PROCEEDINGS OF THE IEEE, VOL. 67, NO.

11, NOVEMBER 1979

Doppler Weather Radar

RICHARD J. DOVIAK, SENIOR
MEMBER, IEEE,

DUSAN S. ZRNIC, DALE S. SIRMANS

SENIOR MEMBER, IEEE, AND

Absmct-The Doppler weather radar and its signah are examined from elementary considerations to show the origin and development of useful weather echo properties such as signal-to-noise ratio (SNR), range correlation, signal statislics, etc. We present a form of the weather radar equation which explicitly shows the echo power loss due to finite receiver bandwidth and how it is related to the range weighting function. Echoes at adjacent range samples have a correlation that depends on receiver bandwidth-transmitter pulsewidth product as well as sample spacing. Stochastic Bragg scatter from clouds is examined, but experimental work is requited to determine if this echo power is lnrger than incoherently scattered power. Section III presents the relation between Doppler power spectrum and the distribution of reflectivity and velocity within a resolution volume. A new formula that relates spectrum width to the shear of radial velocities as well as turbulence? sign? decorrelation from antenna rotation, and signal p buses *I presented. renewed and p r o w s The estimation of power spectral moments ~s of the most commonly used algorithms are discussed. Section V highlights some of the considerations that need to be made for Doppler radar observation of severe thunderstorms. Echo coherency is shown to limit the pulsed Doppler radar's unambiguous range and velocity measurements. S i e and dual Doppler-radar teChniques for wind measurements are reviewed. Observations of thunderstorms show tornado cyclones, and clear air measurements in the boundary layer reveal turbulence and waves.

n N Ns P(7) Pr PRT pr P(rs)

F(rS)

prefilter ampli'tude filter output amplitude of ith scatterer scatterer's weight per unit volume receiver-filter bandwidth, 6-dB width in Hz propagation speed, 3 X 10' m . s-' refractive index structure constant diameter of the antenna system separation of radars for dual radar system structure function of refractive index normalized range weighting function normalized one-way power gain or radiation pattern maximum measured antenna gain; gravitational constant weighting function of resolution volume prefilter echo amplitude, inphase and quadrature components inphase and quadrature phase signal at filter output attenuation rate due t o droplets ( m - I ) ; also an integer gaseous attenuation rate shear along 8 , $, and r directions
Manuscript received November 28, 1978; revised July 30, 1979. This work was partially supported by the FAA under Contract DOTIFA76 WAI-622 (RC360205), the NWS under Contract 8AA80901, the NRC under Contract RC370503, and the ERDA under Contract RD840520. The submission of this paper was encouraged after review of an advance proposal. The authors are with the National Severe Storms Laboratory, NOAA, Noman, OK 73069.

s,,( r , "1
SNR
T TI Ts

wavenumber = 2 r l A a parameter proportional to raindrop's refractive index one-way propagation loss due t o scatter and absorption echo power loss due to finite bandwidth receiver number of echo samples along sample-time axis; mean molecular weight integer; also refractive index thermal noise power number of scatterers echo power of resolution volume centered at i power delivered to the antenna system pulse repetition time point target echo power at the antenna port instantaneous weather echo power (W) mean weather echo power at sample time-range delay rs output of radar receiver range from radar 1, 2 t o grid point range from source t o target or resolution volume location unambiguous range cTs/2 6-dB range width of resolution volume spatial covariance of A (f*) autocovariance at lag TI correlation of samples spaced along range time power spectrum in frequency domain expected echo sample power power spectrum in velocity domain for resolution volume center at T normalized power spectrum signal-to-noise ratio air temperature (K) time lag pulse repetition time (PRT) or sample time interval dwell time to resolve target location in FM-CW radar mathematical symbol representing a pulse: U = 1 when 0 < t < 7 ; otherwise it is zero radial velocity field at a point Nyquist velocity A/4Ts mean Doppler velocities corrected for target fallspeed at data points for radars 1, 2 pulse pair estimate of Doppler velocity mean Doppler velocity at a grid point mean Doppler target velocities measured by radars 1, 2 mean terminal velocity of drops in resolution volume horizontal wind speed

@ 1979 IEEE

DOVlAK et al.: DOPPLER WEATHER RADAR

radial component of velocity (Doppler velocity) prefilter receiver output voltage echo signal voltage weather echo voltage sample at 7 = 7, resolution volume echo samples along sample-time axis cylindrical wind components vertical wind speed vertical velocity of tracers ith scatterer range weight due t o receiver filter reflectivity factor angular coordinate; antenna rotation rate; rate of frequency change in an FM-CW radar air density; phase wind direction range-time sample spacing range over which samples are averaged two-way half-power beamwidth target reflectivity cross section per unit volume (m- I ) angle between incident and scatter direction beamwidth between half-power points of one way antenna p a t t e n radar beam elevation and azimuth angles in horizon coordinates ( 4 = 0 at true north); also angular position of scatterer relative to beam axis radar wavelength (m) structure wavelength wavelength of wind fluctuations perpendicular distance from axis of cylindrical coordinate system backscatter cross section spectrum width due to drop fall speed differences, antenna rotation, shear, and turbulence total spectrum width of Doppler spectrum mean square value of I or Q spectrum widths contributed by shear along 6, 4, and r, respectively second moment of the two-way antenna pattern second moment of the range weighting function pulsewidth time delay between transmitted pulse and the echo sample.

I. INTRODUC~ION
ADARS were developed t o detect and determine the range of aircraft by radio techniques, but as they became more powerful, their beams more directive, receivers more sensitive, and transmitters coherent, they also found highly successful applications in mapping the earth's surface and atmosphere, and their signals have reached out into space t o explore surface features on our planetary neighbors. Recently pulsed Doppler radar techniques have been applied to map severe storm reflectivity and velocity structure with some astounding success, particularly showing, in real time, the development of incipient tornado cyclones [24], [ 4 2 ] , [45]. The radar beam penetrates thunderstorms and clouds t o reveal the dynarnical structure inside of an otherwise unobservable event. This inside look will help researchers understand the life cycle and dynamics of storms. The first detection of storms by microwave radar was made in England in early 1941. An excellent historical review of the early developments in radar meteorology can be found in Atlas' work

[S]. A more recent article appeared in the PROCEEDINGS [1131. Because the angular resolution A6 in degrees (O) at wavelength X is well approximated by A9 2 70 AID where D is the diameter of the antenna system [16], it is evident that remote radio sensing, even at microwave frequencies, is characterized by pool spatial resolution compared to optical standards. One essential distinguishing feature favoring microwaves is its property to see inside rain showers and thunderstorms, day or night. Rain and cloud do attenuate microwave signals, but slightly (for X > 0.05 m) compared t o the almost complete extinction of optical signals. Scattered signal strength can be related to rain intensity, and time rate of change of phase (Doppler shift) is a measure of raindrop radial speed. Development of high power and high gain klystron amplifiers in the 1950's made practical the generation of microwsves that are phase coherent pulse to pulse, a requirement for pulsed Doppler radars if velocities of other than first time around (first trip) echoes are to be measured [89]. Radar signals are phase coherent from pulse to pulse if the distance (or time) between wave crests of successive transmitted pulses is fixed or known. Magnetron oscillators, phase incoherent pulse to pulse, can only be used for Doppler measurements of targets beyond the first trip if provision is made to store phase for time durations longer than the pulse repetition time (PRT). The first reported use of a Doppler radar to observed weather was made by Brantley and Barczys in 1957 [191. A rapid development of Doppler techniques followed. Boyenval [ 171 deduced the drop size distribution of Rayleigh scatterers from the Doppler spectrum while Probert-Jones and Harper [96] used vertically pointed antenna and storm motion to produce a vertical cross section [ 101 . Zenith-pointing Doppler radars can be used to estimate vertical air velocities as a function of height and time, can yield data from which one can sometimes infer the nature of the hydrometeors (snow, rain, or hail), and in some instances, yield data for calculating hydrometeor size distributions [ 1 1 ] . These earliest observations of radial velocities used analog spectrum analyzers or filter banks that have economical utility for, at most, observations in a few resolution volumes. Atlas [4] recognized the utility of scanning storms horizontally t o map radial velocities on a plan-position indicator (PPI) type display and Lhermitte [81] accurately assessed requirements for the development of a viable pulsed Doppler radar. These early investigators foresaw real-time severe storm and tornado warnings from pulsed Doppler observations of storm circulations and their predictions were to be verified a few years later by several investigating teams [24], [ 2 5 ] , [42], [45]. The first remote measurement of tonadic wind speed was accomplished in 1958 by Smith and Holmes [112] using a 3-cm continuous wave (CW) Doppler radar. Real-time reflectivities displayed on PPI have been available to radar meteorologists since the mid-1940's. The PPI shows reflectivity distributions on conical surfaces as the antenna beam sweeps in azimuth at constant elevation angle. But realtime Doppler velocity mapping was a goal that eluded researchers until the late 1960's. Contrary t o reflectivity estimation which only requires echo sample averaging to reduce statistical fluctuations, mean velocity estimation requires sophisticated data processing. Probably the long development and cost of Doppler processors (to estimate velocities simultaneously at all resolution volumes along the beam) lay principally in preoccupation with pursuit of

1524

PROCEEDINGS OF THE IEEE, VOL. 67, NO. 11, NOVEMBER 1979

spectrum measurements, from which the most interesting moments (mean velocity and spectrum width) need t o be extracted. One of the first -Doppler spectrum analyzers that could indeed generate velocity spectra in real time for each contiguous resolution volume is described by Chimera [331, and this machine, called a velocity indicating coherent integrator, processed with a single electronic circuit the echo signals to generate spectrum estimates simultaneously at all resolution volumes. Another machine, called the coherent memory filter (CMF), employing the same principles was developed 1621 for weather radar observations and used by researchers at the Air Force Cambridge Research Laboratories (AFCRL).' This machine produced the first real-time maps of velocity fields on a PPI [ 3 ] . In the early seventies Sirmans and Doviak [ 1081 described a device that generates digital estimates of mean Doppler velocity of weather targets. This device, a phase change estimator, circumvented spectral calculations and digitally processes echoes in contiguous resolution cells at the radar data rate. The need t o obtain the principal moments economically and with minimum variance, and have these in digital format (to facilitate processing and analysis with electronic computers) has led researchers to use covariance estimate techniques popularly known as pulse pair processing described in Section V. Hyde and Perry reported an early version of this method [72], but it was first used by ionosphere investigators at Jicamarca [ 1231. Independently and at about the same time Rummler [102], [I031 introduced it to the engineering community. Soon the advantages of pulse pair (PP) processing became evident, and scientists at several universities and government laboratories began implementing this signal processing technique on the Doppler weather radar [83], [88], [91], [1101. A single Doppler radar maps a field of radial velocities. Two such radars spaced apart to view the winds nearly orthogonally can be utilized to reconstruct the two-dimensional wind field in the planes containing the radials [ 2 ] , [82]. With help of the air mass continuity equation the third wind component can be estimated and thus the total three-dimensional wind field within the storm may be reconstructed. This is most significant as it will enable one t o follow the kinematics during birth, growth, and dissipation of severe storms and thus perhaps understand storm initiation and evolution. It may even provide the answer as to why some storms reach great severity while others under similar conditions do not. Doppler radars are not limited to the study of precipitation laden air. The kinematic structure of the planetary boundary layer (PBL) has been mapped even when particulate matter does not offer significant reflectivity [47]. Coherent processing can often improve the detection of weather echoes [67]. Measurement at VHF [601 and UHF [91, [36] suggests height continuous clear air returns to over 20 km, and experiments with a moderately powerful radar at S band consistently show reflectivity in the first kilometer or two [30]. Although the Doppler radar became a valuable tool in meteorological research, it has not yet been transferred t o routine operational applications. As a matter of fact, several government organizations (The National Weather Service, Air Weather Service, Air Force Geophysical Laboratory, Federal Aviation Administration and the National Severe Storms Laboratory)

are presently engaged in a joint experiment, the purpose of which is to demonstrate the utility of the Doppler radar for severe storm warnings and establish guidelines for the design of next generation weather radars [25]. We anticipate that the new radars with Doppler capability will go in production in the 1980's and believe that this paper will acquaint the electrical engineering community with some specifics of Doppler weather radar, weather echo data processing, and meteorological interpretation.

Fig. 1 shows in a simplified block diagram the principal components of a pulsed Doppler radar. The klystron amplifier, turned on and off by the pulse modulator, transmits a train of high "peak" power microwave pulses having duration T of about 1 ps with spacing at the PRT designated as T,,the sampling time interval. The antenna reflector, usually a parabola of revolution, has a tapered illumination in order t o reduce sidelobe levels. Weather radars measure a wide range of volumetric target cross sections; the weakest (about lo-" m 2 / m3) associates with scatter from the aerosol-free troposphere, the strongest with cross sections (3 X lo-' m-' ) of heavy rain. Needless to say, antenna sidelobes place limitations on the weather radar's dynamic range and can lead to misinterpretation of thunderstorm heights [41] and radial velocity measurements [ 1221. The backscatter cross section ob of a water drop with a diameter Di small compared to h (Rayleigh approximation, i.e., Di < h/ 16) is

' Resently the Air Force Geophysics Laboratory.

where IK l2 is a parameter, related to the refractive index of the water, that varies between 0.91 and 0.93 for wavelengths between 0.01 and 0.10 m and is practically independent of temperature [ l l , p. 381. Ice spheres have ( ~ values ~ of 1 about 0.18 (for a density 0.917 g/cm3) which is independent of temperature as well as wavelength in the microwave region. There is an abundance of experimental and theoretical work that relates particle cross section to its shape, size relative t o wavelength when D i > h / 1 6 , temperature, and mixture of phases (e.g., water-coated ice spheres). These works are well reviewed by Battan [ 111 and Atlas [S] . Were it not for electromagnetic energy absorption by water or ice drops, radars with shorter wavelength radiation would be much more in use because of the superior spatial resolution. Short wavelength (e.g., h = 3 cm) radars suffer echo power loss that can be 100 times larger than radars operated with h 2 10 cm [ 121. Weather radar meteorologists are not only interested in the detection of weather but also need t o make quantitative measurement of target cross section in order to estimate rainfall rate. Thus it is important to consider losses that aregreater than a few tenths of a decibel. Besides attenuation due to rain and cloud droplets, there is attenuation due t o energy absorbed by the atmosphere's molecular constituents, mainly water vapor and oxygen. This gaseous attenuation rate kg is not negligible if accurate cross section measurements are to be made even at = 10 cm when storms are far away (r > 60 km) and beam elevation is low (e, ~ 2 ' ) I S ] . [ The above considerations lead t o the radar equation for a single hydrometeor having backscatter cross section ab, and

DOVIAK e t al.: DOPPLER WEATHER RADAR

SURFACE OF CONSTANT P

#?SE

M ICRWAVE

'STALO'

AMPLIFIER

&&c!~T[pk~
SWITCH

~pki-)pq
AMPLIFIER

AMPLIFIER

DATA PROCESSING

Fig. 1 . Simplified Doppler radar block diagram.

located at angles ( 8 , 4 ) from the antenna axis

mixer outputs are A Io(t, r) = -U(t - 2rlc) cos

fi

(4y

$)
(2.50) (2.5b)

where P, is the power delivered t o the antenna port, g the maximum measured gain, f2(8, 4) the normalized radiation function, and I = exp - [J(kg + k) dr] the one-way loss factor due t o gaseous kg and droplet k (both cloud and precipitation) attenuation. The measured gain, a ratio of far-field power density S(8, 4 ) to the density if power was radiated isotropically, accounts for losses associated with the antenna system (e.g., radome, waveguide, etc.). A. The Doppler-Echo Waveform (Inphase and Quadrature Components) When there is a single discrete target, the echo signal voltage V(t) replicating the transmitted electric field waveform E is proportional to it ;

-A Qo(t, r) = -U(t - 2r/c) sin - -

fi

$)

the inphase Io(t) and quadrature Qo(t) components of the modulating signal. For convenience we ignored losses (i.e., set b = 1) and used a fi factor in (2.5) so that the sum of average power in I and Q channels equals input power A 12 aver' aged over a cycle of the microwave signal. If r increases with time, the phase y = -4nrlX + $ decreases and the time rate of phase change

is the Doppler shfit. It is relatively easy to see from (2.6) that, for usual radar conditions (i.e., 7 s and weather target V(t, r) = A exp [j2nf(t - 2rlc) + j$l U(t - 2r/c) (2.3) velocities of the order of tens of m . s-' , the change in signal where 2r is the total path traversed by the incident and scat- phase is extremely small during the modulating envelope tered waves, A the prefilter echo amplitude, c velocity of light, U(t - 2rIc). Thus we measure target phaseshift over a time, and U = 1 when its argument is between 0 and T, otherwise it T, =Z s from echo t o echo rather than during a pulse is zero. After detection and filtering (to remove the carrier period. Because of this the pulse Doppler radar behaves as a frequency f and harmonics generated in the detection process), phase sampling device; samples are at t = 7, + (n - 1) T, where T, is the time delay between the nth transmitted pulse and its we obtain a signal echo, and is denoted as range time because it is proportional to range (i.e., 7, = 2rlc). It is convenient t o introduce another time scale, designated sample time, wherein time is incremented in discrete T, steps after t = 7,. Echo phase and amplitude changes are usually rnodulat2g signal diffegnce frequency signal examined in sample-time space at the discrete instants (n - l)Ts for a target at range time T,. However, there have been efforts to measure phase change within a time T in order to if receiver bandwidth is sufficiently large. Thus heterodyning eliminate aliasing problems that plague observations of storm and detection serve only to shift the carrier frequency without systems [501. affecting the modulation envelope (for simplicity Fig. 1 shows The receiver's filter (Fig. 1) response is usually a monotonihomodyning wherein STALO frequency is the same as the cally decreasing function of frequency and its width B6 is best transmitted frequency, i.e., fa 0). A Doppler radar usually specified as the frequencies within which the response is larger has two mixers in order t o resolve the sign of the Doppler than one-fourth of its highest level-its 6-dB width [119]. shift; in one the STALO signal is phase shifted by 90' prior to The larger is B 6 , the better is the fidelity of the echo pulse mixing so that the detected and filtered output of this mixer is shape, but noise power increases in proportion to B 6 . Bandequal t o (2.4) but phase shifted by n/2. The actual signal from widths can be adjusted to optimize signal detection perforthe mixer is the real part of (2.4) and for the homodyne re- mance [ 1341 , but optimization causes receiver bandwidth loss ceiver (or after a 2nd mixing step to bring f* to zero) the two that should be part of the weather radar equation [go]. Fur-

