i transformed into 
ie effects. SMAC 
e and particularly 
parameterize the 
omputed for each 
1 relative error on 
id surface NDVI 
rase 1: a constant 
naps at the three 
ints in channels 1 
computed from 
1 
0 
1 
0 
Figure 7: Histogram ofTOA reflectances and surface reflectances in AVHRRJVIS (a) and AVHRR/NIR (b) 
channel, and histogram of NDVI (c) computed from TOA and surface reflectances (July). 
The histograms of reflectances and NDVI are presented in figure 7. In channel 1, the correction of the 
atmospheric effects can either increase the reflectance level when TOA reflectance is high, for instance over 
deserts or decrease it when TOA reflectance is low. In AVHRR/NIR channel, the correction increases 
systematiquely the reflectance level. As a result, surface NDVI is greater than TOA NDVI. Two modes appear 
on NDVI histogram displaying the existence of two distinct states. The first one (around 0.15) is representive 
of bare soils. The second is more spread and is representative of the great majority of plant ecosystems. NDVI 
of vegetated areas varies from 0.2 to more than 0.9. Highest values stand for equatorial and temperate forests. 
One can note that a small peak around 0.02 for NIR reflectance and 0.05 for NDVI appears in the histograms. 
This is probably due to mixed pixel falling on coast-lines. 
Figure 8 shows TOA and surface reflectances for the transect For AVHRR/VIS channel, the relative 
difference varies between 5% and -100 % depending on the TOA reflectance level. It is quite stable for 
AVHRR/NIR channel around 20 %. For NDVI (Fig. 9), the relative difference between raw NDVI and TOA 
NDVI is around 30 to 40 % and greater than 50 % between raw and surface NDVI. 
Figure 8: TOA and surface reflectances in AVHRR/VIS and NIR channels over the 20°E transect (July). 
Figure 9: Raw, TOA, and surface NDVI (left) and relative difference of NDVI (right) over the 20° E transect 
(July). The relative difference is 100*[NDVI(TOA - surf) - NDVI(Raw) ]l NDVI(TOA - surf). 
:s. 
37