1246 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 41, NO.

6, JUNE 2003

Preprocessing EO-1 Hyperion Hyperspectral Data to

Support the Application of Agricultural Indexes
Bisun Datt, Tim R. McVicar, Tom G. Van Niel, David L. B. Jupp, Associate Member, IEEE, and
Jay S. Pearlman, Senior Member, IEEE

Abstract—The benefits of Hyperion hyperspectral data to TABLE I

agriculture have been studied at sites in the Coleambally Irriga- COMMON SPECTRAL FEATURES FOR CROPS AND SOILS
tion Area of Australia. Hyperion can provide effective measures
of agricultural performance through the use of established
spectral indexes if systematic and random noise is managed.
The noise management strategy includes recognition of “bad”
pixels, reducing the effects of vertical striping, and compensa-
tion for atmospheric effects in the data. It also aims to reduce
compounding of these effects by image processing. As the noise
structure is different for Hyperion’s two spectrometers, noise re-
duction methods are best applied to each separately. Results show
that a local destriping algorithm reduces striping noise without
introducing unwanted effects in the image. They also show how
data smoothing can clean the data and how careful selection of in a sun-synchronous orbit at 705-km altitude and is nominally
stable Hyperion bands can minimize residual atmospheric effects 1 min behind the Landsat-7 satellite. Hyperion images 256
following atmospheric correction. Comparing hyperspectral pixels with a nominal size of 30 m on the ground over a
indexes derived from Hyperion with the same indexes derived
7.65-km swath. The platform attitude can vary as the satellite
from ground-measured spectra allowed us to assess some of these
impacts on the preprocessing options. It has been concluded that is rolled to image specific targets. During the first year of
preprocessing, which includes fixing bad and outlier pixels, local operation, Hyperion processing was developed and refined to a
destriping, atmospheric correction, and minimum noise fraction stage (Level 1B1) where well-calibrated data became routinely
smoothing, provides improved results. If these or equivalent pre- available. In this paper, we examine processing and noise re-
processing steps are followed, it is feasible to develop a consistent duction methods that can be applied post-Level 1B1 to provide
and standardized time series of data that is compatible with
field-scale and airborne measured indexes. Red-edge and leaf consistent time series of Hyperion data for an agricultural site
chlorophyll indexes based on the preprocessed data are shown to near Coleambally in New South Wales (NSW), Australia. This
distinguish different levels of stress induced by water restrictions. will be referred to as the Coleambally Irrigation Area Site
Index Terms—Agricultural indexes, Hyperion, hyperspectral, (CIAS). The benefits to that study, and the implications for
image processing. more general agricultural applications, are assessed by focusing
on crop attributes that can be derived from Hyperion using a
range of hyperspectral indexes.
I. INTRODUCTION
Hyperion data have three potential advantages over space-

T HE HYPERION sensor [1], carried by the National

Aeronautics and Space Administration (NASA) Earth
Observing 1 (EO-1) satellite [2], is the first spaceborne hy-
borne multispectral instruments, such as Landsat Enhanced
Thematic Mapper (ETM). First, they can provide an enhanced
level of information for atmospheric correction to derive surface
perspectral instrument to acquire both visible near-infrared reflectance. Second, they can provide access to detailed spectral
[(VNIR) 400–1000 nm] and shortwave infrared [(SWIR) indexes based on surface reflectance. Finally, they can be used
900–2500 nm] spectra. Hyperion is a pushbroom sensor with to integrate (or “bin”) the hyperspectral data consistently to
two spectrometers and a single telescope. The EO-1 satellite is synthetic bands equivalent to any of the broadband sensors
or to bands of yet to be developed instruments. This spectral
Manuscript received July 22, 2002; revised January 1, 2003. This work was “binning” of Hyperion will be dealt with in a separate paper.
supported by the Commonwealth Scientific and Industrial Research Organiza- This paper investigates the first two advantages and how they
tion Earth Observation Centre. The work of T. McVicar and T. Van Niel was
supported by the Cooperative Research Centre for Sustainable Rice Production benefit from careful preprocessing.
under Project 1105. The work of J. Pearlman was supported by the National
Aeronautics and Space Administration Earth Observing 1 Program.
B. Datt and D. L. B. Jupp are with the Commonwealth Scientific and Indus-
II. KEY SPECTRAL INDEXES FOR AGRICULTURAL ASSESSMENT
trial Research Organization Earth Observation Centre, Canberra, ACT 2601, The use of broadband space-based remote sensing is well
Australia.
T. R. McVicar and T. G. Van Niel are with the Commonwealth Scientific established in agriculture for regional and global monitoring
and Industrial Research Organization Land and Water, Canberra, ACT 2601, and assessment [3], as is the use of spectrally more detailed
Australia. information at the laboratory scale [4]. Current research is
J. S. Pearlman is with the Advanced Network Centric Operations, Phantom
Works, Boeing Company, Seattle, WA 98124 USA. using airborne and ground-based hyperspectral data [5]–[7] to
Digital Object Identifier 10.1109/TGRS.2003.813206 scale laboratory techniques to remote platforms.
0196-2892/03$17.00 © 2003 IEEE
DATT et al.: PREPROCESSING EO-1 HYPERSPECTRAL DATA TO SUPPORT THE APPLICATION OF AGRICULTURAL INDEXES 1247

Fig. 1. Typical spectral features occurring in the CIAS. Mean ASD spectra (convolved to 176 Hyperion bands) for the nine sampling sites are shown.

TABLE II
SITES MEASURED BY AN ASD AT CIAS

The work in progress at the CIAS uses field and laboratory soils. A green reflectance peak near 550 nm, photosynthetic
spectral data to calibrate and investigate various derivative and chlorophyll absorption (680 nm), a steep slope in the red-edge
integral indexes. The aim is to find indexes most sensitive to region (700–750 nm), and leaf water absorption near 970 and
plant nitrogen, water, chlorophyll, lignin, cellulose, and other 1240 nm are present in the crop spectra. The soil spectra show
plant chemical substances plus soil chemical factors such as clay clay absorption near 2200 nm, and the stubble spectra show
composition and soil nitrogen content. The major such spectral lignin/cellulose absorption near 2100 and 2300 nm and plant
regions are summarized in Table I. wax/oil absorption near 1720 nm.
A range of ground spectra with a 1-nm resolution has been We have selected six indexes to assess the effects of noise
taken within the CIAS as validation for Hyperion and as infor- management. The wavelengths used in the formulation of these
mation to accompany site sampling. The site means obtained indexes are located within what we will term the “stable” bands
from data taken by an Analytical Spectral Devices (ASD) Field- of Hyperion (see Section V-C). This provides a reasonable
spec spectroradiometer are plotted in Fig. 1 (see Table II). These expectation of stable calculation of these indexes. All Hyperion
spectra have been integrated to the 10-nm Hyperion bands for wavelengths reported in this paper are based on the central
later comparison with Hyperion data and display signatures and wavelengths corresponding to pixel 128. The six indexes are
spectral features typical of green vegetation, dry vegetation, and defined as follows.
1248 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 41, NO. 6, JUNE 2003

