3 - RESULTS AND DISCUSSION 
Price (1990 and Baumgardner (1985) showed that for many soils, the associated reflectance spectra can be 
described with very few parameters. In the same way. Jacquemoud et al. (1993) showed that the bidirectionnal 
properties can also be described with a small set of parameters. However, the very specific soils available in this 
study prevent us to use those results easily. In consequence, we will simply consider the soil background 
reflectance spectra as known in the inversion process. Although not realistic when dealing with applications 
where most of the time soil type can be known, but the surface characteristics such as roughness or moisture 
may vary spatially and with time, this assumption was largely used in many inversion studies (Andrieu and 
Baret. 1993). 
We will first attempt to retrieve all the 6 biophysical characteristics of the canopy: [LAI, 0,, s, 
C ab , C w „ N], from the reflectance spectra observed in the 188 AV1RIS bands. Then, we will compare the 
performances of this inversion to what will be retrieved in the same conditions but using only the 6 TM bands. 
< 
D 
Q 
W 
U 1 
cr 
Figure 2. Comparison between the measured and the spectra simulated in the 188 AVIR1S bands with the 
PROSPECT+SAIL model using the estimated values of input variables retrieved from model inversion. For 
each of the 3 spectra, the solid and broken lines represent respectively the biases (£(p -p)/n, where p and p 
are respectively the measured and estimated reflectances) and the standard deviation (y/fvjp-p) 2 ) /n where n is 
the number of observations) of reflectances. From the top to the bottom, the spectra correspond to: inversion 
from the 188 AVIRIS wavebands to retrieve [LAI, 6 b s, C ab , N]; inversion from the 188 AV1RIS 
wavebands to retrieve [LAI, C ah , C w [, with [6,= 28.6°, . 9 = 0 . 33 , N=1.23]; inversion from the 6 TM bands to 
retrieve [LAI, 6 b s, C ab , C„ N], 
3.1. Using high spectral resolution information. 
3.1.1. Retrieval of the 6 canopy biophysical variables: [LAI, 6 p s, C^, NJ. In this first 
inversion process of the PROSPECT+SAIL model, we will estimate the following 6 canopy biophysical 
variables: leaf area index LAI, mean leaf inclination angle 0] hot spot parameter s, chlorophyll concentration 
Catr equivalent water thickness C^ and leaf mesophyll structure N. The measurement configuration, irradiance 
conditions and, above all, the soil optical properties were assigned to their nominal values. To avoid physical 
inconsistencies, the variables to be estimated were constrained to vary within the following ranges that 
correspond to the minimum and maximum values observed for most canopies: 0.1<L4/<10, 5°<0|<85°, (Ks<l, 
l<C aft <100 pg cm -2 , 0.001<C W <0.1 cm and 1<N<2.5, and. For all inversion, the initial guess was [LAI=3, 0 
1 = 45 °, s=0.5, C ab = 32pg cm -2 , C w =0.0255 cm , N=1.5]. For 8 plots over the % observed, the inversion process 
did not converged. In the following, we will only present results corresponding to a successful inversion 
process. 
The measured spectra are very close to the spectra simulated using the PROSPECT+SAIL 
model with the estimated values of the 6 canopy biophysical variables (figure 2a). The root mean square error 
(rmse) between actual and simulated reflectance values evaluated over all the plots and all the wavelengths is 
185