1526

PROCEEDINGS OF THE IEEE, VOL. 67, NO. 11, NOVEMBER 1979

thermore, filter response is associated with a spatial weight along the range-time axis whose form is just as important t o the radar meteorologist as the antenna pattern weight is along angular directions. We consider these in the weather radar equation (Section 11-C). B. Weather Echo Samples Weather echoes are composites of signals from a dense array of hydrometeors, each of-which can be considered a point target. In this section we show the origin of the statistical properties of the echo samples V(T,) and instantaneous power P(T,) and discuss incoherent, coherent, and Bragg scatter. The echo voltage sample V(T,) is a composite V(T,) =

average larger than the first! We estimate it in the following way. Consider a density function A ( 7 ) that describes the scatterer's weight per unit volume. Then analogous t o (2.7) V(T,) = Jvs A ( 7 ) exp (j4nrlh) dVs (2.9)

C A ~(j4rri/h) exp i

(2.7)

of discrete echoes from all the scatterers, each of which has a weight Ai determined by the scatterer's cross section a b , the radiation pattern f 4(8, $), and receiver bandwidth-transmitted pulsewidth produd B6r. These latter weighting functions determine a resolution volume in space wherein targets significantly contribute t o the echo sample at 7,. The echo sample power averaged over an r - f cycle is proportional t o

which is the Fourier composition of A(?) along the radial direction (for bistatic radars the composition is analyzed along the mirror direction). Hence the intensity of the scattered signal depends directly upon the Fourier content of the scatterer's cross section per unit volume [ 1 18 ] . V is the volume over , which A F ) contributes significantly t o V(T,)-e.g., the resolution volume: see Section 11-Cl . When A ( 7 ) is an random function and can only be described in a statistical way, it can be shown that the sample-time average of P(r,) is

- exp (j4nrlh) d vSI2

where

(2.10a)

is the spatial covariance of A ( 7 ) for a statistically homogeneous medium, p' ~7 - T', and the overbar is a sample-time i 2 , average. To arrive at (2.10) we assumed that the resolution volume size is large compared t o the scales over which R ( 3 ) has significant value. The first term in (2.10a) is due t o fluctuThe above is the instantaneous echo power P(T,) for one transations in the density of scatterers while the second term is mission and N, is the number of scatterers. If scatterers within steady and comes from the mean structure of the density (i.e., a resolution volume move randomly a significant fraction of a specular scattering). If particle positions are uncorrelated, wavelength (e.g., h/4) between successive transmissions, each R ( p ' ) is a Dirac delta function havinga weight so that the first successive echo sample V(rs) (spaced T,) will be essentially term in m a ) is equal t o incoherent scatter [ 1061 . If in uncorrelated. In order t o make measurements of the scatterer's addition A (7)is constant and the radial extent of resolution mean radial speed, the time T, between successive samples volume large compared t o wavelength, steady backscatter from must be small so that contiguous echoes, at fixed delay T,, are the mean structure is negligible. correlated. Scatter in any direction comes from Fourier components The first sum in (2.8) is a constant independent of scatterer's having a structure wavelength h, related t o radio wavelength position and is portional t o the zeroth moment of the Doppler spectrum, whereas the second represents the fluctuating porAs = A/2 sin (8,/2) (Bragg's Law) (2.1 1) tion of the instantaneous power and contains the Doppler velocity information. Although the second sum can be signifi- where 8, is the angle between the incident and scattered-wave cantly larger than the first (it has N,(N, - 1)) contributions directions (8, = n for monostatic radar). While it is customary compared t o N , for the first term) for some echo samples, its t o define Bragg scatter as being from a periodic structure in average over many successive samples (i.e., sample-time average) the mean density profile, one can define a stochastic Bragg approaches zero for spatially uniform distributions because the scatter if it arises from the shape of the correlation function. sample-time average of the exponential term tends t o zero. Thus the first term in (2.10a) contains the incoherent scatter The first sum is then the mean power P(T,). An accurate esti- and stochastic Bragg scatter- Chernikov [ 3 1 ] has determined mate of this term is important because it relates t o the meteo- the relative strengths of Bragg and incoherent scatter and related it to the spatial covariance of cloud liquid water conrological estimates of liquid water in the resolution volume. 1) Incoherent and Bragg Scatter: Because radar meteorolo- tent. He shows conditions of side scatter where stochastic gists relate echo power t o the first term in (2.8), it is impor- Bragg scatter is much larger than incoherent scatter and hence tant t o be aware of the conditions under which the sample- might be important for electromagnetic interference from rain time average of the 2nd term is negligible. The first term showers. Bragg scatter is commonly ignored in studies of precipitation represents incoherent scatter because its power is proportional t o the number of scatterers; the time average of the second backscatter, but it would be significant if the liquid water's term represents coherent scatter [106]. If scatterers are not covariance function indicated scale sizes less than a few meters. independent, but have their positions correlated, we then have Although stochastic Bragg scatter may be negligible for precipa statistically varying nonuniform scatterer density. In this itation backscatter, it is not for echoes from clear air. Indeed case it is possible to have the 2nd term of (2.8) give a time clear air radar echoes are a result of Bragg scatter because A (r)
a

1 -EA;

ijr

AiA; exp [ fin(ri - rk)/h]

DOVIAK et al.: DOPPLER WEATHER RADAR

-t

has fluctuations at scales equal to A/2 although A (F) might be r independent (i.e., no steady echoes). Incoherent scatter from air molecules at microwave frequencies is usually many orders of magnitude smaller than stochastic Bragg scatter. 2) Signal Statistics: The I and Q components of the echo's sample are random variables if scatterers' positions change in an unpredictable way. Because I , Q are comprised of a large number of contributions, each of which is not a significant portion of the whole, we can invoke the central limit theorem [94, p. 2661 t o deduce that I and Q amplitudes have a Gaussian probability distribution with zero mean. Thus I and Q are jointly normal [94, p. 1821 random variables and Davenport and Root [38, p. 1601 have shown that I and Q from a narrowband Gaussian process have zero correlations. Thus the probability distribution of I and Q is Prob (I, Q) =

1.0 20 SO BANDWIDTH PULSEWIDTH PROWCT B,T

-exp (-I2 /2a2 - e2 /2u2) (2.1 2)

1 27raZ

Fig. 2. Receiver signal power loss L, (dB) and normalized 6-dB range width, Z r , / c z , o f the resolution volume versus receiver bandwidthpulsewidth product. Receiver frequency transfer is Gaussian and echo pulse rectangular.

where oZ is the mean-square value of I (equal for Q). Because P(rS) = (IZ + Q?), we see from (2.12) that instantaneous power is exponentially distributed and its mean value is = 20'.
C. The Weather Radar Equation We can now relate the sample-time averaging of echo power P(rs) t o the radar parameters and target cross section. The contribution to average echo power at the filter output from each scatterer is

shown [95] that

where 8 is the 3-dB width (in radians) of the one-way pattern. The range weighting function W(r) can be expressed as a product of a receiver loss factor lo and a weighting function fw (r) whose peak is normalized to unity in order t o have a form analogous to the product of gain squared g2 and pattern function f (8, (I). That is

which can be directly expressed in terms of radar parameters and target cross section through use of (2.2). Thus the sampletime average power at range-time delay rs is where 1, is the echo power loss due t o finite receiver bandwidth. When the receiver has a Gaussian frequency transfer and the transmitted pulse is rectangular, fW(rs, r) contains error funcwhere W(ri) is a range weight determined by the filter band- tions [49]. The numerical integration of (2.1 8) with these width and transmitted pulsewidtt. functions, defined by Abramowitz and Stegun [ I ] , gives a We now consider an elemental volume AV that contains L , = - 10 log I, dependence on the B67 product shown in Fig. many scatterers. The summation of obi over this volume nor- 2. Nathanson and Smith [90] examined the exactly matched malized t o AV defines the target reflectivity q filter receiver (i.e., a sinx/x filter frequency transfer for a rectangular pulse) and deduced I, t o be 1.8 dB. We have assumed a Gaussian receiver frequency transfer characteristic and rectangular pulse shape. These two chosen characteristics Replacing the sum by an integration because particles are as- often approximate those met in practice. For condition B67 = sumed closely spaced compared t o the scale of the weighting 1 corresponding t o a practical matched filter, the numerical functions we have the following form for the weather radar integration of the function wZ(r), shows L , 2.3 dB, or about 0.5 dB more than that obtained with a filter transfer perfectly equation matched to the rectangular pulse. Thus a practical form of the weather radar equation is

--

In the above it is assumed f @,(I)w (r) has a scale (resolution volume dimensions) such that the reflectivity and attenuation can be considered constant over the region which contributes most to F(rs). Range ro is the distance at which w2(r) is maximum and is assumed much larger than the extent over which w2(r) has significant weight. When antenna patterns are circularly symmetric and with Gaussian shape, it can be

where henceforth r is t o be used in place of TO. This extended form of the weather radar equation shows not only the dependency of echo power upon commonly used radar parameters, but also its relation t o receiver bandwidth. Furthermore, in the limit of B6 T >> 1, I , + 1 (see Fig. 2) so (2.19) is in agreement with the ProbestJones radar equation [95] used widely by radar meteorologists.

1528

PROCEEDINGS OF THE IEEE, VOL. 67, NO. 1 1 , NOVEMBER 1979

1 ) The Resolution Volume: The 6-dB width of f 4 ( 8 , 4 ) (the two-way antenna pattern function) is taken to represent the angular width dl of the resolution volume V,. In an analogous manner we define the 6-dB width of to represent the range resolution r6 of V,. The range width r6 versus B6r is also shown in Fig. 2. 2 ) Reflectivity Factor: Radar meteorologists need to relate reflectivity q, general radar terminology for the scatter cross section density, t o factors that have meteorological significance. If scattering particles are know t o be spherical and have ECHO SAMPLE SPACING 8 Ts/T diameters small compared to wavelength (i.e., the Rayleigh approximation), then we can substitute (2.1) into (2.1 5) to Fig. 3. Normalized correlation of echo samples spaced along the rangetime axis. obtain the relation

ft

chosen small (i.e., not "matched" t o r) in order to observe meteorological events in a larger range interval with fewer samples along the range-time axis. This approach becomes where more advantageous when real-time data processing equipment limits simultaneous observations to a few range-time samples (as is sometimes the case for real-time Doppler spectral processors) and pulsewidth cannot be increased. If B6 is "matched" is the "reflectivity factor." Either q or Z can be used in the to T or, as in many meteorological radars, large compared to radar equation, but radar meteorologists have opted for 2. r' , then dimensionally small meteorological events such as When backscattering cross section can not be simply related to tornadic vortices can be missed by samples spaced further than size, e.g, scatterers having large (compared t o wavelength) di- the range extent of the resolution volume or else smaller interameters, or liquid-ice mixes, equation (2.20) is used t o defrne vals wiU be observed causing longer time t o interrogate the the equivalent reflectivity factor 2,. entire storm. When the transmitted pulse is rectangular and The use of reflectivity factor alone really does not relate receiver response Gaussian, the correlation of samples along radar echo power t o meteorologically significant variables such range time is shown in Fig. 3 [149]. The figure shows that as rainfall rate or liquid water content because there is one when B6 is more than twice r' , the echo sample correlation more essential ingredient that needs to be known in addition is principally controlled by pulsewidth, whereas when B6 isless to phase (i.e., IK 12). This is the dropsize distribution. When than 0.5 r-', it is controlled by the receiver-filter 6-dB banddrop size distribution is specified by two moments, Ulbrich width. and Atlas [ I 201 have shown how rainfall rate and liquid water aloft are related to the remotely measurable parameters: reflectivity and attenuation. The Doppler spectrum is a power weighted distribution of the radial velocities of the scatterers that mostly lie within the D. Signal-to-Noise Ratio (SNR) for Weather Targets Consistent with the previous assumptions, it can be shown resolution volume. Not only does the power weight depend on the reflectivity of scatterers, but it also depends on the that the SNR for weather targets depends upon B6r as [49] weighting given to scatterers by the antenna pattern, the transmitted pulsewidth, and the receiver filter. A derivation leading t o a relationship between the velocity and reflectivity fields, the resolution volume weighting function, and the Doppler where Co contains constants pertaining t o the radar. We im- spectrum was first put forward by Synchra [117]. Our derivamediately note that if B6r is a constant, then SNR is propor- tion takes a somewhat different route but neverthelss leads to tional to the square of the transmitted pulsewidth. It can be identical final results. shown that maximum SNR is obtained as B67 -+ 0. Therefore, in contrast to point target measurements, we do not obtain a A. Power Weighted Dism'bution of Velocities maximum of weather SNR at B67 1. However, even though To begin with, we consider scatterers that produce a radial SNR increases monotonically as B~ decreases, resolution ti1 velocity field v(Fl) and a reflectivity field MI). Let the worsens. One can define an optimum B6r as that value which resolution volume center be at a location 7 (Fig. 4) with the maximizes SNR for a given resolution. A more formal deriva- corresponding weighting function I@, TI): tion leads to the following conclusion concerning weather radar design: the optimum system consists of a matched Gaussian filter and pulse which together yield the desired resolution where C1 is a constant that can be obtained from (2.16); also [1341. q now has scales small compared to V, dimensions. E. Correlation o f Echo Samples Along Range Time We locate a surface of constant velocity v @ ] ) and seek the Sample spacing along the range-time axis is usually chosen so total power contribution from scatterers in the velocity range that there are independent estimates of reflectivity and/or v t o v + dv. This contribution will obviously be a summation radial velocity along the beam. Both T and B6 determine the of powers from the volume between the two surfaces v and correlation of these estimates and sometimes B6 is deliberately v + dv. It is convenient to choose for the elemental volume

DOVIAK et al.: DOPPLER WEATHER RADAR

1529

stationary weather phenomena and averaged the respective spectra, he would obtain, on the average, the result given in (3.4). Beware, this does not imply that a Doppler spectrum uniquely specifies the velocity and reflectivity profile in a resolution volume. On the contrary, a variety of reflectivityvelocity combinations may yield identical Doppler spectra. For calculation of the mean velocity and the spectrum width, the normalized y,@, v) version of (3.4) is used.

Note that the integral in the denominator is the total power and can be obtained from the volume integral of (3.3). The sum of (3.3) throughout the volume is proportional to the weighted average reflectivity 5 j p )
Fig. 4. Parameters and the geometr that contribute towards the weather sisal power s ) e c p n ; and " ( 3 ) the reflectivity are and radial velocity fields, r 's the center of resolution volume. The weighting functions in azimuth and range are indicated by dashes.

if)

where

the product dsl ds2 d l where dsl and ds2 are two orthogotangent t o ~ ( 7 ' ~= const (Fig. 4). ) nal arc lengths, at a point 6 , The third coordinate d l is perpendicular t o the surface of v dl = lgrad ~ ( ? ' ~ ) dv.' l(3.2) The integral value in the denominator of (3.7) is obtained from (2.19) for weather radar parameters usually met in practice. Now the mean Doppler velocity is d e f i e d as

The elemental volume contributes an increment of power in the velocity interval v, v + dv proportional to @(v) = qFl)I(T: T1)lgrad v F l ) J - ' dsl ds2 dv. (3.3) Finally, the integral over the. surface A of constant v gives the total power in the velocity range v, v + d v and is, by definition, the product of power spectrum density and dv. That is, F(v) = S ( Z V)dv

which is a combination of reflectivity (power) and illumination function weighted velocity and could be quite different from the I(7, <) weighted velocity. Likewise the velocity spectrum width u,(r) is obtained from

The overbar in (3.4) denotes the mean (unnormalized) power spectrum. This last equation is fundamental and is worthy of more discussion. First the area A consists of all isodop surfaces (surfaces of constant Doppler velocity) on which the radial velocity is a constant v, i.e., it is a union of such surfaces. At each point f; on the surface the reflectivity is multiplied with the corresponding weighting function. The gradient term adjusts the isodops contribution according to their density (i.e., the closer the isodop surfaces are spaced the smaller is the weight applied to the spectral components in the velocity interval between two isodops). So far we have considered a deterministic velocity field and a natural question to pose is how does g(?, v) relate to the velocities of seal scatterers whose relative positions change from pulse to pulse. A heuristic argument about this is as follows. Both point reflectivity 7)(F1) and point velocity v(&) constantly change in time due to relative random motions of air (small scale turbulence). Equation (3.4) which was calculated for deterministic flow is valid on the average for turbulent flow as well. This simply means that if one observed under identical macroscopic conditions a statistically

The relationship between the point velocities v ( i l ) and the power weighted moment F(7) is obtained by substituting both (3.4) and (3.5) into (3.8)

Unlike the pulse volume averaged reflectivity, this is the average of point velocities weighted by both reflectivity and the illumination function. Similarly (3.9) reduces to

and corresponds to a weighted deviation of velocities from the averaged velocity. The mean velocity (3.8) depends on the distribution of scatterers cross section within the resolution volume and its

1530

PROCEEDINGS OF THE IEEE, VOL. 67, NO. 11, NOVEMBER 1979

weighting functions. Thus V(T;) cannot in general be equated t o a spatial mean velocity. However, if wind distribution is symmetrical about the resolution volume center and reflectivity is uniform, then G(i') can be considered a very good approximation of the true radial component of wind at the volume center. But hydrometeor fall velocity must either be insignificant or known (see Section VI-Al). A simplification of (3.4) occurs when the radial velocity field and the reflectivity are height invariant. Then the power spectrum reduces to [ 132 ]

relates the Doppler spectrum S(v) to the power spectrum S(f) via the equation

l a a d v ~ ~ . Y I )ds]- d z l . I ~

(3.1 2)

The inner integral sums the contributions along the line s = s ( x l , y l ) on which v(xl, y l ) is constant. Because ds = [dx: + dy: 1 'I2 the integration is a surface integral with element area ds dzl . Both Mxl, y l ) and lgrad v(xl, Y are independent of z but the illumination function may depend on it. At each point x l , y l along a strip of constant v, the reflectivity is multiplied with the corresponding weighting function. To account for contributions of other infinitesimal strips within the resolution volume, integration is performed along the third (z-axis) dimension. Equation (3.12) was used t o compute spectra of model tornadoes and mesocyclones. These compared well with actual measurements 11321, [ 1331 (see also Section VI-B3). It can be shown with the help of (3.12) that when wind shear and q are constant across V,, the power spectrum follows the weighting function shape. Because Gaussian shape approximates well the range and angular weighting patterns, we may infer, when weather spectra are Gaussian, that reflectivity and radial velocity shear are somewhat uniform within 5 .

