Surface albedo from space: Coupling bidirectional models and remotely sensed measurements

The general purpose of this article is to analyze the ability of land surface bidirectional models to supplement the limited angular sampling provided by Earth-observing sensors in order to normalize shortwave satellite data or to estimate surface albedo. Briefly, our approach relies on the fitting of directional model parameters using a set of satellite measurements, NOAA advanced very high resolution radiometer, in this study. We used the models proposed by Ross [1981], Hapke [1981], Shibayama and Wiegand [1985], Verstraete ed al. [1990], Deering et al. [1990], Roujean et al. [1992a], and Rahman et al. [1993a]. These models are of three types: empirical, geometrical, and radiative transfer based. The comparison of these models with observed reflectances gave similar results, with very close statistics of adjustments (rms error of reflectance around 0.03, i.e., similar to 10%). This means that these models can be used to normalize directional reflectance, provided the reference direction is close to the sampled angular domain. The most attractive reason for using bidirectional models is the assumption that they would allow the prediction of directional behavior in any direction, even if the set of angles used for parameter adjustment is limited. In order to check this assumption, fitted models were used to predict the reflectances that Meteosat would measure at the top of the atmosphere. These predicted reflectances were compared to the reflectances actually measured. This comparison gave satisfactory results for every model, with rms errors of reflectance close to 0.03 (12%). Direct surface albedos estimated from numerical integration of the models exhibit the same general behavior, but some discrepancies exist at large solar zenith angles (>60 degrees) where no observation was available.