• Two measurements acquired during different flight lines but for similar viewing directions show coherent 
values. Some differences, which generally do not exceed 5% of relative value, can be explained from the 
variations in solar zenith angles between the measurements: The flight lasted about 75 mn over our site and 
the solar zenith angle decreased from 46° to 32°, during that period 
Fig. 2 and 3 are the polar representations of the directional reflectance for two dominant cultures of the area, 
sunflower (Fig. 2) and wheat (Fig. 3) at two wavelengths (550 and 850 nm). The order of magnitude of the 
reflectance is as expected for vegetation covers: The reflectance in the visible channel is on the order of 10%. It 
is much larger in the near IR, about 50% for sunflower and 30% for the wheat. The directional signature is 
consistent with surface measurements: the reflectance generally increases with the viewing angle and it is much 
larger for backscattering directions than for forward reflection (Kimes, 1983; Kimes et al., 1985). The minimal 
reflectance is observed in the principal plane, about thirty degrees off nadir, in the forward direction. There is a 
ratio of about 2 between the maximal and minimal directional reflectance for a given target and spectral band. 
The observed directional effects can be explained by geometry and shading considerations, as discussed in 
Kimes (1983). 
Another signature of interest, but of limited angular extension, is the “hot spot” in the antispecular direction. 
In this particular direction, the instrument does not view any shadow and the reflectance is generally maximum. 
The angular extension of the hot spot depends on the canopy geometry as discussed in Jupp and Strahler (1991). 
The observation of the hot spot from the surface is difficult since, unless the radiometer is far from the surface, 
its shadow interfers with the measurement. The airborne POLDER, thanks to its altitude and directional 
resolution, is able to observe the hot spot. This is partly illustrated in Fig. 4 which corresponds to a vineyard 
canopy. The largest measurement (R coz = 54 %) is acquired for the direction the closest (9 V =41°) to the hot spot 
(principal plane, 38.3° zenith angle). Other measurements closer to the nadir (9 V =32°), and further from the 
nadir (9 V =49°) show a much lower reflectance (45.0 % and 47 % respectively). 
Rice fields show a feature that is not observed over other surfaces: a narrow spot of large reflectance (Fig. 5). 
It results from specular reflectance over flooded areas and affects predominantly the direction symétrie to that of 
the hot spot. In Fig. 5, which corresponds to the 550 nm band, the reflectance is 29% at the specular point, and 
only 10% a few degrees away. If we do not consider the specular direction measurements, the rice paddies 
directional signature agrees with that of other surfaces; i.e. a larger reflectance in the backscattering hemisphere. 
Fig. 2 to 5, and others not shown here, demonstrate the ability of the airborne POLDER instrument to 
measure the reflectance directional signature of surface targets. The main advantage of airborne remote sensing, 
when compared to surface measurements, for the evaluation of directional signature is its ability to sample 
surfaces whose typical length scale is larger than one meter. For instance, the directional signature of a vineyard 
or an orchard is not accessible from the surface. Another advantage is the capability of assessing the directional 
signature and their variations at the regional scale. 
4 SPATIAL EXTENSION OF ANGULAR SIGNATURE 
In order to quantify the directional signature of the surface, we make use of a simple directional model. The 
reflectance is assumed to be a linear combination of an isotropic term, and two anisotropic functions derived 
from physical considerations (Roujean et al., 1992b): 
R(Q S , 9 V , <p) = k 0 + *, /,(9 J( 9 V , cp) + k 2 f 2 ( 9,, 9 V , cp) (2) 
The values ko, ki and k2, are determined by a best linear fit over the measurements. Since both functions are 
normalized so that they are 0 for 9 S =9 V =0, ko is the reflectance when both the sun is and the sensor are at 
zenith. Thus, ko is a reflectance corrected for angular effects, i.e. normalized. Similarly, k^/ko and k2/ko provide 
a quantitative estimate of the surface directional signature. 
We performed the linear fit for each wavelength and each pixel independently. The procedure yields the three 
kj and the correlation between the measurements and the model. Fig. 6 is a map of the ko for both 550 and 850 
nm. The covered area is the same as in Fig. 1. The Rhone river can clearly be identified as a wide band of low 
reflectance at both wavelengths. Values are as expected and the spatial structure can be correlated with a land 
use map that was made during the measurement campaign. This result confirms that several directional 
measurements can be used to provide a single normalized value. 
ki/ko and k2/ko provide a quantitative estimate of the directional signature. A major objective is to relate 
these coefficients to surface characteristics. Although their maps do show some structure, it has proved difficult 
to correlate their values to surface coverage indicated by the land use map. 
The correlation between the model and the measurements is very satisfactory as shown in Fig. 7. This figure 
provides a cumulative histogram of the correlation over the study area for 550, 650 and 850 nm. There is a