577 
(Deschamps ei 
1 detector anay 
)HR acqires an 
2 ‘mages, each 
of the radiance 
aages collected 
st. polarization 
are quite large, 
polarization is 
douds in these 
: the cloud, i.e. 
n near infrared 
scattered from 
>le at scattering 
:ulations of the 
w polarization 
ind Liou 1982; 
iko and Travis 
lannels 
¡ported 
0 and 850 nm. 
eneous clouds 
single picture 
unction of the 
zhts the liquid 
ranging from 
p polarization 
1 result of the 
;d from these 
i the cloud top 
g two (lights 
the ability to 
m that ot the 
Accordin' 7 to the previous considerations, we expect that, for moderate viewing zenith angles, the polarized light 
reflected by vegetation canopies and loam soils is small and invariant enough to be handled as a standard boundary 
condition for polarization measurements. In these directions, apart from a systematic correction to be defined (maybe 
from multi-temporal observations), the polarized light should be mainly informative about the atmospheric 
polarization from which the aerosols properties could be retrieved. The surface contribution should predominate for 
grazing observation directions; of course, it must be directly accessible for low enough aerosol loadings. 
In order to test whether polarized light is able to help for separating atmospheric from ground signals, let us consider 
results obtained in the La Crau 1990 campaign. On this occasion, the aerosol loading was moderate, with aerosol 
optical thickness 0.47 at A.=550 nm. The detailed analysis of the results is given in Deuze et al. (1993). POLDER 
pictures in polarized light look much more homogeneous than pictures in total light, which confirms that ground 
polarization is more uniform that ground reflectance. More precisely, surface polarizations are noticeable at wavelength 
¡1=850 nm; the POLDER pictures in polarized light exhibit some contrasts in this channel; but these contrasts vanish 
in channels centered at 650 and 550 nm where the efficiency of the molecular and aerosol scattering increases. Figure 
6A shows a typical BPDF diagram (like in Fig. 5), for measurements obtained over La Crau in the 650nm channel. 
Figure 6B shows that, when normalizing the polarized reflectances to a constant air mass, we reduce the scatter in the 
data which, according to the previous equations, confirms that polarized light is mainly formed by atmospheric 
scattering. The calculated atmospheric molecular scattering is indicated in Fig. 6B; it cannot account for this 
atmospheric signal which is correctly estimated only when taking into account the aerosol contribution; it was derived 
here from ground-based corelative measurements performed during the flight (Deuze et al. 1993). Systematic 
discrepancies between the observed and calculated signals appear for grazing directions; the influence of the surface 
polarization, which is not accounted for in the calculations, becomes very probably noticeable in these directions. 
To estimate the order of magnitude of the ground effect that we neglected in the previous analysis, we can consider the 
measurements performed over the same area one year later, i.e. with nearly the same vegetative cover, but with much 
lower aerosol loading; 0.07 instead of 0.47 for the aerosol optical thickness at h=550 nm. The normalized polarized 
reflectances measured in July 1991 were only a few 0.001 when corrected from molecular scattering, confirming the 
small size of the ground effect. 
Fig. 6. Fig.A: Polarized reflectance Rp, measured above the site of La Crau in channel centered at 650 
nm, during the POLDER flight on June 21, 1990. For each image pixel, Rp has been reported as a 
function of the corresponding scattering angle. Fig.B: same as for Fig.A, but we reported cos9 v Rp 
where 0 V is the zenith viewing angle. 
The very positive results of this experiment were confirmed by other POLDER flights over vegetative covers, 
specially over areas in the Landes. However, the ability of polarization data to detect aerosols above terrestrial surfaces 
is quite certainly limited for some kinds of soil backgrounds. For example, during the Hapex/Sahel experiment, 
POLDER polarized images were acquired on different occasions over the same desertic site. Although the images were 
obtained for very different aerosol loadings, they do not exhibit significant differences.(Breon and Tanre 1994). 
Preliminary analysis of these results seems indicate that this negative result is due to the very large polarized 
reflectance that desertic surfaces exhibit, which is about 3 to 10 times larger than the typical polarized reflectances of 
dense vegetation canopies. 
V-CONCLUSIONS 
First results from the airborne POLDER show that polarized light resulting from reflection from ground surfaces, 
specially from vegetative covers, and from scattering by atmospheric aerosols may be observed from space; the orders 
of magnitude of these contributions and their mean features are in good agreement with our present knownledge of 
these effects, as resulting for a large number of ground based observations.