34 
22Aerosol optical t 
1 e>oo - 
Figure 1: Ground elevation along the transect located at 20° East ( Negative sign is for North latitude). 
Vegetation study are currently made using the Normalized Difference Vegetation Index (NDVI) 
which is the normalized difference of red and near infrared channels. Three different levels of NDVI have 
been used: NDVI computed with i) raw digital counts (initial GVI product), ii) calibrated reflectances (TOA 
NDVI). and iii) surface reflectances ( surface NDVI). 
2.1 G VI 
2.1.1 GV1 product 
Six years (1986 to 1991) of NOAA/AVHRR GVI products (Global Vegetation Index, Kidwell, 1990) have 
been processed in this study. Data include digital counts (DC) for the visible (DC1, [0.58-0.68] pm) and near 
infrared channels (DC2, [0.72-0.98] pm), raw NDVI (DC2-DC1)/(DC1+DC2), two thermal infrared channels, 
solar zenith angle and scan angle. GVI data are available in the Plate Carrée projection, at a weekly time step 
with 15 km x 15 km spatial resolution (1/7 of square degres). Spatial coverage spreads from 75° N to 55° S in 
latitude, -180° W to 180° E in longitude. 
GVI products is obtained after two step process, a sampling of the original full resolution data and a 
weekly compositing procedure based on the maximum difference between DC2 and DC1 (Kidwell, 1990). 
From 1986 to 1991, two satellites were successively in operation (NOAA 9 and 11). Satellite change and 
sensor ageing lead to calibration drift Theses changes are not identical in both AVHRR channels and can 
significantly affect both TOA reflectances and NDVI. Post-launch calibrations over various targets have been 
made and new calibration coefficients provided to users. In this study, reflectances at the top of the 
atmosphere have been calibrated using Kaufman and Holben coefficients (1993). 
2.12 Atmospheric effects in VIS and NIR channels 
Both VIS and NIR AVHRR measurements are sensitive to absorption and scattering by gases and 
aerosols (see Tanré et al., 1992 for a detailled discussion). Weak ozone and water vapor absorption occurs in 
AVHRR/VIS channel whereas a strong water vapor absorption occurs in AVHRR/NIR channel (Fig. 2). Due 
to rapid saturation and nearly constant amount, oxygen absorption in AVHRR/NIR exhibit a flat profile along 
the transect and is not sensitive to angular configurations, which is not the case of water vapor and ozone 
absorptions. Scattering processes are well modelled for molecular scattering which is more important in 
channel 1 than in channel 2. Scattering by aerosol is more difficult to model since the variability of aerosols is 
high and that their characteristics are variables in time and space. Nevertheless, we can assume that aerosol 
optical thickness Cep) is generally larger than 0.05 at 550nm : this value corresponds to a horizontal visibility 
greater than 100km. 
Figure 2 : Variations of total, water vapor, ozone, and oxygene transmittance along the transect in 
AVHRR/VIS (left plot) and NIR (right) channels. Carbon dioxyde transmittance is equal to 1 
At present time, we i 
must rely on a sens 
tested. In the first oi 
0.05 at 550 nm (Th: 
thickness varies as a 
the equator (Fig. 3). 
supposed to select cl 
tropical regions whic 
0 18 • Raytoigh 
^0 16 -AEROSOL 
Figure 3: Variation 
for Case 1 (left) ai 
thicknesses respectiv 
23 Water Vapor C< 
Water vapor climate 
from atmospheric ei 
1973). Initial spatial 
the GVI resolution. 1 
the data around 10° 
even on a mean of 1 
climatology can be 
characterized by a dj 
exists and roughly re 
2.4 Ozone Content 
Globally, the variab 
Therefore, correctioi 
been compiled from 
degrees zonal bands. 
f, 
3 
Figure 4: Variation