientific Assessment of Ozone Depletion: 1991, Report 25. World Meteorological Organization Global Ozone Research and Monitoring Project, Geneva. 2202 STRATOSPHERIC OZONE RECOVERY World Meteorological Organization (WMO) (1995) Scientific Assessment of Ozone Depletion: 1994, Report 37, World Meteorological Organization Global Ozone Research and Monitoring Project, Geneva. World Meteorological Organization (WMO) (1999) Scientific Assessment of Ozone Depletion: 1998, Report 44, World Meteorological Organization Global Ozone Research and Monitoring Project, Geneva. Zellner R (1999) Chemistry of the stratosphere. In: Zellner R (ed.) Global Aspects of Atmospheric Chemistry, pp. 181– 254. Darmstadt: Springer. STRATOSPHERIC OZONE RECOVERY Copyright 2003 Elsevier Science Ltd. All Rights Reserved. Introduction The phenomenon of depletion of the stratospheric ozone layer by human-produced chemicals has been dealt with in other parts of this Encyclopedia (see Ozone: Ozone Depletion Potentials; Ozone as a UV Filter; Photochemistry of Ozone; Role in Climate; Surface Ozone (Human Health); Surface Ozone Effects on Vegetation). We here deal with the realities of recovery of the ozone layer – the reasoning behind the predictions that the ozone layer will in fact recover to a state not necessarily exactly as it was prior to about 1980 (when the effects of ozone depletion emerged) but to a state in which the threat of harmful ultraviolet radiation increases is no longer an environmental concern. The subject of the recovery of the stratospheric ozone layer was dealt with in the WMO/UNEP Scientific Assessment of Ozone Depletion: 1998. The reader is referred to Chapter 12 of that document (see Further Reading) for a detailed discussion of why recovery is expected, the models used to predict the recovery and the conclusions related to when the recovery is expected to be observed. These results will be summarized here and the measurements related to recovery of the ozone layer will be updated. As indicated elsewhere, ozone loss in the polar regions during spring is much more severe than the reduction that has occurred at midlatitudes since about 1980. This is related mainly to the fact that in order for chemical ozone destruction to proceed rapidly, the presence of surfaces for heterogeneous chemistry are required. In most of the global stratosphere, surfaces – as presented by particles – are sparse; however, in the polar stratospheres, winter temperatures are adequately low to condense the small amounts of water vapor and nitric acid vapor, forming polar stratospheric clouds (PSCs). Conditions for rapid ozone depletion occur in association with these clouds and the onset of springtime sunlight following cloud formation during the dark winter. Following discovery of the Antarctic ozone hole in 1985, expeditions to Antarctica in 1986 and 1987 to determine the cause of the springtime ozone depletion resulted in considerable public awareness of the phenomenon. Even now, each austral spring (September–October) finds the Antarctic ozone hole in the news, with reports that either it was ‘not as bad’ as last year or that ‘it was worse than last year’. In addition, recent expeditions to study Arctic ozone loss have indicated the likelihood of major ozone loss in some Arctic springs. These events have resulted in considerable confusion concerning the eventual expected outcome of this phenomenon. In actuality, the year-toyear fluctuations in the severity of the ozone hole have been small in recent years, as can be seen in Figure 1, where the total column ozone as measured at the South Pole during the latter half of October is shown. Adequate sunlight for measurements with the Dobson ozone spectrophotometer is available only after midOctober at the South Pole. The years 1988 and 2000 were exceptions. In those years the polar vortex, in which the winter–spring ozone depletion process is confined, broke up earlier than usual, resulting in less ozone loss when averaged over the October period. 400 Total ozone (Dobson units) D J Hofmann, NOAA Climate Monitoring and Diagnostics Laboratory, Boulder, CO, USA October 15−31 1961−2000 300 200 100 0 NOAA/CMDL 1960 1970 1980 1990 2000 Figure 1 Dobson spectrophotometer total column ozone measurements at South Pole Station for the 15–31 October period since 1961. Reliable data are not available prior to October 15 owing to lack of sunlight for the measurement. STRATOSPHERIC OZONE RECOVERY Total ozone (Dobson units) 500 475 450 425 400 375 350 325 1965 1970 1975 1980 1985 1990 1995 2000 2005 Figure 2 Total column ozone averages for March at latitudes between 631 N and 901 N from TOMS satellite data. (TOMS data courtesy of Dr. Paul Newman, NASA Goddard Space Flight Center.) 5 Ozone loss (%) Although the Arctic stratosphere does not get as cold in winter as does the Antarctic stratosphere, recent springtime breakup of the polar vortex has been delayed in some years. With the presence of PSCs during sunlit hours, Arctic ozone depletion has become more severe. Data from satellite measurements in the Arctic, shown in Figure 2, indicate major ozone losses in recent years, but with a considerable interannual variability. The future of Arctic ozone depletion will depend on a number of factors, including climate change. A predicted colder stratosphere could increase the occurrence of PSCs and tend to stabilize the polar vortex. At midlatitudes, large ozone fluctuations related to transport from the tropics make detection of the much smaller ozone losses observed there more difficult than in the polar regions. Since PSCs do not form at these latitudes, chemical ozone loss depends on the surface area present in the form of stratospheric aerosol particles. Following major volcanic eruptions, sulfuric acid aerosol droplets become important in the heterogeneous chemical process that leads to the enhanced destruction of ozone. Figure 3 demonstrates the degree of ozone loss experienced across midlatitudes of the United States and shows how the fluctuations related to transport rival the losses experienced since 1980. Enhancement in ozone loss in 1992–93 is believed to be related to stratospheric aerosol deposited by the eruption of Pinatubo in June 1991. Since the last ozone assessment in 1998, considerable attention has been given to the role of meteorological variability in ozone trends. While it is likely that a portion of the downward trend in midlatitude ozone is related to changes in transport, the magnitude of such an effect is not known at this time. In addition, current models are not able to capture the past trends in dynamical transport and thus are not expected to be able to predict future trends. The next ozone assess- 2203 0 _5 _10 _15 Average for Fresno/Hanford, CA; Boulder, CO; Nashville, TN; and Wallops Island, VA NOAA/CMDL 1980 1985 1990 1995 2000 Figure 3 Monthly average total column ozone deviations from the pre-1979 mean at four Dobson spectrophotometer stations across midlatitudes of the United States. The large reduction in 1992–1993 is partially related to the Pinatubo volcanic eruption. ment, due in the year 2002, will likely consider this component of the variability in more detail. Global measurements of the chlorine- and brominebearing gases believed responsible for most of the ozone depletion are shown in Figure 4. These data indicate that the combined effective equivalent chlorine (EECl) concentration (all chlorine- and brominebearing molecules are combined by multiplying bromine by 50 owing to its higher reactivity) peaked near the surface in 1994 and was expected to peak in the stratosphere 3–5 years later. Satellite remote sensing measurements of chlorine-containing molecules, derived from human-produced halocarbons, indicate that the concentration of total chlorine reached a maximum in the stratosphere in 1997. Thus there is no reason to expect the Antarctic ozone hole or global ozone depletion to become any worse than at present. Model results suggest that there will be a period of twenty or so years in which stratospheric EECl will decline only slowly and then, following phase-out of CFC replacements (such as HCFCs), will decline more rapidly. Recovery of the ozone layer to pre-1980 levels is not expected until the middle of the twentyfirst century. Model predictions of climate change will delay ozone recovery, especially in the Arctic where a cooler stratosphere would exacerbate ozone depletion. Major volcanic eruptions, which supply aerosol particles to the stratosphere, aid the heterogeneous chemistry of halogen ozone loss and will cause a delay in ozone recovery. Will ozone-friendly replacements for the chlorine and bromine compounds be available by 2020? Clearly, the road to recovery will not be smooth, but it appears that the remedy has been found and it is likely that the phenomenon of stratospheric ozone depletion will not get any worse than at present. But will the ozone layer recover to its former healthy state, and how long will that take? 2204 STRATOSPHERIC OZONE RECOVERY 12.0 ppt 540 10.0 515 CFC-12 8.0 ppt 490 275 HCFC-142b 6.0 ppt 4.0 270 2.0 CFC-11 HCFC-141b 265 0.0 150 5.0 HCFC-22 CH3CCI3 110 CCI4 3.0 ppt ppt H-1211 4.0 130 2.0 90 H-1301 70 1.0 CFC-113 50 1991 1993 1995 1997 1999 0.0 1991 1993 1995 1997 1999 2300 2250 Global EECl ~5% down from peak ppt 2200 2150 2100 2050 1991 1993 1995 1997 1999 Figure 4 Measurements of global chlorine- and bromine-bearing compounds that are included in the effective equivalent chlorine (EECl) calculation shown at the bottom of the figure. Ozone Recovery Defined As in the WMO assessment already cited, the beginning of ozone layer recovery is defined as a measurable increase in ozone toward pre-1980 values. While it is important to note the cessation in the worsening of ozone depletion, which appears to be occurring at the present time, recovery suggests some progress toward a return to previous conditions. Thus the detection of such a recovery is complicated by the requirement to detect a statistically significant ozone increase, above natural variability, that is occurring slowly over a long period of time. Measurement stability and comparability between multiple instruments over a twentyyear or longer period will thus be required to actually detect recovery. These observations will be further confused by occasional volcanic eruptions that will cause ozone depletion to increase for 2–3 years during which the aerosol particles from the eruption slowly fall out of the stratosphere. However, observation of ozone recovery is important because it will show that the implementation of regulations on ozone-depleting substances, established by the Montreal Protocol and its amendments, was an effective course to follow. Modeling Ozone Recovery Efforts to predict the future levels of stratospheric ozone include two-dimensional (2D) chemical models, in which vertical and latitudinal ozone variations are predicted; three-dimensional chemical transport models, with dynamic circulation determined by meteorological analyses, and full three-dimensional (3D) general circulation models (GCMs) that include detailed ozone chemistry. These last, requiring major computer resources, have only recently found applications in the ozone prediction area. In the WMO/UNEP Scientific Assessment of Ozone Depletion: 1998, 2D models from ten modeling STRATOSPHERIC OZONE RECOVERY Total ozone anomaly (%) 0 _2 _4 65˚S _ 65˚N _6 Mean of 10 Models 1 S.D. TOMS Observations _8 _10 1970 1990 2010 2030 2050 Figure 6 Average of ten 2D model predictions (with standard deviations) of global (651 S–651 N) ozone loss compared to TOMS measurements, adapted from the WMO/UNEP Scientific Assessment of Ozone Depletion: 1998. (TOMS data courtesy of Dr. Richard McPeters, NASA Gaddard Space Flight Center.) vortex) with stratospheric cloud processes that result in major ozone depletion. NASA Goddard Institute for Space Studies (GISS) 3D model results, reported in the 1998 Ozone Assessment, used scenarios of future greenhouse gas (carbon dioxide, methane and nitrous oxide) emissions from the 1995 Intergovernmental Panel on Climate Change report and thus include effects on the ozone layer related to climate change. In Figure 7 the GISS model results are compared to TOMS springtime observations for the Antarctic (ozone averages south of 651 S) and the Arctic (ozone averages north of 651 N). The model predicts Arctic ozone depletions rivaling those observed in Antarctica in some years with a large degree of interannual variability. As in the case of the 2D models for global ozone recovery, this model predicts that recovery of the polar ozone layers will not be complete till the 2050 time frame. All the models depend on halogen levels declining as prescribed by the amended Montreal Protocol. This includes future replacement of presently unregulated 4 600 Column ozone (DU) Equivalent chlorine, EECl (ppb) groups were compared. These models typically have resolutions of 5–10 degrees of latitude and 1–2 km of altitude. While gas-phase chemistry is similar in the models, heterogeneous chemistry involving reactions on the surfaces of aerosol particles at low temperatures was represented at varying levels of sophistication. All models used the same reactive halogen distributions as observed up to 1998 and as predicted assuming that the amended Montreal Protocol will be followed. This includes future regulation of the CFC replacements. Figure 5 shows the time history of effective equivalent chlorine (EECl, defined earlier) which was used in the models. This scenario predicts that stratospheric equivalent chlorine will reach 2 parts per billion (ppb), the level at which Antarctic ozone depletion became clearly detectable, in about the year 2050. Figure 6 shows the average and standard deviation of the predictions of the ten 2D models of global (651 S–651 N) ozone loss compared to Total Ozone Mapping Spectrometer satellite measurements. The agreement with measurements is remarkably good. The ozone reduction observed and predicted in about 1992 was related to the Pinatubo volcanic eruption in the Philippine Islands in June 1991. The additional particle surface area deposited in the stratosphere by the eruption exacerbated ozone depletion for one to two years. The model results predict that clear observation of the beginning of recovery of the ozone layer will not be possible until after the year 2020 and that stratospheric ozone will not reach pre-1980 levels till beyond the year 2050. Changes in dynamics such as might have occurred in the past and may occur in the future, for example as related to climate change, are not included or captured in these models. Three-dimensional general circulation models with relatively simple ozone chemistry have been utilized to predict recovery of polar ozone because they are able to generate a realistic winter polar wind system (polar 2205 3 2 1 500 TOMS - Antarctic GISS Model TOMS - Arctic GISS Model 400 300 200 100 2060 0 1975 1985 1995 2005 2015 2025 2035 2045 2055 2065 Figure 5 Global effective equivalent chlorine as measured up to the year 2000 and predicted by emission model estimates. (Adapted from the WMO/UNEP Scientific Assessment of Ozone Depletion: 1998.) Figure 7 Three-dimensional model predictions (Goddard Institute for Space Studies) of polar ozone loss compared to TOMS satellite data. (Adapted from the WMO/UNEP Scientific Assessment of Ozone Depletion: 1998.) 1960 1980 2000 2020 2040 2206 STRATOSPHERIC OZONE RECOVERY 30 Maximum area (106 km2) hydrofluorocarbons, halons, and other bromine-bearing compounds such as methyl bromide. If the emissions of these compounds do not decline as prescribed by the Protocol, because of continued production and/or emission in developing countries who were provided special dispensations in the Protocol, then the predictions of recovery will, of course, not be accurate. 