Actinic Flux and Net Flux Calculations in Radiative Transfer—A Comparative Study of Computational Efficiency

Abstract

The accuracy and speed of three well-known computational techniques (DISORT, the δ?four-stream approximation, and the two-stream approximation), and the matrix inversion method, which is less well known, have been investigated. Results are presented for both broadband actinic and net fluxes over a range of parameters including solar zenith cosine, relative humidity, and altitude for two different surface/aerosol systems: terrestrial and oceanic. The matrix inversion method can only calculate actinic fluxes; therefore, this is the main focus of this paper. Investigations into the comparative accuracy of the four techniques for the oceanic model with and without a cloud layer included are also presented. (DISORT is taken as the benchmark for this research.) Based on results presented here, it is found that for actinic flux calculations, the δ?four-stream approximation is slightly more accurate than the matrix inversion method, and that both are far more accurate than the two-stream approximation. However, for net flux calculations, the δ?four-stream approximation fares better and is clearly the most accurate. The superiority of the δ?four-stream approximation is particularly noticeable for both net and actinic fluxes when a cloud layer is included. In this paper, information is provided to assist modelers in choosing a computational technique that best suits their needs. The relative computational efficiency of the various radiative transfer techniques is also discussed for the benefit of those modelers who seek a compromise between time and accuracy, rather than solely maximal accuracy in a particular technique.