50 
to convert radiance into reflectance. The use of equation (1), applied both to panel and ground measurements 
permit to limit the reference to the panel (pratically difficult to be extensively used during this campaign) and 
instead rely the conversion on a computation of the global irradiance. An other traditional way to obtain ground 
reflectance, which has not been followed during the campaign, is to cross measure on the panel as often as 
possible. By doing so, uncertainties on the aerosol models (such as the refractive index), which certainly affect 
the comparison between laboratory and field measurements, can be avoided. 
4. RESULTS 
The measurement campaign lasted on average two days per measurement area. The sun radiometer was 
operated all day during a measurement campaign, in order to provide the aerosol optical depth, at wavelengths 
given in Table 1, in the morning and in the afternoon, using a standard Langley-Bouguer technique. As 
described in § 3.2, these optical depths serve as inputs in a radiation transfer code to derive the total 
atmospheric transmittance (i.e., the sum of the direct and diffuse irradiances measured at the surface, divided 
by the exo-atmospheric irradiance). One obtains then the surface reflectance p(0 s , 0 V , 0) from the 
measurements of normalized radiances using Eq. (1). Different measurements of surface reflectance have been 
done on each measurement area to extract different parameters: spectral reflectance, BPDF and BRDF at 
different times of the day, temporal evolution of reflectance during a few hours, and spatial variations of 
reflectance. 
4.1. Spectral reflectance 
Measurements of spectral reflectance have been performed at various times and for various viewing 
angle configurations with the REFPOL and ground radiometer instruments (Table 1). We report in Figure 1, as 
an illustration, the measurements of spectral reflectance made with the ground radiometer at nadir over a flat 
area in Algeria 5 on March 6, 1993. The spectral signature is measured when the sun is high (from 11,5 h UT 
to 11,8 h UT), so that the zenithal solar angle variations do not exceed 1°. The sun zenith angle is about 35°. 
Every 20s, a measurement is done which is, then, converted in surface reflectance. Figure 1 represents an 
average of these measurements. The applied calibration coefficients have been chosen to be the average of the 
calibration coefficients derived from laboratory calibration experiments of February 5 and March 22, displayed 
in Table 3. The reflectance increases with wavelength (Figure 1), as expected, and reaches the unexpected high 
value of 0.83 in the shortwave infrared The values of Figure 1 are surprinsingly rather different from those 
obtained by Jaccoberger (1989) on sand dunes at the site of Bahariya (desertic Egypt), who found 0.38, 0.5 and 
0.6 at 550, 650 and 850nm respectively (to be compared to our values, respectively 0.29, 0.55 and 0.67). 
4.2. BRDF and BPDF 
The radiometer-polarimeter REFPOL has been used to acquire several measurements per day of BRDF 
and BPDF. The experimental protocol was as follows. Once a test area was chosen, the car carrying the 
REFPOL instrument was driven along an axis such that when the car stops, the scanning plane of REFPOL 
makes an angle of either 0°, 45°, or 90° with the solar principal plane. This axis was determined before the 
experiment by disposing a number of markers on the test area. Due to driving difficulties on sandy ground, the 
scanning plane orientation differs sometimes from the desired orientation by at most 5°; however, it is 
measured once the car has stopped with an accuracy of the order of the degree. It is important that the test area 
be not altered by car tracks; this implies in practice that one of the observation planes be measured on a test 
area not strictly coincident with the test area used for the two other observation planes. The total acquisition of 
3 observation planes lasted about 30 to 40 minutes, with a corresponding solar zenith angle variations of at 
most 1° around midday and 9° in the morning and in the evening. 
A representative example of the BRDF measurements is shown in Figure 2, which represents a BRDF 
measured in Algeria 5 on March 5, 1993, around noon (0 S = 42°) and in the afternoon (0 S = 64°). The BRDF 
shown in Figure 2 appears under the form of polar plots, where the polar angle represents the relative azimuth 
0 between sun and view directions, and the radius represents the viewing zenith angle. The measurements of 
reflectance have been normalized to nadir reflectances (the BRDF value at nadir is 1). The 0 = 45° 
measurement has been duplicated symmetrically with respect to the 0 = 0° axis, and an isocontouring procedure 
has been processed to interpolate the various measurements. 
Figure 2 shows that when the sun is relatively high (here, 0 S = 42°), the reflectance is nearly Lambertian 
in the forward scattering plane, except at 450nm (not shown) where the reflectance increases by 20% in that 
plane. Some backscattering occurs in the backscattering plane, with an amplitude which decreases steadily with 
wavelength (30%, 20%, 15% and 13% at 450, 650, 850 and 1650nm respectively). Directional effects are more