1) The red-edge (RE) wavelength is the wavelength po- For Hyperion, bands 51 and 110 were used which
sition of the maximum slope in the red-near-infrared correspond to wavelengths of 864.8 and 1245.4 nm,
(700–750 nm) reflectance of vegetation. Several studies respectively.
have shown that the RE wavelength is correlated with 6) The 2200-nm absorption feature [or clay index (CI)]
chlorophyll content in leaves and canopies [8]–[10]. The present in the Coleambally soils results from combina-
RE wavelength is a good candidate for monitoring vege- tions and overtones of hydroxyl absorption in the clay
tation productivity, as well as detecting the onset of stress lattice structure [17]. This feature is present in the re-
and senescence. The RE shifts to longer wavelengths flectance spectrum of kaolinite, a common clay mineral.
during plant growth (increasing chlorophyll content) and CI was calculated as the area of the continuum removed
to shorter wavelengths during stress (chlorophyll decline reflectance spectrum between 2193 and 2213 nm, and
in leaves). used Hyperion bands 204, 205, and 206 corresponding
2) The first-derivative value at the red edge (dRE) defines to wavelengths of 2193.8, 2203.9, and 2214.0 nm. Con-
the value of the maximum slope associated with the RE tinuum-removed absorption feature analysis is a useful
wavelength above. This index can be sensitive to green method for estimating the abundance of compounds
vegetation amount [percent cover and leaf area index from reflectance spectra of samples [18]. The band depth
(LAI)]. Filella and Peñuelas [11] found the area of the and/or area of the continuum removed absorption fea-
red-edge peak to be a strong indicator of LAI. Other tures are calibrated against the concentration of chemical
variants of dRE using the sum of first and second deriva- compounds of interest or compared with contents of
tive values around the red-edge band or derivative green spectral libraries [19], [20]. Such indexes are robust to
vegetation indexes (DGVIs) have also shown strong residual effects following atmospheric correction.
linear relationships with green vegetation cover [12],
[13]. The original Hyperion bands used here to develop
both the RE and the dRE were 33 to 40 or wavelengths III. COLEAMBALLY IRRIGATION AREA SITE:
681.8–753.0 nm. DATA AND IMAGE BASE
3) A leaf chlorophyll index (LCI) developed in [14] was
The CIAS is a 95 000 ha site in southern NSW, Australia
found there to be a sensitive indicator of chlorophyll
(Latitude 34 48 4.3” S, Longitude 146 0 48.96” E, 120 m
content in leaves and was less affected by scattering from
above sea level) that has over 500 farms with large (up to 70 ha)
the leaf surface and internal structure variations. LCI is
flat uniform fields. The climate has high variability, but the
a ratio of relative chlorophyll absorption depths in the
RE (710 nm) and red (680 nm) wavelengths. It works water usage and cropping area are well managed at regional
best at moderate to high chlorophyll levels. The value of and farm levels. Farming practices are mechanized and inten-
LCI increases as leaf chlorophyll content increases, and sive, and landholders are interested in maximizing efficiencies
is calculated as through better management.
The CIAS is a focus of remote sensing research to determine
LCI (1) the information content available from current broadband and
hyperspectral satellite sensors. The CIAS has an excellent base
where, in [14], the reflectance at wavelength
of geographic and agricultural management information plus
850, 710, and 680 nm. The same notation applies
survey information from specific farms. Since December 2000,
throughout the paper. The Hyperion band numbers
a time series of Landsat ETM and coincident EO-1 images have
used to develop the LCI here are 50, 36, and 33. These
been collected over two southern hemisphere summer growing
bands correspond to wavelengths of 854.7, 712.3, and
seasons (2000/2001 and 2001/2002). There have also been a
681.8 nm, respectively.
number of field campaigns to collect spectral and ancillary data
4) The water index (WI) compares the leaf (liquid) water
for calibration and applications research.
absorption band near 970 nm with a reference band at
900 nm, which does not show leaf water absorption [15]. The data for the CIAS on January 12, 2002 include Landsat
The value of WI increases with leaf water content. WI is ETM, EO-1 Hyperion and ALI, plus complementary ground
calculated as (spectral, canopy characteristics, and plant chemistry), airborne
(spectral and geometric), and atmospheric data. An ASD
WI (2) Fieldspec spectroradiometer was used to collect ground-based
spectra for selected crops and fields. HyMap airborne scanner
For Hyperion, bands 55 and 83 were used which cor- data [28], covering the 400–2500-nm spectral range in 126
respond to central wavelengths of 905.5 and 973.1 nm, bands, were also acquired for part of the CIAS on January
respectively. 12, 2002. A CIMEL Electronique CE318–1 instrument was
5) The normalized difference water index (NDWI) is based provided by NASA’s AERONET Program [21] (AERONET
on a leaf (liquid) water absorption band near 1240 nm Instrument Number 69) to measure atmospheric characteris-
and a nonabsorbing reference band near 860 nm [16]. The tics. The application and accuracy of CIMEL and other sun
value of NDWI also increases with leaf water content, and photometers under Australian conditions is discussed in [22].
is calculated as Every January, Coleambally Irrigation Co-operative Limited
acquire a high-resolution (1 : 50 000 scale or 2-m pixel) digital
NDWI (3) air photo mosaic over the CIAS. This provided an accurate
DATT et al.: PREPROCESSING EO-1 HYPERSPECTRAL DATA TO SUPPORT THE APPLICATION OF AGRICULTURAL INDEXES 1249

geographic base for geometric registration of the various image TABLE III
and other data to a common map base. HYPERION 176-BAND SUBSET

The conjunction of these datasets and the nature of the area

(road network and field layout plus little relief) makes Coleam-
bally an ideal location for 1) geometric and sensor alignment
validation, 2) sensor performance validation, 3) sensor intercal-
ibration, 4) atmospheric correction validation, and 5) agricul-
tural applications development. In an initial study addressing
Hyperion geometric accuracy, average rms errors were as small
as 12.9 and 11.6 m, respectively, in the across and along track of indexes and for the direct application of crop models. They
directions. Details can be found in [23]. Recent results also con- also include its utility for directly matching image data to
firm that processing by TRW Corporation to Level 1B1 results spectral libraries, ground data and to plant chemistry mea-
in excellent VNIR and SWIR coregistration. surements. CIAS field spectra have been compared with pixel
The ASD data were measured over nine sites (Table II). At values extracted from atmospherically corrected data using
each site, five ASD spectra were averaged, each spectrum being packages such as HATCH [25], FLAASH [26], and ACORN
internally averaged 15 times. Times and dates are local (GMT + (http://www.aigllc.com/acorn/intro.asp). The packages have all
1100). The Hyperion image was acquired at 10:59 A.M. on Jan- been found to provide measures of the atmospheric water vapor
uary 12, 2002. The location of each ASD site was obtained from that agree in magnitude and variance with measurements made
a real-time differential GPS unit. At four of the sites (two soil by the CIMEL instrument and had generally good agreement
and two stubble fields), 100 m 100 m grids of spectral data with ASD data taken close to the time of the overpass. These
were measured to quantify (inter-) calibration and atmospheric results will be reported separately. Since the different packages
correction performance at the Hyperion scale. These sites repre- gave similar results, we have used the FLAASH package, with
sent the largest, stable dynamic range in reflectivity within the standard options, as the baseline for the work reported here due
study area. At the other five sites, less intensive, more oppor- to its close integration with the Modtran 4 code development.
tunistic spatial sampling occurred on the following two days,
but with a wider range of agricultural covers being measured to B. Hyperion Noise and Implications for Base Processing
evaluate the performance of indexes and their correlation with The Hyperion VNIR sensor has 70 bands, and the SWIR has
crop and soil information. At all sites surface geometry may in- 172 bands providing 242 potential bands. A number of the bands
troduce some variation into the ASD measurements; however, were intentionally not illuminated and others (mainly in the
we have not explicitly corrected the ASD data for sun angle overlap region between the two spectrometers) correspond to
variations due to bidirectional reflectance distribution function areas of low sensitivity of the spectrometer materials. Because
(BRDF) effects. of this, only 198 bands have been provided in Hyperion Level
1B1 products; the unused bands (1 to 7, 58 to 76, and 225 to
IV. IMAGE PROCESSING TO MANAGE DATA 242) are set to zero values by TRW software during the Level
AND PROCESSING NOISE 1B1 processing [27].
The Level 1B1 product provided by TRW [27] includes cor-
A. Basic Image Processing Approach rections that remove dark current bias effects and correction
Since Hyperion operates from a space platform with con- steps for what are termed “bad” pixels remaining in the 198
sequently modest surface signal levels and full-column atmo- bands. The pixel- and band-dependent calibrations have been
spheric effects, its data demand careful processing to manage adjusted over the life of the sensor and represent a well-bal-
sensor and processing noise. The approach taken here involves anced set of calibrations through effective use of solar, lunar,
selecting bands of greatest value (i.e., the best tradeoff between intersensor, and vicarious methods. Among the 198, there are
information content and noise level for the purposes of the ap- four remaining bands in the overlap between the two spectrom-
plication), addressing the vertical striping noise that the VNIR eters. These are VNIR bands 56 (915.7 nm) and 57 (925.9 nm)
and SWIR arrays display, and managing residual and introduced and SWIR bands 77 (912.5 nm) and 78 (922.6 nm). It is usual
noise following a standardized atmospheric correction to ap- to eliminate two of these to obtain 196 unique bands. In the
parent surface reflectance. processing reported here, we have selected as the unique 196
For many forms of data processing such as classification the bands 8 to 57 and 79 to 224. This choice was made, since
with training sets, minimum noise faction (MNF) [24] or the SWIR band 78 has the higher level of base noise after de-
principal component analysis (PCA) [30], and the application striping (described below). This higher base noise level can be
of various indexes, it is not necessary to atmospherically confirmed by analysis of the “dark” images provided with Hy-
correct image data for a single observation. There is also a perion data.
significant risk in using atmospheric correction, as it tends to Atmospheric water vapor bands that absorb almost all of the
amplify noise levels, hence reducing the SNR. However, there incident and reflected solar radiation are easily identified by vi-
are also many situations, which all occur in our work in the sual inspection of the image data or by atmospheric modeling.
CIAS, where it is important to transform hyperspectral data Accepting this as a good criterion for band elimination for land
to apparent surface reflectance. These include its use in data surface applications yields the subset of 176 bands listed in
normalization for temporal studies, for persistent calibration Table III.
1250 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 41, NO. 6, JUNE 2003

