Remote sensing of total direct aerosol radiative forcing from MERIS measurements: A comparison of two different retrieval schemes

The global distribution of aerosol radiative forcing is a large uncertainty in present climate models. Aerosol retrieval from spaceborne sensors can h elp to fill this gap at least over the ocean. For present satellite sensors like AVHRR, MOS, SeaWiFS or POLDER and future sensors, such as MERIS, MODI S and GLI, algorithms have been developed to derive optical and physical ae rosol parameters like optical depth, particle size distribution or aerosol type. Different authors proposed to derive the climatically important aeros ol effect on the upward radiative flux directly from satellite measurements of upward radiance instead of deriving aerosol parameters subsequently use d for the assessment of the aerosol radiative forcing. By such an approach, retrieval errors due to absorption and not well defined phase functions ca n be reduced. In this paper, two different algorithms for the direct retrie val of the top of atmosphere upward radiative Aux in the solar spectral reg ion over the ocean are compared. Based on radiative transfer calculations, a regression algorithm and a Neural Net algorithm have been developed for t he use of present MOS and future MERIS measurements. Sensitivity tests acco rding to sensor noise were performed for both algorithms. The low noise sen sitivity of the Neural Net algorithm makes this type of algorithm more prom ising. The retrieval scheme is applied to MOS scenes of the North Sea and t he North Atlantic. (C) 2000 Elsevier Science Ltd. All rights reserved.