On the derivation of a spatially distributed aerosol climatology for its incorporation in a radiometric Landsat pre-processing framework

The climatology is intended as a fallback strategy for aerosol estimation in bright Southern African savannah ecosystems where the rarity of dark objects decreases the applicability of common image-based aerosol estimation strategies and the widespread burning prohibits the use of a fixed aerosol loading.

We predicted the climatological path reflectance surface by applying a multivariate regression model to all available path reflectance retrievals on basis of the geolocation and the days of the year on which the data were acquired. The resulting predictions are able to successfully model major spatio-temporal gradients of the path reflectance distribution. The prediction error (weighted root mean squared Error at 0.483 µm) was less than 1% reflectance while the prediction itself varied by 4.6% reflectance.