When the spectrum or the autocovariances are known, all pertinent signal parameters can be readily obtained. Neither S(v) nor R(TI) are available; they must be estimated from the ensemble of samples V(T,) spaced T, apart. Because radars observe any resolution volume for a finite time, one is faced with estimating at a given range cr,/2 the spectrum and its parameters from a finite number M of time samples V(nT,). We shall henceforth delete the range-time argument T,, and V(nT,) will designate an ensemble sample at the implicit delay T, and TI increments in steps of T,. Samples V(nT,), often multiplied by a weighting factor W(nT,), are Fourier transformed and the magnitudz squared of this transformation is a power spectrum estimate S(k) commonly known as the periodogram %k) =

, /

1 M-1 W(nT,)V(nTJ exp (-j2nknlM)

n=o

B. Estimating Doppler Power Spectra - In order to measure the power weighted distribution of
velocities, frequency analysis of V(T,) is needed and can be accomplished by estimating its power spectrum. It is important to bear in mind that the frequency analysis is performed along the sample-time axis for samples V(T,) at fixed 7,. Thus we have discrete samples V(nT,), spaced T, apart, of a continuow random process. Next we shall make some general statements concerning spectral analysis of continuous random signals. The power spectrum is the Fourier transform of the signal's autocovariance function.

where TI is a time lag. The autocovariance function of a stationary (statistics do not change during the time of observation) signal is found from the time average

Because V(t) has zero mean, autocorrelation and autocovariance are identical [94]. Note that conservation of power

where k, n are integers. The finite number of time samples from which the periodogram is computed limits velocity resolution and creates an undesirable "window effect." Namely, one may imagine that the time series extends to infinity but is observed through a finite length window. The magnitude squared of the data window transform is referred to as the spectral window and is significant because its convolution with the true spectrum equals the measured spectrum. An illustration of a weather signal weighted with a uniform window and one with a von Hann (raised cosine) window (Fig. 5) shows considerable difference in the spectral domain especially in spectral skirts. Since the von Hann window has a gradual transition between no data and data points, its spectral window has a less concentrated main lobe and significantly lower sidelobes. The resulting spectrum retains these properties and enables us to observe weak signals to over 40 dB below the main peak. T i is very significant when one is hs to estimate the peak winds tornadoes Or other severe weather [I281 within the resolution volume; power in spectral "skirts" due to high velocities is rather weak and would be masked by the strong spectral peaks seen through the sidelobes unless a suitable window is applied. The apparent lack of randomness of coefficients in the spectral skirts for the rectangularly weighted data is due to the larger correlation between coefficients. This correlation is attributed to the strong spectral powers seen through the nearly constant level window sidelobes [ 1281. The example on Fig. 5 is from a tornadic circulation with translation. In this case the broad spectrum results from high speed circulatory motion within the resolution volume. The envelope shape lsin x/x12 is readily apparent for the rectangular window (at negative velocities), and the dynamic range for spectrum coefficients is about 30 dB. This is in contrast t o over 45 dB of dynamic range with the von Hann window which also better defines the true spectrum and the maximum velocity (60 m . s-I). For visual clarity an estimate of the mean power from a 5-point running average is drawn on Fig. 5. Besides the window effect which is intimately tied t o signal processing, there are a number of spectral artifacts due to the

DOVIAK e t al.: DOPPLER WEATHER RADAR

V E W I T Y IN KTERS PER S C E W
r10

-100
W " 7 1 1 1 1

-00
1 1 1 1 1

-so
1 1 1 1

-ro
1

60

80

IOO

RECT
X

VELOCITY IN HETERS PER SECOND

100
1 1

-80
1

-60
I l l
I l

-YO
l

-20
I l l
1 1 1

0
1 1 1

20
I 1 1

Y O
I l l

60
1 1 1

80
1 1 1

100

I
X

DATE 052077 TIME 185350 AZIMUTH 6.1 RLTITUDE 1.952 RANGE 34 . I 5 8 K M GFlTE 03

ELEVATION 3.1 S/N 31 d B

Fig. 5. Power spectra of weather echoes showing statistical fluctuations in spectral estimates denoted by x's. (a) RECT simes spectra of echo samples unweighted whereas (b) HANN signifies samples weighted by a von Hann window. Solid curves are five point running averages of spectral powers. This spectrum is from a small tornado that touched down in Del City, OK, at 35 km from the Norman radar.

radar hardware. These are discussed in several references [126], [129], [131].
C. Velocity Spectrum Width, Shear, and Turbulence

in the radar's spherical coordinate system. The components a: and 02 are related t o the radar and meteorological parameters [891 as

The velocity spectrum width (i.e., the square root of the second spectral moment about the mean velocity) is a function both of radar system parameters such as beamwidth, bandwidth, pulsewidth, etc., and the meteorological parameters that describe the distribution of hydrometeor density and velocity within the resolution volume [S]. An excellent explanation and assessment of each can be found in Waldteufel's work (1 221. Relative radial motion of targets broadens the spectrum. For example, turbulence produces random relative radial motion of drops. Wind shear can cause relative radial target motions as will differences in fall speeds of various size drops. There is also a contribution to spectrum width caused by the beam sweeping through space (i.e., the radar does not receive echoes from identical targets on successive samples). This change in resolution volume Vs location from pulse to pulse results in a decorrelation of echo samples and consequent increase in spectrum width u,. The echo samples will be unconelated more quickly (independent of particle motion inside V,) the faster the antenna is rotated. Thus spectrum width increases in proportion to the antenna angular velocity. If each of the above spectral broadening mechanisms are independent of one another, the total velocity spectrum width a, can be considered as a sum of oZ contributed by each [70]. That is, where a; is due to shear, a: to antenna rotation, 02 t o different drop size fall speeds, and a: to turbulence. The significance of the total width a, for weather radar design is discussed in Sections IV and V. It should be noted that (3.17) does not show a beam broadening term defined by Nathanson [ 8 9 ] because we have elected t o define shear in terms of measured radial velocities

02 = (ado sin 6)'

(3.18)

where 8' is the two-way half-power beamwidth in radians for an assumed circularly symmetric antenna having a Gaussian distribution of power. The width ado is caused by the spread in terminal velocity of various size drops falling relative to the air contained in V,. Lhermitte [80] has shown that for rain, ado is about 1.0 (m s-' ) and is nearly independent of drop size distribution and rainfall rate. The elevation angle 6 is measured to beam center, and a is the angular velocity of the antenna in radians per second. In terms of the usually specified one-way half-power beamwidth dl

The wind shear width term tions, i.e.,

0;

U,

is composed of three contribu(3.21)

+ a& + u;,

where each term is due to radial velocity shear along the elevation, azimuth and radial directions, respectively. Assumptions behind (3.2 1) are that shear is constant within the resolution volume and that the weighting function is product separable along 6, #, and r directions. Let ke, k+ be shears in the 6, # directions and use (3.9) t o obtain

where a$ and a$ are defined as second moments of the twoway antenna pattern in the indicated directions. A circularly symmetric Gaussian pattern has a; = a i = 6 : / 1 6 l n 2 . (3.23)

1532

PROCEEDINGS OF THE IEEE, VOL. 67, NO. 11, NOVEMBER 1979

PROBABILITY THAT WIDTH EXCEEDS ORDINATE VALUE ANT. BEAM WIDTH 8, = .8*

R 135krn

"

01 0.5 1 .

5 1 20 40 60 80 90 95 98 99 99.8 0 CUMULATIVE PROBABILITY X

was less than 4 m e s-', the aircraft experienced only light turbulence in over 50 thunderstorm penetration flights. Accurate estimates of turbulence and shear, as well as rain and hail hazards, should allow safer flights through showers produced by thunderstorms. For most situations, the combination of radar characteristics and meteorological parameters result in negligible contribution from all mechanisms except ke and ke shear of the radial velocity, and turbulence [1071. Because the two angular shears can be determined directly from the angular dependence of the mean radial velocity, the component due to turbulence can be extracted from spectral width. A value of ke shear equal to 1 X lom3s-' is suggested by Nathanson [891, yet NSSL's Doppler velocity fields show kg shears of about 1 t o 2 X s-' which typify mesocyclone regions of tornadic storms.

The pulsed Doppler radar should supply (for each radar resolution volume) three spectral moment estimates of prime importance: 1) the echo power or zero moment of the Doppler spectrum (this is an indicator of water content in the resolution volume), 2) the mean Doppler velocity or the first moFollowing arguments that led to (3.22), one can show that ment of the spectrum normalized t o the zeroth moment, and constant radial gradient of shear kr contributes 3) spectrum width a,, the square root of the second moment dr= a; k; (3.24) about the first of the normalized spectrum, a measure of to the width, where a: is the second central moment of the velocity dispersion within the resolution volume. The Doppler spectrum's zero and second moment can be intensity weighting function in range. For a rectangular transestimated also with incoherent radars employing envelope mitted pulse, Gaussian input filter and under "matched" detectors (1041. By far the mostAused spectzal moment is conditions (i.e., B67 = 1) the last equation reduces to the zeroth or echo power estimate P(7,). The P(7,) values of meteorological interest may easily span a range of lo9 and The width a, due to turbulence is somewhat more difficult often the choice of receiver hinges upon the cost to meet this to model. When turbulence is homogeneous and isotropic large dynamic range requirement. Logarithmic receivers are within the resolution volume, widths can be theoretically quite effective in accommodating such a large dynamic range, related t o eddy dissipation rates [ 521 . thus the Doppler radar may sometimes have a separate logaDoviak et 41. [481 have made measurements of total spec- rithmic channel for reflectivity estimation, whereas a linear trum widths a, in severe tornadic storms and Fig. 6 data channel is well suited for velocity measurements. Moment show a median width value of about 4 m e s-' and about 20 estimates utilize samples of a randomly varying signal and the percent of widths larger than 6 m . s-' . They have deduced confidence or accuracy with which these estimates represent that these large widths are most likely due t o turbulence that the true moments directly depends on the SNR, on the distriis not homogeneous and isotropic suggesting the presence of bution of velocities within the resolution volume, on the energetic eddies of scale size small compared to their radar's receiver transfer characteristics, and on the number of samples = 0.8' and r = 1 processed M. In the case of weather echoes, single sample resolution volume. For these experiments ps; so weather radars, not having better resolution, should estimates have too large a statistical uncertainty t o give meaningful data interpretation. Thus a large number of echo obtain similar width distributions in severe storms. Strauch and Frisch [I161 have measured widths up to 5 samples must be processed to provide t h e required accuracy. To obtain a quantitative estimate of P(r,), samples must be m . s-' in a convective store ( 3 c m wavelength radar, beamwidth 0.9O, range up t o 55 km). It is significant that those averaged over a period long compared t o the echo decorrelamaximums were in the transition region between up and tion time which is the reciprocal of spectrum width. The probability density and moments of the averaged output and downdrafts and close t o the reflectivity core. It is extremely important to relate widths to severe turbu- of the input power estimate can be derived from the known lence so that radars can give reliable measure of turbulence weather echo statistics and the receiver transfer function hazardous to aircraft. Analysis by J. T. Lee at the National [85]. The output signal Q of radar receivers can have one of Severe Storms Laboratory (NSSL) [78] suggests a strong many functional dependencies upon the signal applied to the connection between spectral width and aircraft penetration receiver's input. The problem is t o estimate RrS) from sample measurements of turbulence. Hs data show that when aircraft averages of Q. The estimation is complicated because Q is not i r derived gust velocities exceeded 6 m -s-', corresponding to linearly related to P(rs) (except f ~ square law receiver). That moderate or severe turbulence, the spectral width exceeded is, when mean output estimates Q are used with the receiver 5 m s-' in every case for aircraft within 1 km of the radar iransfer function (i.e., Q versus P(rs)) to obtain estimates resolution volume. Not all storm regions of large spectral P(rs), we generate biases and have larger uncertainty in the width produce aircraft turbulence. Furthermore, when a, estimates P(T,) than if we averaged P(rs) directly [107],

Fig. 6. Cumulative probability of unbiased spectrum widths for echoes from three tornadic storms. Spectrum widths are derived from spectra computed using discrete Fourier transforms of echo samples having SIN > 15 dB. Width estimates from the pulse pair algorithm. Equation (4.4) produce almost identical results.

IV. DOPPLER MOMENT ESTIMATION

DOVIAK e t al.: DOPPLER WEATHER RADAR Because considerable correlation may exist from sample to sample, the variance reduction achieved by averaging is less than it would be for independent samples 1801. The degree of correlation between samples is a function of radar parameters (i.e., wavelength, PRT, beamwidth, pulsewidth, etc.) and the meteorological status (e.g., degree of turbulence, shear, etc.) in the resolution volume. Estimate variance can be reduced, at the expense of resolution, by averaging along range time as well as sample time. Because the resolution volume's range extent is usually small compared t o its angular width, averaging over a range interval Ar usually results in a more symmetrical sample volume with little degradation of the spatial resolution. Range-time averaging the output of a linear or logarithmic receiver introduces a systematic bias of the estimate caused by reflectivity gradients, which will limit the maximum Ar useful for averaging [ l o o ] . Nevertheless, we have a reasonable latitude available in choosing Ar. Only Doppler radar can provide first spectral moment estimates, but at the expense of considerable signal processing. The algorithmic structure of a maximum likelihood (ML) mean frequency estimator is in general unknown, but important special cases have been documented in the literature. For instance, when a pure sinusoid is immersed in white noise, the ML algorithm calls for a bank of narrow-band filters; the center frequency of a filter with maximum output is then the desired estimate 1661. Discrete Fourier transform processing generates, conveniently, a bank of parallel filters but is not used in the ML sense to extract the mean frequency because weather signals have considerable bandwidth. Rather, a straightforward power weighted mean frequency provides the estimate. Miller and Rochwarger [871 and Hofstetter [711 have established the autocovariance argument as a ML estimator for certain conditions. This estimator is popularly known as the PP algorithm. It is ML when pulse pairs are independent, i.e., when the covariance matrix of time samples is tridiagonal with the same off diagonal elements. Also, as shown by Brovko [20], the optimality of PP extends to a firstorder Markov sequence in case the white noise is negligible. Second moment estimators are of necessity more complex and, therefore, their optimum properties are more difficult to establish. Estimates based on Fourier methods and PP processing have proven to be useful, and it is known that for independent PP's, the PP width estimator is ML [71 I , [ 8 7 ] . These two methods of spectral moment estimation are discussed in detail in the remainder of this section. A. Mean Velocity Estimation-Doppler First Moment 1) Fast Fourier Transform: The FFT algorithm is used to evaluate the discrete Fourier transfrom (3.16) (341. Mean velocity calculation by the spectral density first moment usually involves some method of noise and ground clutter removal. More common methods are thresholding by power or frequency [I091 or noise suppression by subtraction of expected noise power from the spectral density coefficient [14]. Performance of two FFT mean velocity estimators is shown in Fig. 7 for Gaussian signal spectra and white noise. 2) Covariance or Pulse-Pair Estimator: The complex covariance and the spectral density constitute a Fourier transform pair and thus by the moment theorem, the moments of the spectral density correspond t o the derivatives of the complex covariance evaluated at zero lag.
[ 1271.
Z

SNR * 0 d B

.6

FFTINS

.02

.04 .06 .OB .I0 .I2 NORMALIZED SPECTRUM WIDTH 0-v/vo

.I4

Fig. 7. Standard deviation of FFT and covariance PP mean velocity estimators at two SNR's. Normalization parameter the Nyquist velocity va and square root of number of samples M e NS signifies estimate after noise power subtraction; 1 5 dB is after application of a 15-dB threshold below spectral peak. Gaussian signal spectrum and white noise are assumed.

An ML unbiased estimator2 of R(Ts) [87]

forms the basis of the algorithm for an estimate of mean velocity ii given by

where 2va = h/2Ts is the unambiguous velocity span (Nyquist interval). The covariance argument is an unbiased estimate of the first moment for symmetrical spectra [141, a condition usually satisfied by meteorological signals. General statistics of covariance estimates for statistically independent sample pairs with a Gaussian signal covariance function and white noise are given by Miller and Rochwarger [871. Equally spaced samples, forming contiguous pairs in which each sample is common to two pairs, of a time signal having a Gaussian spectral density and white noise are treated by Berger and Groginsky [I41 and both correlated and uncorrelated pairs are treated by Zmik [130]. Statistical properties of the covariance argument estimator are also shown in Fig. 7. Satisfactory estimation of mean velocity can be made with input spectrum widths up to about 0.4 of the Nyquist velocity. However, uncertainty of the estimate increases rapidly for larger widths, requiring long dwell times for quantitative estimates. This can be seen from the exponential growth of variance at large widths [ 130 I . VAR [CPP1 =

X2 exp [(4na,~,/X)~I
32n2MT:

) { ( ;

+2

(1 - exp I8(2noUTs/h)' I ) +

43: S r1 ~ T

In addition to performing well with populations having wide widths, the PP estimator is superior (in terms of estimate standard deviation) to the FFT at low SNR (Fig. 7). One of 'This is strictly true if successive pairs give independent estimates of R(Ts). It has not been shown that similar properties ensue for correlated sample pairs.

PROCEEDINGS OF THE IEEE, VOL. 67, NO. 11, NOVEMBER 1979

NoRMALlED TRE SPECTRUM WIDTH,

GN,
WORYILIZED SPECTRW
WIDTH, Q,IV,

Fig. 8. Expectation of spectrum width by FFT-TH (normalized to Nyquist velocity) versus true width for simulated Gaussian spectra and white noise; a sliding threshold equal to SNR was applied below the spectral peak.

Fig. 10. Standard deviation of width estimate by PP from contiguous PP's. At large and narrow widths the SD increases rapidly. Range of normalized input widths where estimates are precise is from about 0:06 to 0.6. Because this is a perturbation solution, it takes a large number of pairs M for the results towards the origin to be valid. acually at zero width the standard deviation is proportional to rather than M-S [ 1301.

is the signal power estimate obtained from the complex video signal after subtraction of the known noise power N. Statistics of this estimator for independent PP's was examined by Rummler [ 103I . A related estimator
NORMAL SPECTRUM WIDTH

G'Va

Fig. 9. Standard deviation of width estimate by FFT-TH versus true spectrum width.

does not depend on spectrum shape provided that the width is the major advantages of this estimator, though, is the ability sufficiently smaller than the Nyquist interval. Velocity spectra to operate on pairs of samples as opposed to the equally associated with weather echoes have a wide range of spectrum spaced pulse train required for straightforward FFT analysis. widths, but since it has been our experience that shapes are It is worth mentioning that the PP algorithm is a special mostly Gaussian, version (4.4) is recommended because it case of Burg's maximum entropy method; i.e., when the eliminates the asymptotic (i.e., M - t m ) bias caused by the weather signal is modeled as a first-order autoregressive pro- finite difference approximation for the derivatives. Perturbacess, the equation that locates the spectrum peak from the tion analysis on Gaussian spectra shows both (4.4) and (4.5) forward and backward prediction coefficients is the same as have identical variances and very similar number of sample ( M ) (4.21, [231. dependent bias. For the most part, the bias is not serious because it is proportional t o M-' , as is the variance. Fig. 10 B. Spectrum Width Estimation-Second Moment About illustrates the standard deviation of width (4.4) when the autothe Mean covariance is calculated from contiguous pairs. Note again 1 ) FFT Width: A spectrum width estimate is the square that coherency limits the useful range of input spectrum width root of the second moment about the mean velocity. In t o about 0.6 of the Nyquist velocity, while noise increases practice this computation usually involves some type of noise errors at narrow widths [1301. ._Very similar results are obsuppression. Thresholding the spectrum by power tends t o tained if noncontiguous or independent pairs are considered. bias systematically the width estimate low since part of the signal spectrum as well as noise is removed [ 1101. The ex- C. Errors in Estimated Moments pected width estimate by FFT is given in Fig. 8 as a function Besides inherent uncertainties due to the stochastic character of true width with Gaussian spectra and a threshold at the of weather signal, the moments are subject to biases generated SNR below the spectrum mode. The standard deviation of the at various stages of processing and errors due to extraneous FFT width estimate is shown in Fig. 9. 2 ) Covariance Techniques for Estimating Width: Like the spurious signals. Jitter in the oscillator chain broadens the mean frequency, the second spectral moment can be estimated spectrum and so does the clipping prior t o or in the analog-tohs directly without recourse to Fourier transform. For Gaussian digital converters. T i and other nonlinearities generate spectra and when PP's in autocovariance calculations are harmonics. DC offsets and line frequency pickup bias the independent, one can show using the results of Miller and mean and width but can usually be controlled by proper Rochwarger that the following spectrum width estimate is design and ground clutter filters [ I 101. Imbalances in the phase and amplitudes of the video signals create undesirable ML [87] image spectra. If the amplitudes are balanced t o within 10 percent and the phase to better than 5 percent, the image peak is more than 25 dB below the signal [ 1291.