The wavelengths removed correspond to strong atmospheric since launch and to handle specific cases where it affects the de-
water vapor absorption bands between 1356 and 1417 nm, 1820 rived products.
and 1932 nm, and above 2395 nm. They are commonly removed
from ASD or other spectrometer data (including Hymap) taken C. Balancing Vertical Stripes in the VNIR and SWIR Arrays
under natural light conditions. They are not sampled in the spec- The statistics of the detector arrays can be studied by accumu-
tral ranges selected in the HyMap airborne scanner [28]. For the lating mean, variance, minimum, and maximum data for each
work presented here, the 176 bands provide the primary input pixel in each band over the lines of an image. As discussed
band set for processing and interpretation. above, a vertical stripe is said to occur where the statistics in-
The 176 Hyperion bands still display additional effects that dicate that the image information is likely to be valid (that is,
arise from the pushbroom technology. In this paper, the spatial the pixel is not “bad”) but with significantly modified gain and
elements of one line of an image will be called “pixels” and offset. We are also assuming that such gains and offsets are rel-
are the same as “samples” in other usage. They sample the land atively stable over a collect—but not necessarily between col-
surface in the across-track direction. As the pushbroom moves lects. A general approach to removing vertical stripes with these
forward, a given pixel creates a “column” of data or an along- characteristics is then similar to methods used in the past to bal-
track (vertical direction) dataset. Each column in a single band ance horizontal stripes in mirror scanner images by histogram
corresponds to a single detector in one of the arrays. equalization [29] or to flatten images affected by limb bright-
For pushbroom instruments, a poorly calibrated detector ening or to balance detectors in airborne pushbroom sensors
in either the VNIR or SWIR arrays will leave a vertical [34]. That is, histogram moments, such as the means and vari-
“stripe” in a displayed image band. The most extreme of ances of the columns in each band, are used to balance the sta-
these (which contain little or no valid data) are identified as tistics of the arrays to those of a reference histogram.
“bad” pixels and interpolated in the Level 1B1 processing. The pixel balancing introduced here is different in that it may
In more recent NASA processing [35], these pixels are listed be done either “globally” or “locally.” In global balancing, the
but left unmodified allowing users to select different methods statistical moments of each column are modified to match those
to handle or replace the pixels. Sometimes additional “bad” for the whole image for each band. In the local approach, ref-
pixels occur due to infrequent pixel saturation, which makes erence moments are estimated locally. However, if an affected
the correction for sensor “echo” inoperable and at other times pixel is quite extreme but not “bad” in the sense defined above,
the extra “bad” pixels seem to be associated with events in the its presence can adversely affect the local statistics. It is, there-
space environment such as high-energy particles. Fortunately, fore, best to treat such outliers separately in an initial pass. The
there are not many “bad” pixels. The current Level 1B1 “bad” initial outlier pass used here is local but uses median statistics
pixel list contains about 17 individual pixels out of 242 256 and thresholds rather than means to identify the pixels to adjust
pixels with two cases where a block of bands for a single pixel rather than modifying all of the column data to statistics affected
needs to be fixed. There are also a number of independent by the outliers.
detector offsets that reflect the structure and read-out process of Mathematically, let be the mean of the detector at the th
the arrays. These are primarily removed in the dark-correction pixel position for band and be the corresponding within-
although some effects remain that are significant for dark column standard deviation over the data sensed as the image is
targets—such as water bodies—but do not affect the indexes acquired—or on some group of lines in the image. Moreover,
being discussed here. let and be corresponding “reference” values for these
Hyperion also displays pixel pushbroom effects of a less se- moments. The differences between local and global methods
vere but still problematic nature. These are pixels where the and the outlier or general image passes occur in the way the
data calibration is effectively spatially and possibly temporally reference values are established.
varying at scales determined mainly by array structure rather Formally, the approaches can all be expressed as finding a
than the earth’s surface structure. At one level, a single detector gain ( ) and an offset ( ) for each detector such that the
in one of the arrays may have an anomalous calibration. If such values of the image data for sample , line , and band
pixels have no association with other detectors they will create are modified to
pixel-level and pixel-independent vertical stripes in the image
data. Spatial correlation can come from within-array interac- (4)
tions and may be related to the read-out process. Correlated
blocks of stripes seem more prevalent in the SWIR array than where the gain and offset are computed by
the VNIR array.
Hyperion also has low-frequency array effects such as those
due to spectral “smile. This can arise from the optical techniques (5)
used to spectrally disperse the input imaged “slit” over the de-
tector arrays. It creates a variation in central wavelength and That is, generally speaking, the gain setting controls the
bandwidth across the swath of the sensor—or in a single image within-column standard deviation after processing and the offset
line. The spectral “smile” effect is fully described in other pa- controls the mean. It follows that if the reference standard
pers in this special issue, and while well characterized prior to deviations are the same as the actual image values, the result
launch [27], it is the subject of on-going investigations (such as is an additive change with no alteration of the within-column
in [33]) to establish whether there have been changes to its form standard deviation.
DATT et al.: PREPROCESSING EO-1 HYPERSPECTRAL DATA TO SUPPORT THE APPLICATION OF AGRICULTURAL INDEXES 1251

The global method takes the reference mean to be the total

image mean and may also take the reference standard deviation
to be the whole image within-column standard deviation

(6)

The local methods involve either outlier detection and replace-

ment or the use of local smoothing filters. For outlier detection
(assuming “bad” pixels have been previously fixed) it is pos-
sible to compute

test (7)
(a) (b) (c) (d)
where “ ” indicates a local median of selectable neighbor- Fig. 2. MNF Bands 1 and 15 (a) and (b) before and (c) and (d) after global
hood. Outlier pixels are those where “test” is above a specified destriping. The 176-band selection from Hyperion was used for the MNF.
threshold. Pixels with anomalous standard deviations are also
identified by a similar formula. The outlier destriping is applied
as an initial pass to the identified bands using median values as
the reference values.
Local destriping proceeds when the outliers have been
treated. It uses as reference values

(8)