20 10 0 Observing the Recovery As discussed in WMO assessment, there are a number of reasons why it is likely that the earliest evidence for recovery of chemical ozone depletion will come from Antarctica. The main reason is that the depletion magnitude is large, with about two-thirds of the ozone layer lost each spring. This magnitude is considerably larger than natural variability, which makes detection of recovery at midlatitudes, where the ozone deficit is only of the order of 5% difficult. In recent years, ozone has been totally destroyed in the heart of the ozone hole region at 15–20 km. In these regions, more chlorine and bromine are activated than is required to destroy all the ozone available. Thus this region is not expected to be an early indicator of the beginning of ozone recovery. However, at both the horizontal and vertical boundaries of the ozone hole region the phenomenon is not saturated and thus presents perhaps the best opportunity for early detection of the beginning of recovery. The horizontal extent of the ozone hole can best be observed by satellite instruments such as the Total Ozone Mapping Spectrometer (TOMS), which detects total column ozone by observing ultraviolet radiation being reflected off the surface or off clouds. It has been customary to use the 220 Dobson unit (DU) contour to define the outer boundary of the springtime Antarctic ozone hole as this is the value at which a steep gradient in ozone exists with substantial depletion internal and minimal depletion external to the 220 DU contour. Figure 8 shows the magnitude of the area interior to this contour, averaged for the springtime period 9 September to 13 October, as a function of time since TOMS satellite measurements began in 1979. In recent years this parameter has been in the range of 22 (72)  106 km2. This is equivalent to the area poleward of about 661 S latitude. All but the tip of the Antarctic Peninsula lies internal to this area, so that the ozone hole defined in this manner covers essentially all of Antarctica. At the boundaries of the depletion region, stratospheric temperatures are not as cold as internal to the boundary and thus polar stratospheric clouds, which 1980 1985 1990 1995 2000 Figure 8 Geographical area of the 220 Dobson unit contour over Antarctica between 9 September and 13 October from TOMS satellite data. (TOMS data courtesy of Dr. Paul Newman, NASA Goddard Space Flight Center.) provide the surfaces required for the heterogeneous chemistry do not form as readily. Thus, assuming that temperatures will not change substantially with time (climate change related to greenhouse gas increases is expected to cool the stratosphere but will have larger effects on the Arctic stratosphere than on the Antarctic because the Antarctic stratosphere is already very cold), the area enclosed by the springtime 220 DU contour at maximum depletion should be a sensitive indicator of the beginning of ozone recovery as halogens begin to decline in the stratosphere. For example, values below 20  106 km2 have not been observed since before 1990 and would be an indication of the beginning of ozone hole recovery. The vertical extent of the ozone hole can be observed with balloon-borne instruments and has been monitored annually since 1986 at the South Pole. Figure 9 shows vertical ozone profiles measured at the South Pole during the ozone hole maximum depletion at the beginning of the current continuous measurement period in 1986 and during the ozone hole period in 2000. These profiles are compared to ones measured at the South Pole prior to the advent of the ozone hole phenomenon during the 1967–71 period. The progression toward total ozone destruction in the 15–20 km region is clear. It also is clear that ozone depletion has progressed to higher altitudes during the 1986–2000 period, with a sharp top to the ozone hole at about 21 km in 1986 and about 24 km in 2000. Barring major temperature trends in this region, the top of the ozone hole should begin declining in altitude as halogens begin their decay. Another parameter that will be sensitive to halogen decay is the rate of ozone loss in the main ozone loss region (12–20 km) during September, the period when ozone is declining rapidly. Figure 10 shows the rate of ozone loss in September measured with balloon-borne instruments at the South Pole since 1986. In recent years the value has been about 3.170.4 DU per day. STRATOSPHERIC OZONE RECOVERY 160 Column ozone (DU) 30 25 29 Sep. 2000 98 DU _ 2.6 DU day _1 120 80 Year 2000 40 7 Oct. 1986 158 DU 0 (A) J F A M M J J A S O N D 4 15 _ Ozone loss rate (DU day 1) Altitude (km) 20 October average 1967−1971 282 DU 10 5 3 5 10 15 1 1985 Ozone partial pressure (mPa) Figure 9 Ozone vertical profiles obtained with balloon-borne ozonesondes at South Pole Station at the time of maximum ozone depletion in 1986 (when continuous measurements began) and in 2000. The recent measurements are compared to those made during the 1967–71 period. Pre-1990 values were in the range of 2.270.4 DU per day. The detection of the recovery of Arctic ozone loss is expected to be more difficult because models suggest that the worst is yet to come in the Arctic owing to the more dynamic situation in the Arctic compared to the Antarctic (see Figure 7). The interannual variability in the degree of springtime Arctic ozone depletion will be much too large to allow any simple observations of the beginning of recovery of the Arctic ozone layer. At midlatitudes, detection will be difficult as well. Here the small signal compared to natural variability makes difficult the detection of an ozone increase. Statistical models that include past trends and 2D chemical model predictions suggest that recovery may be detected at southern midlatitudes prior to northern midlatitudes. As indicated in Figure 6, the detection of a meaningful increase in ozone levels is not predicted by the chemical models before about 2030. Further complications arise from uncertainties in the global levels of some chemicals, for example, methane, which reacts with chlorine and thus reduces ozone depletion. In recent times the methane growth rate has declined 1990 1995 2000 Figure 10 (A) South Pole Station ozone between 12 and 20 km (the region of maximum ozone depletion) as a function of time during the year 2000, with a determination of the ozone loss rate during September. (B) September ozone loss rate (Dobson units per day) at South Pole Station since measurements began in 1986. substantially and if it does indeed cease to increase, as proposed, then the recovery of ozone will be further delayed by 10 or so years (see Figure 11). Finally, since the models do not capture the dynamic variability observed in lower stratospheric circulation, which in some analyses can account for almost onehalf of the ozone loss at northern midlatitudes since 1979, the observed recovery could be either slower or faster than predicted by chemical models, depending 2 0 Ozone loss (%) 0 September ozone loss rate 2 (B) 0 2207 _2 _4 _6 Model - 98 Assess. Baseline Methane Constant After 1995 TOMS Observations _8 _ 10 1980 1990 2000 2010 2020 2030 2040 2050 Figure 11 A two-dimensional model prediction of global (651 S– 651 N) ozone loss showing the effect of holding methane concentrations constant after 1995. TOMS measurements are also shown. (Adapted from the WMO/UNEP Scientific Assessment of Ozone Depletion: 1998; TOMS data courtesy of Dr. Richard McPeters, NASA Goddard Space Flight Center.) 2208 STRATOSPHERIC WATER VAPOR on whether the dynamical factor is increasing or decreasing. Summary and Conclusions In summary, while chemical models indicate that the maximum ozone depletion will occur within the next two decades, uncertainties related to emission scenarios of greenhouse gases and climate change make estimates of the beginning of ozone layer recovery unreliable. Even in Antarctica, where it is believed that the earliest and least ambiguous observation of the beginning of ozone recovery will be possible, the unambiguous detection of the beginning of the recovery of the ozone layer will not occur until well into the next century, beyond the maximum loading of ozonedepleting gases. It is quite clear that the atmosphere will be in a different chemical and thermal state when equivalent chlorine levels drop to pre-1980 levels in the stratosphere, making precise predictions of ozone recovery impossible. Barring major volcanic eruptions during the next decade, a cessation of the downward trend in midlatitude ozone, now only hinted at, should be observed and would be a harbinger of the coming recovery. See also Ozone: Ozone Depletion Potentials; Ozone as a UV Filter; Photochemistry of Ozone; Role in Climate; Surface Ozone (Human Health); Surface Ozone Effects on Vegetation. Further Reading Anderson J, Russell JM III, Solomon S and Deaver LE (2000) Halogen occultation experiment confirmation of stratospheric chlorine decreases in accordance with the Montreal Protocol. Journal of Geophysical Research 105: 4483–4490. Dlugokencky EJ, Masarie KA, Lang PM and Tans PP (1998) Continuing decline in the growth rate of the atmospheric methane burden. Nature 393: 447–450. Farman JC, Gardiner GG and Sahnklin JD (1985) Large losses of total ozone in Antarctica reveals seasonal ClOx/ NOx interaction. Nature 315: 207–210. Hofmann DJ, Oltmans SJ, Harris JM, Johnson BJ and Lathrop JA (1997) Ten years of ozonesonde measurements at the south pole: implications for recovery of springtime Antarctic ozone. Journal of Geophysical Research 102: 8931–8943. Hood LL (2000) Trends in lower stratospheric circulation and their effects on column ozone trends at northern midlatitudes during the 1979–1998 period. Proceedings of the Quadrennial Ozone Symposium, Sapporo, 2000, pp. 49–50. Sapporo: Hokkaido University. Hood LL, McCormick JP and Labitzke K (1997) An investigation of dynamical contributions to midlatitude ozone trends in winter. Journal of Geophysical Research 102: 13079–13093. IPCC (1995) Climate Change 1995, The Science of Climate Change (Houghton JT et al., eds). Cambridge: Cambridge University Press. Montzka SA, Butler JH, Myers RC, et al. (1996) Decline in the tropospheric abundance of halogen from halocarbons: Implications for stratospheric ozone depletion. Science 272: 1318–1322. Montzka SA, Butler JH, Elkins JW, et al. (1999) Present and future trends in the atmospheric burden of ozonedepleting halogens. Nature 398: 690–694. Shindell D, Rind D and Lonergan P (1998) Increased polar stratospheric ozone losses delayed eventual recovery due to increasing greenhouse gas concentrations. Nature 392: 589–592. Weatherhead EC, Bishop L, Hollandsworth Frith SM, et al. (2000) Detecting the recovery of total column ozone. Journal of Geophysical Research 105: 22201– 22210. WMO (1999) Scientific Assessment of Ozone Depletion: 1998, World Meteorological Organization, Global Ozone Research and Monitoring Project, Report No.44. Geneva: WMO. STRATOSPHERIC WATER VAPOR J E Harries, Imperial College of Science, Technology and Medicine, London, UK Copyright 2003 Elsevier Science Ltd. All Rights Reserved. Introduction The Earth is indeed the ‘water planet’. So many of the key conditions and properties here on Earth are determined by this enigmatic and vital substance. Not least of the mysteries involving water is that the stratosphere, that region of the atmosphere between roughly 12 and 50 km, is extremely dry. The concentration of water vapor, expressed as a relative fraction of the total mass of air at any given altitude, is in the range of ‘parts per million’ (ppm; 10  6), in other words, only a few molecules in every 106 are water molecules. This contrasts with humidities up to 10 000 times higher in the troposphere. Nevertheless, this small concentration is profoundly important. It is one of the fascinations of the study of stratospheric humidity that, while this extreme aridity and the STRATOSPHERIC WATER VAPOR overall mechanisms causing it have been known for more than half a century, the detailed understanding of precisely how this state is maintained remains elusive. This extreme dryness was first discovered during high-altitude research flights in Canberra aircraft over the United Kingdom, beginning as long ago as 1943 and continuing for many years thereafter. In the frostpoint hygrometer, a mirror surface is cooled until the ambient humidity causes a frosting of the surface: knowing the temperature at which this happens, and the pressure of the local air allows the humidity to be calculated. Using this device, scientists from the British Meteorological Office measured frost-point temperatures at the tropopause (the boundary between the troposphere and stratosphere) of 215 K, which equates to a mixing ratio of about 55 ppmv (parts per million by volume, expressing the relative number of molecules of water and air in a given volume). At altitudes about 2 km above the local tropopause, mixing ratios of 3 ppmv were observed. Further analysis, including the use of studies of the way in which radioactive decay from nuclear tests spread around the world, led to the formulation of the Brewer–Dobson theory to explain this dryness. It was postulated that a slow overturning of the whole atmosphere, with air rising above the warm equatorial region, passing through the very cold tropopause, moving poleward in the stratosphere, and then sinking at higher, colder latitudes, could cause dessication of the air as it rises through the cold tropical tropopause. The tropopause temperatures in tropical regions were known to be very low (well below 220 K) and quite capable of producing this degree of dryness. This basic theory of the humidity of the stratosphere has survived and today forms the basis of our understanding, although we now realize that there are many important details that modify this model. The humidity of the stratosphere is important because the amount of water vapor determines important aspects of the planetary radiative energy balance through the strong cooling to space from water vapor. This contributes to determining the temperature of the stratosphere, which then affects the dynamical circulation of the upper atmosphere. Moreover, water vapor provides the source of the hydroxyl radical, OH, which takes part in a number of stratospheric chemical processes, and the influence on temperatures affects chemistry through temperaturedependent reaction rates. An understanding of how the distribution of water vapor is controlled, and of how this distribution might change in future, is therefore important in a variety of scientific and environmental problems, as we shall see below, not least in determining the role of water vapor in controlling changes in our climate. 2209 Observations of Stratospheric Water Vapor Following the pioneering work in the United Kingdom, measurements at higher altitudes (to 30 km and above) were made in the United States using highaltitude balloon technology. At first, this work indicated a much wetter stratosphere, and something of a controversy brewed up. However, it was soon realized that the balloon data were being contaminated by moisture carried up by the balloon itself, and since then, a long series of very accurate measurements, also using the frost-point hygrometer principle, have been reported by US scientists working for the Naval Research Laboratory and subsequently the National Oceanic and Atmospheric Administration, (NOAA). These showed that the humidity of the stratosphere remained low, with only a small possible increase, up to an altitude of about 30 km. These measurements have continued from the late 1960s up to the present day, first over Washington DC and subsequently over Boulder, Colorado. These measurements have given support to the possibility of significant long-term trends in stratospheric moisture, existing over decades. We shall come back to this later. A step up in sensitivity, which allowed measurements at much higher time and space resolution to be made, came with the invention of an alternative measurement technique using the resonance fluorescence hygrometer. This device uses the spectroscopic property of the water vapor molecule that if it is illuminated by ultraviolet radiation from a lamp at a certain frequency, it will reemit radiation in measurable quantities and with an intensity that is proportional to the relative amount of water vapor to air molecules in the line of sight. This device has now been widely deployed on aircraft and balloons around the world, and has given rise to a much more finely detailed knowledge of how water vapor is distributed around the globe. It was using this device that the ‘hygropause’, a minimum in mixing ratio some 2–3 km above the local tropopause, was discovered in tropical regions. This discovery gave a clue to more detailed mechanisms of how the dryness of the stratosphere is controlled, as we shall see later. For a more global view of how stratospheric water vapor is distributed and might vary, however, satellite techniques inevitably became crucial. Measurements of stratospheric water vapor were first made by the NASA Limb Infrared Monitor of the Stratosphere, (LIMS) on the Nimbus 7 spacecraft launched in 1978. This device, which employed sensitive cooled detectors in space, detected emission from stratospheric water vapor from one of its infrared vibration– rotation bands (the n2 band centered at 6.3 mm). 