DOVIAK e t al.: DOPPLER WEATHER RADAR

COMPLEX MULTIPLICATION PRODUCTS TRUNCATED INDIVIDUALLY TO 1 BlrS 0

SUMMATION OF REAL AND IMAGINARY COMPONENTS

SUMS SCALED TO 9 BITS SUCH THAT LARGER, IXLI, IS 128 S X L S255, OR 8 LSB'S OF SUM CONVERSION TO SIGNED BINARY DIVISION OF SMALLER BY LARGER

T
I

- 7 BIT

QUOTIENT

;T
TRUNCATION AT W R POINT AY

VECTOR LOCATED TO WITHIN ONE OCTANT; 3 BIT VECTOR LOCATED IN OCTANT ; 7 BIT ADDRESS, 8 BIT OUTPUT VECTOR LOCATED IN COMPLEX PLANE

Fig. 1 1 . A flow diagram for hardware implementation of the covariance mean frequency estimator.

D. Hardware Implementation o f Covariance Mean Estimators One aspect common to any hardwired implementation is the digital quantization. Reduction of the length of input or output words or any intermediate numerical result has the advantage of decreasing the amount of hardware and thus both cost and complexity. One method of implementing the covariance mean estimator in hardware is just straightforward expansion of the algorithm (4.1) into a series of real operations (Fig. 11). To select digital processing parameters it is convenient t o describe the quantization effects statistically frcm which bias and variance due to quantization can be easily modeled. The bias, for fixed point arithmetic with a 6 bit word and the usual assumption of uniform density across the digital class, is zero for round off and digital class width 2-b, for truncation. The quantization variance is 2-Zb/12. Word length is established on the basis of quantization variance relative to the inherent signal variance at each point in the computation. See the following example. Consider the flow diagram of Fig. 11. Two of the digital parameters which should be selected primarily from required performance (meteorological) criteria are input word length, on the basis of expected signal dynamic range, and output word length such that this quantization standard deviation (SD) is small compared to the required SD of the Doppler estimate. The output word specifies the arc tangent table output increments and thus the table input increments. Internal n word lengths are then specified by these two variance boundaries. For the scheme and word lengths on Fig. 11, the truncation that dominates quantization variance occurs in the digital multiplier. The product is truncated to make the quantization SD compatible with estimate SD. Care must be taken in selecting this product length to preserve the dynamic range of the input because a significant estimate bias (comparable to output estimate SD) at low level signals appears when the product is truncated to the input word length. . I example for the Fig. & 1 1 scheme is on Fig. 12. This bias increases sharply as the product length is truncated t o the input word length. A mea-

SIGNAL RMS= 8
ffv/Va'D3125

Fig. 12. Bias of mean velocity estimate with truncation of complex product terms to input word length. Results apply to both full precision calculations after the multiplier and hardwired calculator.

NP=2

aw Fig. 13. Ratio of standard error of mean velocity estimate with truncated product terms ON to standard deviation of estimate with product of 16 bits and full precision, o,,p. NH is the difference (number of bits) between input work length and product terms in hardware scheme. Np is the difference with fl precision. ul

sure of the variance introduced by product truncation is reflected in the ratio of the SD of mean frequency estimate

PROCEEDINGS OF THE IEEE, VOL. 67, NO. 11, NOVEMBER 1979

-va
NP"

"PP
516
C M ~ V

TIMF ,16178

FYI-E

L753&7

7777

T ~ D F

nrlr

31 '77

Fig. 14. Scattergram of hardware pulse pair velocity and power spectrum derived velocity for the same resolution volumes in a storm SIN power ratios are between zero and IS dB.

with truncated hardwire logic to the estimate SD using full computer precision (Fig. 13). The sharp increase in estimate SD for product truncation from 10 bits to 8 bits coupled with the bias increase at this point indicates that a product word length of input word plus two bits is a good compromise between data load and quantization variance for this scheme. An example of the performance of NSSL's hardwired PP processor is shown in Fig. 14. This is a scattergram of mean velocities in thunderstorms estimated by the hardwired processor versus the mean velocity at the same radar pulse volumes as derived by FFT analysis. The SNR for this data set is between 0 dB and 15 dB, and the computer labels along the axes are the offset binary numbers used in the numerical data handling. Velocity is given on the auxiliary axis where 34 m . s-' is the Nyquist velocity of NSSL's Doppler radar.

This section examines limitations, due to range-velocity ambiguities, in pulsed Doppler radar observations of severe storms and presents techniques to mitigate these restrictions. Radar waveform designs [39], formulated to remove ambiguities when targets are discrete and finite in number or when cross sections do not span a large dynamic range, do not work well with weather targets which are distributed quasi-continuously over large spatial regions (tens t o hundreds of kilometers), and whose echo strengths can span an 8 0 d B power range.

A. Ambiguities To illustrate the point, Fig. 15 is a conglomerate of thunderstorm celk as seen by a WSR-57 radar. One of the cells produced a tornado that was tracked with a nearby Doppler radar (see Section VI-B2b). The Doppler radar's unambiguous range extends only to 115 km (Fig. 16) as opposed to the 900 km on the WSR-57. Although the Doppler radar range is ambiguous beyond 115 km, the tornado producing cell at 150 km in this particular case is not completely overlaid (obscured) with echoes from other trips (cT,/2 intervals). Furthermore, because the radar is fully coherent from pulse to pulse, velocity measurement is possible within unobscured regions of this second-trip storm. Target range becomes ambiguous when true range exceeds r, = cT,/2, but these ambiguities can be resolved if different trip echoes are not overlaid so as to be scrambled. Target velocities are ambiguous because we cannot distinguish between real Doppler shifts and those aliases spaced in frequency by the pulse repetition frequency. Thus the range velocity product

typifies the ambiguity resolution capabilities of conventional (i.e., uniform pulse spacing) Doppler radars. The equation shows the advantage of longer wavelengths, but other factors may control this choice.

DOVIAK et 01.: DOPPLER WEATHER RADAR

X OF VELOCITIES IN I ms-' INTERVALS

'"1

CST I

Fig. 15. WSR-57 PPI display of thunderstorm cells on a tornadic day (181704 CST April 19, 1976). Gray shadings (dim, bright, black, dim, etc.) represent dB2 levels differing by about 10 dB2 starting at 17 dBZ. Range marks are 100 km apart, elevation angle = 0.0'. Unambiguous range is 910 kin. The boxed area outlines a tornadic storm cell whose mesocyclone signature was detected in real time by NSSL's Norman (NOR) Doppler radar which is nearly colocated with the WSR-57.

Fig. 17. Relative frequency of occurrences (percentage) of mean Doppler velocity for three tornadic storms. Data samples are uniformly spaced through most of the convective cell. Note the large spread of radial velocities which needs t o be measwedunambiguously.

tion of the exponential term because sampling was ignored) was derived by Denenberg 1401 for variance in mean velocity estimates computed from Doppler spectra. This leads one to consider the inequality

or, in terms of the unambiguous range

Fig. 16. Same storm system (1 81635 CST) seen with the Doppler radar having 115-km unambiguous range. 10-log Z brightness categories ) start at 10 dBZ and increment at about 10-dBZ (dim, bright. steps. The 10-log Z scale applies only to fust-trip echoes. Some range-overlaid echoes can be recognized by their radially elongated shape. The box outlines the same area as in Fig. 15. Range marks are 20 km apart. Part of the tornadic storm is obscured by a nearby (30-60-km range) storm.

..

B. Echo Coherency Considerations In principle one can choose Ts large enough so no second or higher order trip echoes would ever be received, but this choice is Limited in that signal samples spaced Ts apart must be correlated for precise Doppler shift measurement. Correlation exists when [51

as a condition to maintain signal sample correlation. Width estimate variance also has an exponential increase when corrections are made to account for bias [1301. Requirement (5.3) means that o, limits the largest unambiguous range for a given wavelength, whereas (5.1) restricts ra only if ambiguities due t o velocity aliases need to be suppressed by choosing a large va. As shown below, there are methods to resolve velocity aliases (provided a, is sufficiently small (see Section V-D4)). Thus (5.3), a necessary condition to maintain signal sample correlation or echo coherency, must dictate r, for a chosen h or vice versa, not (5.1). Spectrum width o, is dictated by the weather. Unless spectrum widths are less than a few meters per second, it is unlikely that 10 cm or smaller wavelength radars wl eliminate (i.e., by having an ra > 500 il km) overlaid storm echoes.

where o, is the velocity spectrum width of echoes at range r. Condition (5.2) merely states that Doppler width should be much smaller than the Nyquist interval X/2T,. Correlation decreases appreciably when 2a,Ts/X 2 (2n)-' and the variance in mean Doppler velocity estimates u^ increases exponentially as can be seen from (4.3). A similar formula (with the excep-

C. Dism'bution o f Velocify in Severe S t o m s Histograms of mean velocity shown in Fig. (17) illustrate what is t o be expected from severe storms when viewed by a narrow beam (0.8') antenna [48 I . Some 20 000 sample points from resolution volumes near ground to about 10 km are used, and the estimates are from PP's. The center of the velocity distributions (Fig. 17) are displaced relative to one another and to zero due, in part, to storm motion. More important than the mean motion, or peak radial speeds, is

PROCEEDINGS OF THE IEEE, VOL. 67, NO. 11, NOVEMBER 1979

PULSE POSITIONS

Fig. 19. A possible spacing o i s a ~ p l i n g gates for a staggered PRF system t o obtain covariance R,,R, estimates at two different lags.

power can be decreased, possibly by as much as 20 dB [44]. This may make insignificant the occurrence of pesky overlaid echoes. Fig. 18. Signal diagram for orthogonally polarized (V, H)samples of a A disadvantage of spaced pairs is that a longer time will be pulse pair; Pt is transmitted power and P , received mean power. required t o collect sufficient number M of sample pairs so that , , Weather-type targets are assumed to produce the echo pattern of P. velocity estimate variance is reduced t o acceptable limits. However, one does not need the same M as in the case of unithe larger than 50 m . s-' spread in velocities for any of these formly spaced pairs because sample pairs are less correlated storms. Conventional Doppler radars may require an un- [ 1301 . Furthermore, by changing the carrier frequency beambiguous velocity span of at least +35 m s-' in order t o tween pairs of pulses, we can space pairs closer, thus signifilimit velocity aliasing t o a few percent or less. cantly increasing the number of sample pairs without worsening obscuration. Another disadvantage of this technique is that D. Extension of Maximum Unambiguous Range and Velocity ground clutter canceler design is more comFcated. In the absence of practical methods t o simultaneously elimi4) Staggered PRT: A velocity e s t F a t e v 1 is obtained from nate range-velocity ambiguities, several schemes have been PP estimates of complex~ovariance at lag T,, while R1 is devised t o alleviate the aliasing problem. A list of those more derived from covariance R2 for pairs sp^aces 5,. TO decrease promising follows. estimate variance, covariance estimate R1 (or R 2 ) is an average 1 ) Random Signal Radar: This radar uses random modula- of M sample pair covariance estimates (R i (or R i))(Fig. 19). tion of a CW signal [291 and requires continuous calculations Because c l , Vh2 are associated with different Nyquist coof the autocorrelation for each range of interest. This exten- intervals, velocity aliases may give simcantly different sive signal processing, the lack of any "clear area" [39], and estimates-differences that can be attributed t o the different the need for two antennas has limited the use of this radar in aliases [48]. Mean velocity aliases can be resolved so long as meteorological research. the expected difference E[$, ] remains unambiguous. 2) Phase Diversity: If an r,, v, combination can be found However, errors inA reso,lving aliases may occur because one which wl provide the desired range coverage with acceptable cannot estimate E[v2 - v1 ] with zero variance [ 1 1 1 1. il To illustrate what can' be gained with staggered PRT, convelocity aliasing, the phase of multitrip signals with equally spaced transmitted pulses can be decorrelated by retention in sider the two unambiguous Nyquist velocities val = 16 m . s-' the local oscillator, transmitter phase information for only and v = 24 m s-' . The maximum unambiguous Nyquist , one system period while transmitter phase is randomly shifted (composite) velocity va becomes f 48 m . s-' . Now a single from pulse t o pulse. Alternately the transmitter phase history PRF radar with a Nyquist velocity vm = f 3 5 m s-' has an unmay be shifted by one or two system periods in another re- ambiguous range of 100 km if wavelength is 10 cm and 50 km ceiver and measurements thus made in the second or third-trip if it is 5 cm. On the other hand, the staggered PRT radar regions. Echoes outside the selected trip are then incoherent would have the unambiguous range increased by a factor of and when overlaid one over the other only produce an effec- v,/v, 1.5 or an increase in area by 2. This improvement is tive SNR decrease that does not bias mean velocity estimates. obtained at the expense of dwell time. In summary, a staggered PRT technique can significantly inThis technique is especially suited for good magnetron transcrease the unambiguous range, and the maximum unambigumitters and has been used on several systems 183], [27 ] . 3) Spaced Pairs: The use of spaced PP's of orthogonally ous velocity vm can easily be made large so that velocity aliaspolarized samples (Fig. 18) for weather radars can reduce ing is of little importance. As of now, there is no conclusive study t o determine whether occurrence of overlaid echoes [441. This technique has been successfully adapted t o radars measuring ionospheric motions spaced pairs (with or without polarization diversity) would [ 123I . Also, the Wave Propagation Laboratory has imple- give Doppler radars sufficient immunity from velocity-range mented it (without orthogonal polarization diversity) in their ambiguities in severe storms and whether this approach is more 3-cm Doppler radar [27]. (or less) advantageous than staggered PRT, phase diversity, or The first advantage of the method is that, with T2 sufficiently random signal radar. Because the intrapair spacing TI in the large, overlay is limited t o targets in contiguous range invervals space pair technique is forced t o be small t o achieve the same (cTl 12) or trips. That is, first- and second-trip targets can have unambiguous velocity as obtained with staggered PRT, we overlaid echoes as well as second- and third-trip targets, etc., undoubtedly expect a larger occurrence of scrambled echoes but there is no echo overlay for targets in first- and third-trip for the spaced pair method. intervals or second and fourth, etc. Second, all overlaid echoes 5) Dual-Sampling Technique: Even though obscuration of > are incoherent if T2 is sufficiently large (i.e., T2 > h/4uV). Doppler signatures by overlaid multiple trip echoes might not Thus they do not bias velocity estimates and only serve t o be eliminated, there is reason t o present the observer a velocity decrease the effective SNR. When the pulses of a pair are data field wherein range t o datum is unambiguous and velocity orthogonally polarized, (polarization diversity), overlaid echo values are credible (i.e., are not in error due to scrambled

s2

DOVIAK et al.: DOPPLER WEATHER RADAR

1539

trip echoes is small if ra > 130 km, but there could be a significant improvement in decreasing obscuration of secondtrip storms. In any case, an advantage of incorporating a staggered PRT is the automatic de-aliasing of velocity estimates.

