Towards a Generalized Approach for Correction of the BRDF Effect in MODIS Directional Reflectances

We propose a new method for the estimation and the correction of directional effect on satellite derived surface reflectance time series. The correction assumes that the shape of the Bidirectional Reflectance Distribution Function (BRDF) varies much more slowly than the reflectance magnitude. The BRDF shape is quantified by two parameters R and V which account primarily for surface scattering and volume (canopy) scattering processes. Assuming a constant shape throughout the year permits an easy inversion of the two parameters that are used for a normalization of the reflectance time series to a standard observation geometry. Using this method, the corrected time series of data from the Moderate Resolution Imaging Spectroradiometer show much less high frequency variability than the original input data. A refinement of the method accounts for a slow variation of R. and V, linearly with the Normalized Difference Vegetation Index (NDVI) accounting for vegetation dynamics, and produces an even better correction of directional effects and noise reduction in the time series. The nonlinear short-term variations in the reflectances, which are interpreted as "noise," are strongly reduced, by a factor of two to four. The reduction in noise in the NDVI time series is less, as the index already implicitly corrects for the directional effects. Nevertheless, the noise in the NDVI time series is reduced by typically 30%-40%. Although the main output of processing using the method is the corrected reflectance, estimates of R and V are also generated. Global and regional map of these parameters show coherent patterns consistent with landcover.