2210 STRATOSPHERIC WATER VAPOR we are gradually unraveling the mysteries of this enigmatic substance in our upper atmosphere, as we shall see below. The Mean Water Vapor Distribution of the Stratosphere Figure 1 gives a representation of the annual mean distribution of water vapor, based on the most recent measurements from the UARS satellite. The main points to note are the strong gradients in mixing ratio at the tropopause; a slow increase in mixing ratio with height; a minimum of mixing ratio immediately above the tropical tropopause; and shallow minima above the polar tropopause in both hemispheres, but particularly the southern pole. We will now discuss current ideas that account for this mean distribution. Mean Meridional Circulation The Brewer–Dobson theory was mentioned earlier. This accounts for the extreme dryness of the stratosphere as due to a slow meridional circulation with rising air over the tropics and descending air at higher latitudes. The very low temperatures found at the tropical tropopause cause a dessication of the air as it passes from tropical tropopause to overlying stratosphere. Above the tropopause, as the air rises, the mixing ratio increases due to conversion from methane, CH4 (see below). Air at extratropical latitudes is made up of dry air spreading nearly isentropically along contours of constant potential temperature, mixing with air that has risen to the stratopause, increasing in mixing ratio as it rises, and which has 0.3 6.0 1 5.5 5.2 40 3 4.8 4.4 30 10 4.0 30 20 3.6 400 3.6 100 10 90° S 60° S 30° S Pressure (hPa) 50 Height (km) The intensity of this emission is proportional to the atmospheric temperature, and the water vapor concentration: Knowing the temperature from separate measurements allowed the water vapor concentration to be determined globally as a function of altitude and position. This was a very exciting development, which employed the new technique of limb-sounding that increased the precision of stratospheric measurement by aiming the instrument sideways, toward the limb of the atmosphere, where the stratosphere is exposed against the cold, dark background of space. Nimbus 7 was notable for another reason: this program introduced the idea of assembling international ‘Experiment’ or ‘Science’ teams to assist in the development and scientific exploitation of the experiments on board the satellite, a method adopted in almost all satellite experiments since. Many other notable experiments to measure stratospheric water vapor have been undertaken since. Also on Nimbus 7 was the British instrument called the Stratospheric and Mesospheric Sounder, which used another new technique, pressure modulation radiometry, to detect and measure water vapor (among other gases). This technique actually carried a sample of water vapor in a cell on board, as a type of ‘calibrator’ of the detected infrared emissions. A very beautiful Fourier transform spectrometer called ATMOS (Atmospheric Trace Molecule Spectroscopy) was flown several times on the Space Shuttle in the 1980s and 1990s, providing highly accurate spectral information about water vapor and many other stratospheric molecules. Long-term measurements were initiated using shorter-wavelength visible and near ultraviolet observations of water vapor absorption, using the Stratospheric Aerosol and Gas Experiment (SAGE). More recently, the Upper Atmosphere Research Satellite (UARS) has operated from 1991 until the present day (mid-2001) and has carried several experiments that measured stratospheric water vapor. Perhaps the most successful of these is the Halogen Occultation Experiment (HALOE), which has provided near-global measurements for a decade. HALOE measures the absorption of infrared solar radiation by stratospheric water vapor, using the LIMS technique of staring through the limb of the atmosphere, in this case as the Sun rises or sets behind the atmosphere. Another new sensor, the Microwave Limb Sounder, operated in the millimeter wave part of the spectrum. All these new data have given us a completely new perspective on the distribution and variability of water vapor in the stratosphere. Many other satellite and non-satellite-borne experiments have given new measurements that have helped us to understand more and more about the behavior of stratospheric humidity. We still face many puzzles, but 300 0° 1000 30° N 60° N 90° N Latitude Figure 1 Annual zonal mean water vapor mixing ratio (ppmv) from HALOE and MLS data by height and latitude. Contour interval is 0.2 ppmv. Thick dashed line is the tropopause. Thick solid line is the 400 K potential temperature (isentropic) surface. (Source: SPARC 2000.) STRATOSPHERIC WATER VAPOR then traveled poleward, sinking as it cools. This makes for a complex situation at these latitudes, with laminae of dry and moister air overlying each other. The two routes give rise to air that has different ‘ages’ in the stratosphere. The rapid isentropic transport gives rise to a dry layer in the lower stratosphere over much of the globe. It is now also believed that mixing of air caused by planetary wave activity in the lower stratosphere is important, especially at mid-latitudes and in winter, when such wave activity is at its strongest. Stratospheric Photochemistry In the stratosphere, the high intensity of short-wave solar radiation means that methane and molecular hydrogen can be photolyzed, to release active hydrogen compounds. The methane is oxidized to produce roughly two molecules of water vapor for every one molecule of methane that is destroyed. As a consequence, the quantity c ¼ 2½CH4
þ ½H2 O
may be regarded as a quasi-conserved quantity, which can help analysis of observed fields. Lifetimes of both methane and molecular hydrogen are over 100 years at the tropopause, a few years at 30 km, and a few months at 40 km, owing to the increasing solar flux at short wavelength. Thus, air can be ‘tagged’: low values of [CH4] and high values of [H2O] in lower stratospheric air indicates that the air has been transported down from higher altitudes. Tropical Tropopause Dehydration It is an amazing fact that, despite decades of research, we are still unsure about the precise mechanism(s) that cause the ‘cold trap’ phenomenon at the tropical tropopause. The tropical tropopause temperature around the globe is not low enough everywhere to produce the observed very low values in the stratosphere. A variety of mechanisms have been postulated that could account for the observations. These include the suggestion that there are preferred longitudes around the Equator, for example over Micronesia, where vertical convection is particularly strong and where, as a result, tropopause temperatures are especially low. This gave rise to the notion of a ‘stratospheric fountain’, one of the more picturesque concepts of recent years! Also, considerable work, notably in the United States, has been done on the effect of very deep convective clouds that might penetrate the local tropopause for a limited period: These clouds usually have ice clouds associated with the top and downwind side of their ‘anvils’. It is thought that both dessication and increase in moisture in the lower stratosphere could arise from such a mechanism. Other studies have considered the exist- 2211 ence of more permanent ‘cirrus veils’ in the region of the tropopause, the formation of which causes a drying of the air as it supplies water to form ice: Subsequent descent of the veil, due to radiative cooling, can leave behind a dry layer of air. It now seems likely that many of these mechanisms are operating simultaneously. Thus, the mean meridional circulation is augmented by longitudinal variability in convective energy, by local cloud penetrations of the tropopause, by cirrus veils, and probably by other processes as well. Furthermore, these may be operating on quite small space and time scales, and what we observe is the net effect averaged over time and space. For example, the combined effect of many cumulonimbus clouds penetrating the tropopause may be a significant contributor to the dryness of the lower stratosphere. However, capturing such a mechanism in a model that may have a horizontal spatial resolution of over 100 km is difficult. Polar Dehydration Within the stratospheric vortices that form around each of the poles in winter, temperatures can fall to extremely low values (e.g., 180 K) and, of course, the air in the vortex is very dessicated under such conditions. In the north, the strength of the vortex, the degree of cooling within the vortex, and the consequent degree of dessication are not thought to be sufficient to affect the annual mean. However, in the south, the dessication is very significant and is thought to decrease the annual mean, even though it is only a seasonal effect. Below the 400 K potential temperature surface, in spring and summer, dehydration can affect mid-latitudes, but the effect on the rest of the stratosphere is minimal. Troposphere–Stratosphere Exchange at Mid-Latitudes In the mid-latitude lower stratosphere, the extreme dryness of the stratosphere must be maintained against the relatively very high humidity of the tropopause just a few kilometers away under the tropopause. At these latitudes, of course, there is not the strong vertical convection to maintain the cold trap mechanism and, indeed, tropopause temperatures are warmer on average than at lower latitudes. What happens? First, the air in the stratosphere is, on the average, subsiding from higher up and so maintains an appropriate level of moisture, though there is also ‘leakage’ through the tropopause at mid-latitudes. The most likely route is from the tropical high tropopause, along isentropes, passing through the break in the tropopause that often exists at mid-latitudes (as a result of deformation of 2212 STRATOSPHERIC WATER VAPOR the tropopause into ‘folds’, caused by various tropospheric dynamical features such as low-latitude troughs). The influence of high topography, such as the Himalayas and the Tibetan Plateau is also thought to influence isentropic flow from tropical upper troposphere into mid-latitude stratosphere. Thus, explaining the mean distribution of water vapor in the stratosphere involves an understanding of global mean circulation on the one hand, and on the other a range of detailed tropopause-level processes on scales from a typical depression to an individual cloud. The best interpretation at present involves a range of processes, covering all these scales, with no simple picture of a single determining process emerging. since no barrier to the motion exists. Later in the year the humidity of the lower stratosphere rises as the tropopause becomes warmer, and the lower stratosphere becomes generally moister. This happens particularly in the northern subtropics, possibly due to the effect of the Himalayas and the Tibetan Plateau. In mid-latitudes there is evidence from other observations of an annual cycle in the lower stratosphere. For example, balloon measurements over Boulder, Colorado, show a maximum in summer and a minimum in mixing ratio at about 15 km in March and at about 18 km in September. This phenomenon is consistent with the idea of rapid isentropic flow from the tropics to mid-latitudes. The annual cycle in the tropics gives rise to an annual variation of mixing ratio, and a phase lag in time as the prevailing humidity at the tropopause ascends to higher altitudes. This phenomenon has been named the ‘tape recorder’, for reasons that become clear from Figure 3. These observations, taken from the HALOE experiment demonstrate the power of satellite data for obtaining a global-scale perspective on processes. The figure shows how, below 20 km, the tropical humidity follows the annual cycle seen in Figure 2 earlier. At higher altitudes, up to 30 km, the peaks and troughs occur later, owing to the finite vertical motion from the tropopause upward. The ‘signal’ of the tropopause variations is imposed on the ‘tape’ as a record of humidity with height. An interesting use of this result was to examine the mixing times at different altitudes, which tend to wash out these signals. This study suggested a strong mixing The largest and most prominent variation in lower stratospheric water vapor is the seasonal cycle. Figure 2 uses data from the HALOE experiment to illustrate tropical and mid-latitude seasonal variability. The data are averaged over the years 1991–1999, and mixing ratio values below 3.4 ppmv are shaded. In tropical latitudes, an annual cycle in the mixing ratio is due to the annual cycle in tropopause temperature in the tropics. Low mixing ratios form in December– January near the tropical tropopause, caused directly by the ‘cold trap’ working at low temperatures. These low mixing ratios spread quickly poleward in an isentropic sense until in April they cover the range 601 S–601 N: This isentropic motion is, of course, fast 60° N 0 4.00 3.6 30° N 3.2 0° 4.80 4.00 Latitude 0 4.8 0 4.4 0 5.20 5.6 0 Seasonal and Annual Variability 3.60 30° S 4.40 3.20 60° S J F M A M J A J Month S O N D J Figure 2 Latitude–time evolution of water vapor mixing ratio on the 380 K isentropic surface derived from seasonal cycle fits of the HALOE data. (Source: SPARC 2000.) STRATOSPHERIC WATER VAPOR 2213 60 0.3 6.0 6.0 5.6 5.2 40 1 3 8 4. 4.4 30 4.0 4.0 20 10 4.0 4.0 3.6 30 Pressure (hPa) Height (km) 50 3.6 100 J A J O J A J O J Figure 3 Altitude–time evolution of water vapor mixing ratio over the Equator, from HALOE, derived from seasonal cycle fits to the data. (Source: SPARC 2000.) between the tropics and mid-latitudes in the lower stratosphere below about 20 km, and a region much more isolated from the mid-latitudes between about 20 and 30 km. Nonseasonal Variability In addition to the seasonal variability driven by the annual variation in solar flux, there are other important processes that cause variability on other time scales in the humidity of the stratosphere. Understanding these is an important part of understanding how the climate of our planet might vary with time. The Quasi-Biennial Oscillation (QBO) The QBO has been known to exist in the stratosphere for a number of years. It is a reversal of the mean zonal wind in the lower equatorial stratosphere, with a mean period of about 28 months (though this period varies significantly), and is probably driven by vertically propagating waves from the troposphere. The vertical motion of air through this region is affected by the direction of the mean flow (positive shear of easterly momentum is associated with above normal ascent and vice versa). There are associated temperature perturbations in the lower stratosphere of a few kelvin. Because of these variations in vertical ascent, the water vapor distribution with height is affected, and a small but significant QBO signal can be detected in the mixing ratio at a given altitude, once the seasonal variability has been subtracted from a suitably long data set. Here, of course, long-term satellite data come into their own. QBO signals between 0.2 and 0.5 ppmv have been detected, for example, in the long-term HALOE data set. Madden–Julian Oscillation (MJO) The tropical Pacific region exhibits a phenomenon called the Madden–Julian Oscillation (MJO), a tropical intraseasonal variation with a period of about 30–60 days that has a signal in a number of tropospheric fields. The MJO has a strong effect on tropospheric humidity, and recent work has highlighted the fact that there is a weak signal in the variability of the lower stratosphere, around 100 hPa. This signal is no more than a few tenths of a kelvin, but is nevertheless significant. Long-Term Trends For time scales longer than the QBO F that is, longer than about 2 years F it is of great interest to know whether there are trends that might be significant in terms of the climate of the Earth. In order to examine long-term trends, it is obviously necessary to obtain data sets with the required longevity, accuracy, and precision. In the stratosphere, only two data sets have all the required properties, the frost-point hygrometer balloon measurements over Boulder, and the HALOE satellite measurements from the UARS spacecraft. The former cover the time period 1981 to the present, and the latter the period 1991 to the present. Balloon Measurements over Boulder, Colorado A consistent series of balloon frost-point measurements has been made over Boulder over a 20-year period, and these have been analyzed to determine trend statistics. Measurements have been made on a roughly monthly basis, and heights up to about 28 km have been accessible using the balloon system available at Boulder. The results of this study are given in Table 1. The detected increase of about 1% per year is statistically highly significant at all levels above 16 km, up to 28 km. 2214 STRATOSPHERIC WATER VAPOR Table 1 Stratospheric mixing ratio trends measured above Boulder, Colorado, 1981–2000 Altitude range (km) Gradient of mixing ratio (%/year) Uncertainty in gradient (%/year) 1.3 1.0 1.0 0.4 0.2 0.3 0.2 3 hPa 0.1 0.0 −0.1 Trends from HALOE The HALOE experiment on the UARS satellite has made measurements of a number of stratospheric constituents from 1991 to the present day. These include water vapor (H2O) and methane (CH4). Thus, the trends of both water vapor and the hydrogen parameter defined above, c ¼ 2½CH4
