36 
2.5 Digital elevation model 
The ground elevation map (Edwards, Etopo5, 1989) which has a 5 minute resolution, has been transformed into 
pressure map and then degraded to the GVI resolution. 
3.METHOD : Correction of the atmospheric effects 
We used the SMAC method (Rahman and Dedieu, 1994) to correct the GVI from atmospheric effects. SMAC 
method is based on the 5S radiative transfer code (Tanffi et al, 1990). The model is easy to use and particularly 
well adapted for processing large temporal series of satellite data. Analytical functions parameterize the 
absorption and scattering processes for both gases and aerosols. Their coefficients have been computed for each 
channel. Performance of SMAC compared to 5S is good, only a few percent for the maximum relative error on 
TOA reflectance computation. 
4. RESULTS AND DISCUSSION 
4.1 Results at global scale 
Surface reflectances have been computed using SMAC and climatologies described in §2 and surface NDVI 
derived from that surface reflectances. The input parameters correspond to those defined in the Case 1: a constant 
aerosol optical thickness (0.05) and climatologies of water vapor and ozone contents. NDVI maps at the three 
steps of the processing are shown on figure 6. The first one is NDVI computed from digital counts in channels 1 
and 2. The second is NDVI computed from TOA reflectances, and the last one is NDVI computed from 
atmospherically corrected reflectances. Sea is masked at step 2 and 3 of the processing. 
Figure 7: Histogi 
channel, 
The histogi 
atmospheric effects 
deserts or decrease 
systematiquely the i 
on NDVI histogram 
of bare soils. The se 
of vegetated areas \ 
One can note that a 
This is probably dui 
Figure 8 si 
difference varies h 
AVHRR/NIR chant 
NDVI is around 30 
Figure 6 :NDVI evolution along data processing. Top : Raw NDVI; Middle : NDVI 
computed from TOA reflectances; Bottom : NDVI computed from surface reflectances.