822 
3.3. Validation over land 
Over land, we need the emissivity 645 . It varies in general from 0.94 to 1.00 following the different kinds of 
target. As suggested by Kaufman and Remer, 1994, we can use a midrange emissivity of 0.97. Another 
assumption may be to represent the emissivity as a function of surface greenness following the formula of Van 
de Griend and Owe, 1993, which gives £45 from the NDVI (channels 1 and 2). The formula of the surface 
reflectance in channel 3 has been tested assuming the emissivity derived from the NDVI and using the FIFE-87 
data set. The results of the retrieved surface reflectance are plotted Figure 10 where the points called 
"Measurements" are not real measurements but are the value of the surface reflectance retrieved from 
MODTRAN simulations using all AVHRR data, radiosondes, surface temperature measurements. In this figure 
we compare our results with this given by the formula developed by Kaufman [Kaufman and Remer, 1994]. 
Figure 10: Retrieved surface reflectance in channel 3 using the FIFE-87 data set 
5 - APPLICATIONS 
In this part, we present some applications of the surface reflectance in channel 3 for specific studies. We first 
look at the advantage to use channel 3 for the surface properties retrieval. We use the formula to detect the 
Dense Dark Vegetation in channel 1 and to perform an aerosol optical thickness retrieval over these targets. At 
last we study a desertic aerosol event, and show how channel 3 can be used in desertic aerosol retrieval. 
5.1. Surface properties retrieval. 
Kaufman and Remer pointed out a possible high correlation between the AVHRR channels 1 and 3. Using the 
FIFE-87 data set, we compare in Figure 11 the surface reflectances in channels 1 and 3 versus the view zenith 
angle. The surface reflectances of channel 1 were computed for TOA signal using 6 S [Vermote et al, 1994]. 
Figure 14 confirms the good correlation between channels 1 and 3 and thus shows the potentiality to BRDF 
studies using channel 3. 
This figure also shows the limitations of surface reflectance formula. We can see a disagreement for view zenith 
angles greater than 50°, but that be can improved when computing mo, mi and m 2 (equation 05). Currently, only 
angles from 0° to 50° has been used. 
Figure 11: Comparison between the surface reflectance in channel 1 (from FIFE-87) 
and the surface reflectance in channel 3 (from eq 06).