þ ½H2 O
, can be determined. These data have been analyzed by a number of groups in the United States and the United Kingdom, with broadly similar results. Strong increases in both H2O and c have been detected (of between 0.05 and 0.15 ppmv y  1: i.e., a few percent per year) between about 20 and 50 km for both species, for the years 1991–1996. However, the trend detected from an analysis of the years 1996–1999 was statistically indistinguishable from zero. This represents a very significant change in the long-term variability of stratospheric humidity, and indicates that the trend detected between 1991 and 1996 may be due to a particular episodic event, for example, the eruption of Mt. Pinatubo in 1991, which is known to have affected the stratosphere profoundly from 1991 to at least about the end of 1992. Figure 4 shows the de-seasonalized mixing ratio anomalies in c from HALOE data at three levels: 31, 10, and 3 hPa. Solid circles show data for the Southern Hemisphere, and open circles for the Northern Hemisphere. The smooth lines are the results of the use of a smoothing filter to remove the effects of the QBO (see above), which is clearly seen in the 10 hPa data. From these studies, we can conclude that long-term trends in stratospheric water vapor may occur at the level of 1–2% per year, but that events on the time scale of 5 years can produce changes of similar magnitude. We shall see below, however, that even small changes of this order are significant as far as the radiative properties of water vapor are concerned. −0.2 Mixing ratio (ppmv) 16–18 20–22 24–26 0.2 0.1 0.0 −0.1 −0.2 10 hPa 0.4 0.2 0.0 −0.2 31 hPa −0.4 1991 92 93 94 95 96 97 98 99 2000 Year Figure 4 De-seasonalized mixing ratio anomalies in c from HALOE data, at 31, 10 and 3 hPa. Solid circles: Southern Hemisphere; open circles: Northern Hemisphere. (Source: SPARC 2000.) tions in the literature indicate that a fixed increase in stratospheric mixing ratio of 0.7 ppmv could decrease temperatures by up to 3–7 K (in the spring). This change in temperature would have two effects. The first, is to decrease Arctic ozone columns owing to the temperature sensitivity of some of the reactions involved in the ozone balance. Second, the increase of water vapor mixing ratio would raise the saturation temperature required for the formation of polar stratospheric clouds. Thus, an increase in water vapor amounts could have a significant effect on Arctic ozone depletion. According to other recent studies, a further effect of increasing stratospheric humidity might be an added radiative forcing of the Earth’s surface of about 0.2 W m  2, about 25% of the ‘standard’ water forcing, and about 5% of the total forcing thought to be due to CO2 doubling, including feedbacks. Significance of Long-Term Trends of Stratospheric Water Vapor for Climate Water vapor in the stratosphere is a significant greenhouse gas that, by virtue of its temperature, provides a very significant cooling to space. Calcula- Acknowledgement The author is particularly indebted to the authors of Worlds Climate Research Programme Report No. 113 SURFACE LAYER MEASUREMENTS OF TURBULENCE (SPARC, 2000), which provided very valuable background in the writing of its article. See also Arctic Climate. Climate: Overview. Climate Variability: Seasonal to Interannual Variability. Global Change: Upper Atmospheric Change. Middle Atmosphere: Planetary Waves; Quasi-Biennial Oscillation. Observations for Chemistry (In Situ): Resonance Fluorescence; Water Vapor Sondes. Observations for Chemistry (Remote Sensing): IR/FIR; Microwave. Satellite Remote Sensing: Water Vapor. Stratospheric Chemistry and Composition: Hydrogen Budget. Further Reading Brewer A (1949) Evidence for a world circulation provided by the measurements of helium and water vapor distributions in the stratosphere. Quarterly Journal of the Royal Meteorological Society 75: 351. Gille JC and House FB (1971) Limb sounding of the stratosphere. Journal of the Atmospheric Sciences 28: 1427–1442. Harries JE (1976) Stratospheric water vapour. Reviews of Geophysics and Space Physics 14: 565–575. Harries JE (1995) Earthwatch: The Climate from Space. New York: Wiley-Praxis. Harries JE, Russell JM, Tuck AF, et al. (1996) Validation of measurements of water vapor from the halogen occulta- 2215 tion experiment (HALOE). Journal of Geophysical Research 101: 10205–10216. Holton J (1992) An Introduction to Dynamic Meteorology. London: Academic Press. Mote PW, Rosenlof KH, McIntyre ME, et al. (1996) An atmospheric tape recorder: the imprint of tropical tropopause temperatures on stratospheric water vapor. Journal of Geophysical Research 101: 3989– 4006. NASA: Upper Atmosphere Research Satellite web site: http://uarsfot08.gsfc.nasa.gov/ NASA: Earth Science Enterprise Programme: http:// www.earth.nasa.gov/ Newell RE and Gould-Stewart S (1981) A stratospheric fountain? Journal of the Atmospheric Sciences 38: 2789– 2796. Oltmans SJ, Vomel H, Hofmann DJ, et al. (2000) The increase in stratospheric water vapor from balloonborne, frostpoint hygrometer measurements at Washington, DC, and Boulder, Colorado. Geophysical Research Letters 27: 3453–3456. Rosenlof KH, Oltmans SJ, Kley D, et al. (2001) Stratospheric water vapor increases over the past half-century. Geophysical Research Letters 28: 1195–1198. Salby ML (1996) Fundamentals of Atmospheric Physics. London: Academic Press. SPARC (2000) Assessment of Upper Tropospheric and Stratospheric Water Vapour. World Climate Research Programme Report No. 113. World Meteorological Organisation, Geneva. SURFACE LAYER MEASUREMENTS OF TURBULENCE N O Jensen, Risø National Laboratory, Roskilde, Denmark Copyright 2003 Elsevier Science Ltd. All Rights Reserved. Introduction This article focuses on in situ measurements for the study of turbulence in the atmospheric surface layer. Specifically it deals with the eddy correlation calculations of the vertical fluxes of momentum, sensible heat, and latent heat, as well as fluxes of chemical trace constituents such as carbon dioxide (CO2) and other important greenhouse gases, ozone, and gaseous nitrogen compounds. Measuring techniques that are built on aircraft platforms are not considered here. Such techniques can provide area averages of the measured quantities on a horizontal scale of the choice of the experimenter, in contrast to the tower-based techniques that we deal with in this article, which have a ‘footprint’ given by the experimental situation (height of observation and turbulent intensity in the boundary layer). Nor shall we deal here with remote sensing techniques, be they ground-based or airborne. These are typically based on the detection of backscattered energy from a light source (light detection and ranging, lidar), microwave transmitter (radio detection and ranging, radar) or loudspeaker (sound detection and ranging, sodar). The first two operate over a relatively long range while the latter is limited to the scale of the atmospheric boundary layer. Using these devices it is possible to derive wind as well as temperature information. The acoustic backscatter devices (sodars) have become widespread for operational use in airports for ‘nowcasting’ of wind conditions. 2216 SURFACE LAYER MEASUREMENTS OF TURBULENCE Turbulence in the Atmospheric Surface Layer Turbulence in the atmospheric boundary layer is manifested by the eddy motions that provide the transport from the atmosphere to the surface and vice versa. Thus measuring the fluxes and analyzing them in relation to environmental conditions is one of the important tasks of all micrometeorological research. The goal is to devise robust relationships between emanating fluxes and more amenable parameters such as those that can be calculated from numerical models, based on solutions to the equations of fluid motion (Navier–Stokes equations). The need for these kinds of experimental studies derives from the fact that relevant analytical solutions cannot be found, but can only be investigated numerically with a lot of assumptions based on direct experimental evidence (parameterization). However, the numerical procedure imposes a certain gridpoint resolution. This resolution is limited by two factors: the size of the total domain that is to be resolved, and the power of the computer available. Currently the weather and climate models covering major parts of the globe operate with horizontal grid cells of order 100 km. This calls for good parameterizations (the purpose of the experimental studies based on the instruments described here); how to aggregate the fluxes from the various patches onto a grid resolution of 100 km is another matter. Micrometeorology provides an estimation of the fluxes in conjunction with measurements of the relevant vertical profiles of the mean quantities. The classical example is the momentum flux (loss of kinetic energy from the atmosphere) compared to the vertical wind gradient. The modern method of flux determination builds on the concept of covariance between the current vertical velocity, w, and the concentration of interest, c. The ‘instantaneous’ product of wðc  hciÞ ¼ wc0 , where c0 is the ‘instantaneous’ deviation from the mean concentration hci, averaged or summed over a suitable period, for example 30 min. This is by definition the vertical transport of c. This technique is also called the eddy correlation method. A common pitfall is in the correct determination of w. In geophysical flows the vertical velocity is not necessarily perpendicular to the geopotential, but rather perpendicular to the stream surface of the local air flow. For the above-mentioned eddy correlation calculations, it is therefore necessary to measure the instantaneous horizontal wind components in order to define the directions of the streamlines. The question of the measurement of c0 is considered later in this article. In the surface boundary layer the turbulent motions occur on a wide spectrum of scales (both time and space). When a large eddy, perhaps on the scale of the depth of the entire boundary layer (H ffi 1 km), moves past the observation point it can produce a lengthy perturbation on the mean velocity on a time scale of t ¼ H=hui, where hui is the mean wind speed somewhere up in the boundary layer. This would typically be of the order of several minutes. However, vertical motions determining the fluxes are generally on shorter time scales. In order to determine appropriate fluxes, both c and w must be recorded at sufficiently short time scales (rapid) to account for all the relevant motions. Examination of w, i.e., how it varies and the frequency distribution of this variation as seen from the fixed (tower) observer, shows that the variation of w is constant with changing frequency from low frequencies until a certain point. This point, or kink, is closely associated with the production scale of the eddies, which is in turn closely related to the height above the ground of the measuring instrument. At higher frequencies the variation in w falls off (as f 5=3 , according to Kolmogorov, where f is the frequency of the fluctuations). The fall-off in variation intensity is associated with the approach to isotropic turbulence as opposed to the scale at which it is produced. The fluctuations finally vanish at scales on the order of 1 cm, where viscosity smears all fluctuations. When we deal with fluxes, i.e., covariances, it is important to recognize the w spectrum. When we get into the Kolmogorov range proper, perhaps one decade beyond the above-mentioned kink, the turbulence becomes isotropic and therefore the remaining eddies carry no net flux. In consequence we need sensors that can resolve w and c at frequencies up to about 10 Hz at typical observation heights within the surface layer. In this article we first deal with wind sensors and subsequently sensors for temperature (heat flux) and humidity (evaporation). Finally, some details about fluxes of other species are discussed, including a technique (relaxed eddy accumulation, REA) that avoids the need for a rapid c sensor. Fast-Response Wind Sensors For measurements of the vertical wind component a number of devices have been developed, including vertically aligned light propellers and various types of drag anemometer. The latter is based on the relationship between the force on a body, e.g., a golf ball, and wind velocity. The sensing element is usually based on the bending of the supporting beam detected by a SURFACE LAYER MEASUREMENTS OF TURBULENCE 2217 strain gauge (a filament in which electrical resistance changes with strain). Both devices have frequency response problems. These occur in the propeller because of finite starting and stopping velocities and inertia, and in the drag anemometer because of eigenmode vibrations (Kármán vortices). In addition, most propeller and drag anemometers distort the wind flow with their bulk. Another instrument that has been used to measure wind speed in the atmosphere is the hot-wire anemometer. This instrument is optimized for laboratorytype flows (very small size and very rapid frequency response). The principle is that a metal wire is heated by an electric current. The more it is cooled by the (perpendicular) wind, the more current is needed to maintain the temperature of the wire at the set level. Since it is a fine-scale instrument it is only suited to very detailed studies of the atmospheric turbulence – like the eddy cascade towards smaller scales and final dissipation. Few studies have used this instrument in micrometeorological research, and it has not been developed into more rugged designs. The instrument that has won the most widespread use in micrometeorology is the ultrasonic anemometer (Figure 1). This is because it has no fragile or moving parts and does not significantly interfere with the flow, if properly designed. It builds on the principle of the propagation of sound. A transmitter sounds a pulse (typically of 100 kHz, i.e., ultrasonic) and a receiver some distance l away detects it some time t later. This time depends on the speed of sound, c, as well as the local instantaneous wind velocity. If this is done in the opposite direction as well the wind velocity can be derived. The precise relationship is given by eqn [1]:   l 1 1 va ¼  ½1
2 t t0 where va is the wind velocity component along the transmitter–receiver axis. Figure 2 provides an explanation of the principle. Here vn denotes the wind velocity normal to the transmitter–receiver axis. Fast-Response Temperature Sensors Fast-response temperature sensors are used in connection with the covariance or eddy correlation technique to determine the vertical heat flux in the atmospheric surface boundary layer. As mentioned above, the requirement is that the thermometer has a rapid response, i.e., it reacts on temperature fluctuations on a short time scale, of order 0.1 s (10 Hz). One option is to use very thin metallic wires whose heat capacity is low enough for them to follow the Figure 1 Ultrasonic anemometer (Metek USA-1 3D). Note that there are three sets of opposite microphones/transmitters, none of which is along a vertical path. However, from these three components it only requires simple geometry to calculate vertical and horizontal wind components. ambient temperature quite closely. The signal then consists of variations in the electrical resistance of the wire. An alternative method is again to use the sonic principle. If the reciprocals of the flight times as shown in eqn [1] are instead added, then one obtains a measure of the speed of sound (eqn [2]):   l 1 1 c¼ þ 2 t t0 ½2
