Cloud statistics-based estimation of land surface albedo from AVHRR data

Cloud cover constitutes a major challenge for the surface albedo estimation using AVHRR data. The demand of pixelwise high accuracy cloud masking based on only single AVHRR images without any augmenting information is not realistic in all possible conditions of cloud fraction and cloud type on any land cover type. Another approach to tackle cloudy conditions is presented in this study. In the areas studied (Finland, Switzerland, the Netherlands) cloudy broadband albedo distributions were constructed for AVHRR data of one month and 15 km x 15 km area. The peak of the distribution, which is caused by the clear sky conditions, correlated well with the corresponding surface albedo distribution obtained from pyranometer measurements at Sodankyla, Payerne and Cabauw masts. Using the half and ¾ height points of the peak it was possible to estimate the surface albedo with quite good accuracy using a simple physically motivated formula.