The notation “ ” used here indicates a local mean with se-

lectable neighborhood.
Global equalization to remove the striping is mathematically
simple but enforces a very severe constraint of uniform column
statistics on an image. Gradients in image radiance and distinct
patches of different cover types are common in images, and it is
therefore, rare for the vertical column statistics of an image to (a) (b)
be well balanced across a whole line in the land surface infor- Fig. 3. MNF bands 1 and 5 of the difference between global destriped and
mation. An apparent advantage claimed for global equalization original radiance image. The 176-band selection from Hyperion was used for
has been that it seems to remove low-frequency effects such as the MNF.
those due to spectral “smile.” This comes at the cost of changing
the spectra as discussed more fully in Section IV-D. [Fig. 2(a)]. The subsequent MNF bands are soon dominated by
In our experience, the particular choices of method and set- vertical striping [Fig. 2(b)].
tings that work best are different for the VNIR and SWIR ar- If global destriping is applied to this image, the result of ap-
rays. They also tend to vary with environmental conditions and plying the MNF transformation [Fig. 2(c) and (d)] indicates
between images with distinctly different land covers (such as that both the broad low-frequency effect and the local stripes
forests, crops, water and deserts). The CIAS was treated as a have been reduced and the image is clean. Moreover, by se-
single land cover for this study but the local neighborhoods were lecting stable bands that reduce residual atmospheric effects
different in the VNIR and SWIR arrays as discussed in Sec- and globally destriping, the MNF becomes clear of the effects
tion IV-D. Other land covers may well need to be stratified. noted above to about 20 transformed bands. This is a signifi-
cant number of features for such data as will be discussed in
D. Testing the Effects of Destriping Using the MNF Transform Section V-B.
The effects of destriping and band selection on Hyperion were However, examination of the differences between the original
tested using the MNF transformation [24]. The MNF technique and destriped data indicates that not only are the pixel to pixel
responds to interactions between the spatial structure of the data stripes and the broad low-frequency VNIR effect removed, but
and that of the noise when the noise has strong spatial structure. also midrange frequencies related (in this case) to field sizes and
This is the case with the image striping. The MNF cannot gen- the balance of crop and fallow fields in vertical columns. This
erally be used to filter out such effects, since they merge with is illustrated in Fig. 3 where the MNF Bands 1 and 5 of the dif-
image information but it can illustrate them clearly, as shown ferences between the global destriped and raw or “undestriped”
in Fig. 2(a) and (b) where the MNF images of the radiance radiance images are presented. Removing these field-scale spa-
data are shown. The first MNF band has a strong spatial gra- tial patterns can alter the spectral and spatial characteristics of
dient that corresponds to the spectral “smile” in the VNIR array the data. For example, the predominance of watered rice in one
1252 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 41, NO. 6, JUNE 2003

(a) (b) (c) (d)

Fig. 4. MNF bands 1 and 5 of the difference between local destriped and original image. The 176-band selection from Hyperion was used for the MNF.

vertically (north/south) extended area [see Fig. 3(b)] depressed Consequently local destriping with different widths in the VNIR
the red region over the whole length of the image to the point and SWIR is the recommended approach in the current applica-
where the atmospherically corrected reflectance was negative in tion with “smile” being handled in a different way such as has
the rice fields! been described in [33].
The low-frequency “smile” effect is displayed clearly in the
difference image MNF 1 (see [33] for more explanation) while V. MANAGING IMAGE NOISE FOLLOWING
MNF 5 shows a variety of midfrequency effects that are due to ATMOSPHERIC CORRECTION
fields and not to noise at either the low- or high-frequency end
of the scale. These represent informative spatial information that Atmospheric correction generally reduces the SNR in the
has been removed by the process. data due to its transfer of the uncorrelated noise to the resulting
A local method in which the mean and variance used as refer- spectra and also due to the introduction of spectral effects
ence moments were obtained by local averages over five pixels through differences between the model used and the actual
removed pixel-to-pixel scale stripes and not the midfrequency- atmosphere. The latter includes effects in Hyperion due to
scale land cover effects. However, it retained the low-frequency spectral “smile” not being taken into account by currently
“smile” effects in the image data. The result is shown in Fig. 4 available packages. The management of noise in this situation
where Fig. 4(a) and (b) shows the MNF 1 and 5 for the differ- is dependent on the application, the section of the spectrum
ence image and Fig. 4(c) and (d) shows the MNF 1 and 15 of of most interest, the environmental conditions during the data
the locally destriped bands. The low-frequency “smile” effect acquisition, and many other factors and is currently the focus
is now retained as MNF 1 in the data but the lower order MNF of continuing research. We have, therefore, used only a few
is clear of the local striping and also retains the midfrequency of the known and established steps as examples against which
field-scale information. to judge their benefits to agricultural measurements. These
are reference spectrum smoothing, MNF smoothing, and band
By examining the statistics of the VNIR and SWIR arrays
selection.
separately, it is found that the majority of the striping effects in
the VNIR are distinct, mainly independent and persistent within
and between scenes. They are easily and best removed by a local A. Reference Spectrum Smoothing
destriping with a narrow window (such as five pixels). There are The residual noise after atmospheric correction includes two
also stripes that occur in blocks of pixels and seem to be related specific types. One is the sensor and processing noise that
to the read-out of the array. These blocks of stripes, which occur contributes to the inherent (and reduced) SNR of the data.
more often in the SWIR than the VNIR, are best handled by It is generally uncorrelated spectrally and spatially or has
using a much wider local filter. Local filters with widths of up only local correlations that reflect more the geometry of the
to 41 pixels were used to reduce these effects in the SWIR in sensor than the spatial correlations of the scene. The other
this paper. Block striping in the VNIR seems to be isolated to a type includes systematic effects that relate to the differences
few specific bands and these are best treated separately. between the modeled and real atmospheres at the time the
With the local approach it is necessary to investigate and data were collected.
understand the nature of the low-frequency pushbroom effects In principle, the second could be estimated if you had a
separately from the vertical striping. Global destriping seems range of actual ground surface reflectance values and compared
to remove the VNIR “smile” effect as well as the stripes but them with the output from the atmospherically corrected data.
also alters midfrequency spatial effects in the data and so is Systematic and persistent differences may be identified and
generally not a good processing step. Local destriping leaves removed from the image data. Such cleaning actions, based
the “smile” effect but only removes pixel-to-pixel scale stripes. on known spectra, have often been used to improve the results
DATT et al.: PREPROCESSING EO-1 HYPERSPECTRAL DATA TO SUPPORT THE APPLICATION OF AGRICULTURAL INDEXES 1253

Fig. 5. Mean reflectance spectrum from the Hyperion reflectance image

(dashed line) fitted using a linear combination of mean ASD spectra from nine
sites (solid line). ASD spectra were integrated to the Hyperion bandpasses.

of atmospheric correction. Since convenient ground spectra of

this type are rare in operational data processing, a number of
alternative, related approaches have been developed. These are
typified by the EFFORT approaches available in ENVI and
other packages [31]. In that approach, typical image spectra are
obtained and fitted by smooth polynomial or spline functions
and the residuals are assumed to be the noise. After that pass,
the procedure is much the same as if reference spectra were
available.
To provide an example of the general approach for our
study—without attempting to define the best operational
method—we fitted the overall mean spectrum of the atmo-
spherically corrected Hyperion image by a linear (mixture)
Fig. 6. Hyperion (dashed line) and HyMap (solid line) MNF comparison. (A)
combination of the means of the field spectra measured at the Data mean spectra over a common area. (B) MNF values on a Log scale.
nine sites (Fig. 5). In this way, all of the shapes and features
that we expected to find in the ASD data were preserved in the high SNR MNF components. However, this provided no basis
fitted model and a global residual noise factor could be defined. for assessing the significance of the 20 MNF bands except
The ratio of the mean to the model provides a multiplier to through visual inspection. To provide a benchmark, MNF of
clean up the spectra. This is similar to a directed EFFORT. the 126 bands of atmospherically corrected HyMap data [28]
acquired at the same time was used for the same area as a
B. MNF Smoothing coregistered Hyperion data based on the 176-band set. The 176
The uncorrelated, or at most locally spatially correlated, Hyperion bands cover a similar range to the 126 HyMap bands
noise is not reduced by a general approach like reference but with a finer spectral resolution—10 nm compared with
spectrum smoothing. The PCA or MNF transformations are 15 nm full-width half-maximum and spacing. The atmospheric
often used for this purpose, as they do not degrade the spa- correction for the two datasets used the same basic parameters
tial resolution. The MNF transformation extracts information except for the flying height and other platform-dependent
dimensions relative to an assumed noise structure in the data settings.
[24]. The components corresponding to low SNR and unstruc- Fig. 6(A) shows the total image means in the common area
tured spatial statistics can often be eliminated from the data by imaged by both Hyperion and HyMap. The greater level of
“putting the transformation back together” without them—or residual noise visible in the Hyperion mean plot is higher
an “inverse” MNF transformation. than can simply be ascribed to the narrower bandwidth. With
It is useful to have some idea of the significance of the advantages of broader bands, higher signal levels and higher
number of transformed bands retained in the data for the intrinsic SNR, the HyMap data seems to achieve 25 to 30
inverse MNF transformation. As noted above, global destriping significant MNF bands [Fig. 6(B)]. Taking 25 to 30 HyMap
using the 176-band selection seemed to provide about 20 MNFs as a base potential for resolving MNFs (i.e., the limit of
MNF bands that are free of the spatially structured striping environmental SNR in this application) it is a good achievement
noise. A similarly striping-noise free set was obtained after to remove noise to achieve 20 MNFs from Hyperion data. The
local destriping but with the low-frequency effects retained as similarity between the HyMap data in Fig. 6(A) and the model
1254 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 41, NO. 6, JUNE 2003