except for a small error of order va ðva =2cÞ. From the gas law of thermodynamics c is related to the absolute 2218 SURFACE LAYER MEASUREMENTS OF TURBULENCE y y (vnt ′, vat ′) B′ A′ A′ ct ′ l B A x A ct (vnt, vat) x Figure 2 The sketch to the left shows how a spherical sound pulse propagates under the influence of the ambient wind where the center of the pulse moves from A to B while it is detected at the receiver A0 (and vice versa). The sketch to the right defines the variables of importance. temperature through eqn [3], from which the virtual temperature, T, can be found. c¼ pffiffiffiffiffiffiffiffiffiffi gRT ½3
Here g and R are thermodynamic constants that depend on the molecular composition of the gas mixture. In normal atmospheric air the temperature derived from this method will slightly depend on, or needs correction for, the content of water vapor in the air. In principle this is also true for other trace constituents, such as CO2 , but in practice the effect is negligible. Fast-Response Humidity Sensors From the above, it would appear that the sonic method could also be used to derive air humidity in an otherwise known atmospheric air mixture, provided the true T is measured in the traditional way. In practice the signal is not large enough for this technique to be used in normal atmospheric conditions. In a more traditional way, it is possible to measure humidity fluctuations sufficiently rapidly in order to determine the evaporation from the Earth’s surface (water, ground, or vegetation) by using the psychrometer technique: a thermometer is covered by a wet ‘sock’, and the amount of evaporation, and thus cooling of the thermometer, depends on the dryness of the ambient air. For flux purposes the thermometer must be of very thin wire as described above and the sock should have a similar low heat capacity. Methods like the dew-point mirror have not been developed for flux purposes because of the large thermal mass of mirror and condensate. The more common techniques are based on optical principles. These are mainly the Lyman a hygrometer and the infrared hygrometer. The Lyman a hygrometer, as the name suggests, operates at the a line in the hydrogen molecular spectrum. However, the lenses in the optics are made of magnesium fluoride, which is transparent for light at this wavelength. This material is very sensitive and the lenses corrode quite quickly when exposed to normal atmospheric air (humidity!). The infrared technique is in more widespread use. It is based on the broader absorption bands of H2O molecules in this part of the light spectrum. Still another method is based on light from a krypton lamp. Some designs have the optical path in the open, similar to the sonic methods. In other designs the optical path is enclosed and the sample is pumped through the enclosure. Such designs need to allow for damping of fluctuations in air lines and phase lag. Fluxes of Chemical Trace Constituents As mentioned above, it is possible to measure the flux of any trace constituent using the eddy correlation method. The chief limitation is the requirement for a rapid time response of the detector or chemical analysis apparatus. All current methods are based on optical techniques – either directly or indirectly. The infrared absorption technique, already discussed in the context of humidity measurement, is quite common for the measurement of CO2. In fact both measurements are often integrated in the same instrument. Another direct method is based on tunable diode lasers, where the wavelength of the laser light is adjusted to an absorption line that is specific to the molecule (or compound) of interest. This device needs SURFACE LAYER MEASUREMENTS OF TURBULENCE 2219 Figure 3 A sonic anemometer and other flux instruments mounted on a mast in a real micrometeorological setting. The box to the left of the mast at about the height of the sonic anemometer contains a REA control system. very strict temperature control of the light emitter as well as the detector (requires cryostatic devices) and is therefore quite difficult to operate in the field. Other more indirect optical methods are built on the detection of light emission. A classical technique is flame photometry, where the gas is passed through a hydrogen flame and the light emitted from the excited molecules of interest (particular wavelengths) is measured. Other quite widespread methods are based on specific chemical reactions that emit light (chemiluminescence). These are available for fast ozone, NO, and NO2 detectors. A relatively new method that has a lot of potential is the so-called relaxed eddy accumulation (REA) technique. In its simplest form it consists of a pump, a fast two-way valve, and two collector reservoirs (for example Teflon bags or Tenax tubes), together with the vertical channel of a sonic anemometer to control the valve (see Figure 3). When the air motion is upwards the air is led to the ‘up’ reservoir, and vice versa. After a suitable period of time (say 30 min), the concentration of all compounds of interest in the two reservoirs can be determined by conventional slow response analyzers. The flux is then determined from eqn [4], F ¼ bsw ðcþ  c Þ ½4
where cþ and c are the concentrations in the ‘up’ and ‘down’ reservoirs respectively, sw is the standard deviation of the fluctuations in the vertical wind velocity, and b is an empirical coefficient of order 1. The coefficient b is not a constant, though, but depends on the statistical properties of the turbulence and thus, for example, indirectly on the atmospheric stability. There are many variations to the actual design. Online analyzers can replace the bags and analyzers based on a differential principle are ideal. Introducing an interval for the vertical wind velocity (a so-called dead band) in which air is not sampled in either of the reservoirs has the advantage of reducing the activity of the valve and increasing the difference between cþ and c . If the dead band is not kept as a constant but made variable as a fixed fraction of sw , then b turns out to be effectively constant. The value of b is determined by using eqn [4] for a flux F which can also be measured directly by eddy correlation, and then assuming that b is similar for all compounds. The big advantage of this technique is that it opens up for flux measurements of a host of compounds because the chemical detectors do not need to have a fast response. See also Boundary Layers: Observational Techniques In Situ; Surface Layer. Land–Atmosphere Interactions: Trace Gas Exchange. Observations for Chemistry (In Situ): Gas Chromatography; Resonance Fluorescence. Parameterization of Physical Processes: Turbulence and Mixing. 2220 SYNOPTIC METEOROLOGY / Forecasting Further Reading Busch NE (1973) On the mechanics of atmospheric turbulence. In: Haugen DA (ed.) Workshop on Micrometeorology, pp. 1–65. Boston: American Meteorological Society. Dobson F, Hasse L and Davis R (eds) (1980) Air–Sea Interaction, Instruments and Methods. New York: Plenum Press. Hasager CB and Jensen NO (1999) Surface-flux aggregation in heterogeneous terrain. Quarterley Journal of the Royal Meteorological Society 125: 2075–2102. Jensen NO and Busch NE (1982) In: Plate EJ (ed.) Atmospheric Turbulence. Engineering Meteorology, pp. 179–213. Amsterdam: Elsevier. Kaimal JC and Finnigan JJ (1994) Atmospheric Boundary Layer Flows, Their Structure and Measurements. Oxford: Oxford University Press. Lenschow DH (ed.) (1986) Probing the Atmospheric Boundary Layer. Boston: American Meteorological Society. SYNOPTIC METEOROLOGY Contents Forecasting Weather Maps Forecasting D Mansfield, National Meteorological Center, Bracknell, UK Copyright 2003 Elsevier Science Ltd. All Rights Reserved. Introduction This article will consider the role of the human in forecasting for middle or high latitudes where the weather is dominated by synoptic-scale disturbances. The role of the human in forecasting the weather, be it for the next few hours or for up to a week ahead, has changed enormously over the last 30 years. Long gone are the days when the forecaster relied on empirical rules and some very basic dynamics to predict the next day’s sea-level pressure pattern and hence the weather. For some time this part of the task has been carried out for the forecaster, and, with increasing accuracy, by numerical weather prediction (NWP) models. For forecasts up to 36 h ahead serious errors in the predicted surface pressure and upper wind patterns are rare. One important advantage that human forecasters still have over the numerical model is their ability to interpret cloud or moisture patterns from satellites in terms of weather systems. Although the forecaster cannot normally expect to ‘beat’ the computer at predicting the pressure pattern over a large area, there is still scope for local adjustments based on an assessment of the accuracy of the initial conditions upon which the numerical forecast is based. NWP models are less accurate when predicting the actual weather elements such as precipitation amount and type, cloud amounts, fog, etc. The forecaster’s role has increasingly become that of interpreting and refining raw NWP products, especially in terms of weather elements. However, there are still a few occasions when numerical guidance can go seriously wrong and the forecaster must continually monitor the NWP output for signs of this and be prepared to modify the whole forecast if necessary. There are many different roles required of forecasters, depending on who their customers are. They may be providing central guidance on the synoptic-scale evolution to other (local) forecasters, or providing forecasts for the general public (most often via the media), to the military, to civil aviation or to other commercial customers and on a variety of time scales. One common aspect of all these roles is timeliness. A weather forecast, particularly a short-range forecast, is a very perishable commodity, and even forecasts for several days ahead may be subject to adjustment after 12 h, when the next set of NWP products are produced. It is normally 2–3 h after data time before NWP products become available to the forecaster and there is often a further chain of processing and briefing before the forecast reaches the customer. As NWP models continue to improve and mesoscale and single–site models enable more accurate prediction of local weather, the ability of the forecaster to SYNOPTIC METEOROLOGY / Forecasting 2221 add value to the numerical guidance will continue to decrease, at least on average. The forecaster’s role for most of the time will become that of interpretation. However, there will still be rare occasions when the NWP models produce large errors. Although not making much impact on skill scores such as rms errors of mean sea-level pressure (MSLP), these are likely to be associated with rapidly developing systems that may produce life-threatening extreme weather events, and it is in recognizing these occasions that forecasters will continue to prove their worth. by the numerical analysis scheme and the new analysis t þ 0 h field of the next model run will be a closer fit to the observations. However, if the difference between the observations and the background field is large, the observations may be rejected by the quality control procedures. In some centers such as the UK Met Office, it is possible for the forecaster to intervene to assist the quality control scheme make the correct decisions, to add weight to crucial observations in the assimilation scheme, and even to invent ‘bogus’ data where satellite or radar imagery suggest the NWP background is in error, but where there are no real observations in the area to correct this (see Figure 1 for an example). Central Guidance Most national weather services of developed countries have a central guidance center whose role is to provide an interpretation and assessment of the latest NWP products and to issue warnings of any expected severe weather likely to be a threat to life or property. In many cases this is extended to guidance on the actual weather details expected so as to ensure that all forecasts issued by different offices of the national weather service are consistent. Because of the time taken to disseminate the guidance, this is usually for the period from about 6 h ahead to perhaps 5–7 days ahead. Analysis The first step is for the forecaster to analyze the current situation. Up until a few years ago this would normally have involved hand drawing of surface-pressure maps and upper air height contours. Nowadays computerdrawn ‘first guess’ charts (usually a 3 or 6 h forecast from a previous model integration) are nearly always close enough to reality for the forecaster to use these along with surface and upper air observations and satellite and radar imagery to recognize the dominant weather systems and processes at the current time. Most NWP models are global in extent, but for shortperiod forecasts, the forecaster will normally restrict his or her interest to the forecast region and an area upstream, though this may be fairly large (typically the whole of the North Atlantic for European forecasters and most of the North Pacific for those in the United States or Canada). Using conceptual models of these processes and systems, the forecaster then compares satellite and radar imagery and surface and upper air observations with the computer-drawn charts in order to assess the accuracy of the NWP first-guess fields. If there is a discrepancy between the NWP field and the observations, the forecaster will be alerted to a possible problem with the subsequent forecast. In most cases, if the difference is small, it will be corrected Diagnostics Actual weather elements such as low cloud, fog, surface temperature, and some details of the precipitation, particularly showers, are less well forecast by NWP models than the basic pressure patterns. In order to be able to add value to the raw forecast in these areas the forecaster has to understand the dynamics of the large-scale environment in which the smaller-scale processes are embedded and the way in which the different scale processes interact. Forecasters have access to many diagnostic fields from the NWP models to help them in this task. As well as surface-pressure maps and upper air contour charts, most commonly used are model relative humidity (as a proxy for cloud) for comparison with satellite imagery, and vorticity, vorticity advection (Figure 2A), thickness (a measure of the mean temperature between two levels in the atmosphere), and thickness advection, at various heights, to monitor the two most important aspects of large-scale flow. Wind strength (Figure 2B) or wind vectors or barbs are also useful in delineating model jet cores that can be compared with satellite imagery. Jet cores are often apparent on infrared images as a linear contrast between bright areas of cold high cloud on the warm side of the jet and dry areas of subsided air on the cold side. These features are even more apparent in water vapor images. In differentiating between moist and dry regions of the middle troposphere, these images give information about the atmosphere in cloud-free regions and, through the associated changes in humidity, can indicate ascending and descending motion associated with developing weather systems before this becomes apparent in other imagery. Since potential vorticity (PV) has become available as a diagnostic from most NWP models, the strong relationship between water vapor imagery and the distribution of potential vorticity is becoming increasingly used as a tool to check the initial conditions and 2222 SYNOPTIC METEOROLOGY / Forecasting Figure 1 An example of a VDU display used to check NWP background field with observations and satellite imagery, in this case t þ 6 h background field (blue contours) and t þ 0 analysis (red) MSLP compared with surface observations and infrared satellite image over part of the central North Pacific. Important observations are arrowed. An observation from a drifting buoy (A) has a pressure of 997.0 