49 
arm attached to one of our cars, in such a way that the observation plane is at about 1 meter from the vehicle 
front, and perpendicular to the car axis. This instrument was used to provide the BRDF (Bidirectional 
Reflectance Distribution Function) and BPDF (Bidirectional Polarization Distribution Function) of the sites. 
The angular resolution in the observation plane is about 2,5°. Also, when looking upwards, the instrument 
measured the sky radiance and polarization. 
3.2. Calibration 
The sun radiometer was calibrated using the Langley-Bouguer technique. This technique was applied on the 11 
data sets we collected during the campaign. We added to those measurements two extra data sets collected 
during two clear days: one on April 4th, 1993, on White Sands (New Mexico) and one on La Crau (South-East 
of France) in June 1993. All the offsets are corrected from the variation of the sun-earth distance. The turbiest 
day was excluded and we averaged the other values to obtain the conversion factor to get the direct 
transmission; the coefficient of variation of the obtained conversion factors for all collected data sets is around 
4 percent in all the bands. 
Several methods are applied to convert the digital counts of the ground radiometer and of REFPOL into 
normalized radiance (corresponding to an incident irradiance equal to n). First at LOA, we used an integrating 
sphere to calibrate the two radiometers just before and just after the campaign. Then, in Tucson (Arizona), after 
the campaign, the calibration laboratory of P. Slater (University of Arizona) was used to calibrate the 
instruments by reflection on a calibrated panel. These calibrations used a lamp as a source. 
Other methods used the sun as a source. Both instruments were used viewing at nadir a halon panel. The 
bidirectionality of the panel reflectance was measured at LOA with the ground radiometer, using the technique 
reported in Jackson and Slater(1986). Around noon, the panel is rotated to vary the incident angle and a sun 
shadower used to derive the direct sun irradiance. These bidirectional measurements are converted into 
bidirectional reflection coefficient p using the panel albedo provided by the manufacturer. Measurements of the 
normalized radiance L above the panel, related to p by : 
P( e s , e v , 4 » = L [ dv T ^~ (i) 
cos0 s T(0j) 
where T is the total transmittance (i.e., the global irradiance normalized by the exo-atmospheric irradiance), 
provide the desired instrument calibration since p, T and 0 S are known. T is computed using a transfer code 
(see Deuzé et al., 1989), using the method of successive orders of scattering, with as inputs the optical 
thicknesses (derived from the barometric pressure for the Rayleigh and from the extinction measurements for 
the aerosols) and an aerosol model (a Junge size distribution associated to the Angstrom coefficient and a 
standard refractive index for continental material). The gaseous absorption is derived from a climatology of 
absorbing gas constituants and the 5S radiative trabsfer code (Tanré et al. 1990). 
Table 3 reports the calibration coefficients obtained both for laboratory measurements as well as for field 
measurements for the ground radiometer. The digital counts have to be multiplied by .00001 and by these 
coefficients to get the normalized radiances (dimensionless). 
cal i brat 
location 
on 
date 
method 
550 
wavelen 
650 
gth (nm) 
850 
1650 
LOA 
01/03 
sphere 
8.14 
12.6 
6.34 
14.2 
LOA 
05/02 
sphere 
8.75 
13.4 
6.77 
15.2 
LOA 
22/03 
sphere 
8.59 
13.1 
6.59 
15.0 
Tucson 
05/04 
panel 
8.88 
13.6 
6.83 
14.8 
LOA 
06/10 
sphere 
8.60 
13.0 
6.60 
14.7 
Sahara 
03/03 
panel 
8.02 
11.9 
6.38 
13.9 
Sahara 
03/03 
panel 
7.85 
12.0 
6.42 
13.7 
Sahara 
03/05 
panel 
8.57 
11.8 
6.27 
14.7 
Maximum relative dispersion 
(±%) 
-6,8 
+7,3 
+4,7 
+4,7 
Table 3 - Ground radiometer calibration 
Table 3 shows that the dispersion of coefficients relative to their mean values is at most 7%. Note 
however that the interband calibration accuracy (not shown in Table 3) is better than 1%, which proves very 
satisfactory. We notice that laboratory and field measurements present the same dispersion with a systematic 
bias between the two data sets. Other field measurements have to be reduced to check that and to have a better 
estimate of the field calibration. In.fact, the field calibration is also a way to verify the suitability of equation (1)