Fig. 7. Estimated sensitivity measure of radiances to changes in location of atmospheric features or Hyperion band centers. Continuous line is the measure and
the blocks are the 155 bands selected for “stability.”

based on fitted site spectra in Fig. 5 supports the assumption of TABLE IV

low residual noise in the HyMap data. 155 STABLE HYPERION BANDS

To apply the dimensional noise reduction in practice, it is

best to handle the VNIR and SWIR data separately and com-
bine them after the processing. In the following examples, for
the inverse MNF transformation we used 12 components in the
VNIR and eight in the SWIR. These choices allowed rejection
of some low SNR residual array effects and were more conser-
vative than the 20 MNFs from the combined set that were used
in the comparison with HyMap. Handling the VNIR and SWIR The results were averaged over the cases of varying atmospheric
data separately provides greater capacity to manage the noise levels of water vapor and aerosol and then integrated over the
due to its different structure in the two arrays. Hyperion band passes. The result obtained is partly a function
of the conditions at the site and the time of overpass up to the
C. Band Selections That Avoid Residual Atmospheric Noise variation introduced by the varying atmospheric data. The final
Effects value is plotted in Fig. 7.
The residual noise introduced by atmospheric correction can There was a level of arbitrariness about the overall level
be reduced by methods such as those described above. However, chosen to indicate instability. Initially, a value of 0.25 was
an alternative or complementary approach is to eliminate bands selected and then some bands at the edge of atmospheric
that are most likely to contain unstable atmospheric artifacts features were added where there were some important known
above a certain threshold. features to pursue beneath the image noise. Based on this index,
To quantify the choice, an index of band sensitivity to atmo- a subset of 155 bands (plotted in Fig. 7) has been used as a set
spheric effects (such as variations in absorber profile as well as of “stable” bands with respect to residual atmospheric feature
interactions with the sensor) was defined by using the derivative effects and used to develop indexes and measurements that are
of model radiance as a function of wavelength. In effect, this free of the effects they create. Table IV presents the resulting
measure is most sensitive to locations where a change in band choices by original band number and wavelength.
center will create the largest differences in radiance. Specifi- These bands are clearly located in the middle of atmospheric
cally, atmospheric models were computed for the CIAS based windows and also cover the areas of response of the ETM and
on a clear day as the base and with variations in aerosol and ALI band passes, and thus the selection retains the ability to con-
water vapor to provide the range. These were combined with struct intersensor comparisons. Obviously there can be modifi-
an average surface reflectance computed with a wavelength ( ) cations for specific purposes but in most cases we are finding
step of 1 nm and converted to radiance ( ) at the sensor. that using these 155 bands provides a simple but effective pro-
Numerical differentiation was used to compute the quantity cessing step that avoids, rather than reduces, residual noise. All
the indexes used in Section VI used selections from the stable
155 Hyperion bands. However, Fig. 7 also shows that the RE
Sensit (9)
indexes may need special attention.
DATT et al.: PREPROCESSING EO-1 HYPERSPECTRAL DATA TO SUPPORT THE APPLICATION OF AGRICULTURAL INDEXES 1255

Fig. 8. Mean January 12 Hyperion spectra for the nine sampling sites of Fig. 1. The within-site variation is used to evaluate the indexes.

VI. AGRICULTURAL APPLICATIONS: MEASUREMENTS is reported in [33]. Also, since the most obvious VNIR striping
FROM A SPACE PLATFORM occurs toward the western side of the image, there was little or
The benefits of noise management, including the destriping no visible VNIR striping near the sites. However, the ubiquitous
and smoothing operations, were evaluated by comparing the SWIR vertical striping was present everywhere in the image.
values of selected spectral indexes from the geolocated ASD For completeness, the comparisons were made on Hyperion
measurements with values from corresponding Hyperion pixels images processed in nine ways. First, atmospheric correction,
(Section VI-A). These data do not provide a complete design using FLAASH, was applied to the original and globally and
for the effects discussed (and other effects not fully discussed locally destriped (Section IV-C) radiance images. These are
in this paper such as the “smile” and surface BRDF) but do pro- denoted as 1) Undestriped, 2) Global destriped, and 3) Local
vide important insights. The mean Hyperion reflectance spectra destriped. Using these three atmospherically corrected images
corresponding to the nine ASD site means of Fig. 1 are shown in as inputs to an “MNF-inverse MNF” transformation (using
Fig. 8. These Hyperion spectra show similar reflectance ampli- 12 MNFs for the VNIR and eight for the SWIR as described
tude and spectral features to the ASD spectra of Fig. 1 but with in Section V-B) to smooth the data produced 4) Undestriped
more noise. The numbers of samples involved in these means MNF smoothing (MNFS), 5) Global destriped MNFS, and
varies from site to site and are provided in Table II. 6) Local destriped MNFS. Again the outputs (1)–(3) were used
The purpose of the analysis is twofold. First, it is important to as inputs to the reference spectrum smoothing (RSS) technique
know if the information measured using field spectrometry can (introduced in Section V-A) to create 7) Undestriped RSS,
be retrieved from space data, and second, which noise manage- 8) Global destriped RSS, and 9) Local destriped RSS.
ment strategies increase the consistency between the datasets? Following this, the sensitivity of spectral indexes to crop
The dataset can partly answer these questions, but it is impor- stress was used as a test of application by comparing Hyperion
tant to keep both its strengths and limitations in mind during the and ASD derived values for two rice bays known to have
analysis. different levels of water stress (Section VI-B) from field work
The data were obtained at a variety of sites on and during and landholder survey.
days following the EO-1 overpass on January 12, 2002 (see Sec-
tion III). To provide a consistent comparison, the ASD spectra A. Narrowband Indexes: Performance and Stability
were convolved to equivalent Hyperion bands. The sites were
chosen to represent different covers, yet were relatively close The six indexes introduced in Section II were used to investi-
together for logistical reasons. Being near the center of the Hy- gate the relationship between the ASD and Hyperion data. In the
perion image the variations due to spectral “smile” are small VNIR region, three were explored: 1) RE, 2) dRE, and 3) LCI.
and so effects due to the “smile” cannot be assessed using these In the SWIR region the clay hydroxyl absorption feature in soils
data. The “smile” is better addressed by comparing Hyperion near 2200 nm was selected. In the VNIR/SWIR region, the two
and HyMap data over the common area of the CIAS, and this leaf water content indexes WI and NDWI were used.
1256 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 41, NO. 6, JUNE 2003

