BI-DIRECTIONAL REFLECTANCE MODELING AND IT’S INVERSION TO RETRIEVE WHEAT BIOPHYSICAL PARAMETERS
Loading...
Date
2011
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
IARI, DIVISION OF AGRICULTURAL PHYSICS
Abstract
Accurate and repetitive estimation of vegetation biophysical parameters at regional
scales are required for a large number of ecological, meteorological and agricultural
applications. Anisotropic reflectance behaviour in remote sensing observations
provides more information than nadir reflectance for distinguishing features and
deriving their biophysical information at regional scales. Canopy radiative transfer
models are based on explicitly defined relation between biophysical variables and
canopy anisotropic reflectance which can be inverted to derive canopy parameters.
This study describes (a) the measurements and analysis of bi- directional reflectance
anisotropy of wheat, (b) rigorous validation of canopy radiative transfer model
PROSAIL5B, and (c) retrieval of wheat biophysical variables of leaf chlorophyll
(Cab), LAI, canopy chlorophyll (CCC), and leaf wetness (Cw) by inversion of
PROSAIL using different inversion approaches. The study reconfirms the strong and
consistent anisotropic patterns of wheat reflectance in VIS and NIR regions in
response to change in sun-target-sensor geometry and the magnitude was highest in
the principal plane. The study found that this anisotropic pattern extends equally in
SWIR wavelength region also. The PROSAIL model simulated spectra was in good
agreement with the observed spectra for all the view zenith and azimuth angle
combinations used in the experiment. The model simulations also showed very good
match in the principal plane, the region of highest anisotropy, except in the VIS band
where little underestimation was found in the back scattering direction at higher view
zenith angles.The model performed best in the NIR region followed by SWIR and
maximum relative error was observed for VIS region. Over the whole optical region,
model simulations showed an average error of 27 percent and this average error was
higher (~33%) in nadir view position than in off-nadir view position. The inversion
approaches implemented were: a look up table with best solution (LUT-I), a look up
table with best 10% solutions (LUT-II) and an artificial neural network (ANN)
approach. All the approaches could estimate the biophysical variables by capturing
variability in their observed values, though accuracy of estimation varied among
three parameters. Approach LUT-II outperformed other two approaches indicating
that a set of best 10% solutions is a better strategy while ANN was worst performer.
In most of the cases, the parameters were underestimated signifying the limitation in
PROSAIL model to accurately simulate the full range of crop reflectance. In all the
inversion approaches the general order of estimation accuracy was LAI > Cab > CCC
> Cw. The range of Cw was very narrow and none of the approaches could estimate
variations in it. Performance of inversion was comparable for IRS LISS-III and
LandSat ETM+ broadband reflectances in optical region. The findings of this study
may help in generating operational crop biophysical products using IRS LISS-III
sensor by space agencies across the world.
Description
T-8411