662 
Figure 3 - The polarized reflectance as measured at 650 nm by POLDER over three different Fields 
presenting the same geometry. The measurements are plotted versus the view angle which is affected of a 
negative values for observations in backscattering. 
o 
o 
o 
o 
0 
o 
c 
© 
o 
® 
0 
■o 
© 
o 
Q. 
150 
100 - 
50- 
0 — 
-60 
o ■+■ 
o* 
°x 
sunflower; 
* vine : 
o wheat ■ 
+ o o 
40 -20 0 20 40 60 
view angle 
3.2 Atmospheric corrections 
We need now to devellop atmospheric correction for the polarized radiance in order to 
decontaminate the POLDER data and to extract the surface signature. The formulation of the polarized 
radiance is conducted in the same way than in the 5S code ( Tanre et al,1989) trying to indentify all the 
contribution to the signal These corrections are based on the following decomposition of the signal: 
( 1 ) 
where the measurement p*p corresponds to the atmospheric term p a p and to the ground contribution pSp 
suitably attenuated on the direct to direct path (|i s and |0. v are the cosine of the SZA and VZA). The 
Rayleigh scattering , with an optical thickness 8 r , is quite isotropic: the diffuse ligth reflected by the 
ground is then unpolarized and on the upward path, the molecular scattering also depolarizes the ground 
contribution. For the aerosols, with an optical thickness ô a , the process is quite identical except for the 
large forward peak which corresponds to about 30 percent of the scattering. The corresponding photons are 
non polarized and are re-injected in the direct beam which explain the factor 0.7. Figure 3 justifies the 
equation (1). 
The reverse equation of equation (1) to derive the ground contribution is quite obvious. A slight 
modification has to be introduced because of the altitude of the aircraft (3000 m). On the surface-to-sensor 
path, all the aerosol optical depth is accounted for assuming their tropospheric nature while the Rayleigh 
contribution is exactly introduced. The validation of the atmospheric correction is illustrated on figure 4 
where ground-based measurements over the sunflower field are compared to corrected POLDER data. 
3.3 An agricultural inventory based on the polarized reflectance 
The NDVI mainly provides informations on the LAI while the distribution of the polarized 
radiance is more subject to the canopy structure. If the NDVI does not depend strongly upon the view 
angle, the polarized radiance is very sensitive to this parameter. We then need to normalize the 
measurements to a same geometry. We did that assuming that the polarized reflectance is described 
following Herman and Rondeaux by: 
Ppol - 
( 2 ) 
where R(9) is the Fresnel coefficient in polarized reflectance (in other words, the polarization ratio for a 
natural incident beam equal to one), computed for a standard value of m = 1.50 for the vax of the leave; 0 
the scattering angle. F(0), the function of structure of the canopy, is the unknow parameter. We corrected 
the measurements from the geometrical parameters (p s and p v ) and from R(9) and extrapolated F(0) for 
the specular direction to re-compute using equation (2) a normalized value of the polarized radiance for 
this direction.