38 
42 Sensitivity study 
We have attempted to assess the impact of the variability of atmospheric parameters on surface reflectance 
retrieval and NDVI computation. The sensitivity study investigates the most variable parameters (aerosol 
optical thickness and water vapor content) and also the impact of variable ground elevation. The variability of 
each parameter is tested independantly of the others. Results are compared to a reference case (Case 1). 
4.2.1 Sensitivity to aerosol optical thickness 
Figure 10 presents absolute difference when NDVI is computed from surface reflectances corrected for a 
constant aerosol optical thickness ("tp = 0.05 at 550nm, Case 1) and for a latitude dependant optical depth 
(Case 2). As expected, the maximum difference of NDVI is located in the tropical zone where Tp is two and 
three times greater than its value in Case 1. 
Figure 10: NDVI difference (NDVI2 - NDVI1) involved by different amount of aerosol, over the 20° E transect 
(July). NDVI1 and NDVI2 refers to Cases 1 and 2 respectively. 
4.22 Sensitivity to vapor water content 
According to Faizoun and Dedieu, (1993), the temporal variability of the water vapor content is +/- 1 g cm'2 
around the monthly mean climatological value in tropical regions where contents are variables and high. To 
be close to observed variability of water vapor content, we modified the water vapor content by + or - 20 % 
and assessed the impact of the fluctuation on NDVI (Fig. 11). In this case, the optical thickness is constant and 
set to those of the Case 1. 
Figure 11: Relative difference of NDVI along the 20° transect in July when water vapor content is modified 
by +20% and -20% around climatology. 
The relative difference is expressed by 100.0*(NDVI-NDVl20%)l NDVI. 
Relative variation on NDVI induced by a +/-20 % change on water vapor content is globally low, less than 5 
%. The highest variation occurs in areas where the NIR reflectance is large, since only channel 2 is strongly 
dependant on water vapor absorption. 
423 Taking into account ground elevation 
SMAC model has been modified to take into account ground elevation as a new input parameter. First, we 
modified 5S subroutines as specified in Teillet and San ter (1991) and then adapted SMAC parameterization of 
optical thickness (Rayleigh and aerosols) and gazeous transmissions as a function of pressure. Rayleigh and 
aerosols optical thicknesses defined in Case 1 are now supposed to be optical thicknesses at sea level. For a 
given location, the; 
simulation is referei 
When ground eleval 
0.008 for molecule: 
case, only oxygen l 
water climatology 
difference of 0.05 a 
0 18 R«yt*tgh 
0 18 --AEROSOL 
$0 -80 -40 
Figure 12: Molecul 
optical thicknesses 
for relief (Case 3). 
43 Temporal mom 
One may wonder if 
an example, Figure 
nearly four years 
atmospheric correcl 
Although part of th 
study show that ad 
variability are i) res 
scan angle; iii) surf 
for long term studie 
Figure 14: Tempon 
5. CONCLUSION 
Correction of atmt 
stratospheric aerosc 
of NDVI, when c< 
atmospheric charac 
uncertainities are d 
elevation effects ai 
temporal signal, bu