VI. OFSERVATIONOF WEATHER

Fig. 20. Dual sampling technique, where R, ,R, ,R , , . . ,are covariance measurements (at equal lags) whose average is used t o derive mean Doppler velocity estimates. We depict only fust and second-trip echoes and assume T2 = 2T1. Clearing period T2 removes multipletrip echoes from reflectivity estimation in a contiguous T2 interval.
A A A

echoes). This can be accomplished by taking echo reflectivity samples during an interval T2 sufficiently long to remove, for practical purposes, all overlaid echoes and have this sampling period interlaced with another whose PRT is short enough to allow coherent measurements for velocity estimates (Fig. 20). By interlacing the velocity estimation period MTl with one for reflectivity (Tz), we can have nearly colocated resolution volumes for velocity and reflectivity measurements. The figure shows one block of samples that contain M = 3 covariance estimates and one reflectivity estimate (for each range bin). To reduce velocity and reflectivity estimate variance, we need to average covariance and reflectivity estimates for K blocks. In order t o have all n-trip echoes samples (n = 2 in Fig. 20) in one Tl period, sampling should start in the nTl interval because the nth multiple-trip echo wl not appear il until then. Dual sampling provides echo reflectivity measure without range ambiguities so we can determine, through comparison of echo powers, those velocity data that are significantly contaminated by scrambled multiple-trip echoes and eliminate them from observation. Furthermore, we can assign correct ranges to the surviving valid velocity data. Such a dual-sampling system is in operation at NSSL (N= 7, K = 8, TI = 768 ps, T2 = 4T1) and velocity fields displayed in real time are not range ambiguous. The most persistent obscuration to plague the dual sampling radar is caused by ground clutter echoes overlaid onto the second trip (i.e., ground clutter seen just beyond ra) as well as ground clutter within the first trip. Ground clutter obscuration can be lessened with cancelers and by displacing (through changes in TI) the secondtrip ground clutter ring away from the storm of interest; otherwise, we have a permanent ring of about 10-15 km in range wherein the Doppler radar is blinded. A closely related method of increasing the range t o which reflectivity can be resolved unambiguously is to transmit coherent signals at two different frequencies o ~ol each at dif, ferent PRT's so that simultaneous reception is possible. The long PRT yields large unambiguous range for reflectivity estimates while the high PRT is used for velocity estimation. This technique and its signal processing are analogous to the dualsampling technique described above: its advantage is a reduction of the acquisition time and possibility for better canceler design. The dual-sampling mode can accommodate a staggered PRF during the velocity estimation period to allow increase both in ra and v, at the expense of increased data acquisition time (i.e., less scans-about 30 percent less-per observation period) for given velocity estimate accuracy. These might be important gains for a dual-sampling radar, but an issue is whether obscuration is significantly decreased. Studies with simplified storm models suggest that the decrease in obscuration of first-

The great utility of storm observation with centimeter wavelength pulsed Doppler radar derives from its capacity to map reflectivity (q) and mean radial velocity C inside the storm's shield of clouds. A three-dimensional picture of a single storm takes about 2 to 5 min of data collection time not only because of antenna rotation limitations, but also because a large number of echoes from each resolution volume needs to be processed in order to reduce the statistical uncertainty in the q and iT estimates. Although storm structure can change significantly during this' period with distortion of the radar image of the true reflectivity and velocity fields, highly significant achievements have been made in depicting the structure and evolution of the thunderstorm. But the meterologically interesting variables are not q nor v, but parameters such as rainfall rate (on the ground) and wind. Pulsed Doppler radar most often measures the radial speed of hydrometeors, not air, and in certain situations such as vertically directed beams, these speeds can differ significantly from the radial component of wind. Likewise, surface rainfall rate estimates are not easily related to q, and often radar reflectivity measurements are supplemented by surface rain gages [ l 8 ] , [371. Incoherent radars map q but, if the radar's resolution volume is sufficiently small and reflectivity estimates are accurate, these radars can track prominent reflectivity structures t o map winds that steer cells [35]. However, Doppler radar measures, practically instantaneously, velocities in each resolution volume and hence can provide better resolution of the velocity field. A . Dual Doppler Radar A single Doppler radar maps a field of velocities that are directed toward (or away from) the radar. A second Doppler radar, spaced far from the first, produces a field of different radial velocities because the true velocities are projected on different radials. The two radial velocity fields can be vectonally synthesized to retrieve the two-dimensional velocities in the plane containing the radials [2]. It is customary to accomplish the synthesis on common grid points to which radar data are interpolated. Radial velocities in each of the radar's resolution volumes surrounding a grid point are not measured simultaneously but are separated in time up to the few minutes required for each radar to scan the common volume. The respective resolution volumes are also usually quite different in size and orientation. Nevertheless, useful estimates of wind can be made on scales of air motion large compared to the biggest resolution volume dimensions if the velocity field is nearly preserved over the period required for data collection. Targets such as water drops having small mass quickly respond to horizontal wind forces and faithfully trace the wind. Stackpole [I141 has shown that, for energy spectrum of wind scales following a -513 law to at least 500 m, more than 90 percent of the rms wind fluctuations are acquired by the drops if their diameters are less than 3 mm. When radar beams are at low elevation angles, target terminal velocities (i.e., the steady-state vertical velocity relative to the air) give negligible error in the radial

PROCEEDINGS OF THE IEEE, VOL. 67, NO. 11, NOVEMBER 1979

Fig. 21. Cylindrical coordinate system used for dual radar data analysis. The radars are located at points 1 and 2, and G8, 2a are the unit normals defming direction of the three orthogonal velocity components.

&,

'CHW

EL RENO

m' -"

CIMARRON

* MEEKER

4
PRAGUE

Fig. 22. Map of central Oklahoma with reflectivity field (shaded areas) and winds relative to the ground at a height of 1 km. Velocity scale vector is shown along eastern corner of box. This velocity is the maximum at this height. Reflectivity contours (dashed lines) are labeled as log Z (from Ray et al. [ 9 7 ] ) .

wind component. At high elevation angles these velocities need to be considered. 1 ) Reconstruction o f Wind Fields: The dual Doppler radar technique t o derive quasi-horizontal winds was first illustrated by Lhermitte [82] and later extended by Frisch et al. [53] t o display all three Cartesian wind components. Wind field determination is greatly simplified if the synthesis is performed in cylindrical coordinates with an axis being the line connecting the two radars. That is, radial velocities at data points (centers of resolution volumes) are interpolated to nearby grid points on planes having a common axis (the COPLAN technique) [82]. Cartesian wind components can be derived from these synthesized cylindrical components. Although one could solve directly for Cartesian wind components, this necessitates a solution of an inhomogeneous, hyperbolic partial differential equation t o derive vertical wind [2]. The cylindrical coordinate system is illustrated in Fig. 21. The mean Doppler velocity needs t o be corrected for the scatterers terminal velocity relative t o the air in which they are located. The corrected radial velocity estimate is
v l , z = V I , ~ vt +

where v I m Z are the mean Doppler target trelocities.measured by radars 1,.2 at data points. To estimate vt for each resolution volume, one could use the empirical expression [7]

v t = 2.65z0-ll4

):(

[ m .i l ]

(6.2)

sine

(6. )

where the parenthetical term is a correction suggested by Foote and duToit [51] t o account for height dependent air density 7 , and Z is the reflectivity factor. This relation well represents experimental data over a large range of Z (i.e., 1 < Z < 10' rnm6 m-3) for regions of liquid water, but large errors, up t o several meters per second, in v, estimates can be caused by erroneously relating regions of hail with a ur, Z relation appropriate for liquid water. Usually there is little or n o information t o identify these regions uniquely, and errors in vertical wind w, can result. However, it has been shown for typical arrangements of storms relative t o the two radar placement, that the error in w, is significantly smaller than errors in vt [46 I. The estimated radial velocities v,., of the ar can be interi polated t o uniformly spaced grid points in planes at angle cx t o the horizontal surface containing the baseline. Interpola-

DOVIAK e t al.: DOPPLER WEATHER RADAR

1541

tion filters the data and reduces variance [46]. The cylindrical wind components in the p, s plane are related t o Tl, &, as

where i , are the interpolated Doppler velocities of air. , The wind component w,, normal t o the plane, is obtained by solving the continuity equation in cylindrical coordinates:

factor Z within a resolution volume, 3) use of incorrect vt, Z relationship, 4) inaccuracies in resolution volume location, 5) increase of vertical velocity variance with height owing t o error in derivative estimates in the continuity equation, 6) nonstationarity of the storm during a data collection scan, and 7) echoes received through sidelobes that contaminate signals associated with the resolution volume. How these errors affect the estimates of horizontal and vertical wind is discussed in [46].

with an appropriate boundary condition [97]. density 7 is given by 7 = 70exp[-gMp sin a/(RT)I

The mass (6.6)

and g is the gravitational constant (9.8 m - s-'), M the mean molecular weight of air (29 grnol-I), T the absolute temperature (K), and R the universal gas constant (8.3 14 J . mol-' . K-' ). Appropriate values of yo, T can be obtained from surface site and upper air soundings. The Cartesian components of wind can then easily be obtained from w p , wS,w,. 2 ) Observation o f a Tornadic Storm: The first dual Doppler radar observations of a tornadic storm were made on April 20, 1974, with NSSL's 10-cm radars. These radars provide a large unambiguous range and velocity capability well suited for observation of the large and severe thunderstorms that frequent the high plains of the United States. Fig. 22 locates the two radars, one at Cimarron airfield (CIM), Oklahoma City and the second (NRO) at NSSL's headquarters, Norman, OK. It also shows the horizontal wind field synthesized using a slight modification of the above outlined scheme [97] .3 Streamlines have been drawn in addition to velocity vectors, whose length is proportional t o wind speed. The curvature in the streamlines shows appreciable local vorticity in the region near the grid point (30.0, 24.0). To view the storm's kinematic structure at several altitudes the mean wind at each height is subtracted from the wind vector at each grid point. This perturbation velocity field is displayed at two heights in Fig. 23(a). Cyclonic circulation is apparent at the grid point (30, 24), where vorticity was noticed in Fig. 22. Inflow into the tornadic cyclone is shown at an altitude of 3 km, and divergence and outflow are apparent at 7-km height. High reflectivity factor (60-dBZ) regions are located on the downwind side. These velocity fields are in general agreement with present storm models, particularly in the weak echo region where both imply a strong updraft northeast of the circulation (Fig. 23(b), X = 3 3 , Y = 27.0 km). A downdraft (Fig. 23(b), X = 33, Y = 22.5 km) is found to the southwest of the circulation. 3) Errors in Synthesized Wind Fields: The wind fields derived from dual Doppler radar measurement have errors that arise from several sources. Some of these are: 1) variance in the mean Doppler velocity and Z estimates due t o the statistical nature of the weather echo, 2) nonuniform reflectivity
'Terminal velocity corrections were obtained from interpolated reflectivity factor values and a slightly different vt, Z relationship was used.

B. Observations with a Single Doppler Radar Although the Doppler radar measures only the radial wind component, its spatial distribution can signify important meteorological events such as tornado cyclones. Moreover, high straight winds, if not across the beam, can be measured as well as turbulent regions. Thus a single Doppler radar offers good promise for severe weather warning and in our view will most likely become the operational tool of the National Weather Service in the near future. 1) Linear Wind Measurements- Velocity Azimuth Display: When the antenna beam is scanned in azimuth @ while elevation angle 8 is fixed, the radial velocity has the q3 dependence
v, = w sin

0 + v h cos 0 cos (@ 6 - n) -

(6.7)

where 6 is the wind direction, w the vertical velocity, and v h the horizontal speed of tracers in the resolution volume. The w and vh velocities are readily computed from data in a velocity azimuth display (VAD) [84] under the assumption that air is in pure translation. (The VAD is a display of radial velocity at a single range location versus azimuth.) Then (6.7) has a sinusoidal dependence on q3; thus amplitude and phase of the sine curve are measures of vh and 6 at the height r sin 0 of the sampling circle. Vertical motion produces a dc offset of the sine wave. However, when wind is not horizontally homogeneous, equation (6.7) is no longer purely sinusoidal. Caton [28] showed how divergence can be determined from VAD data. Browning and Wexler [22] carried the analyses even further by assuming the wind field was well represented by a linear velocity field [68, p. 1981 over the circle of measurement. Under this assumption there are four basic fields of motion that convey air: pure translation, vortical, divergent, and deformative. Fourier analysis of (6.7) for linear wind reveals that of these four motions only the vortical one cannot be measured by the VAD method. The average component of (6.7) is proportional t o mean horizontal divergence DIV v h plus mean E , that is

where E is the average of vertical velocity on the sampling circle of radius r . The first harmonic component gives v h and 6, and the second measures deformation. By inserting the mass continuity equation into (6.8), we can then solve for vertical wind if we have an estimate of target terminal velocity averaged over the circle. If the target is refractive index fluctuations (see Section VI-C), then clearly terminal velocity is zero. Thus a single Doppler radar can measure the three components of wind averaged over a sampling circle of radius r and produce a vertical profile of wind. 2 ) Severe Storm Cyclone Observations and Presentations: Because the radar maps the distribution of Doppler velocity

1542

PROCEEDINGS O F THE IEEE, VOL. 67, NO. 11, NOVEMBER 1979

3 KM

20 .o

n/s

(b) Fig. 23. (a) Perturbation wind and reflectivity factor (log 2) at 2 different altitudes. Mean wind at each altitude is shown in lower right comer. Velocity scale is in upper right comer. Distances are in km from an origin s = -1 0 , p = 15 km ( = 0 is the midpoint of the base s line) and abscissa is parallel t o s. (b) reflectivity and velocity fields in the x, z plane for two planes. The mean horizontal velocity in each vertical plane was removed and is indicated in the upper left of each plane. Reflectivity contours are labeled as log Z (from Ray e t al.
[971).

inside the storm, significant meteorological events (unseen from outside) such as tornado cyclones should produce telltale signatures. Donaldson 1421 stipulated criteria whereby a vortex can be identified from single radar observations. Briefly, there must be a localized region of persistently high 25 X s-' azimuthal shear (i.e., the velocity gradient along an arc at constant range) that has a vertical extent equal to or larger than its diameter. It can be shown that nontranslating cyclones have isodops forming a symmetric couplet of closed contours with equal number of isodops encircling positive and negative velocity maxima (Fig. 24). If the inner portion of the vortex is a solidly rotating core, its tangential velocity linearly increases with radius t o a maximum. Outside this maximum, the velocity decreases (roughly) inversely with the radius. The isodop contours of such a combined Rankine vortex are circular sec-

tions connected with straight lines (Fig. 24). This pattern has been observed many times, and an example is shown in Fig. 25(a) for a tornadic storm that did considerable damage to Stillwater, OK, in 1975 [21], [24], [1331. Fig. 25(b) shows contours of Doppler U, for the same storm, and we immediately see the striking correlation of large U, with significant radial velocity shear. Regions of large a, may also indicate the presence of strong turbulence. a) PPI weather display: Reflectivity factor is now routinely displayed by the National Weather Service radars on the PPI scope and by some television stations on a color PPI display. While reflectivity cannot be reliably used for tornado detection, it has proved valuable for hydrological studies and severe weather warnings. Those warnings are primarily based on reflectivity values, storm top heights, and sometimes on circulatory features or hook echoes.

DOVIAK e t al.: DOPPLER WEATHER RADAR

ANTENNA PATTERN

, . l " e
2rt TO RADAR

Fig. 24. Plan view of idealized isodop pattern for a stationary modified Rankine vortex located at range large compared t o vortex diameter. B is Doppler velocity normalized t o peak tangential wind. Radar is located towards the bottom of the figure. Resolution vdume, antenna and range weighting functions are depicted. The angular tilt a determines radial inflow (a < 0) or outflow (a > 0).

Because PPI scopes are commonly used with radars, it is no surprise that weather researchers have begun displaying velocity fields on them. One of the early displays was obtained when a mean velocity processor of the National Severe Storms Laboratory was mated t o the PPI [ 1081. In this case brightness level signifies a particular radial velocity category (e.g., Fig. 26(b)). If the wind and reflectivity were uniform over a full 360' of the scanned space, the display would show intervals of radial velocity as angular sectors of different brightness for each velocity category. Strobed brightness as a function of azimuth is used t o differentiate positive from negative velocities (unstrobed brightness). NSSL's real-time Doppler velocity processor was operational during the Spring of 1973, generating the first velocity contour maps of mesocyclone signatures. The Marlow tornadic storm on June 4 produced a particularly large mesocyclone, and its reflectivity and isodop signatures a e clearly shown in i Fig. 26. This storm's reflectivity structure exhibits a hook echo feature suggesting mesocyclonic circulation, severe weather, and tornadoes. Although the hook echo is identified usually as an appendage to the storm, it may, as in this case, be found within the storm. Average storm height during data L.!"I , I , j Q'Y!~ ' 2a.w ' P.W OISTRNCE%Fin HORRCM~~PR) Yl .m ' I 1 . Mm collection was 16 km; motions of signature and storm were x SPECTRUM WIDTH equal (280j13 m - s-I), with both considerably t o the right @) of mean environmental wind (250/12.5 m .s-I). East of Fig. 25. (a) Doppler velocity field for the Stillwater s l m n at 1.5 km the high reflectivity core, a low reflectivity notch extends height. Grid spacing is 4fO m. Velocities are in m.s and contours (isodops) are in 5 m . s steps. The mesocyclone is centered at well into the storm. Storm motion relative t o the radar is Y = 97 km, X = 36 km. The other shear region from 94 k m north, slight, and therefore, isodop patterns indicate radial velocity 40 krn east t o the bottom of the field was identified from dual relative to the storm. The nearly symmetric isodop signature Doppler data t o be the low-level boundary (gust front) between storm inflow t o the east of the shear line and outflow t o the west. (b) Conindicates circularly symmetric cyclonic rotation consistent tours of constant spectrum width at 1.5 krn above ground. Values with that inferred from the reflectivity pattern. equal t o or larger than 6 m.s-' in steps of 2 m.s-' are displayed for A signature pattern for circularly symmetric convergence visual clarity. Large widths farther north are where the tornado mesocyclone formed; the other region of large width is embedded in is similar t o the vortex pattern but rotated clockwise by the gust front. Lnterpolation grids are spaced at 400 m. 90' [451. The reflectivity spiral suggests convergence, a view supported by the clockwise angular displacement of isodop maxima about the vortex center (190, 8 2 km). Color tion volume, and large second moment magnitudes have been displays of reflectivity, velocity and spectrum width allows, judged potentially important as tornado signatures. Suffiin real time, an easier quantitative evaluation and better ciently intense azimuthal shear is the feature used t o locate resolution of cyclones. the tornado vortex signature (TVS); this has been correlated b ) The multimoment Doppler display: Large changes in with many tornadoes [21]. Large spectrum width is another the first Doppler moment from resolution volume t o resolu- distinguishing feature of a tornado predicted by Atlas [4] and

PROCEEDINGS OF THE IEEE, VOL. 67, NO. 11, NOVEMBER 1979

Fig. 26. Storm reflectivity (a) and isodops (b) displayed on PPI a t 2115 CST. The elevation angle b 1.9', range marks correspond t o 60, 80, 100 km. Reflectivity factor categories are: dim (<21 dBZ), bright (21-31), black (31-44), dim (44-57), and bright (>57 dBZ). Velocity categories are dim <13 ms-'), bright (13-21), and bri&test (>?I).' Positive radial velocities are angularly strobed in brightness. Mesocyclone signature b between 194'-203' and 73-90 km.

Fig. 27. The multimoment Doppler display of a mesocyclone. Each arrow contains information of the 3 principal Doppler spectrum moments for a resolution volume. For interpretation of arrows see insert in upper right comer (arrow length is proportional t o received power, arrow direction to velocity and arrowhead size t o Doppler spectrum width). Abscissa is azimuth and ordinate scale denotes range (km) from radar. Housekeeping information is at top of screen.