hPa, nearly 13 hPa lower than the t þ 6 background. The satellite image supports the idea of a more rapidly deepening low than suggested by the NWP field, but the ship observation (C) to the south looks unrealistically low, though the 30 knot (15 m s  1) northwesterly wind supports the idea of a deeper low to the north. A pressure of 1008.0 hPa looks more likely than 1000.8; mistakes in coding frequently lead to this sort of error. To help the assimilation scheme a bogus observation (B) of 1002 hPa and a 40 knot (21 m s  1) southerly wind has been inserted to the southeast of where the low center was estimated to be. As a result, the t þ 0 NWP background pressure is a much better fit to the real and bogus observations and more in line with the forecaster’s interpretation of the satellite imagery. (Reproduced by permission of the Met Office.) early stages of a forecast. Dark (dry) areas in the image are associated with high PV in the upper troposphere, while in dynamically active regions, particularly when cyclogenesis is taking place, contours of PVon a quasihorizontal surface such as a pressure surface or isentropic surface curve anticyclonically over areas of ascent and developing cloud. One problem is picking a suitable surface on which to display the PV, as the area of interest is usually just below the tropopause and the associated pressure and potential temperature will vary with the season, current weather situation, and geographical location. Using the fact that PV increases sharply across the tropopause from around 110  6 m2 s  1 K kg  1 (1 PV unit; PVU) to around 6 PVU in the lower stratosphere the problem can be avoided by plotting the height of a PV surface (usually 2 or 1.5 PVU) that is always close to the tropopause. An example is shown in Figure 3. On some occasions the relationship between the PV and water vapor imagery can be confused or misleading, particularly in the very early stages of cyclone development. If the development is initially taking place in the low to middle troposphere, the image may show the pattern of ascent and descent before it influences the PV distribution at higher levels. However, another diagnostic, so-called ‘pseudo-imagery’, is becoming available to the forecaster to cope with these problems. The radiance at the top of the atmosphere in the water vapor channel is computed from the numerical model values of temperatures and humidity and can be displayed either as an image or as contours of brightness for comparison with the real imagery. An example is shown in Figure 4. The comparison of the model PV and the real image (Figure 4A) shows a possible problem south-west of Portugal where the numerical model high PV is associated with low radiance in the water vapor image, but the pseudo-imagery (Figure 4B) also shows low radiance in this area, confirming that this is due to convective cloud penetrating into the otherwise dry upper troposphere. However, near the center of the image, to the south-west of the Azores, a small PV maximum also corresponds to a region of low radiance in the real image, but in this case the dry, dark area in the pseudo-image extends south to SYNOPTIC METEOROLOGY / Forecasting 2223 40 38 36 34 32 30 28 26 24 22 20 18 16 14 12 10 522 528 534 528 540 546 552 552 558 564 558 564 598 516 528 534 540 546 570 576 582 (A) 980 80 75 70 1000 1000 65 1000 60 55 1020 1020 1040 50 45 1040 40 1060 35 30 1080 (B) Figure 2 Examples of model diagnostics used to interpret and understand NWP output. (A) 500 hPa geopotential height contours and absolute vorticity (colors). The forecaster can quickly see where areas of large vorticity advection contribute to ascending and descending motion in the model. (B) 250 hPa geopotential height contours and wind strength (colors). Regions of maximum wind strength (jets) can be compared with indications of jet axes from satellite imagery and important dynamical regions at the entrances and exits to jets and their relative strengths quickly assessed. (Images courtesy of the NOAA-CIRES Climate Diagnostic Center, Boulder, CO, USA from their website at http://www.cdc.noaa.gov/.) coincide with the PV maximum, suggesting a small error in the model in this area. The pseudo-image gives a much closer comparison with the actual image, but still has to be used with caution and is best used in conjunction with the PV comparison. Apparent discrepancies between the two images may be due only to poor model simulation of the relative humidity, which may be unimportant in the subsequent developments. On the other hand, a close fit can also be misleading, as many NWP models assimilate water vapor radiances. Any adjustment to the NWP radiance is mostly through the humidity, so it is possible that this may mask an underlying problem with the dynamics. Interpretation Most global models will distinguish between and display different types and phases of precipitation, i.e., steady rain or snow from large-scale ascent and showers due to local convection. However, it is still necessary to refine the NWP output in these areas. For example, in most models, showers cannot be advected from their source region and therefore stop abruptly and unrealistically at windward coasts in winter as the air transfers from over the warm sea to over cold land. The extent to which showers penetrate inland will depend not only on local orography, not fully resolved by the numerical model, but also on the large-scale vertical motion. An important aspect of the precipitation in winter is the boundary between, rain, snow, or freezing rain (ice storms). There is a very fine balance between these different types of precipitation when the low level temperature is close to 01C, depending on the initial vertical profiles of temperature and humidity, and the balance between thermal advection and the cooling of the air by evaporating or melting precipitation, which in turn will depend on the precipitation rate. Any of these physical processes may be inadequately parameterized, but the correct forecast of the type of precipitation is crucial in issuing timely warnings of severe weather. The forecaster must use his or her experience and knowledge of any weaknesses in the NWP models to try to add value to the forecast. Although NWP models indicate the possibility of strong winds and heavy showers, the forecaster still has to distinguish those occasions with the potential for severe weather, such as violent thunderstorms, tornadoes, hail, and downslope winds, which are not directly forecast by numerical models, and issue advanced warnings. In regions where such severe weather is common, short-period detailed forecasts are issued locally using specialized models, radar, and other forecast aids. Correction When comparing model fields over the first few hours of a forecast with observations and satellite and radar imagery, the forecaster often finds small discrepancies such as not enough or too much rain in the NWP output, fronts or rainbands too fast or slow, or depressions not quite deep enough and hence winds not strong enough. The forecaster can then apply appropriate adjustments to the NWP forecast, 900 . 750 . 2224 SYNOPTIC METEOROLOGY / Forecasting 750. 90 0. 10 50 . . 750 750. .6 60. 900 . 1050 . 1300 . 1350 . 80. 900. 900. 80. 1050. 1050. 1200. 80. 1200. 80. 1350. Figure 3 An example of a water vapor image for part of North America overlaid with contours of the height of the PV 5 2 PVU surface (red) and wind strength on the surface (blue). The strong gradient of PV 5 2 height at the foot of the figure and maximum wind speed corresponds to the edge of the bright area showing that the model jet is correctly positioned while the minimum in PV 5 2 height just to the rear (north) of the dark area over the south-east of the Great Lakes shows that the cold trough (area of high upper level PV) in the model is also correctly positioned – reassuring for the forecaster as in the subsequent forecast theses two features interact to form an intense depression over the Atlantic Ocean. (Image and model fields from the French ARPEGE model supplied by Meteo-France Forecast Laboratory.) assuming that these errors persist through the forecast period or decay or grow in a simple manner. This technique is effective up to 24 h or perhaps 36 h ahead. The adjustments are usually made in terms of written or verbal guidance or by the adjustment of single time forecast pressure charts, but techniques are becoming available to adjust electronically the NWP fields in a dynamically consistent manner at all time frames before the output is disseminated to other users. Very rarely the NWP initial conditions may be so seriously in error that the forecaster has to disregard the model guidance and use his or her own synoptic and dynamic knowledge to make a new forecast. Figure 5 shows a satellite image for the North Atlantic on 23 December 1997. Conceptual models of cyclogenesis suggest that the cloud area (a) indicates a rapidly developing depression, whereas the NWP field showed only a very weak circulation (b). Although not a bad fit to the available surface observations, the analysis and subsequent forecast were considered completely inadequate, a theory confirmed by the development of the cloud area in the next 2–3 h. The forecaster overrode the NWP model to forecast a deep depression just west of Ireland 24 h later (Figure 6). The manual forecast depression was 16 hPa deeper than the unmodified forecast and only 3 hPa higher than the actual depth, though marginally displaced, and enabled the forecaster to give timely warning of damaging winds over parts of Ireland and the UK. Medium Range Even after 24 h, different models sometimes show significant differences in detail of weather patterns. An example is shown in Figure 7. Although the forecast pressure pattern (Figure 7A) is hardly different in the two models, apart from a deeper trough over Tennessee, there are large differences in the predicted rainfall over the west coast of the United States, over Mexico, and particularly over the south-eastern states (Figure 7B). The forecaster has to use his knowledge of the strengths and weaknesses of the two models to help decide which is more likely to be correct. Beyond about 36–48 h ahead, errors in the initial conditions or those due to imperfections in the numerical models have grown such that forecasts from different initial times or by different forecast centers normally begin to differ. It is no longer possible to predict with confidence details or exact timings of weather events, though the general evolution and type of weather systems likely to be experienced can usually be predicted out to 4 or 5 days, and sometimes beyond this. SYNOPTIC METEOROLOGY / Forecasting 2225 Figure 5 Infrared satellite image over part of the North Atlantic for 1200 UTC 23 December 1997 overlaid with first guess NWP t þ 6 h MSLP field (contours every 4 hPa). The area of cloud (a) indicates a rapidly developing depression, whereas the model has only a weak depression (b). (Reproduced by permission of the Met Office.) Figure 4 (A) An example of a water vapor image overlain with contours of PV (blue, interval 0.7 PVU) and geopotential height (yellow, interval 120 m) on the 300 hPa surface. (B) Corresponding pseudo water vapor image with the same fields superposed. (Image and numerical model fields supplied by Servicio de Técnicas de Analisis y Predicción (STAP), Instituto Nacional de Meteorologia (INM) Madrid, Spain.) At this range it is no longer possible to extrapolate errors in the initial conditions, nor is it possible to beat the models at a dynamical forecast, but there is still a role for the human forecaster. Most large forecast centers exchange raw NWP output with one or two other centers for use as backup, so that the forecaster can usually compare output from two to three or more different model integrations for their region of interest in the same format on screen. In addition to this, the output from many global models is available via the Internet, so that the forecaster may have available as many as 10 different models to choose from. At the same time, several centers around the world are addressing the problem of the uncertainty in the initial conditions and the subsequent error growth by running ensembles of forecast with slightly different initial conditions in an attempt to cover all the possible evolutions of the real atmosphere. In spite of the inherent uncertainty, many customers still require a categorical forecast. The mean of the ensemble of different forecasts is on average more skillful, at least in rms terms, than an individual forecast because it averages out the less predictable smaller-scale features, but by its nature is very bland and does not give a good indication of the actual weather. The forecaster must use his or her judgment and synoptic experience to select most likely evolution or ‘blend’ elements from different models, the socalled deterministic forecast. However, it makes more sense to couch forecasts at this range in terms of probabilities. Even with the deterministic forecast, this is done to some extent by the confidence placed in the forecast. It is important to try and convey this in public service forecasts. If a large, slow-moving anticyclone covers the region, the forecaster may be almost 100% confident of dry weather, but in a more changeable spell of weather, even though the most probable forecast is for a transient ridge of high pressure to bring a dry day, possible errors in timing could mean that there is still a 50% chance of rain. Having decided on the most likely evolution, it is important to convey the degree of uncertainty associated with this, particularly when issuing guidance to other forecasters so that they can couch the forecast for their customers in suitable terms. For forecasts of point probability, normally expressed in terms of the likelihood of a threshold being exceeded, such as wind speed of gale force or more, the 2226 SYNOPTIC METEOROLOGY / Forecasting Figure 6 Infrared satellite image for area over and west of UK for 1200 UTC 24 December 1997 overlaid with 24 h numerical forecast MSLP (blue, dashed contours) and 24 h forecast modified by the forecaster (red contours), contour interval 8 hPa in each case. The satellite image clearly suggests a deep depression and surface observations confirmed that the modified forecast, of a depression of 976 hPa was much more accurate than the NWP version (990 hPa). (Reproduced by permission of the Met Office.) ensemble forecast can give a direct estimate. However, ensembles based on a single model, in spite of perturbations to the physics within the model as well as to the initial conditions, still do not cover the whole spread of possible outcomes. An ensemble of different models such as accessed by the forecaster via the Internet often has greater spread (though it is not uncommon for all models to agree but still differ from reality). There is also still the problem that the models may not accurately predict the weather elements, in spite of having the correct pressure pattern. There is therefore still scope for the forecaster to add value to probability forecasts of individual weather elements, though time will necessarily limit this to a few crucial parameters at only a few geographical locations. A logical conclusion is then for these corrections to the ensemble forecast to be applied to an appropriate degree at surrounding locations. Specialist Forecasts Aviation Forecasts for aviation again rely very heavily on NWP guidance. They can be divided roughly into three types of forecast.  High-level significant weather forecasts. These are forecasts of conditions near the tropopause where jet airliners fly.  