

Wang, CX; Yang, P; Nasiri, SL; Platnick, S; Baum, BA; Heidinger, AK; Liu, X (2013). A fast radiative transfer model for visible through shortwave infrared spectral reflectances in clear and cloudy atmospheres. JOURNAL OF QUANTITATIVE SPECTROSCOPY & RADIATIVE TRANSFER, 116, 122131. Abstract A computationally efficient radiative transfer model (RTM) for calculating visible (VIS) through shortwave infrared (SWIR) reflectances is developed for use in satellite and airborne cloud property retrievals. The full radiative transfer equation (RTE) for combinations of cloud, aerosol, and molecular layers is solved approximately by using six independent RTEs that assume the planeparallel approximation along with a singlescattering approximation for Rayleigh scattering. Each of the six RTEs can be solved analytically if the bidirectional reflectance/transmittance distribution functions (BRDF/BTDF) of the cloud/aerosol layers are known. The adding/doubling (AD) algorithm is employed to account for overlapped cloud/aerosol layers and nonLambertian surfaces. Two approaches are used to mitigate the significant computational burden of the AD algorithm. First, the BRDF and BTDF of single cloud/aerosol layers are precomputed using the discrete ordinates radiative transfer program (DISORT) implemented with 128 streams, and second, the required integral in the AD algorithm is numerically implemented on a twisted icosahedral mesh. A concise surface BRDF simulator associated with the MODIS land surface product (MCD43) is merged into a fast RTM to accurately account for nonisotropic surface reflectance. The resulting fast RTM is evaluated with respect to its computational accuracy and efficiency. The simulation bias between DISORT and the fast RTM is large (e.g., relative error > 5%) only when both the solar zenith angle (SZA) and the viewing zenith angle (VZA) are large (i.e., SZA > 45 degrees and VZA > 70 degrees). For general situations, i.e., cloud/aerosol layers above a nonLambertian surface, the fast RTM calculation rate is faster than that of the 128stream DISORT by approximately two orders of magnitude. (C) 2012 Elsevier Ltd. All rights reserved. DOI:
ISSN: 00224073 