Lhermitte [ 8 1] (see Fig. 5). The various signatures of mesocyclones and tornadoes may be revealed at once by using three separate displays or, as described below, with a single display called a multirnoment Doppler display (Fig. 27). To present simultaneously the three principal Doppler moments for each resolution volume, a field of arrows is displayed where arrow length is proportional t o the logarithm of echo power, arrow direction to velocity, and arrowhead size to spectrum width (see insert on Fig. 27 and [26]). Zero velocity is a horizontal arrow pointing right and nonzero velocities are proportional to the angle between the arrow

direction and its zero position. A horizontal arrow pointing left corresponds to the Nyquist velocity (f 34 m s-'). As the velocity increases beyond f 3 4 m . s-', the arrow rotates smoothly through the Nyquist limits and appears as a lower velocity of opposite sign (e.g., 38 m . s-' appears as -30 m . S-'). In these displays, the radar is always toward the bottom of the figure so that arrows in the upper half of the circle denote flow away from the observer, whereas arrows in the lower half denote flow towards the radar. Thus the field of arrows in Fig. 27 illustrates quite nicely the signature of circulation (centered at 1 8 7 ' - ~ z 70-km range) or con-

DOVIAK e t al.: DOPPLER WEATHER RADAR

1545

TORNADO CYCKNE SIGNATURE clDIUUOE P W A

* l O l v n ~
I-5mi.-

Fig. 28. Mesocyclone signature track for the Falconhead tornado showing damage path location relative t o the cyclone signature center (April 19, 1976). The reflectivity field of the storms for that day is shown on Figs. 15 and 16, where the boxed areas contain the storm that produced the only tornado on this day.

vergence (188' Az and 75-km range). The roadial component of storm motion (SM: 10 m . s-' from 225 ) has been subtracted from all velocities. A tremendous advantage is obtained with Doppler radar because it can sort out among many storms the ones that have intense circulation and hence potential for tornado development. Fig. 15 shows a large s t o m system composed of many individual convective cells. The multimoment display depicts the principal moments in a sector of space that can be placed over any storm so that the principal moments can intat metebe simultaneously examined for evidence of sgicn orological phenomena. Each storm can be systematically interrogated and for the example shown in Fig. 15 only the storm outlined by the box produced a tornado. This tornado's mesocyclone signature was tracked for almost an hour, and Fig. 28 shows the position of the signature relative to the damage path. Even though the storm was in the Doppler radar's second trip, the beamwidth was sufficiently small (0.8') for tracking at ranges to 170 km when the signature was first noticed. Although the Doppler radar had an unambiguous range of 115 km on this day, storm distribution was such that none were range-overlaid onto this mesocyclone, thus allowing an unobscured measurement of its velocity signature. The overlaying of storms due to small unambiguous range (see Section V) associated with Doppler radars can result in obscuration of signatures. 3) Doppler Spectra o f Tornadoes: In 1961 Smith and Holmes reported a tornado spectrum that was obtained with a CW Doppler radar [ 1 121. It was twelve years later that a tornado was first observed by a pulsed Doppler radar [1321. ~adar views principally that portion of circulation which lies within the resolution volume so that tracers moving with the same velocities contribute t o spectrum components ac-

cording to their reflectivity, isodop density, and antenna pattern iUumination (see Fig. 24). Only those targets whose spectral power is above the receiver noise level may be positively identified. However, it has been our experience, that tornadoes in a resolution volume offer enough reflectivity, due to debris and hydrometeors, that a large velocity span can be observed. Presently it is not known whether tracers moving at the peak tornado wind speed can be resolved. When centered on the' beam axis, the Rankine vortex model predicts a bimodal spectrum which was verified experimentally several times [ 1321, [ 1331. Shown on Fig. 29 are Rankine model vortex spectra matched to data by a least squares fit. The examples of spectra are from the Stillwater maxitornado. Deduced from the fitting are maximum velocity of 92 m . s-' and tornado diameter of 300 m. The deduced maximum velocity is larger that the radar's unambiguous velocity (k34.4 mas-'); therefore, aliasing was introduced in the model spectrum and the estimates are indirect. Two spectra closest to the tornado were simultaneously least squares fitted. Simulated model vortex spectra and real spectra show very good agreement not only for two where the fit was made but also for adjacent gate locations (Fig. 29). Resolution volumes corresponding to any of these simulated spectra are assumed t o have uniform reflectivity within the volume. Differences in echo power from each resolution volume are accounted for by forcing each simulated spectrum to have power. equal to its matching data spectrum. Asymmetry of spectral peaks (Az 21 .lo; range 104.1 36 km) about zero velocity suggests that targets were centrifuged outward with a velocity of 13m.s-'. In view of the variety of displays and signatures associated with tornadoes, it is natural to ask which technique is most promising for their detection. A project was established involving the Environmental Research Laboratories, National Weather Service, Air Weather Service, and Air Force Geophysical Laboratory t o conduct experiments that should provide some answers. Operations were conducted during the Spring of 77 and 78. It became apparent that mesocyclone circulations showed very nicely on the color display of radial velocities. Multirnoment display in conjunction with the color display proved most suitable for TVS recognition. Using criteria discussed in Brown et al. [2 11, the project scientists were able to detect a number of tornadoes [25]. As a matter of fact, all tornadoes that occurred within a range less than 115 km (the radar's unambiguous range) were detected. The average lead time was about 20 r i . It was established that nn circulation starts at mid levels (6-8 km) and works its way toward the ground. At further ranges signatures of small tornadoes are lost due to poor resolution; however, large destructive tornadoes were detected up to 240 km using the 0 . 8 ~ beam of NSSL's Doppler radar. Spectrum width and shear alone have not been reliable indicators of tornadoes since turbulent areas in storms exhibit large widths and could be easily mistaken for tornadoes. Based on those experiences, i t is believed that an operational tornado detection system will involve interaction between human operators and an automated scheme whereby the velocity pattern of a tornado cyclone is recognized.
C. Pulsed Doppler Observation o f Clear Air Wind Fields

Whenever turbulence mixes air in which there are gradients of potential temperature and water vapor density, the turbulence causes spatial fluctuations in the refractive index n . The

1546
m m m
RIIM

PROCEEDINGS OF THE IEEE, VOL. 67, NO. 11, NOVEMBER 1979

' m md

I m

1r4

la.!=

1 " Rl'E IM 16

rn n.
14.

fm a.1 f

rn n.

..
'a'
1.

4. I . 'a.

'a. 4.

.
4.

a.

j/
4
I

J,L-,-.
4

,
I . .

. .

. I

. ..
=

%'CiR~*rrtDsF&d

t7
C = o.38~-l/1. i

Fig. 29. Spectra from three consecutive azimuthal and range locations; Stillwater tornado. Dots show spectrum estimates from recorded time series data weighted with a von Hann window. Solid Lines are three-point running averages. Dashed lines are simulated spectra. The mean square difference between data and simulated spectra is simultaneously minimized for two spectra (Az-21.1' and range of 103.5 and 104.1 km). Resolution volume depth is 150 m, range gate spacing 600 m, and antenna beamwidth 0.8'. The tornado is located between the two upper middle gates. Tornado parameters obtained from these two fitted spectra are used to compute the remaining 7 simulated spectra Height above ground for these spectra is 640 m.

Thus if we know the structure constant of refractivity, we can, through use of (6.1 1) and (2.19), determine the echo power scattered by refractive index irregularities. Echoes from clear air have been seen almost from the inception of radar observations. These angel echoes were first mystifying, but often were actually associated with birds and insects. Clear air echoes, not related to any visible object in the atmosphere, were conclusively proven t o emanate from refractiveindex fluctuations through use of multiwavelength radars at Wallops Island [63]. Simultaneous measurements of refractive index fluctuations and reflectivity corroborated this finding [ 1 0 5 ~ . Of course, radar studies were preceded by an enormous amount of measurement using tropospheric-scatter DAn ([n(r + Ar) - n(r)12) (6.9) communication links which often depend upon clear air refractive index fluctuations t o provide reliable wide-band cirwhere ( ) denotes ensemble average. Whenever DAn is as- cuits between distance points [981 (see also a special PROsumed independent of r (i.e., the statistical properties of An CEEDINGS OF THE IRE, Scatter Propagation Issue, Oct. are spatially uniform) we refer t o the refractivity fluctuations 1955). In the 1960's ultrasensitive incoherent radars were used t o as being locally homogeneous. Tatarski [ 1 181 has shown for scales within the inertial subrange of atmospheric turbulence remotely detect and resolve clear air atmospheric structure (i.e., scales from a few millimeters t o tens or sometimes a and these studies are well reviewed by Hardy and Katz [64]. These radars 'showed meteorological phenomena such as confew hundred meters) that vective thermals [73], [75], sea and land breeze [86], and Kelvin-Helmholtz waves [69]. Doppler processing of coherent radar echoes can improve where C: is the refractive index structure constant and is re- target detection by at least an order of magnitude [67] and

fluctuations are small (e.g., one part in a million). Nevertheless, sensitive microwave radars detect the very faint echoes returned from these irregularities .in what otherwise (without turbulence) would be a smoothly changing n with negligible backscatter. in temperature, and pressure need t o be described in a statistical manner. Thus correlation and its Fourier transform, the power spectrum, are used t o characterize the spatial variability of n. Tatarski [ 1181 related the velocity spectrum of turbulence scales t o the correlation and spectrum of refractivity scales. Furthermore, Tatarski demonstrated that, there is a hierarchy of that prevail in turbulent flow, those scales equal t o A12 contribute most to backscatter. The parameter needed to obtain the tion from statistically inhomogeneous media is the structure function DAn defmed as

lated to reflectivity 17 as [ 63 ] (6.1 1)

DOVIAK et al.: DOPPLER WEATHER RADAR

1547

hence medium resolution weather radars could have a detection capability matching that often associated with large aperture (18-27-m diameter) antennas used with incoherent radars. Furthermore, a coherent radar provides a way in which ground clutter can be distinguished from moving atmospheric targets and allows data acquisition at closer ranges, thereby taking advantage of the r-2 dependence in echo power. Chadwick et al. [30] have monitored C: values in the planetary boundary layer and have concluded after one year of observation that winds can always be measured to several hundred meters height with moderately sensitive radars. Clear air wind measurement has practical signifkance because pulsed Doppler radars under development by the FAA also could measure wind shear hazards near airports for all weather Fig. 30. Rofde of average structure constant cn2 due to temperature fluctuations Tersus height h. Smooth limes are minimum radar conditions [79]. detectable cn at range R using NSSL's Doppler system as configured Harrold and Browning [65] found that radar can delineate on 4/19/76 (dashed line) and for system if reconfigured (solid line) to the upper limit of convection, prior to precipitation, showing provide lower system noise temperature and higher transmitted power. (From Ochs and Lawrence (931.) that it is deeper in some regions than others. Some of these areas of deep convection persist for several hours and, if showers develop, they occur w t i , and only within such ihn regions. Important economic advantage can be achieved if fractive index structure constant was consistently above radars can serve the dual purpose of locating weather hazards lo-'' m-2/3 over land and ocean within the measured height interval. A sample profile of averaged C due to temperature : to aircraft and predicting the location of incipient showers. Mapping boundary-layer wind over large areas has double fluctuations over land in November 1971 is shown in Fig. 30. significance for severe studies because 1) it allows early o b A bistatic radar was used to resolve weak scatter in the upper servation of thunderstorm development-hence, storm genesis trdposphere from strong scatter in the lower 1-2 km of the can be followed from the very beginning of cumulus develop troposphere [43]. These measurements inferred a continuum : ment; 2) the capability of Doppler weather radars to map of scatterers up 7 km having a C value that could be as large clear air wind makes possible monitoring thunderstorm out- as lo-'' m-'I3 which compares with those values at 3 km reported by Ochs and Lawrence. Crane's [36] monostatic flow and inflow which is usually precipitation free. : Strongest fluctuations in refractive index occur where radar analysis for several da s of observations show that C turbulence mixes large gradients of mean potential tempera- is larger than 1 X 10-l7 m-2 for heights up to 15 km in the ture and specific humidity [ 1251. Gage et al. [54] measured clear air. These results suggest Doppler weather radars with the height distribution of forward scattered signal strength minimum detectable structure constant of 10-l7 m-'I3 might to show good correlation between it and gradients of poten- be able to monitor continuously the wind throughout the tial temperature at high altitudes where water vapor con- troposphere. 2 ) Frequency Modulated Continuous Wave Radar: The FM tributions can be ignored. The tropopause is a region where potential temperature always increases with height, and it but CW radar transmits a continuous sinusoidal signal whose was first detected by radar in 1966 [6]. More recently Van frequency f = a t is linearly swept with a period T,. A target Zant et al. [ 12 11, using backscatter results from a vertically at range r returns a signal which is mixed with that being pointed VHF radar, obtained consistent agreement between transmitted, to produce a difference frequency f A proporC rawinsonde inferred C: and radar measured : for the clear tional to (2ralc) (see Section IEA). Thus fa is a measure of target range, and ranging accuracy is determined by the a r above the moist boundary layer. i For many years ionospheric scientists have been employ- dwell time Td available to measure f A . For near targets Td ing high power VHF and UHF radars t o observe the ion& is nearly equal to the time T, required to change the microsphere. Recently these radars at Jicamarca, Peru [124]; wave frequency by A f =aT,. The decisive advantage of Arecibo, Puerto Rico; Chatanika, AK [8]; and Lindau, Ger- FM-CW radar is that range resolution can be increased simply many [ 101I have lowered their sights to examine echoes in by spanning a larger frequency A f without decreasing average the nonionized stratosphere and troposphere. A 6-m VHF transmitted power. Spatial resolution of the order of 1 m can be achieved pulsed Doppler radar specifically designed for tropospheric studies was recently assembled at Sunset, CO, and is making easily with an FM-CW radar to show detailed structure of continuous observations of winds in the troposphere [60]. wave phenomena in the clear air planetary boundary layer Some advantages of the VHF radar are that it can sometimes (PBL). High resolution PBL probing with FM-CW radar was differentiate between scatter from hydrometeors and re- first demonstrated by Richter [991 and clear air echoes fractive index fluctuations [61], and it might be able to usually are found in layers, often with wave-like structure detect coherent scatter (second term of (2.10a)) from mean 1571. refractive index gradients at inversion layers [55]. The FM-CW radar is sufficiently coherent for nearby targets. I ) Height DLtribution of Refractive Index Structure Con- For the first time, Strauch et al. [ 115 I exploited this property stant: In 1971 Ochs and Lawrence [931 made temperature to measure Doppler shift of FM-CW radar echoes from the fluctuation measurements using a sensor mounted on an air- optically clear boundary layer. Linear sawtooth modulation craft and sounded the atmosphere in the lower 3 km. Their of the microwave frequency gives the FM-CW radar characwork demonstrated that the temperature contribution to re- teristics that have analogy to the pulsed Doppler radar: the

7'

PROCEEDINGS OF THE IEEE, VOL. 67, NO. 11, NOVEMBER 1979

m';.?,
+ "-4
\
-/

I
/

G"AL WPREA
AREA

'.-//

Fig. 31. PPI sector scan of echo power from clear air seen by the NRO Doppler radar. Bright area of high power is aligned roughly parallel t o mean wind and bands are spaced about 4 km apart. Range marks are 2 0 km apart. EL = 1.6' (April 2 7 , 1 9 7 7 , 1 4 4 7 CST).

Fig. 32. The dual Doppler-radar area (dashed lines) within which the angle subtended by the radials from the Cimarron (CIM) and Norman (NRO) radars lies between 30' and 150". The outlined box is the region wherein Doppler velocities were synthesized for detailed wind analyses. The wind speed and direction is a mean over 1.24-km depth of the PBL.

reciprocal of A f can be considered analogous t o an effective pulse length for the pulse radar, while T, is equivalent t o the PRT. Because the FM-CW's peak-to-average power ratio is unity, we immediately infer that resolution can be increased without changing the effective per-pulse transmitted energy; not so with pulsed radar where decrease in 7 reduces transmitted pulse energy. Therefore, per-pulse SNR is decreased in proportion to the square of pulsewidth, whereas the equivalent SNR of the FM-CW radar decreases linearly with effective pulse width (i.e., A f - I ) . T i unquestionable advantage of hs the FM-CW vanishes when considering pulsed Doppler radar and coherent echoes. As a matter of fact, the pulsed radar may prove more advantageous [67]. 3) Observations of Turbulence and Roll Vortices: Clear air winds in the planetary boundary layer (PBL, i.e., surface t o heights of about 1.5 km have been synthesized from dual Doppler radar measurements [59], [76], wherein chaff was dispensed over large areas to provide suitable echo levels detection and processing). Doviak and Jobson [47] showed first results of two Doppler radar synthesized wind fields in the PBL clear air where only the diffuse and intrinsic scatterers of the medium were used as targets. They observed mean wind fields at low height to have qualitative agreement with mean wind measured (with conventional anemometers) near the surface. On April 27, 1977, a day marked by strong nondirectional shear and curvature in the wind profile, NSSL's Doppler radar echo power measurements showed evidence of clear air convective streets (Fig. 31), an observation that should signify the presence of roll vortices. First radar detection of clear air thermal streets was reported by Konrad [74]. Fig. 32 locates the radars at CIM and NRO as well as a444-m meteorologically instrumented tower. The 25 km X 25 km region (solid lined area) is where synthesized dual Doppler radar winds were analyzed in detail and is referred to as the primary synthesis region. However, reasonably accurate winds can be synthesized from Doppler data in the entire area enclosed by the dashed lines. The winds were fairly uniform from the southwest on this day, but there were small perturbations from the mean wind having a magnitude of about one order less than the mean wind itself. As is evident in Fig. 32, the x direction and u component of wind are along the mean wind and the y direction and u component normal to the mean wind.

Reflectivity is estimated from digitally averaged log-video samples at 762 contiguous resolution volumes (spaced 150 m) along the beam. Doppler velocities are measured for each of these volumes using the PP autocorrelation algorithm. Radial velocity estimates were obtained from 255 contiguous PP samples. Data are acquired at the rate of about 5 radialsls, and with the rotation rate of 2'/s, there are about 2 independent data pointslbeamwidth (0.8'). Sectors stepped in elevation and scanned in azimuth by each radar encompass the primary synthesis region. Each beam scanned the same sector six times, starting at 0.5' elevation, and tilted up in 0.5' increments to 3.0'. With each radar, the tilt sequence of sector scans was begun at the same time. The synthesized wind field and its perturbations from the mean are shown in Fig. 33. Berger and Doviak [13] have examined the spatial spectra S(K), where K is the wavenumber of the synthesized winds on this convectively dry day, and have compared results with those spectra obtained from anemometers located on the tall tower. The spectra follow a 513 power law in the wavelength A range of 1 to 8 km in agreement with spectra of tower winds. Spectral analyses of clear air longitudinal wind fluctuations using chaff and a single Doppler radar have been reported by ~ h e r n i k o v t al. [32]. Their spectra, extending to e 3-km wavelengths, also show a 513 power law dependence in agreement with results on this day. However, O'Bannon [92] shows spectra on another day when such a power dependence is not evident. A sample set of spectra, multiplied by K and plotted on a semilogarithmic paper t o show turbulence intensity per waveband, is displayed in Fig. 34. The v component spectra in the y direction have a peak at 4-km wavelength which persisted over the one hour of data collection. Various theories (e.g., [58], [77] ) suggest that roll vortices tend to form parallel to the mean wind in a strongly heated PBL having a large unidirectional shear. When vertical profiles of horizontal velocity have curvature, Keuttner [77] predicts rolls to have a horizontal spacing of 2.8 times their depth. The strong 4-km wave in t h e y direction of the u component might be the convective rolls predicted by theory. Fig. 35(a) depicts perturbation winds at one of six levels synthesized from a tilt sequence of each Doppler radar. A bandpass filter was applied in the y direction to emphasize

DOVIAK et al.: DOPPLER WEATHER RADAR 27 R R 1377 P

HEIGHT

DURL DOPPLER UINDS 1'4'46'45-1'4'4935 CST

ionPs
=.51
.

1.0 Kn
. $

LI-L----A-

144645 144935 CST 27 APRIL 1977

----ID km 1.25 --

AGL

--=-------. ----------7 -~-------01

.-----------_C__------

":12.s'

------------------- '-pi' 'i&.b '4.0. . b ' 'i4.i. -4.0 'L2.i' 'G.S 2

? - - - - - - - - - - - - - - d - - - C
I -

27 RPR 1377
HEIGHT

ionps

1.0 Kn

X-DISTANCE l KM 1

-i.5'

'3.5

(a) DURL DOPPLER UINDS 1'4'46'45-1'44935 CST

.S=MERN FLOII

Fig. 34. A plot of KS(K) versus log K which shows distribution of power (1446:45-1449: 36 CST). Note large power in velocity fluctuations at wavelength A = 4 km.

