820 
Over sea, we can assume an emissivity in the channel 4 and 5 equals to the unity (Takashima and Takayama, 
1981). Figure 4-a shows the result of the computation for a typical case of sunglint (emissivity £3 of 0.80 and 
0.94). 
Over land, the emissivities in channel 4 and 5 may be quite different from the unity. For the time being, we 
assume no spectral differences between the emissivity in channels 4 and 5, £ 45 . In Figure 4-b, we show 
computations for £45 equal to 0.96 and for 2 emissivities in channel 3 (0.90 and 0.94) 
In these two figures, we can see a good correlation between T^-T 4 and T 4 -T 5 with a non linear effect. We then 
define a relationship as follow: 
T^-T 4 = mo + mi.(T 4 -T 5 ) + m 2 . (T 4 -T 5 ) 2 (05) 
where mq, mj and m 2 are function of £ 3 , £ 45 . Once we have T^, and using the equations (03) and (04), we get the 
surface reflectance: 
n[B 3 (T 3 )-B 3 (Tp] 
p3 “ E s cos( 6 s ) x 3 (0 v ) T 3 (0 S ) (06) 
3.2. Validation over sea 
Using £45 equal to 1.00, we computed T 3 for 3 solar zenith angles (0°, 30° and 60°) and for the conditions 
described for our theoretical data set. 
Figure 5 shows the theoretical retrieval of the surface reflectances in the case of sunglint observations. The 
observed discrepancy is less than 10% in RMS for a solar zenith angle of 60° and decreases to less than 5% for 
an angle of 0 °. 
True 
Figure 5: Theoretical retrieval of surface reflectances in channel 3 
using (06) for sunglint cases. 
We tested the formula with 5 AVHRR scenes over Pacific Ocean between 1985 and 1990. First we computed 
images of the reflectance in channel 3 and then we compared with theoretical computations of sunglint using 
Cox and Munk parametrization [1954]. Except for desertic aerosols the effect of aerosols is very low in channel 
3 and the principal parameter is the wind speed. Figure 6 plot the reflectance over the sunglint and we compare 
with computations for 3 wind speeds. The measurements follow the theoretical computations for a wind speed of 
5m/s. 
In order to further validate the approach, we computed the signal in channel 1 and 2 assuming a wind speed of 5 
m/s and considering 3 optical thicknesses (0.1, 0.2 and 0.3 at 0.55 mm). The aerosol model used is a maritime 
(which is a good approximation in the middle of the ocean) and the ozone content is given by TOMS. As we can 
see Figures 7, the optical thickness of 0.1 fits the measurements in the two channels with a good agreement.