Low-level significant weather forecasts. These forecast conditions up to around 10 000 ft (3048 m), used by ‘general’ aviation, e.g., private pilots, small local airlines, military aircraft, couriers, etc.,  Terminal airfield forecasts (TAFs). These are forecasts of surface wind and weather elements at specific airfields. Upper-level significant weather charts are produced centrally by centers designated by the international Civil Aviation Authority to display jet streams, the level of the tropopause, and any high-level aviation hazards, in an agreed format, and are usually valid for fixed times 18–24 h ahead and are updated every 6 h. An example is shown in Figure 8. At normal flight levels, the weather does not often present a serious hazard to modern airliners. The main concerns for airlines are the temperature and wind speed, which will affect fuel consumption. These are generally forecast very well by NWP models and most companies take direct NWP forecasts of winds in digital form for use in flight planning. Only very rarely will a forecaster see the need to correct the NWP winds. The main hazards at these levels are thunderstorms and clear-air turbulence, though it is also the responsibility of weather services to track and warn of volcanic ash. SYNOPTIC METEOROLOGY / Forecasting 2227 576 570 564 558 552 546 540 534 528 522 516 510 504 498 492 486 480 474 1000 992 1024 1016 1024 1016 1024 10 1 1016 1016 1024 1024 1024 1024 1016 1016 576 570 564 558 552 546 540 534 528 522 516 510 504 498 492 486 480 474 1024 1024 1016 1024 1016 1016 1024 992 10 1032 1016 1016 1024 1024 24 1016 (A) 3 2.5 2 1.75 1.5 1.25 1 0.75 0.5 0.4 0.3 0.2 0.1 0.01 3 2.5 2 1.75 1.5 1.25 1 0.75 0.5 0.4 0.3 0.2 0.1 0.01 (B) Figure 7 (A) Comparison of 24 h forecasts from two NCEP models, valid 0600 UTC 22 February 2001. Solid contours are of sea-level pressure, every 4 hPa. The colours represent the thickness layer between the 1000 and 500 hPa, a measure of the mean temperature of the lower troposphere. (B) Comparison of the forecast rainfall accumulations for the same forecasts as in (A) for the 12 h up to 0600 UTC. (Images provided by the NOAA-CIRES Climate Diagnostic Center, Boulder, CO, USA from their website at http://www.cdc.noaa.gov/.) Areas of thunderstorm activity are reasonably well forecast by the NWP models, but it is still necessary for the forecaster to check and sometimes correct details such as cloud top height. Forecasters must also use their experience to decide whether thunderstorms are likely to be isolated, in which case they are not considered a hazard, or embedded in other cloud layers so that they cannot be easily detected, or difficult to avoid due to their spacing or due to being in a line. Occasionally the models may misplace or miss areas of thunderstorms altogether, especially in the tropics, where a series of recent satellite images may be a better guide to activity over the next 24 h. Clear-air turbulence occurs in areas of strong wind shear, normally around jet streams. It is not associated with cloud and therefore cannot be detected in advance, and can be sufficiently violent to cause injury or even death to passengers or aircrew if not restrained by seat belts. However, it is a very intermittent phenomenon and impossible to forecast precisely at present. NWP models provide an indication of regions of strong vertical or horizontal shear where turbulence is likely to occur, but this is a necessary rather than a sufficient condition. Forecasters can add value by using conceptual models of the type of airflow most likely to lead to actual severe turbulence to refine the forecast. Areas where a risk of moderate or severe turbulence is expected are marked on the significant weather charts along with the range of heights over which the hazard is expected to extend. However, most of the time, aircraft flying through these areas experience no serious problems. For this reason pilots encountering severe turbulence make an immediate report, which is relayed to the forecaster, who then issues a more definite forecast in the form of a SIGMET. This is a text forecast, which is disseminated with maximum priority to all aviation users, so that any following aircraft may take avoiding action or at 2228 SYNOPTIC METEOROLOGY / Forecasting PGEE07 181800 KKCI 300 WAFC WASHINGTON FIXED TIME FORECAST CHART ICAO AREA A SIG WX FL 250–630 VALID 18 VTC 19 JUN 2001 300 CB IMPLY MOO OR SEY TURE. ICE AND HAIL HET IN FLIGHT LEVELS. ALL SPEEDS IN KNOT CHECK SIGMETS FOR VOLCANIC ASH 90 60 300 L 340 450 FL360 30 420 ISOL L EMBO CB 450 XXX 120 ISOL 450 EMBO CB 420 XXX ISOL EMBO CB 450 XXX 500 ISOL EMBO CB 420 XXX 80 500 90 340 XXX 330 260 M 15 300 L 450 500 30 30 460 L 60 ISOL EMBO CB 450 XXX 0 FL33 500 390 G 450 10 500 340 XXX 0 32 FL 0 34 FL 450 60 0 33 FL ISOL EMBO CB 300 XXX ISOL EMBO CB 300 XXX 30 FL330 120 60 0 32 FL 340 300 L 60 370 XXX ISOL EMBO CB 300 XXX FL320 60 10 500 SOUFRIERE HILLS 16.7N 62.2W 530 500 30 ISOL EMBO CB 420 XXX ISOL EMBO CB 300 XXX D ISOL Figure 8 Part of a high-level aviation significant weather chart, produced by World Area Forcast Center Washington, valid 1800 UTC 19 June 2001. Surface fronts are shown by conventional symbols, and jet streams (wind speed 480 knot) by the magenta-colored arrows, with the maximum wind strengths in red (each barb represents 10 knot, and the solid triangles 50 knot). Yellow dashed lines outline areas of forecast moderate or severe clear air turbulence. The height range (in hundreds of feet) over which the turbulence may occur is indicated by the associated text. The green scalloped areas denote areas of significant thunderstorm activity, with details of flight levels affected in the associated boxes. (Provided by the NOAA-CIRES Climate Diagnostic Center, Boulder, CO, USA from their website at http:// www.cdc.noaa.gov/.) (1 knot 5 5.144 44  10  1 m s  1.) least ensure seat belts are in use. If a forecaster is sufficiently convinced of a high risk of severe turbulence he or she may issue a SIGMETwithout any actual aircraft reports. SIGMETs are also issued when there is very high confidence in the forecast of other hazards such as embedded thunderstorms, line squalls, and severe low-level turbulence, or when these events are observed. Low-level significant weather forecasts are also mostly produced centrally, but nationally rather than regionally, and are normally valid for shorter periods, typically up to 9–12 h ahead, though planning forecasts are produced from some centers for up to 36 h ahead. As well as the thunderstorms and turbulence (though in this case normally low-level turbulence due to strong winds flowing over the Earth’s surface), low cloud (especially where it covers hills) and icing are the main forecast parameters. Although NWP provides a framework of the positions of frontal zones, areas of convection, etc., parameters such as the amounts and base of low cloud, visibility, and the likelihood of icing are not well forecast numerically, and the forecaster relies more on experience and extrapolation of present conditions, subject to any changes in the large-scale conditions indicated by the NWP models. The third type of forecast, the TAF, is normally valid for 9 h ahead, although at major airports, forecasts of up to 24 h ahead are provided to give airlines an idea of likely risk of long-haul flights being diverted. Forecast parameters are wind speed and direction, cloud amount and height of base and visibility, plus any weather conditions that may be a hazard, such as thunderstorms, hail, snow, freezing rain, and mist or fog, though the latter are also implied by the visibility. These forecasts have traditionally been produced locally on site by forecasters who have a great deal of experience of the peculiarities of the particular airfield, and are based largely on extrapolation of local or upstream conditions after allowance for diurnal changes, and effects of local topography and an idea of the synoptic-scale development. This is still true in many cases, especially at military airfields, but SYNOPTIC METEOROLOGY / Forecasting 2229 improved local detail and better estimation of actual weather parameters from mesoscale numerical models has meant that it is becoming possible for these forecasts also to be produced centrally with a single forecaster responsible for the TAFs for a dozen or more airfields. always detailed enough to represent these effects on the wind speed and direction or sea state, nor do they adequately represent small-scale changes in sea surface temperature likely to have an important impact on mist or fog formation. Local Forecasts Marine Forecasts Detailed forecasts of surface wind speed and direction, visibility, and sea state are normally provided for up to 24–36 h ahead. For the high seas, well away from land, numerical models provide a good estimate of all but visibility, though it may be necessary for the forecaster to make some adjustments in accordance with any central guidance on the perceived accuracy of the latest NWP forecast. Visibility is estimated, within broad limits, from knowledge of the source of the air mass, the air temperature relative to the sea surface temperature (will the air be cooled by the sea to form mist or fog?), and consideration of recent ship reports in the same air mass. Coastal forecasts rely slightly more on local knowledge and interpretation of numerical output around complex coastlines, as numerical models are not 10 14 18 22 The local forecaster has to consider the numerical model output and any corrections that may be made to this in the central guidance, then adjust the forecast for any small-scale effects due to local topography that may not be fully resolved by the numerical model. He or she will be concerned with the combination of these effects on weather parameters such as rainfall rate, rain–snow boundaries, cloud amount and sunshine, temperature, fog, and how they vary across the region. The forecaster will also have to understand the synoptic-scale dynamical processes taking place in order to make sensible adjustments to the numerical forecast. Accurate forecasts of temperature at individual sites are crucial for forecasts of fog, frost, and snow in winter, and for showers, and in particular thunderstorms, in summer. The central guidance may indicate 26 30 300 400 500 CCL–! LCL–!600 b a 12 16 20 700 800 900 1060 Figure 9 Radiosonde ascent from Denver, Colorado, at 1200 UTC 29 April 2001. The vertical scale is logarithmic in pressure and approximates to height. The diagonal pale blue lines are temperature in 1C, every 101. The solid red line shows the temperature profile and the green line the profile of humidity mixing ratio. The blue line represents the temperature curve of a parcel rising from the surface without any mixing with its environment and with an initial temperature of 201C. The surface pressure is around 830 hPa because the station is over 5000 ft (1640 m) above sea level. (Environmental curve provided by the NOAA-CIRES Climate Diagnostic Center, Boulder, CO, USA from their website at http://www.cdc.noaa.gov/. The parcel ascent curve was added by the author.) 2230 SYNOPTIC METEOROLOGY / Weather Maps that these weather elements are to be expected over a broad area, but the local forecaster must estimate the risk at individual locations. There are several semiempirical models available to forecasters to estimate diurnal temperature changes that can be used to check and refine NWP output. The forecaster will use profiles of temperature and humidity from local radiosonde ascents, plotted on an aerological diagram, to diagnose the type and height of any cloud layers and assess the likely developments due to diurnal changes in temperature. The forecaster will then compare the basic profile and his or her analysis with those from the numerical model, and in particular assess the potential for shower or thunderstorm development. An example is shown in Figure 9. The solid red line shows the temperature profile and the green line the humidity mixing ratio, or dew point temperature. The temperature at the surface is colder than the air just above as this a night time profile, but as the temperature at the surface rises during the day the temperature will become warmer than that just above the surface and the air will begin to rise. Unsaturated air will cool as it rises, following the red dashed lines. At the same time the humidity mixing ratio will remain constant and the dew point will therefore follow the yellow dashed lines. In the example the surface temperature must reach 201C before the air can rise sufficiently to reach the condensation level and form convective (cumulus) clouds. The temperature will fall along the blue line until the air parcel becomes saturated at the point A, after which it will cool more slowly due to the release of latent heat as the water condenses in the cloud. In an unmixed parcel the temperature would now follow the dashed green lines. The continuing blue line therefore gives the maximum height to which a parcel could rise. In practice, except in the core of a large cloud, mixing with drier air outside the cloud will lead to re-evaporation of the some of the cloud water and the parcel will cool more quickly and air parcels would be unlikely to rise beyond the point B, as they would then be cooler than the surrounding air. However, the forecaster would have to consider if the air at this level is likely to be cooled by the large scale motion or if local conditions could lead to warmer or moister conditions near the surface. A small change may allow the air to rise all the way to the tropopause causing the formation of heavy showers or thunderstorms (unlikely though in the example, as the surrounding air is very dry and mixing would cool the parcel back towards the environment temperature). The forecaster would check his or her prediction of showers with the NWP output. If no showers were indicated, the forecaster would need to ask why. Does the model surface temperature reach the required value? Does the model represent the observed sounding adequately in its initial conditions? The formation or otherwise of even severe storms may hinge on the detail of a shallow layer of higher temperature, which may not be resolved by the numerical model. See also Aviation Weather Hazards. Cyclogenesis. Satellite Remote Sensing: Cloud Properties; Precipitation. Synoptic Meteorology: Weather Maps. Turbulence and Mixing. Weather Prediction: Data Assimilation; Ensemble Prediction. Weather Maps R Reynolds, University of Reading, Reading, UK Copyright 2003 Elsevier Science Ltd. All Rights Reserved. Introduction Weather maps provide an irreplaceable, succinct summary of a wide variety of atmospheric phenomena and characteristics. They are a critically important working tool for the operational and research meteorologist, both of whom become very familiar with ways of illustrating features of interest that can be immediately appreciated by their peers. In a sense, weather satellite and weather radar images are maps of aspects of the weather. However, this section will not deal with remotely sensed fields, but with maps that portray weather-related features at the surface or in the upper air. History Surface Weather Weather maps are nothing new. Synoptic meteorology is concerned with understanding relatively largescale weather-producing disturbances like frontal SYNOPTIC METEOROLOGY / Weather Maps 2231 depressions, tropical cyclones, and anticyclones – features that have a horizontal scale of many hundreds to a few thousand kilometers, and a lifetime counted in days rather than hours. The earliest instrumental weather records began soon after the invention of the thermometer and barometer. Robert Hooke’s manuscript from 1664, kept at the Royal Society in London, summarized the keeping of such early observations. He did in fact appreciate the value of a network of weather observations, although it would be many decades before that became a reality. The late seventeenth and early eighteenth centuries saw the simultaneous use and gradual expansion of thermometry and barometry to many western European countries and America. This was also the age of sail, with many navies plying the Atlantic Ocean and other oceans. It was the English natural philosopher Edmond Halley (1656–1742) who published the very first map to depict an atmospheric variable, in the Philosophical Transactions of the Royal Society of 1686. This was strictly a climate map, to illustrate the mean surface winds over the tropical regions of the Atlantic and Indian Oceans (Figure 1). The stimulus for producing the first weather map proper lies with a letter written to the Annalen der Physik in 1816 by Professor Heinrich W. Brandes of Breslau University (now Wroczaw University). He suggested that a daily surface weather map could and should be produced for part of the period from 1781 to Figure 1 Halley’s map of trade winds and monsoons. (From Halley E (1686) An historical account of the trade winds, and monsoons, observable in the seas between and near the tropicks, with an attempt to assign the physical cause of the said winds. Philosophical Transactions of the Royal Society 16: 153–168.) 