TABLE V
SUMMARY STATISTICS R , F-TEST PROBABILITY (P ), MEAN (Mn), AND STANDARD DEVIATION (Std) FOR SIX INDEXES FOR HYPERION
PREPROCESSED USING NINE OPTIONS. Mn AND Std ARE ALSO GIVEN FOR ASD

Indexes from VNIR, SWIR, and combined VNIR/SWIR The sampling strategy at the CIAS for the soil and stubble
spectral regions are needed to provide tests for noise reduction sites acquired on January 12, 2002 was designed to enable
due to the different noise statistics in data acquired by Hype- scaling to Hyperion pixels. However, the rice sampling was
rion’s two spectrometers. The performance of all six indexes is, less intensive and includes more spatial variation (and hence
however, expected to be relatively stable, since the wavelengths higher standard deviation) due to the scale difference between
used in their formulation all lie within the 155 “stable” bands site and Hyperion data.
we defined for Hyperion. 2) SWIR Clay Index: CI values were calculated from 55 ge-
1) VNIR Indexes: The RE and dRE values were calculated olocated ASD spectra of soils (Soil_1, Soil_2, NN_Soil, and
from Hyperion by fitting a third-order polynomial and returning M_Soil in Table II) and from the 55 georeferenced Hyperion
the wavelength and first-derivative value at the inflection point pixels. Where ASD data fell into the same pixel the values were
using a generic IDL_ENVI module developed by the CSIRO averaged, there are 55 unique Hyperion pixels associated with
Division of Exploration and Mining, Mineral Mapping Tech- the 73 ASD soil measurement sites presented in Table II. The
nologies Group. ASD and Hyperion index values were compared using linear re-
The indexes were calculated at the 20 sites for rice (Rice_1, gression (Table V). The results in Table V show that both global
Y_Rice, and G_Rice in Table II) using geolocated ASD spectra and local destriping (cases 2 and 3) slightly increased the
and the corresponding georeferenced Hyperion pixels from values compared to the undestriped image (case 1). Since the
the reflectance images. The ASD and Hyperion derived index region used is in the “stable” area of the SWIR, Hyperion bands
values were compared using linear regression. Summary 204, 205, and 206 used in the Clay Index did not show heavy
statistics are provided in Table V. striping in the radiance data; thus, the impact of destriping was
In most cases (the dRE excepted), the Hyperion values cor- not large and would be more significant for other SWIR bands.
relate to a good level with ground-based ASD values. This is an The largest increase in was observed with MNF
encouraging result, since one objective is to use ground-based smoothing, as shown for cases 4, 5, and 6 in Table V. This result
models or calibrations with the Hyperion data. The dRE is the indicates that base noise has a major impact in the SWIR region
index most affected by the Hyperion “smile” and this is taken compared with stripes and that MNF smoothing reduces this.
up separately in [33]. In this case, as discussed in Section V-B, the MNF smoothing
The local destriping, MNFS and RSS smoothing all produced in which the VNIR and SWIR are handled separately is also
similar values to that of the undestriped image. This is to be removing residual array effects. The RSS smoothing (cases 7,
expected, as the VNIR SNR is high, and there was no major 8, and 9 in Table V) did not show any change in for this
VNIR vertical striping in the central region of the Hyperion index as may be expected from Figs. 5 and 6 and from the way
image. However, it is significant that the lowest values for the index is defined. Note that in the best results the means
each index were obtained for the global destriped images (cases and standard deviations of image and ASD data are similar as
2, 5, and 8 in Table V). This is consistent with the findings in would be expected due to the design of the sampling at the soil
Section VI-A, that global destriping alters the data. sites that aimed to bring the data to the same spatial scale.
It is important to keep in mind the number of samples 3) VNIR–SWIR Indexes: The values of WI and NDWI were
involved in these comparisons (20 in this case) as well as calculated from 20 geolocated ASD spectra of rice (Rice_1,
the relative standard deviations of the image and ASD data. Y_Rice, and G_Rice in Table II) and from the 20 georeferenced
DATT et al.: PREPROCESSING EO-1 HYPERSPECTRAL DATA TO SUPPORT THE APPLICATION OF AGRICULTURAL INDEXES 1257

TABLE VI indexes in the Hyperion data. The lower bay with higher chloro-
MEANS AND STANDARD DEVIATIONS OF RE, LCI, dRE, WI, AND NDWI FOR phyll content shows significantly higher RE and LCI values
HYPERION AND ASD MEASUREMENTS IN THE UPPER (WATER STRESSED),
AND LOWER (NON-WATER STRESSED) BAYS compared with the upper bay. Chlorophyll measurements were
taken with a Minolta SPAD-502 meter (produced by the Soil-
Plant Analysis Development (SPAD) Division, Minolta Camera
Company, Osaka, Japan). On January 14, the lower bay had an
average of 43.89 (SD for four sample locations), and the
upper bay had an average of 17.00 (SD for four sample
locations). At each of the eight locations (four in the upper bay
and four in the lower bay) 30 individual “SPAD” measurements
were made. The SPAD data are in relative units that are known
to be highly correlated with leaf nitrogen concentration.
For RE and LCI indexes, Hyperion and ASD show similar
values and trends between the bays. There was close agreement
Hyperion pixels. The ASD- and Hyperion-derived index values between ASD and Hyperion values in the lower bay, which had
were then compared using linear regression (see Table V for sta- a uniform green rice canopy. For the upper bay, Hyperion values
tistics from WI and NDWI indexes, respectively). were higher than the ASD values, mainly due to the heteroge-
As can be seen from Table V, the values were nearly zero neous mix of green rice and yellow senescing weeds and the
for most cases. The reason for the low values is almost certainly greater spatial averaging in the Hyperion signal. The dRE values
the low range of variation in the WI and NDWI values for the 20 were higher for the lower bay in both ASD and Hyperion mea-
ASD rice samples. This is shown by the image and site standard surements. This relates to the higher green leaf cover and den-
deviations. The influence of atmospheric water vapor in the 973- sity in the lower bay compared to the upper bay which had more
and 1245-nm Hyperion bands was expected to create variation senescing weeds in it and therefore lower dRE values. However,
but does not seem to have done so. Relatively few conclusions as in the previous examples, the dRE in the space data has a
on the effects of various processing options can be drawn for reduced variation in response than in the ASD values—despite
these two indexes except to note that the absolute values ob- the fact that the ASD data are integrated to Hyperion bands. The
tained from the ASD spectra and the image are very close. To NDWI and WI values in the lower bay, for both the Hyperion
assess whether the variations are as well related as the means and ASD data, were higher than the values for the upper bay
requires a dataset with more underlying variation. indicating consistently greater leaf water content in the greener
canopy.
B. Mapping the Onset of Stress Due to Water Restrictions On the basis of these results, it appears that one or a number
The images processed in the best of the selections discussed of these indexes taken together could provide useful maps of
above were then used in a simple but very useful mapping ap- the state of water stress at the time of the EO-1 overpass. If the
plication. Due to reduced water allocations and high water de- stress were to become more severe, it may be expected that the
mands in the early growing season, some growers reduced or leaf water indexes would become more informative and if the
stopped irrigation much earlier than normal. Irrigation to one of stress progressed to senescence it may be expected that indexes
the rice fields (Farm 51, field H) was stopped ten weeks earlier not used here but based on the leaf waxes and oils and cellu-
on December 28, 2001. For this “dewatered” field, water was lose features (see Table I) would track the progression into that
managed by directing water from three bays at the same eleva- stage. These tools are, therefore, ready for analyzing hyperspec-
tion to the fourth bay at a lower elevation. The landholder at- tral time series as acquired for CIAS.
tempted to grow a successful crop in the lower bay by directing
water from the other bays.
VII. CONCLUSION
Due to this management strategy, the lower bay was less
stressed than the upper bays on January 12, 2002. Geolo- The consistency and degree of relationship among the in-
cated ASD spectra were acquired in the lower (greener, labeled dexes derived from the space platform and those from field data
G_Rice) bay and in one of the upper (yellower, labeled Y_Rice) provide the entry point to intercalibration between laboratory,
bays on January 14, 2002. Since water stress is associated ground-based, airborne, and spaceborne data. However, the re-
with a decline in chlorophyll, it would seem possible to use sults from this research indicate that preprocessing including
hyperspectral indexes to map the variation. The question is how band, or feature, selection, atmospheric correction, local de-
well can such mapping be undertaken at a regional scale from striping, and noise management such as that provided by min-
the EO-1 space platform? Mean and standard deviations of four imum noise fraction smoothing were needed to achieve this.
representative pixels at which geolocated ASD (see G_Rice Destriping is an area of current focus and development for
and Y_Rice in Table II) and georeferenced Hyperion derived Hyperion data processing. The work reported here supports
values were extracted for the suite of indexes are presented in the use of local methods to remove the stripes with the degree
Table VI for the upper bay (water stressed) and the lower bay of locality (or window size) dependent on the array (VNIR or
(not water stressed). SWIR), the land cover, the environmental conditions and the
Despite the restricted statistics, the differences in chlorophyll application. The consistency of the size of window used in the
content between the two bays are captured by the RE and LCI local destriping algorithm needs to be studied further. It seems
1258 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 41, NO. 6, JUNE 2003

