MODELING THE BIDIRECTIONAL REFLECTANCE DISTRIBUTION FUNCTION OF MIXEDFINITE PLANT CANOPIES AND SOIL

An analytical model of the bidirectional reflectance for optically sem i-infinite plant canopies has been extended to describe the reflectanc e of finite depth canopies with contributions from the underlying soil . The model depends on 10 independent parameters describing vegetation and soil optical and structural properties. The model is inverted wit h a nonlinear minimization routine using directional reflectance data for lawn (leaf area index (LAI) is equal to 9.9), soybeans (LAI, 2.9) and simulated reflectance data (LAI, 1.0) from a numerical bidirection al reflectance distribution function (BRDF) model (Myneni et al., 1988 ). While the ten-parameter model results in relatively low rms differe nces for the BRDF, most of the retrieved parameters exhibit poor stabi lity. The most stable parameter was the single-scattering albedo of th e vegetation. Canopy albedo could be derived with an accuracy of less than 5% relative error in the visible and less than 1% in the near-inf rared. Sensitivity tests were performed to determine which of the 10 p arameters were most important and to assess the effects of Gaussian no ise on the parameter retrievals. Out of the 10 parameters, three were identified which described most of the BRDF variability. At low LAI va lues the most influential parameters were the single-scattering albedo s (both soil and vegetation) and LAI, while at higher LAI values (> 2. 5) these shifted to the two scattering phase function parameters for v egetation and the single-scattering albedo of the vegetation. The thre e-parameter model, formed by fixing the seven least significant parame ters, gave higher rms values but was less sensitive to noise in the BR DF than the full ten-parameter model. A full hemispherical reflectance data set for lawn was then interpolated to yield BRDF values correspo nding to advanced very high resolution radiometer (AVHRR) scan geometr ies collected over a period of nine days. The resulting retrieved para meters and BRDFs are similar to those for the full sampling geometry, suggesting that the limited geometry of AVHRR measurements might be us ed to reliably retrieve BRDF and canopy albedo with this model.