639 
A linear regression analysis of soil reflectance data measured by the CIMEL radiometer and 
predicted by the model was used here. The reflectance data corresponding with positions of the radiometer for 
which the instrument cast a shadow on the observed soil surface, were eliminated from the analysis. 
3. RESULTS 
The accuracy of the soil reflectance distribution in the view zenith angle function generated by the model was 
tested on an untypical soil surface. The surface created by pebbles makes easier a geometrical description of the 
soil surface for reflectance modeling. The average proportion between the vertical (b) and the horizontal radii 
(a) of the pebbles was 0.56. The average pebble area index (AI) measured from the photographs within the 
radiometer field-of-view (FOV) at the nadir was 0.56, and the average relative distance between the pebbles 
(d/a) was 2.37. The ratio (f) was evaluated by substituting different values to the model and looking for the 
values which gives the highest correlation coefficient and the lowest root mean square between the model 
generated and observed soil reflectance data. The value of the f equals 0.20 was found in the way for all the 
three channels. 
The regression analysis was performed separately for the three channels, using 169 pairs of 
data representing the soil surface under different illumination conditions. The analysis yielded the highest 
coefficient of determination r 2 = 0.94 for the SX2 channel and the lowest one, r 2 = 0.88 for the SX3 channel 
(Fig. 2). The relative reflectance factor may be predicted for the channels SX1 and SX3 with a mean deviation 
(rms) from the measured reflectance data of about 0.07 - 0.08, and 0.06 for the channel SX2. 
The coefficient of determination and the root mean square (rms) computed separately for 15 
solar positions for the three channels are presented in Table 1. For all these three channels the measured 
reflectance curves generally show a similar fit to those predicted by the model when the solar zenith angles are 
lower than 50° (Fig. 3). The precision of the fit is between 89% and 97%. For higher solar zenith angles 
(SZA) than 50° the precision for the SX1 and the SX3 dicreases to 82%. The best correlation demonstrate the 
data for the channel SX2 for which the coefficient of determination does not reach the lower value than 93%, 
even for higher SZA than 50°. The relative rms difference between the generated and measured reflectance 
factor for SZAs lower than 50° was less than 0.06 - 0.09 for all the channels. When solar zenith angles were 
higher than 50° the root mean square for them reached values lower than 0.14. 
Table 1. Coefficient of determination (r 2 ) and root mean square (rms) for measured and predicted reflectance 
data for different ¿Illumination conditions defined by the solar zenith angle (SZA) and the solar azimuth angle 
(SAA). N is the number of the data included in the analysis. 
SZA 
SAA 
N 
SX1 
SX2 
SX3 
r 2 
rms 
r 2 
rms 
r 2 
rms 
25.2 
174.0 
11 
0.92 
0.07 
0.97 
0.04 
0.93 
0.06 
25.5 
192.4 
11 
0.94 
0.06 
0.98 
0.03 
0.96 
0.04 
27.7 
150.1 
12 
0.93 
0.07 
0.96 
0.05 
0.96 
0.06 
28.8 
214.8 
11 
0.96 
0.05 
0.98 
0.05 
0.96 
0.05 
31.6 
135.2 
11 
0.89 
0.09 
0.95 
0.06 
0.91 
0.07 
33.7 
230.5 
12 
0.94 
0.06 
0.97 
0.04 
0.96 
0.05 
36.5 
123.2 
12 
0.93 
0.09 
0.96 
0.08 
0.91 
0.09 
40.3 
243.5 
11 
0.97 
0.05 
0.98 
0.04 
0.97 
0.05 
41.3 
114.8 
11 
0.91 
0.09 
0.93 
0.08 
0.89 
0.09 
46.8 
253.0 
11 
0.95 
0.07 
0.97 
0.05 
0.95 
0.06 
48.3 
105.2 
11 
0.95 
0.05 
0.95 
0.06 
0.89 
0.09 
54.1 
98.5 
11 
0.93 
0.07 
0.94 
0.08 
0.82 
0.10 
57.3 
264.8 
11 
0.90 
0.07 
0.95 
0.06 
0.88 
0.08 
62.5 
90.2 
12 
0.91 
0.08 
0.94 
0.09 
0.85 
0.12 
66.0 
266.8 
12 
0.84 
0.14 
0.93 
0.11 
0.83 
0.14 
The largest incompatibility of the measured data with the predicted ones refers to the 
forwardscattering range of the relative reflectance factor. It is caused by a fact that specular features of the soil 
reflection are disregarded in the modelling.