likely that the window used for data acquired by the VNIR REFERENCES
array will normally be much smaller than that used for data
acquired by the SWIR array. This is a function of the different [1] J. Pearlman, S. Carman, C. Segal, P. Jarecke, and P. Barry, “Overview
of the Hyperion imaging spectrometer for the NASA EO-1 mission,” in
noise patterns in data resulting from the two arrays. Proc. IGARSS, Sydney, Australia, 2001.
It seems that the current standard and advanced atmospheric [2] S. G. Ungar, “Overview of EO-1, the first 120 days,” in Proc. IGARSS,
correction methods provide useful normalizations of the data Sydney, Australia, 2001.
[3] M. S. Moran, Y. Inoue, and E. M. Barnes, “Opportunities and limitations
and can certainly be recommended for consistent analysis of for image-based remote sensing in precision crop management,” Remote
time series such as that of the CIAS. However, it is clear from the Sens. Environ., vol. 61, pp. 319–346, 1997.
work reported here that there will normally be significant (and [4] P. L. Nagler, C. S. T. Daughtry, and S. N. Goward, “Plant litter and soil
reflectance,” Remote Sens. Environ., vol. 71, pp. 207–215, 2000.
generally increased) levels of residual noise in the data after any [5] E. Boeg, H. Soegaard, N. Broge, C. B. Hasager, N. O. Jensen, K.
atmospheric correction with consequent reduction in SNR if it Schelde, and A. Thomsen, “Airborne multispectral data for quantifying
is based on current Level 1B1 data. For this reason, extra data leaf area index, nitrogen concentration, and photosynthetic efficiency
in agriculture,” Remote Sens. Environ., vol. 81, pp. 179–193, 2002.
cleaning and feature selection was performed and will need to be [6] I. B. Strachan, E. Pattey, and J. B. Biosvert, “Impact of nitrogen and envi-
performed by all users of the data. It also appears that most of the ronmental conditions on corn as detected by hyperspectral reflectance,”
options for noise management are best applied after atmospheric Remote Sens. Environ., vol. 80, pp. 213–224, 2002.
[7] M. E. Martin and J. D. Aber, “High spectral resolution remote sensing
correction. Local destriping is an exception. of forest canopy lignin, nitrogen, and ecosystem processes,” Ecol. Appl.,
This paper has not specifically addressed issues that are cur- vol. 7, pp. 431–443, 1997.
rently at the forefront of discussions surrounding atmospheric [8] D. N. H. Horler, M. Dockray, and J. Barber, “The red edge of plant leaf
reflectance,” Int. J. Remote Sens., vol. 4, pp. 273–288, 1983.
correction of Hyperion. These include issues such as spectral [9] P. J. Curran, J. L. Dungan, B. A. Macler, and S. E. Plummer, “The
“smile” effects [33] and adequacy of transmittance codes. We effect of a red leaf pigment on the relationship between red edge
have rather selected a very standard baseline for correction. It and chlorophyll concentration,” Remote Sens. Environ., vol. 35, pp.
69–76, 1991.
is quite likely that there will be added levels of noise reduc- [10] A. A. Gitelson, M. N. Merzlyak, and H. K. Lichtenthaler, “Detection of
tion from advances in this area in the near future and that these red edge position and chlorophyll content by reflectance measurements
will include more sophisticated handling of the spectral “smile.” near 700 nm,” J. Plant Physiol., vol. 148, pp. 501–508, 1996.
However, it seems from the investigation in Section VI-A that [11] I. Filella and J. Peñuelas, “The red edge position and shape as indicators
of plant chlorophyll content, biomass, and hydric status,” Int. J. Remote
the indexes being used in this paper (with the possible excep- Sens., vol. 15, pp. 1459–1470, 1994.
tion of the dRE) can be used successfully without such further [12] Z. Chen, C. D. Elvidge, and W. T. Jansen, “Description of derivative-
advances. based high spectral-resolution (AVIRIS) green vegetation index,” Proc.
SPIE, vol. 1937, pp. 43–54, 1993.
In summary, the best noise management strategy found here [13] C. D. Elvidge and Z. Chen, “Comparison of broad-band and
was based on calibrated, locally destriped and atmospherically narrow-band red and near-infrared vegetation indices,” Remote Sens.
corrected data. Following this, MNF smoothing was used based Environ., vol. 54, pp. 38–48, 1995.
[14] B. Datt, “A new reflectance index for remote sensing of chlorophyll con-
on separate MNF transformations of the stable bands of the tent in higher plants: Tests using Eucalyptus leaves,” J. Plant Physiol.,
VNIR and SWIR data followed by inverse transformations re- vol. 154, pp. 30–36, 1999.
taining only the high MNFs with significant land surface in- [15] J. Peñuelas, J. Piñol, R. Ogaya, and I. Filella, “Estimation of plant water
concentration by the reflectance water index WI (R900/R970),” Int. J.
formation. The steps we have described can provide an effec- Remote Sens., vol. 18, pp. 2869–2875, 1997.
tive processing pathway to generate indexes from apparent sur- [16] B. C. Gao, “NDWI—A normalized difference water index for remote
face reflectance images for the time series of EO-1 data that has sensing of vegetation liquid water from space,” Remote Sens. Environ.,
vol. 58, pp. 257–266, 1996.
been acquired for the CIAS. Because the data can be closely re- [17] G. P. Asner and D. B. Lobell, “A biogeophysical approach for automated
lated to ground-based hyperspectral data, the standardization al- SWIR unmixing of soils and vegetation,” Remote Sens. Environ., vol. 74,
lows monitoring and measurement of crop growth from a space pp. 99–112, 2000.
[18] R. N. Clark and T. L. Roush, “Reflectance spectroscopy: Quantitative
platform based on field-scale calibration. The aim is to assess analysis techniques for remote sensing applications,” J. Geophys. Res.,
changes in plant chemistry, including changes in nitrogen levels vol. 89, pp. 6329–6340, 1984.
such as those developed for forests in [32], and to model crop [19] R. N. Clark, T. V. V. King, M. Klejwa, G. Swayze, and N. Vergo,
yields. “High spectral resolution reflectance spectroscopy of minerals,” J.
Geophys.Res., vol. 95, pp. 12 653–12 680, 1990.
[20] R. N. Clark, G. A. Swayze, A. Gallagher, N. Gorelick, and F. Kruse,
“Mapping with imaging spectrometer data using the complete band
ACKNOWLEDGMENT shape least-squares algorithm simultaneously fit to multiple spectral
features from multiple materials,” in Proc. 3rd Airborne Visible/Infrared
The Hyperion data were processed by TRW and provided by Imaging Spectrometer (AVIRIS) Workshop, 1991, JPL Pub. 91–28, pp.
NASA. B. Holben (AERONET Group) provided the CIMEL 2–3.
installation. Thanks to M. Bramston, A. Tiwari, R. Robinson, [21] B. N. Holben, T. F. Eck, I. Slutsker, D. Tanre, J. P. Buis, A. Setzer,
E. Vermote, J. A. Reagan, Y. J. Kaufman, T. Nakajima, F. Lavenu, I.
and G. Robertson (Coleambally Irrigation Cooperative Lim- Jankowiak, and A. Smirnov, “AERONET—A federated instrument net-
ited) and the landholders for their cooperation and support, G. work and data archive for aerosol characterization,” Remote Sens. Env-
Beecher and B. Dunn (NSW Department of Agriculture ) for iron., vol. 66, pp. 1–16, 1998.
[22] R. M. Mitchell and B. W. Forgan, “Aerosol measurements in the Aus-
their real-time differential GPS unit, J. Angus (CSIRO) for the tralian outback: Intercomparison of sun photometers,” J. Atmos. Oceanic
SPAD meter, D. Graetz (CSIRO) for field equipment. Particular Technol., vol. 20, pp. 54–66, 2003.
acknowledgment is due to J. Lovell, S. Campbell, and E. King [23] T. R. McVicar, T. G. Van Niel, and D. L. B. Jupp, “Geometric validation
for computing and field support and the Mineral Mapping and of Hyperion data acquired by Earth Observing 1 satellite at Coleambally
Irrigation Area,” Commonwealth Scientific and Industrial Research
Technology Group of CSIRO for access to their MMTG-A set Organization, Canberra, Australia, CSIRO Land and Water Tech. Rep.
of software. 46/01, 2001.
DATT et al.: PREPROCESSING EO-1 HYPERSPECTRAL DATA TO SUPPORT THE APPLICATION OF AGRICULTURAL INDEXES 1259