BURL DOPPLER

WINDS

(b) Fig. 33. (a) Dual radar synthesized wind field at 1.0 k m above ground. (b) Wind with its mean removed. Synthesized wind fields were lowpass fiitered once in X and Y direction with a 3-point Shuman fdtm [131.

the 4-km wave feature for visual display. A low-pass fitter was applied in the x direction along which no dominant wavelength was noted. Fig. 3 5(b) is a vertical cross section at X = 25.5 km perpendicular to the mean wind. Vertical velocities were derived by integrating the mass continuity equation using wind fields from the six horizontal surfaces. Vertical grid spacing is 250 m. Readily apparent are counter-rotating vortices (roll vortices) having approximately 4-km wavelength whose maximum vertical velocities are of the order of 1 m s-'. Furthermore, the ratio of roll spacing t o height is 2.6 in good agreement with that predicted by theory [77]. Gilmer et al. [56]

Kg. 35. Horizontal (a) and vertical (b) cross sections of bandpass filtered wind data that highlights the clear air roll structure seen in the spectra displays of unfdtered data.

PROCEEDINGS OF THE IEEE, VOL. 67, NO.

27 APRIL 1977

11, NOVEMBER 1979

144810 - 145140 CST

HEIGHT = I.Okm

-2

X-LAG

(Kt4 I

X-LAG I K M I

FTg. 36. Cross correlation coefficients c (as a function of horizontal spatial lag) of (a) the v component (in the direction of mean wind) for two wind fields, at 1.0 km AGL, synthesized 3% min apart. The mean wind speed is 14.9 m-s-' and the medium times for the two fields are 1448:lO CST and 1451 :40CST (b) the v component (transverse to the mean wind). A spatial lag corresponding to mean wind advectioil during the 3% min time interval maximizes both C(U) and c(v).

analyzed aircraft gust probe data collected on this day during the time of the radar observation. They also detect a prominent peak in power density at a wavelength of about 4 km in the y direction. 4 ) Time Correlation o f Clear Air Wind Perturbations: The cross correlated (in space and time) wind fields (synthesized 3-1/2 min apart) at 1.0 km above ground level (AGL) show that correlation is maximized for a translation equal to mean wind advection (Fig. 36) proving that these convective kinematic features advect with the mean wind. O'Bamon [92] has traced the time evolution of clear air eddies using wind synthesized from a sequence of dual Doppler-radar data acquired 30 s apart. His results demonstrate that clear air eddy fields of kilometer scales have a time scale of at least 10 min and, as with results on April 27, 1977, correlation is maximized when eddy winds are displaced to account for advection by the mean wind.

VII. CONCLUSIONS The introduction of Doppler frequency shift measurement capability into weather radars has opened new horizons for exploration by atmospheric scientists. The astounding success achieved with these radars in detection of thunderstorm cyclones well in advance of tornado formation should cause incorporation of coherent systems in new radars used for operation by the national services. Advances in digital signal processing and display techniques have allowed economical development of real time presentation of the three principal Doppler spectral moments. The techniques are here and constantly improving, and we believe that important new scientific disclosures and new operational applications in the areas of wind measurement and tornado detection are forthcoming. Doppler radars at centimeter wavelengths do not have a sufficiently large velocity-range ambiguity product ra ua to match that required to observe, without obscuration, severe convective storms. There is no comprehensive data and docu-

mentation that show the full extent of the .problem nor is there any foreseeable solution. Thus it appears that storm observers will have to accept some limitations in Doppler radar weather measurements. Dual Doppler-radar observations of the kinematic structure of severe storms and the planetary boundary layer (PBL) agree with theoretical models but much investigation is still required. The Doppler weather radar shows promise of greatly increasing our knowledge of thunderstorms and t e planetary boundary h layer on scales not before possible. Furthermore, we can monitor -cant mesoscale phenomena which are of importance to a r traffic safety, air pollution control, and (peri haps most important) we may be able to see the triggering impulses of severe storms that each year cause such destruction. Increased power and sensitivity of weather radars may soon result in the meteorologist being able to observe the wind structure and its evolution throughout the troposphere. The important advances in meteorological observations brought forth by the application of Doppler techniques to uue. weather radars will continute in the f t r . However, there is room for further improvement in the radar system to reduce the deleterious effect of ambiguities while lessening data acquisition time for observation of severe storm convection, shear, and turbulence in clear or precipitation laden air.

The authors appreciate the support they have received from their colleagues in the Environmental Research Laboratories (ERL), Air Force Geophysics Laboratory, National Weather Service, Energy Research and Development Administration, National Research Council, and Federal Aviation Administration, not only in the preparation of this paper, but also for the continuing development of Doppler weather radar technology. We are particularly indebted to Dr. R. Strauch of the Wave Propagation Laboratory (ERL) and a reviewer whose careful study of this manuscript has significantly improved

DOVIAK e t al.: DOPPLER WEATHER RADAR

the text. We are indebted to Ms. Joy Walton for her efficient and accurate typing and careful editing of the manuscript.

127) W. C. CampbeU and R. C. Strauch, "Meteorological Doppler

1281 [ I ] M. Abramowitz and I. A. Stegun, Handbook of Mathematical Functions, Nat. Bur. Standards, Appl. Math. Ser. 55, 2nd [2] [3] (41 15) [6] (71 (81 1291 [30]

[9]

(101

[ 11 ]
[I21 [13]

1141 11 5 1
[ 16 1

(171 [I81 [19]

[20] 1211 122 1

[ 23 1
(241

1251

1261

printing, Supp. Doc., U.S. Gov. Printing Office, Washington, DC, 1964. L. Armijo, "A theory for the determination of wind and precipitation velocities with Doppler radars," J. Atmos. Sci., vol. 26, pp. 570-573,1969. G. M. Armstrong and R. J. Donaldson, Jr., "Plan shear indicator for real time Doppler radar identification of hazardous storm winds," J. Appl. Meteorol., vol. 8, pp. 376-383, 1969. D. Atlas, "Radar analysis of severe storms," Meteorol. Monogmphs, Amer. Meteorol. Soc., Boston, MA, 02108, vol. 5, no. 27, pp. 177-222, 1963. -, "Advances in radar meteorology," in Advances in Geophysics, Landsberg and Mieghem, Eds. New York: Academic, pp. 31 7-478, 1964. D. Atlas, K. R. Hardy, K. M. Glover, I. Katz, and T. G. Konrad, "Tropopause detected by radar," Science, vol. 153, pp. 11101112, 1966. D. Atlas, R. C. Srivastava, and R. S. Sekhon, "Doppler radar characteristics of precipitation at ver~ical incidence," Rev. Geophys. Space Phys. vol. 2, pp. 1-35, 1973. B. B. Balsley, "The use of sensitive coherent radars t o examine atmospheric parameters in the height range 1-100 km," in Preprints 1 8 t h Conf. Radar Meteorol., Amer. Meteorol. Soc., Boston, MA 02108, pp. 190-193, Mar. 1978. B. B. Balsley, N. Ciasnos, D. T. Farley, and M. J. Baron, 'Winds derived from radar measurements in the arctic troposphere and stratosphere," J. Appl. Meteorol., vol. 16, pp. 1235-1239, Nov. 1977. L. J. Battan, "Some observations of vertical velocities and precipitation sizes in a thunderstorm," J. Appl. Meteorol., vol. 3, Pp. 41 5-420, 1964. -, Radar Observation of the Atmosphere. Chicago, IL: Univ. Chicago Press, 1973. B. R. Bean and E. J. Dutton, Radio Meteorology, Nat. Bur.Standards, Monograph 92, Supp. Doc., U.S. Gov. Printing Office, Washington, DC 20402, 1966. M. I. Berger and R. J. Doviak, "Dual Doppler-radar observations of wind fields in a dry convective boundary layer," in Preprints Fourth Symp. Meteorol. Obxrv. I n s m . , Denver, CO Amer. Meteorol. Soc., Boston, MA pp. 2 19-226, 1978. T. Berger and H. H. Groginsky, "Estimation of the spectral moments of pulse trains," presented a t the Int. Conf. Infonn. Theory (Tel-Aviv, Israel, 1973). L. V. Blake, "Prediction of radar range," in Radar Handbook, MemU I. Skolnik, Ed. New York: McGraw-Hill, 1970, ch. 2. M. Born and E. Wolf, Principles of Optics, 2nd ed. New York: MacMillan, 1964. E. N. Boyenval, Echoes from precipitation using pulsed Doppler radar," in Proc. 8 t h Weather Radar Conf., Amer. Meteorol. Soc., Boston, MA 02108, pp. 57-64, 1960. E. A. Brandes, "Optimizing rainfall estimates with the aid of radar," J. Appl. MeteoroL,vol. 14, pp. 1339-1345, 1975. J. Q. Brantley and D. A. Barczys, ' S o m e weather observations with continuous wave Doppler radar," in Proc. 6th Weather Ra&r Conf., h e r . Meteorol. Socl, Boston, MA. 02108, pp. 297-306,1957. 0. Brovko, "The structure of a maximum likelihood center frequency estimator," Tech. Internal Corresp. No. 2313.00/042, Hughes Aircraft Co., May 16, 1977. R. A. Brown, L. R. Lemon, and D. W. Burgess, "Tornado detection by pulsed Doppler-radar," Mon. Weather Rev., vol. 106, no. 1, pp. 29-38, Jan. 1978. K. A. Browning and R. Wexler, "A determination 0 kinematic : properties of a wind field using Doppler-radar, J. Appl. Meteorol., vol. 7 no. 2, pp. 105-1 13, 1968. I. P. Burg, "Maximum entropy spectral analysis," presented a t the 37th Annu. Meeting Soc. Exploration Geophysicists (Oklahoma City, OK, Oct. 1967). D. W. Burgess, "Single Doppler radar vortex recognition: Part I-Mesocyclone signatures," in Preprinb 1 7 t h Conf. Ra& Meteorol., Amer. Meteorol. Soc., Boston, MA 02108, pp. 97103, Oct. 1976. D. W. Burgess, J. D. Bonewitz, and D. R. Devore, "Joint Doppler operational project: Results year 1 ," in Preprints 1 8 t h Conf: Radar Meteorol., Amer. Meteorol. Soc., Boston, MA 02108, pp. 442-448, Mar. 1978. D. W. Burgess, L. D. Hennington, R. J. Doviak, and P. S. Ray, "Multimoment Doppler display for severe storm identification," J. AppL Meteorol., vol. 15, pp. 1302-1306, 1976.

[31]

1321

1331 134) 135) [36] (371 [38] [39] 1401 1411 [42] 1431 1441 1451

(461

[47] 1481 [49]

[ 501
151 ] [52j [53] [541

radar with double pulse transmission," in Preprints 1 7 & Conf. Radar Meteor., Arner. Meteorol. Soc., Boston, MA 02108, pp. 42-44, Oct. 1976. P. A. F. Caton, 'Wind measurement by Doppler radar," Meteorol. Mag., vol. 92, pp. 213-222, 1963. R. B. Chadwick and G. R. Cooper, "Measurement of distributed targets with the random signal radar," IEEE Trans. on Aerosp. Elecrron Syst., vol. AES-8, pp. 743-750, Nov. 1972. R. B. Chadwick, K. P. Moran, and G. E. Morrison, "Measureclimatology, in Preprints 18th Conf. Radar ments toward a cnZ Meteorol., Amer. Meteorol. Soc., Boston, MA 02108, pp. 100103, Mar. 1978. A. A. Chernikov, "Radio-wave scattering in clouds and the radar equation," in Radar Meteorology: Proceedings of the Third AllUnion Conf. (1968), V. V. Kostarev, A. A. Chernikov, and A. B. Shupyatskii, Eds., translated by Don Hardin. Jerusalem, Israel: Keter, 197 1. A. A. Chernikov, Yu. V. Mel'nichuk, N. 2.Pinus, S. M. Shmeter, and N. K. Vinnichenko, "Investigations of the turbulence in convective atmosphere using radar and aircraft," Rad. S c i , vol. 4 , no. 12, pp. 1257-1259, 1969. A. M. Chimera, "Meteorological radar echo study," Final Rep. Contract AF33(616)-6352, Cornell Aerosp. Labs., Buffalo, NY, 1960. J . W. Cooley and J. W. Tukey, "An algorithm for the machine calculation of complex Fourier series," Math Comp., vol. 19, pp. 297-301,1965. R. K. Crane, "Automatic cell detection and tracking," Special Issue on Radio Meteorol., I E E E Tram. Geosci. Electron, Oct 1979. -, "Stratospheric turbulence analysis," Final Rep AFGL-TR0207, Air Force Geophys. Lab., Hanscom AFB, MA 01 731, 1977. K. C. Crawford, "The design of a multivariate mesoscale field experiment," Ph.D. dissertation, Univ. of Oklahoma, Norman, OK 73069,1977. W. B. Davenport and W. L. Root, An Introduction to the Theory of Random Signals a n d Noise. New York: McGrawHill, 1958. G. W. Deley, 'Waveform design," in Radar Handbook, Merrill I. Skolnik, Ed. New York: McGraw-Hill, 1970, pp. 3-1-347. J. N. Denenberg, "The estimation of spectral moments," Rep. Lab. Atmos. Probing, Dep. Geophys. Sci., Univ. Chicago and Dep. Elec. Eng., Illinois Inst. Tech., 82 pp., 1971. R. J. Donaldson, Jr., "A demonstration of antenna beam errors in radar reflectivity patterns," J. Appl. Mereorol., vol. 3, pp. 61 1-623, Oct. 1964. -, "Vortex signature recognition by a Doppler radar," J. Appl. Meteorol.,vol. 9, pp. 661-670, 1970. R. J. Doviak, J. Goldhirsh, and A. R. Miller, "Bistatic radar detection of high altitude clear-air atmospheric targets," Rad. Sci. V O ~ . 7, pp. 993-1003, 1972. R. J. Doviak and D. Sirmans, "Doppler-radar with polarization diversity,"J. Atmar. Sci., vol. 30, pp. 737-738., 1973. R. J. Doviak, D. Burgess, L. Lemon, and D. Sirmans, "Doppler velocity and reflectivity structure observed within a tornadic storm," J. Rech. A m o s . , vol. VIII, nos. 1-2, pp. 235-243, Jan.June 1974. R. J. Doviak, R. S. Ray, R. G. Strauch, and L. J. Miller, "Error estimation in wind fields derived from dual-Doppler radar measurement," J. Appl. Meteorol., vol. 15, no. 8, pp. 868-878, 1976. R. J. Doviak and C. T. Jobson, "Dual Doppler-radar observations of clear air wind perturbations in the planetary boundary layer," J. Geophys. Re*, vol. 84, no. C2, pp. 697-702, 1979. R. J. Doviak, D. Sumans, D. Zmif, and G. B. Walker, "Considerations for pulse-Doppler radar observations of severe thunderstorms," J. Appl. Meteorol., vol. 17, no. 2. pp. 189-205, 1978. R. J. Doviak and D. S. Zrnif, "Receiver bandwidth effect on reflectivity and Doppler velocity estimates," J. AppL Meteorol., vol. 18, no. 1 , pp. 69-76, 1979. R. W. Fetter, "Real time analog Doppler processing (RANDOP)," in Preprints 16th Conf. Radar Meteorol., Amer. Meteorol. Soc., Boston, MA, pp. 153-555, 1975. G. B. Foote and P. S. duToit, "Terminal velocity of raindrops aloft," J. Appl. Meteorol., vol. 8, pp. 249-253, 1969. A. S. Frisch and S. F. Clifford, "A study of convection capped by a stable layer using Doppler radar and acoustic echo sounders," J. Atmos. Sci.,vol. 31, pp. 1622-1628, 1974. A. S. Frisch, L. J. Miller, and R. G. Strauch, 'Three-dimensional air motion measured in snow," Geophy. Res. Letr., vol. 1, pp. 86-89, 1974. K. S. Gage, W. P. Birkemeyer, and W. H. Jasperson, "Atmospheric stability measurement at tropopause altitudes using forward-scatter CW radar," J. Appl. Meteorol., vol. 12, pp. 12051212,1973.