2232 SYNOPTIC METEOROLOGY / Weather Maps It appears however, that this significant pioneering work was never summarized in map form – until Brandes’s proposal that led him to plot 365 maps for 1783. These maps have never been found however; they were not a part of his article on 1783’s weather in the Annalen of 1819. However, there are published reconstructions of his weather map for 6 March 1783 that illustrate both the simultaneous distribution of the departure of pressure from average and surface winds (Figure 2). In 1831, the American James P. Espy (1785–1860) organized a committee from his base in the Franklin Institute in Philadelphia to collect weather data: in 1834 the Joint Committee on Meteorology was formed by the Franklin Institute and American Philosophical Society with Espy as chairman. The first American weather map based on widespread observations appeared in an 1837 issue of the Journal of the Franklin Institute (Figure 3). The prime problem with all these early endeavors however was that the maps could be drawn only after the event. They did, though, offer meteorologists at least some insight into the scale of synoptic features, how pressure and wind appeared to be related, and how pressure features moved and evolved. The full utility of the weather map had to wait for a truly momentous event for the world as a whole, when Samuel Morse connected Washington and Baltimore by electric telegraph in 1844. This brilliant development paved the way for the ultimate ‘live’ mapping of weather observations and so to Figure 2 Brandes’s weather map for 6 March 1783 (reconstruction). (Reproduced with permission from Wilhelm Trabert (1905) Meteorologie und Klimatologie. Leipzig: Franz Deuticke). L 1792: this was the era of the Meteorological Society of the Palatinate, based in Mannheim, Germany. The society fostered the science during this period when observations were taken three times a day, collected from 39 people across 18 mainly European countries. .O IO NTAR Albany Boston Medfield Middleto wn Batavia Silv er L R LE d n gto B al t im Richmond B. Staunton ar w a ke ti sp e Alexandria C he C i n cin n a la o re Emmetsb OC Gettysb De Springfield n shi Wa EA N Foxburg F l u sh i n g e B. ATLANTIC C l e a v e la n . rk N. Yo Phi lad el p hi a IE Figure 3 Espy’s weather map for 20 June 1836. (From Espy JP (1841) The Philosophy of Storms. Boston: Charles C. Little and James Brown.) SYNOPTIC METEOROLOGY / Weather Maps 2233 producing up-to-the-minute weather maps. By 1860, the Smithsonian Institution in Washington, DC had organized the electronic transmission and display of current weather reports from some 45 companies in the United States. The details were presented on a large map on public display in the institution. In Europe, some few years earlier, the world’s first same-day weather maps were being offered to the public gaze at the Great Exhibition of 1851. This was held in the Crystal Palace, situated at that time in Hyde Park in London. From 8 August to 11 October the public could purchase a copy of the day’s weather map for the British Isles (Figure 4). From about 1863, the Daily Weather Map Company of the Strand, London, offered monthly sub- scriptions to maps of British and Irish weather (Figure 5). Across the Channel, Jean Joseph Le Verrier (1811–77), director of the Paris Observatory, founded a daily weather summary for mainly France in 1858. From September 1863, the bulletin of the Paris Observatory included a daily weather map. On 1 April 1875, the London newspaper The Times initiated the presentation of a daily weather map to a much wider public. It published a chart of 8 a.m.’s weather for the previous day over the British Isles and parts of continental Europe that included plotted details of temperature, wind direction, ‘weather’, sea state, and analyzed isobars. The first regularly published daily weather map in the United States appeared in the New York Daily Graphic on 9 May 1879, although this initiative lasted Figure 4 Daily weather map, Great Exhibition, London, 1851. (Met Office Library.) 2234 SYNOPTIC METEOROLOGY / Weather Maps only a few years. By the final decade of that century however, many daily papers around the world had incorporated a weather map. The gradual increase in the number of nations publishing ‘government’ weather maps led to the desirability of some sort of internationally accepted standard way to depict or symbolize the broad range of surface observations. Although such a standard was accepted at the International Meteorological Congress in Vienna in 1873, it was not to be globally accepted for some decades. By 1891 at least, some 18 countries – mainly in Europe – were publishing government-service synoptic weather maps. As the surface weather network expanded, so did the area covered by weather maps. From 1 January 1914, the US Weather Bureau published surface weather maps for the entire Northern Hemisphere routinely. After World War I, many of the world’s weather services were producing their own analyses on this scale. In postwar Norway a group of brilliant scientists led by Vilhelm Bjerknes (1862–1951), now known as the ‘Bergen School’, worked on the analysis of weather changes associated with the passage of traveling synoptic-scale disturbances in that region. The group refined some of the work from the previous century in which the notion of pulses of warm and cold air in the extratropics had been discussed. They developed the concept of warm and cold fronts, and the more general structure of the commonplace mobile frontal cyclones. Figure 5 Daily Weather Map Company’s map accompanying their promotional material (c. 1863). SYNOPTIC METEOROLOGY / Weather Maps 2235 to represent the circulation at 3500 and 10 000 feet (1067 and 3048 m), so that early ‘upper air analyses’ were made available before the year 1900. It was as early as 1903 that Bigelow began the publication of daily barometric pressure charts for the two constant levels above and for that of mean sea level (Figure 7). A supplement to these observations was provided by the use of kites. These sensed pressure, temperature, humidity, and wind speed up to a height of some 3 km but were replaced in the 1930s by pilot balloons. Such Figure 6 Vertical cross-sections and plan view of an open wave frontal depression. (Reproduced with permission from Bjerknes J and Solberg H (1922) Life cycle of cyclones and the polar front theory of atmospheric circulation. Geofisiske Publikationer 3(1): 3–18.) Jacob Bjerknes (1897–1975) and Halvor Solberg (1895–1974) published a highly significant ‘map’ of such a system in 1922 (Figure 6). Another member of the school, Tor Bergeron (1891–1977), proposed the term ‘occlusion’ and the currently used symbols for the three types of surface fronts. In addition he suggested the used of slightly different symbols for upper fronts. This exceptional work on the structure and evolution of extratropical frontal cyclones – and the location and representation of such fronts on surface weather maps – formed the basis for how all the world’s weather services located these critically important features. The methods have moved on, so that in the UK Met Office, for example, objective schemes are utilized for the automatic positioning of fronts – or at least for providing useful guidance to the analyst. Upper-Air Weather During the 1890s, Frank Bigelow of the then US Weather Bureau composed wind charts for three levels covering the contiguous United States. He did so by analyzing surface anemometer data along with lower and upper cloud drift observations supplied from 140 telegraphic stations. The cloud information was taken Figure 7 Barometric pressure analyses for mean sea level, 3500 feet (1067 m), and 10 000 feet (3038 m). (Reproduced with permission from Bigelow FH (1903) IV. The mechanism of countercurrents of different temperatures in cyclones and anticyclones. Monthly Weather Review 31: 26–29.) 2236 SYNOPTIC METEOROLOGY / Weather Maps 180q 160q W 160qE 140qE 10 10 qN 20 ˚N 20 H H 10 10 100qE 1020 L H L H L H H H 30 10 H 80˚ N L 60˚ N L H H L 1020 20 ˚N H 40q W 140q E 0q H 20q W 0q 160q E 180q H H 40qE 160qW 140qW H H 02 0 H H H H H L L H H 1020 L 80qS 1010 L (B) L 40q E 80qW H L L 1010 H L H H H 1020 L H H H H 20qS 60qW H 0q L H L 40qS HL H L 1000 H L H 0q L H H 10 10 60qS 100 0 L L 60q E L L 80q E H H 100qW H L L L 120qW H 10 10 L L H H H 100q E 101 0 H1 10 20 H 1010 H H 20qE L H L H H H 120q E H H H H H H 10 10 H 60qE H 40˚N H H (A) H H L 1010 H 80qE H H H 20 qN L L H 80q W 60q W H H H H L H H H H 100q W H 120qE H 30 10 10 10 L 990 0 0 10 H L 10 20 L 120q W 1020 140q W H 20q E 0q 20qW 40qW Figure 8 Northern and Southern Hemisphere mean sea-level pressure analyses. (Reproduced with permission from the European Centre for Medium-Range Weather Forecasts.) SYNOPTIC METEOROLOGY / Weather Maps 2237 60qW 40qW 20qW 0q 20qE 40qE L L 10 10 L H H L 1020 H 60qE 10 10 60qW 10 10 60qE L H 60qN 10 20 0 100 L H 0 100 L 0 101 H L 40qE H 40qW H L 10 10 40qN H L 1020 H 20qW 102 0 0q 20qE Figure 9 Predicted mean sea-level pressure and 24-h accumulated precipitation (shaded). (Reproduced with permission from the European Centre for Medium-Range Weather Forecasts.) balloons were tracked optically to enable mapping of wind direction and speed at a variety of heights – so long as cloud didn’t mask the observer’s view. This significant problem was overcome during the next two decades with the advent and gradual spread of radio-tracked balloons known as radiosondes. They transmitted ‘live’ data to their ground launch station from their pressure, temperature, and humidity sensors, and were tracked by radar to derive wind information. Today there are some 600 to 650 unevenly scattered radiosonde stations globally that report pressure, wind, temperature, and humidity twice daily. Modern Surface Weather Maps It is true to say that the way in which the surface weather features like highs, lows, and fronts are represented on today’s analysis charts is not very different from those of the interwar years of the twentieth century. There has been an extension of the symbolism to include frontogenesis and frontolysis, and greater appreciation by forecasters of the variety of, for example, cold fronts. The advent of weather satellites and radars has aided our knowledge of the variety of frontal structure – and this appreciation has led to the need to inform forecasters of the differences, for example, between ‘ana’ and ‘split’ cold fronts. The latter would be ideally represented on a weather map by a surface cold front symbol and a leading upper cold (humidity) front symbol. The use of supercomputers in weather analysis and forecasting has opened up a massive array of surface weather representations that are mapped automatically. The ‘classic’ map of mean sea-level isobars is still produced as a global analysis. Figure 8 exemplifies these for the larger part of the Northern and Southern Hemispheres, from the European Centre for Medium-Range Weather Forecasts (ECMWF). Larger-scale maps illustrate predictions of the mean sea-level pressure field, as well as a derived field. Figure 9 is a map of the prediction of mean sea-level pressure, valid at 12 UTC at the end of a 48 hour forecast, and of total precipitation during the last of the two days (00 to 24 UTC). The isobars thus provide a snapshot of likely conditions at one instant, while the precipitation patterns are expressions of the rain, drizzle, or snow that 2238 SYNOPTIC METEOROLOGY / Weather Maps Figure 10 Probability of predicted 24-h precipitation exceeding (A) 1 mm, (B) 5 mm, and (C) 10 mm. (Reproduced with permission from the European Centre for Medium-Range Weather Forecasts.) is forecast to fall from disturbances over the whole 24-h period in question. As an extra aid to operational meteorologists, it is possible also to indicate the likelihood that the 24-h precipitation total will exceed some critical value. Figure 10 highlights three such criteria: the probability (5%, 35%, 65%, and 95%) that the total fall over 24 h will exceed 1 mm, 5 mm, and 10 mm. Similarly, Figure 11 illustrates the same probability values for the 10 m wind speed to exceed 10 m s 1 and 15 m s 1. The latter falls just inside the category of ‘gale’. Modern Upper-Air Maps As with weather maps for the surface, the range of charts available for the representation of upper-air features has increased dramatically over the last decade or so. There are still the ‘traditional’ synoptic maps. At ECMWF, for example, the predicted height field for the 500 hPa surface is charted as it has been for many decades (Figure 12). The contemporaneous thermal field at 850 hPa illustrates the large-scale waves of relatively warm and cold air in the lower troposphere. As with some surface phenomena, the model can provide indications of the extent of, for example, predicted 850 hPa anomalies that are greater than 74 K and 78 K (Figure 13). An innovative synoptic map produced by ECMWF is that of the predicted cloud cover for low, medium, and high levels. Such maps are provided in daily time steps, valid at 1200 UTC, and in essence give a broad SYNOPTIC METEOROLOGY / Weather Maps 2239 H H L H H L H H 5 35 H 5 (A) (B) Figure 11 Probability of predicted 24-h 10 m wind speeds exceeding (A) 10 m s the European Centre for Medium-Range Weather Forecasts.) 60qW 516 52 8 _ 16 40qW 20qW 0q 20qE _ 16 1 40qE 60qE . (Reproduced with permission from _ 8 528 52 8 _ 8 L and (B) 15 m s H _8 60qW 1 552 H H 60qE 540 60q N 0 L 528 0 0 2 55 2 55 540 564 40qE 40qW 20qW L 4 56 _8 6 57 8 58 _ 8 564 2 55 40q8N 57 6 0q 20qE Figure 12 Predicted 850 hPa temperature (deg C) and 500 hPa height (dm). (Reproduced with permission from the European Centre for Medium-Range Weather Forecasts.) 2240 SYNOPTIC METEOROLOGY / Weather Maps H L 35 5 H H 35 65 H H 5 L H L (A) (B) L 5 35 65 95 5 L H 35 H 95 65 L H L 35 H 955 L 6 H L L H (C) (D) Figure 13 Probability of predicted 850 hPa temperature anomalies: (A) less than 8 K, (B) less than 4 K, (C) greater than 4 K, (D) greater than 8 K. (Reproduced with permission from the European Centre for Medium-Range Weather Forecasts.) indication of what a satellite image might look like at each of these times. They act as another indicator for the forecaster – of whether a particular weatherproducing system is composed of deep cloud or shallow cloud, for instance (Figure 14). It is not possible to illustrate the truly enormous range of weather maps available to today’s forecaster. The fact is that some charts have stayed the same over many decades, and for good reason. The mean sealevel isobaric and frontal analysis has stood the test of time: there have been extensions of the symbolism used as our knowledge of the variety of fronts has improved. There has not been the need to abandon these representations for something better. Similarly, the standard upper-air isobaric height analyses are still among the working charts that operational meteorologists use. What is different nowadays, however, is that a whole host of analysed or forecast fields that are a particular forecaster’s favorite can be called up at the press of a button. What might be chosen can depend on the situation at hand, and must of course be used profitably within the strict time confines of the operational forecast cycle. This is the critical change to the production and utility of weather maps today. It is that they can be provided automatically, rapidly, for a larger range of basic or derived fields, and can be overlain with satellite images, for example. This ‘‘richness’’ can not only aid the forecaster’s day-today operations but also gradually improve their knowledge and understanding of the phenomena to hand. SYNOPTIC METEOROLOGY / Weather Maps 2241 60qW 40qW 20qW 0q 20qE 40qE 60qE 60qE 60q W 60qN 40qE 40q W 40qN 20q W 0q 20qE Figure 14 Predicted cloud cover: low (yellow), medium (red), and high (blue). (Reproduced with permission from the European Centre for Medium-Range Weather Forecasts.) See also Further Reading Frontogenesis. Fronts. Observation Platforms: Balloons. Radar: Cloud Radar; Precipitation Radar. Radiosondes. Satellite Remote Sensing: Cloud Properties; Precipitation; Surface Wind; Temperature Soundings. Synoptic Meteorology: Forecasting. Weather Prediction: Ensemble Prediction. Monmonier M (1999) Air Apparent: How Meteorologists Learned to Map, Predict and Dramatize Weather. Chicago and London: University of Chicago Press.