DEL
ARTS OF
along with IRS -
coefficients. The
nodel parameters
sses are extracted
asthan along with
the three MODIS
ngle independent
the BRDF model
; have been made
r of parameters,
face BRDF from
-empirical BRDF
-angular data sets
rical models, the
| of trigonometric
which are derived
some physical
of semi-empirical
000).
ng Spectrometer)
ion System)-AM
nds including red
patial resolutions,
JDIS is 2300 km
it is capable of
ar reflectance of a
üned on separate
the EOS-MODIS
s including LAI,
) the user, using a
d methodology of
rithm Theoretical
sular information
egistered MODIS
sed along with a
ree BRDF model
k and Li-Sparse
ngth regions, at a
BD). These three
ne and geometric
IAPRS & SIS, Vol.34, Part 7, *Resource and Environmental Monitoring", Hyderabad, India,2002
components of radiation reflected from the target respectively,
and have physical interpretation in terms of underlying surface
parameters such as LAI (Leaf Area Index), leaf reflectance, soil
reflectance etc. The isotropic (angle independent) term
represents surface reflectance with sun at zenith and nadir view,
while the volume and geometric terms represent the anisotropic
contributions to surface reflectance. While semi-empirical
BRDF models have been used quite widely to normalize multi-
angular remotely sensed data of wide swath sensors to a
common geometry (Leroy and Roujean 1993), there have been:
relatively few attempts to explore the inherent information
content of the model coefficients (Roujean et al, 1997).
In this paper, an attempt has been made to analyze empirically
the relation between MODIS BRDF model coefficients in the
red and near infrared wavelengths with different land cover
classes as well as LAI field measurements carried out at the
Central State Farm, Suratgarh, Rajasthan, India.
2. STUDY AREA AND DATA USED
2. Study Area
The study area consists of parts of Western India, viz. Gujarat
and Rajasthan. The area is mostly semi-arid with an average
annual rainfall of 20 to 25 cms. During the month of December,
the period under study, wheat was in the crown root initiation
stage while mustard and gram crops were at flowering stage.
2.2 Data Used
A 10 deg by 10 deg tile MODIS BRDF product (MOD43BI
version 003) covering the study area (-1200 *1200 km) was
used along with the corresponding IRS WiFS and LISS-III data.
The geographical bounds of the study area are 20 deg N — 30
deg N and 69 deg E — 80.deg E. The details are given in
Tablel.
Table 1: Satellite sensors used in the study
S.No | Satellite/ Date of Wave- Spatial
sensor acquisition length resolution
(Path/ (nm) (m)
Row)
1 EOS/MO Dec 01- 620-670, 925.6
DIS 015, 01 841-876
BRDF
product
2 IRS- 14 Dec 01 | 620-680, 180.0
1D/WiFS 770-860
(P94-R53)
3 IRS- 11 Dec 01 | 620-680, 239
1D/LISS- 770-860
III(P93-
R50)
3. METHODOLOGY
3.1 Extraction of BRDF Model Coefficients
As mentioned above, the MODIS BRDF product (MOD 43B1)
available in HDF-EOS format in ISIN (Integerised Sinusoidal)
Grid consists of three BRDF model coefficients, viz., the iso-
tropic (fiso), Ross-Thick (fvol) and Li-Sparse (fgeo) coeffi-
cients representing the angle-independent, volume and geomet-
ric contributions to surface reflected reflectance. These
coefficients are given for each of seven spectral bands and three
broad bands and shape parameters representing surface anisot-
ropy in the red and NIR wavelengths are also available in the
product. The BRDF model coefficients as well as the shape pa-
rameters corresponding to red and NIR bands were extracted
from the MODIS BRDF product and converted into GeoTiff
format in geographic latitude-longitude coordinates using the
Modis Reprojection Tool (MRT).
3.2 Comparison with IRS-1D WiFS data
The IRS-1D WiFS data of spatial resolution 188 m was
resampled to ~ lkm resolution using relevant modules of
ERDAS-IMAGINE software operating on SGI workstation.
The degraded WiFS image was registered to MODIS image
using a number of GCP's distributed throughout the image. The
maximium error in the scan and pixel directions was less than a
pixel in scan and pixel directions. Various land cover classes
such as sand, crops (mustard), shallow water and deep water
were identified on the WiFS image and corresponding three
BRDF model coefficients along with three shape parameters
were noted from the MODIS BRDF product for further
analysis.
3.3 Field measurements of LAI
Field measurements of LAI were carried out during December
14-17, 2001 at the Central State Farm, Suratgarh using the
LICOR-2000 Plant Canopy Analyser. The Central State Farm
spans an area of ~ 12, 000 ha with an average field size of 250
* 250 m. The dominant crop in the rabi season is gram followed
by mustard and wheat crops. For each field, measurements
were carried out at four locations, and for each location six A
and B readings of LICOR-2000 were taken to obtain
representative LAI values for that location. The field LAI was
computed as an average of LAI values of the four locations.
Measurements were carried out for seven fields including
mustard, gram and wheat crop. The measured LAI values
ranged from 1.152 to 3.235 with combined standard deviation
range of 0.198 to 0.90.
3.4 Comparison with IRS-1D LISS-III data
A sub scene (540 lines * 540 pixels) containing the Central
State Farm, Suratgarh, Rajasthan farm was extracted from the
IRS 1D LISS-III data and was registered to a georeferenced
LISS-IH scene of the same area with sub-pixel accuracy
(RMSE=0.19 and 0.26 in the pixel and scan directions
respectively). Using the gain and offset of the IRS LISS-III
sensor along with extraterrestrial irradiance in red & NIR
wavelengths and solar zenith angle, an apparent reflectance
image was generated in the red & NIR wavelengths.
Atmospheric path radiance effects were taken into account
approximately by subtracting deep-water body reflectance at
both wavelengths (~ 6.6 % in red and 5.2 % in NIR) and
subsequently the NDVI image was generated. The fields where
LAI measurements were carried out were delineated in the
image, a linear regression relation was developed between
NDVI, and field measured LAI, which is given below:
LAI = 3.0486 * NDVI — 0.8256 (N=6, R*=0.81) (1)