1552 [ 5 5 ] K. S. Gage and J. L. Green,Evidenceforspecularreflection from monostatic VHF radar observations of the stratosphere, Rad. Sci.,vol. 1 3 , n o . 6 ,pp. 991-1001, Nov.-Dec. 1978. [ 561 R. 0.Gilmer, R.E. McGavin, and R. F. Reinking, A small aircraft probe gust system studies for of boundary layer convectionandtransport,in PreprintsFourthSymp.Meteorol. Observ.Znstru., Amer.Meteorol.SOC.,Boston, MA, pp. 426432, Apr. 1978. [ 571 E. E. Gossard, J. H. Richter, and D. R. Jensen, Effect of wind shear atmospheric on wave instabilities revealed by FM-CW radarobservations,in Boundary-LayerMeteorology, vol. 4. Dordrecht, The Netherlands: Reidel, 1973, pp. 1 1 3-1 3 1 . [ 581 E. E. Gossard and W.R. Moninger, The influence of a capping inversion on the dynamic and convective instability of a boundary layer model with shear, J. Atmos. Sci, vol. 32, pp. 21112124,1975. [ 591 E. E. Gossard and A. S. Frish, Kinematic models of a dry convective boundary layer compared with dual Doppler-radar observations of wind fields, in Boundcrry-LuyerMeteorology, vol. 10. Dordrecht,TheNetherlands:Reidel, 1976, pp. 31 1-330. [ 6 0 ] J. L. Green, J. M. Warnock, R.H. Winkler, and T. E. VanZandt, Studies of winds in the upper troposhere with a sensitive VHF radar, Geophys. Res. Letts.,vol. 2, pp. 19-21, Jan. 1975. [ 61 ] J. L. Green, R.H. Winkler, J. M. Warnock, W. L. Clark, K. S. Gage, and T. E. VanZandt, Observations of enhanced clear air reflectivity associated with convective clouds, in Preprints 18th Con$ Radar Meteorol., Amer. Meteorol. Soc., Boston, MA, pp. 88-93, Mar. 1978. [ 6 2 ] H. L Groginsky, The coherent memory fdter, Electron Prog., . Raytheon Co.,vol. 9, no. 3. pp. 7-13, 1965. [ 6 3 ] K. R. Hardy, D. Atlas, and K. M. Glover, Multiwavelength backscatter from the clear atmosphere, J. Geophys. Res., vol. 71, no. 6 , pp. 1537-1 552, 1966. [ 6 4 ] K. R. Hardy and I. Katz,Probingtheclearatmospherewith highpower,highresolutionradars, Proc. ZEEE, vol. 57, pp. 468-480, 1969, [ 6 5 ] T. W. Harrold and K.A. Browning, Identification of preferred areas of shower development by means of high power radar, Quart. J. Roy. Meteorol. SOC.,vol. 97, no 413, pp. 330-339, July 1971. [ 6 6 ] C. W. Helstrom, StatisticalTheory of Signal Detection. New York: Pergamon, 1968, [ 6 7 ] L. Hennington, R. J. Doviak, D. Sirmans, D. Zmii, and R. G. Strauch, Measurements of winds in the optically clear air with microwave pulse-Doppler radar, in Preprints 17th Con$ Radar Meteorol., Amer. Meterol. SOC., Boston, MA, pp. 342-348, Oct. 1976. [ 6 8 ] S. L. Hess, Introduction to Theoretical Meteorology. New York: Holt, Rinehart, and Winston,1959. (691 J. J.Hicks and J. K. Angell,Radarobservationsofbreaking gravitational waves in the visually clear atmosphere, J. Appl. Meteorol.,vol. 7, pp. 114-121, Feb. 1968. [ 701 W. Hitschfeld and A. S. Dennis, Turbulence in snow generating cells, Sci. Rep. MW-23, McGill Univ., Montreal, p-Q., Canada, 31 pp., 1956. [ 711 E. M. Hofstetter,Simpleestimatesofwakevelocityparameters, M.I.T. Lincoln Lab., Lexington, MA, unpublished Tech. Note 1970-11, 1970. [ 7 2 ] G. H. Hyde and K. E. Perry, Doppler phase difference integrator, M.I.T. Tech. Rep. 189, 1958. [ 7 3 ] I. Katz, Sea breezestructureasseenbyradar,Appl.phys. Lab., Johns Hopkins Univ. Memo BPD 66U-25, 1966. of clear airconvectivecells, [ 7 4 ] T. G. Konrad,Thealignment Proc. Intern Con$ Cloud Phys., Toronto,Ont.,Canada, pp. 539-543, 1968. [ 7 5 ] -, The dynamics of the convective process in the clear air as seen by radar, in Preprints 14th Con$ RadarMeteorol., Amer. Meteorol. SOC., Boston,MA, DR. 57-60, 1970. (761 R. A. Kropfli and N. M. Kohn,Persistenthorizontalrollsin the urban mixed layer as revealedbydual-Dopplerradar, J. AppL Meteorol., vol. 17, pp. 669-676, May 1978. [ 771 J. P. Kuettner, Cloud bands in the earths atmosphere, Tellus XXZIZ, v01. 4-5, pp. 404-425, 1971. 1781 J. T. Lee, Application of Doppler weather radar to turbulence measurements which affect aircraft, Final Rep. #FAA-RD-77145, Syst. Res. Development Serv., FAA, Washington, DC 20590, 1977. [ 7 9 ] J. T. Lee and C. Goff, Gust front wind shear and turbulenceconcurrent aircraft and surface based observations, in Preprints 7th Con$ Aerosp. Aeronaut. Meteorol. Symp. Remote Sensing from SateZIites, Melbourne, FL, h e r . Meteorol. SOC., Boston, MA, pp. 48-54. 1976. [SO] R. M. Lhermitte, Motions of scatterers and the variance of the mean intensity of weather radar signals, Sperry-Rand Research Center, 5RRC-RR-63-57, Sudbury, MA, 1963.

PROCEEDINGS O F THE IEEE, VOL. 67, NO. 1 1 , NOVEMBER 1979 (81 I -, Doppler radars as severe storm sensors, Bull. Amer. Meteorol. Soc., vol45, pp. 587-596, 1964. [ 821 -, Dual-Doppler radar observations of convective storm circulation, in Preprints 14th Con5 Radar Meteorol., Amer. Meteorol. SOC., Boston, MA, pp. 153-1 56, 1970. [ S 3 ] -, Real processing meteorological time of Doppler radar signals, in Preprints 15th Con$ Meteorol., Radar Amer. Meteorol. SOC., Boston, MA, pp. 364-367, 1972. [ 841 R. M. Lhermitte and D. Atlas, Precipitation motion by pulse Doppler radar, in h c 9th Con$ o. Weather Radar, Amer. Meteorol. SOC.,Boston, MA, pp. 218-223, 1961. [ 8 5 ] J. S. Marshall and W. Hitschfeld, Interpretation of the fluctuating echo from randomly distributed scatterers,Can. J. Phys., VOl. 31, pt. I, pp. 962-995, 1953. [ 861 J. H. Meyer, Radar observations of iand breeze frontsat Wallops Island, Virginia, in Preprints 14th Con$ RadarMeteor., Amer. Meteorol., SOC.,Boston, MA, pp. 61-67, Nov. 1970. [ 871 K. S. M i e r a n d M. M. Rochwarger, A covariance approach t o spectral moment estimation, IEEE Trans Inform, Theory, VOI. IT-18, pp. 588-596, 1972. [ 8 8 ] E.A. Mueller and E. J. Silha, Unique features of the CHILL radar system, in Preprints,lBth Con$ Radar MeteoroL, Amer. Meteorol. Soc., Boston, MA, pp. 381-382, Mar. 1978. [ 891 F.E. Nathanson, Radar Design Principles. New York: McGrawHill, 1969. L. Smith, A modified coefficient for [ 9 0 ] F. E. Nathanson and P. theweatherradarequation,in Preprints1Sth Con$ Radar Meteorol., Amer. Meteorol. SOC., Boston, M k pp. 228-230, Oct. 1972. I911 L. R. Novick and K. M. Glover,Spectralmeanvarianceestimation via pulse pair processing, in Preprints 16th Con$ Radar Meteorol., Amer.Meteorol.SOC.,Boston, MA, pp. 1-5,1975. (921 T. OBannon, A studyof dual-Dopplersynthesizedclearair wind fields, in Preprints 18th Con$ Radar Meteorol., Amer. Meteorol. SOC., Boston, MA, pp. 65-69, 1978. [ 9 3 ] G. R. Ochs and R. S. Lawrence, Temperature and Cnz profiles measured over landandoceanto 3 km abovethesurface, NOAA Tech. ERL Rep. 251-WPL 22, Boulder, CO, 39 pp., 1972. [ 9 4 ) A. Papoulis, Probability, Rnndom Variables, Stochastic and Procesws. NewYork: McGraw-Hill, 1965. The equation in meteorology, [ 9 5 ] J. R. Robert-Jones, radar Quart. J. Roy. Meteorol. SOC., vol. 88, pp. 485-495,1962. I961 J.R. Probert-Jones and W. G.Harper,Verticalairmotionin in Proc. 9th Weather showers as revealedbyDopplerradar, Radar Con$, Amer Meteorol. SOC., Boston, MA, pp. 225-232, 1961. [ 9 7 ] P. S. Ray, R. J. Doviak, G. B. Walker, D. Sirmans, J. Carter, and B. Bumgarner, Dual-Doppler observation of a tornadic storm, J. Appl. MeteoroL,vol. 14, no. 8, p. 1521-1530, 1975. I981 P. L. Rice, A. G. Langley, K. A. Norton, and A. P. Barsis, Transmission loss predictions for tropospheric communication circuits, Nat. Bur. Standards Tech. Note 1 0 1 , vols. 1-2, Supp. Doc., .US. Gov. Rinting Office, Washington, DC 20402, 1967. I991 J. H. Richter, High resolution troposheric radar sounding, Rad. Sci.,vol. 4 , pp. 1261-1268, 1969. 1001 R. R. Rogers, effect variable The of target reflectivity on weatherradarmeasurements, Quart. J. Roy. Meteorol. Soc., VOl. 97, pp. 154-167, 1971. 1011 J. Rottger and C.H. Liu,Partialreflection and scattering of VHF radar signals from the clear atmosphere, Geophys. Res. Lett., vol. 5 , no. 5 , pp. 357-360, May 1978. 1021 W.D. Rummler, Introduction of a new estimator for velocity spectral parameters, Tech. Memo -684121-5, Bell Telephone Labs., Whippany, NJ, 1968. 1031 -, Accuracy of spectralwidthestimatorsusingpulse pair waveforms, Tech. Memo, Bell Telephone Labs., Whippany, NJ, 1968. 1041 W. Rutkowskiand A. Fleisher,R-meter: An instrumentfor measuringgustiness, MZT WeatherRadar Res. Rep. No. 24, 1955. 1051 Adwnces in RadioResearch, J. A. Saxton, Ed. vol. 1. New York: Academic, 226 pp., 1964. 061 A. J. F. Siegert and H. Goldstein, Propagation of Short Radio Waves, D. E. Kerr,Ed., vol. 13, MIT Rad.Lab.Series, 1964, Appendix B. 071 D. Sirmans and R. J. Doviak, Meteorological radar signal intenNSSL-64, 1973. sity estimation, NOAA Tech. Memo ERL OS] -, Wed-Doppler velocity isotach displays of storm winds in real time, J. Appl. Meteorol.,vol. 12, no. 4, pp. 694-697, 1973. 091 D. Sirmans and B. Bumgarner,Numericalcomparisonof five meanfrequencyestimators, J. AppLMeteoroL, vol. 14, pp. 991-1003,1975. 101 -, Estimation of spectral density mean and variance by covariPreprints 16th Con5 Radar ance argument techniques, in

DOVIAK e t al.: DOPPLER WEATHER RADAR

1553

Meteorol., Amer. Meteorol. Soc., Boston, MA, pp. 6-13, Apr.

1975.

[ 11 1 ] D. Sitmans, D. Zmib, and B. Bumgarner, "Estimation of maximum unambiguous Doppler velocity by use of two sampling rates," in Preprints 17th Con$ RadarMeteoroL, Amer. Meteorol. Soc., Boston, MA, pp. 23-28, 1 9 7 6 . [ I 121 R. L. Smith and D. W. Holmes, "Use of Doppler radar in metedrological observations," Monthly Wearher Rev., vol. 8 9 , pp. 17, 1961. [ I 1 3 1 P. L. Smith, Jr., K. R. Hardy, and K. M. Glover, "Applications of radar t o meteorological operations and research," Proc. ZEEE, vol. 6 2 , pp. 724-745, June 1 9 7 4 . [ I 1 4 1 3. D. Stackpole, "The effectiveness of raindrops as turbulence sensors," in Proc. 9th Weather Radar Conf., Amer. Meteorol. Soc., Boston, MA, pp. 2 1 2 - 2 1 7 , 1 9 6 1 . ( 1 1 5 1 R. G. Strauch, W. C. Campbell, R. B. Chadwick, and K. P. Moran, "FM-CW boundary layer radar with Doppler capability," NOAA Tech. Rep. ERL 329-WPL 3 9 , Boulder, CO, 1 9 7 5 . ( 1 1 6 1 R. G. Strauch and A. S. Frisch, "Doppler radar measurements of turbulent kinetic energy dissipation rates in a northeastern Colorado storm," J. Appl. Meteorol., vol. 15, pp. 1012-1017, 1976. [ 11 71 J. Sychra, "Relation between real and pulse volume averaged fields of reflectivity and velocity," Canadian Meteorol. Service Res. Reps., 1 9 7 2 . [ 11 81 V. I. Tatanki, Waw Propagation in a Turbulent Medium (translation by R. A. Silverman). New York: McGraw-Hill, 1 9 6 1 . [ I 1 9 1 J. W. Taylor and J. Mattern, "Receivers," Radar Handbook, Merrill I. Skolnik, Ed. New York: McGraw-Hill Book Co., 1 9 7 0 , Ch. 5. [ I 2 0 1 C. W. Ulbrich and D. Atlas, "The rain parameter diagram: Methods and applications," J. Geophys. Res., vol. 83, no C3, pp. 1 3 1 9 - 1 3 2 5 , 1 9 7 8 . ( 1 2 1 ] T. E. VanZandt, J. L. Green, D. S. Gage, and W. L. Clark, "Vertical profiles of refractivity turbulence structure constant: Comparison of observations by the sunset radar with new theoretical model," Rad. Sci., vol. 1 3 , pp. 819-829, 1 9 7 8 . [ I 2 2 1 P. Waldteufel, "An analysis of weather spectra variance in a

tornadic storm," NOAA Tech. Memo ERL NSSL76, 1 9 7 6 . [ I 2 3 1 R. F. Woodman and T. Hagfon, "Methods for the measurement of vertical ionospheric motions near the magnetic equator by incoherent scattering," J. Geophys. Res. Space Phys., vol. 7 5 , pp. 1205-1212, Mar. 1 9 6 9 . R. F. Woodman and A. Guillen, "Radar observations of winds and turbulence in the stratosphere and mesosphere," J. Atmos. Sci., VOI. 31, pp. 4 9 3 - 5 0 5 , 1 9 7 4 . A. M. Yaglom, "On the local structure of the temperature field in a turbulent flow," Dokl. Akad. Nauk SSSR, vol. 6 9 , p. 7 4 3 , 1949. G. W. Zeoli, "If versus video limiting for two-channel coherent signal processon," ZEEE Tram. Inform. Theory, vol. IT-1 7 , pp. 579-587, Sept. 1 9 7 1 . D. zrnid, "Moments of estimated input power for finite sample averages of radar receiver outputs," ZEEE n a m . Aerosp. Electron. Sysf.,vol. AES-11, pp. 1 0 9 - 1 1 3 , 1 9 7 5 . -, "Estimated tornado spectra and maximum velocity statistics," Final Rep. Grant No. 04-5-022-17, 1975. -, "Signal-to-noise ratio in the output of nonlinear devices," ZEEE Tmns. Inform. Theory, vol. IT-2 1, pp. 662-663, Nov. 1975. -, "Spectral moment estimates from correlated pulse pairs," ZEEE Tmm. Aerosp. Electron. Syst., vol. AES-13, pp. 344354, 1 9 7 7 . D. ~ r n i dand B. Bumgarner, "Receiver chain and signal processing effects o n the Doppler spectrum," in Preprints 16th Conf. Radar MeteoroL, Amer. Meteorol. Soc., Boston, MA, pp. 163-168, Apr. 1 9 7 5 . D. ~ r n i h and R. J. Doviak, "Velocity spectra of vortices scanned with a pulse-Doppler radar," J. Appl. Meteorol., vol. 14, pp. 1531-1539,1975. D. Zrnid, R. J. Doviak, and D. W. Burgess, "Probing tornadoes with pulse-Doppler radar," Quart. J. Roy. Meteorol. Soc., vol. 1 0 3 , vol 4 3 8 , pp. 707-720, 1 9 7 7 . D. Zrnif and R. J. Doviak, "Matched f l t e r criteria and range weighting for weather radar," IEEE Trans. Aerosp. Electron. Syst., vol. AES-14, pp. 925-930, 1 9 7 8 .

11241 [I251 I1261 [I271 11281 11291 [I301 (1311

[I321 [I331 11341

> "
A

*
,

r - a
,

Z>< r

Fig. 1 5 . M S H 57 1'1.1 d l ~ p l d r . thut1der5tr1rrr1 rrlls. on .I ~ c ~ r n a d iday nf c (181704 (-ST k p r d I 1 J 7 6 ) . <;r:~) slladlnp5 (dim, hrlghr, htack, d i m , etc.) rrprefent d B Z levels differing I??.31mut I 0 d R I a t d r t ~ n r t a 1 1 (IBL. K s n p r mark3 are 100 k n ~ :lp.Jrt, clcvnrivn analc = 0.0". Onarnbieuous rnncr Ir V l U km. ltte hilucd 3 r r ~ oufllnes a tornadic storrri cell w h u ~ e n e l ~ c y c l d n esicnature w3s r l e t e c t r ~ ltn r e d lime hv r N SSL's Yorm;!rr (YO H 1 Doppler r ~ t l ~ -r t l c h nearly colncated w l l l ~ ~ 1 i% the iVSR-57.

Fi6. 16. Same storm svstern ( 1 8 1 6 3 5 CST)seen wnh the I)oppler radar having 1 15-km unnmhr~uc~us range. IO.lug Z hriglllness categclries (dim, b r ~ g h t . ) starf at 10 da/ snd increment nt ahnut 10-dl3Z stepr. Tlie 10-Iog I wale applreq only tu furt-trip r c t ~ o m . Some range-overldid achnes can be r e c o g n ~ ~ eby Z?lzfr radialfv e l o n p tcd d shape. The llot nutllnes rhr same area :+sill Fik. 15. Range marks a r e ?U A m apart. Part of the tornnllxc storm i s obscured Ilv a nrarhy (30-60-kmrance) storm.

..

1?g. 26. Storm reflectivity In) and s n d o p ~ displayed on PPT a t 3 1 1 5 (b) CS?. The e l e w t ~ o n arlple i s I .sL.range n~nrkq cnrrrsponrl to 6 0 . 80, are: t 0 0 km. Reflectint!, tnctor c ~ t e ~ o r i e s dim ( < 2 I clBZ1, hrigllt (2 r -311, hlach (31 -44). dim {44-5?). and htip_hl ( > 5 7 d H 7 ) . Velvc ~ t ycatcetrrie~ are d r n ~< 1 3 ms-'1, R r ~ g h t ( 13-21). and hriphtcct ( 2 2 1 ). f'o>itive rad131 ve1wcifies a r e angularly strnhed in hriirhfncrs. Vrsncycltrnr. xiknrltrrrr is bctxvrcn 1 9 4 5 - ~ ~ 3 ' 73-'30 km. nnrl

i t

2 - I I t . r ; ~ ~ l l l ~ m tn n ~ t r r I lorly>l~.r1 ~ t - 1 ~ 11 5 4 b 1 3 n w \ o c \ t ki111t- t l i r l b d r r u w camtalnk rnforrrl tlion ot t h c 3 principal L)~)]?plrr hpecrrutn rnornenls t o r a re%nll~tl#lrl o l . t ~ r n[.'or ~ntcrpretartc~rl arrr>rv Grc ~~ r. 01 lnscrt. In Trpyrr r i e h t corner (arrow lengtl~i-, pruport~onaltrr received power, arrow rtiricrirm tn vcloc~tyant1 arrowl~etlds l r t t o Doppler specttum mtIt11). ..thsci~sa is 3nrnuth and ordrnate scale cleni~t~nf nt range (kmf fi.r>rnradar. H r l u s c l c e e p l ~ ~ ~ o r n ~Ion ir n t t clp nf hcree11.

Fis. 31. PPI

<ccta,r ,,an

ut

t.cl~r)IT<,\\,.r

t r r * r ' ~ t I , Ir t ~ r w;rb

bv the h R C )

B r l ~ r h t,lre:r r > t trtcl~ yoxtrer Ir slrrncd rtlughly p a r s l l ~ 4 tn nicsn winri and fland? rtre sp:u ed ahour 4 k:il a p m . I!anpe mark% are 2 0 kmapart. EL : 1.6 (Tcprll 23, 1977. 1417 CSTI.

Iloppler radlr.