pronounced when the sun is low on the horizon (here, 9 S = 64°), and reflectance increases with 0 V , by 10% to
50% in the forward scattering plane (depending on wavelength), and 20% to 50% in the backscattering plane;
the dependance of reflectance on azimuth <b is weaker than for 0 S = 42°. These signatures resemble only
partially those obtained by Kimes (1983) or Escadafal and Huete (1991) on bare grounds. In fact, their
measurements show that reflectance is smaller in the forward direction than at nadir, whereas the opposite
holds in our measurements. However, these measurements and ours are alike in the sense that they show a
significant backscattering component. Note that although the directional signatures are very weakly dependant
on wavelength; this dependance is still perceivable; this may not simplify the modelling of p(X, 0 S , 9 V , <(>)
mentioned in the introduction, which cannot be represented by the simple product of a function of wavelength
and a function depending only on the geometric angles.
The BPDF (that is, the distribution of polarization rate as a function of viewing angles) corresponding
to the same case as that of Figure 2, is shown in Figure 3. If we except, for the time beeing, the case 450nm
(not shown), this figure shows that the polarization rate is remarkably small in the backscattering plane, less
than 1,5 %. It increases when the viewing direction approaches the forward scattering direction and when 0 S
increases, but remain anyhow of rather small amplitude, less than 8 %. Note that the distribution of
polarization rates is rather insensitive to spectral wavelength. At 450nm, the polarization rate has the same
shape as in the other spectral bands, but has a larger magnitude (15 - 20%). Nonetheless, we have verified that
the profiles of polarized reflectances (product of reflectance and polarization rates), plotted as a function of
zenith viewing angle in the principal plane, are in fact very similar for all spectral bands including 450nm.
4.3. Temporal evolution of reflectance at hourly time scale
We have performed several experiments consisting of making during several hours continuous measurements
of reflectance with the ground radiometer viewing at nadir the same area during the recording. These
measurements may be very helpful to cross calibrate BRDF measurements obtained by REFPOL at different
times.
Figure 4 shows the evolution of the surface reflectance measured in Algeria 5 with the ground
radiometer in the four spectral bands from the sunset (sun zenithal angle: 70°) to midday (zenithal angle: 37°)
on March 6, 1993. The reflectance decreases with 0 S by about 20 % for all the 4 spectral bands. A weak
discontinuity appears near 8,5 hUT which is due to cirrus perturbation. The order of magnitude of directional
effects found in that campaign (20 % - 30 %, see § 4.2) may be compared with the seasonal and hourly
variations of the METEOSAT-4 visible signal, found to be 5 % and 17 % respectively by Cosneffoy et al (1993)
for Algeria 5. More work is needed to ascertain whether these results are consistent.
4.4. Spatial variability
The spatial variability of each site at a scale length of a few tens or hundred meters is estimated with the
ground radiometer which views different areas at nadir, around midday to minimize directional effects. The
different areas represent a sampling of a surface whose typical size is about 2x2 km 2 . On each site, we
compute the standard deviation of the different surface reflectances measured on each area, and also the
maximum deviation, divided by the average reflectance. The results are presented on Table 4.
Sitc
Mean zenithal angle (°)
Band
Standard deviation
(%)
Maximum
deviation (%)
550
4,7
10,7
Algeria 5
38,5
650
6,1
17,6
850
6,4
17,5
1650
8,3
26,3
550
7,8
17,3
Algeria 4
37,7
650
6,4
15,8
850
6,7
16,3
1650
7,7
19,3
550
7,4
20,1
Algeria 3
34,0
650
7,8
21,5
850
6,9
20,6
1650
7,0
23,5
Iat>lg 4 - Standard and maximum deviation, in relative values, of nadir measurements of reflectance at different
places in a 2 x 2 km 2 area