[24] A. A. Green, M. Berman, P. Switzer, and M. D. Craig, “A transformation Tim R. McVicar received the B.S. (honors I) from the University of Queens-
for ordering multispectral data in terms of image quality with implica- land, Brisbane, Australia, and the Ph.D. degree from the Australian National
tions for noise removal,” IEEE Trans. Geosci. Remote Sensing, vol. 26, University, Canberra, Australia, in 1988 and 2001, respectively.
pp. 65–74, Jan. 1988. In 1989, he joined the Division of Water Resources, Commonwealth Scien-
[25] A. F. H. Goetz, Z. Qu, B. Kindel, and M. Ferri, “Atmospheric correction tific and Industrial Research Organization (CSIRO) Land and Water, Canberra,
of Hyperion data and techniques for dynamic scene correction,” in Proc. Australia. He is currently a Senior Research Scientist in the Integrated Catch-
IGARSS, vol. III, Toronto, ON, Canada, 2002, pp. 1408–1410. ment Management Research Directorate with CSIRO Land and Water, and is a
[26] T. W. Cooley, G. Anderson, G. W. Felde, M. L. Hoke, and A. J. Co-Investigator for the Australian team of the National Aeronautics and Space
Ratkowski, “FLAASH, A MODTRAN-4 based atmospheric correction Administration’s EO-1 Science Validation Team. His research interests include
algorithm, its application and validation,” in Proc. IGARSS, vol. III, time series remote sensing for ecohydrological and agricultural applications.
Toronto, ON, Canada, 2002, pp. 1414–1418.
[27] P. Barry, EO-1/Hyperion Science Data User’s Guide. Redondo Beach,
CA: TRW Space, Defense & Inform. Syst., 2001.
[28] T. Cocks, R. Jenssen, A. Stewart, I. Wilson, and T. Shields, “The Tom G. Van Niel received the A.A.S. degree from Monroe Community Col-
HyMap airborne hyperspectral sensor: The system, calibration and per- lege, Rochester, NY, and the B.S. and M.S. degrees from Utah State University,
formance,” in Proc. 1st EARSEL Workshop on Imaging Spectroscopy, Logan.
Zurich, Switzerland, Oct. 1998, pp. 37–42. He is currently an Experimental Scientist with the Integrated Catchment Man-
[29] B. K. P. Horn and R. J. Woodham, “Destriping landsat MSS images by agement Research Directorate, Commonwealth Scientific and Industrial Re-
histograms modification,” Comput. Graphics Image Process., vol. 10, p. search Organization Land and Water, Canberra, Australia and is involved in the
69, 1979. Measurement and Mapping Program for the Cooperative Research Centre for
[30] K. G. Joreskog, J. E. Klovan, and R. A. Reyment, Geological Factor Sustainable Rice Production, Wagga Wagga, Australia. His research interests
Analysis. Amsterdam, The Netherlands: Elsevier, 1976, p. 178. are in spatio-temporal analysis as well as error and uncertainty of spatial data.
[31] J. W. Boardman, “Post-ATREM polishing of AVIRIS apparent re-
flectance data using EFFORT: A Lesson in accuracy versus precision,”
in Summaries of the 7th JPL Airborne Earth Science Workshop, vol. 1, David L. B. Jupp (A’90) received the B.S. degree from Adelaide University,
1998, JPL Pub. 97–21, p. 53. Adelaide, Australia, and the Ph.D. degree from Flinders University, Adelaide,
[32] N. C. Coops, M. L. Smith, M. E. Martin, S. V. Ollinger, and A. A. Held, Australia, in 1969 and 1973, respectively.
“Predicting Eucalypt biochemistry from HYPERION and HYMAP im- He is currently with the Commonwealth Scientific and Industrial Research
agery,” in Proc. IGARSS, Toronto, ON, Canada, 2002. Organization Earth Observation Centre, Canberra, Australia. His areas of in-
[33] D. L. B. Jupp, B. Datt, T. R. McVicar, T. G. Van Niel, J. S. Pearlman, J. terest have included scene and image modeling, BRDF modeling, hyperspec-
Lovell, and E. G. King, “Improving the analysis of Hyperion red edge tral data analysis, applications, technology transfer, and commercialization of
index from an agricultural area,” in Proc. SPIE Conf. Remote Sensing remote sensing. Applications include coastal zone and inland waters, water re-
Asia 2002, Hangzhou, China, Oct. 2002. sources applications, land surface temperature for applications to soil moisture,
[34] A. F. H. Goetz, G. Vane, J. Solomon, and B. N. Rock, “Imaging spec- and water balance estimation. International collaboration has included work
trometry for Earth remote sensing,” Science, vol. 228, pp. 1147–1153, with groups in the United States, Southeast Asia, and China. He is Principal In-
1985. vestigator for the Australian membership of the National Aeronautics and Space
[35] L. Ong, private communication, 2002. Administration EO-1 Science Validation Team.

Jay S. Pearlman (M’84–SM’03) received the

B.S. degree from the the California Institute of
Technology, Pasadena, and the Ph.D. degree from
University of Washington, Seattle.
Bisun Datt received the B.S. degree in physics and mathematics, the PG Dipl. in He is currently the Chief of Science and Applica-
physics, and the M.S. degree in physics from the University of the South Pacific, tions for the Advanced Network Centric Operations,
Suva, Fiji, in 1985, 1992, and 1996, respectively. He received the Ph.D. degree Phantom Works, Boeing Company, Seattle, WA, and
in remote sensing from the University of New South Wales, Sydney, Australia, is actively involved in remote sensing development
in 1999. and applications. Previously, he was with TRW, Re-
He is currently with the Commonwealth Scientific and Industrial Research dondo Beach, CA, and was Deputy Program Manager
Organization Earth Observation Centre, Canberra, Australia. His main research and Scientist for the Hyperion instrument on EO-1. In
interests are in hyperspectral remote sensing of vegetation and soils with spe- addition to instrument development, his activities include applications of remote
cific applications in agricultural crops and forest biochemistry. He is a Co-In- sensing to agriculture and forestry using both airborne and space-based systems.
vestigator for the EO-1 Hyperion calibration/validation activities in Australia, Dr. Pearlman is an Associate Editor of the IEEE TRANSACTIONS ON
through Australian membership of NASA’s EO-1 Science Validation Team. GEOSCIENCE AND REMOTE SENSING.