Assessing the accuracy and efficiency of longwave radiative transfer models involving scattering effect with cloud optical property parameterizations

A radiative transfer model (RTM) that accurately and explicitly accounts for both absorbing and scattering effects requires a substantial amount of computational effort. In the longwave (LW) spectral regime, the atmosphere is optically opaque and absorbs a large portion of terrestrial thermal radiation. To alleviate the computational burden, clouds are assumed to be only absorptive in most general circulation models (GCMs) and their scattering effects are neglected. Using parameterizations of cloud bulk single-scattering properties derived from the latest cloud optical property models for satellite remote sensing, this study analyzes the numerical accuracy and efficiency of a variety of LW RTMs . The approaches considered in this study include the absorption approximation (AA), the absorption approximation with scattering parameterization (ASA), the two-stream approximation (2S), a hybrid two- and four-stream approximation (2/4S), and the discrete ordinate radiative transfer (DISORT) with multiple streams. These approximations are benchmarked against 128-stream DISORT calculation. After evaluating the full ranges of ice and water cloud optical and microphysical properties using these RTMs, we find that neglecting LW scattering effect causes simulation errors as large as +/- 15% by using the AA method. Among these RTMs. the 2/4S method provides an optimal balance between computational efficiency and accuracy, leading to the maximum cloud emissivity errors within +/- 5%, 25th to 75th percentile errors about +/- 1%, and almost zero net bias. Therefore, the 2/4S approximation is computationally accurate and yet affordable option for incorporating cloud longwave scattering effect into the radiation schemes used in weather and climate models. (C) 2019 Elsevier Ltd. All rights reserved.