A fast radiative transfer model for hyperspectral remote sensing

Recent advances in the use of hyperspectral data in remote sensing have imposed new challenges for radiative transfer (RT) models. The RT model must be computationally efficient, yet have appropriate accuracy and an extensive application range to ensure reliable retrievals under all conditions. Towards this goal, we report on the development of a fast RT model that can be applied to situations involving multiple cloud layers or for a cloud that is vertically inhomogeneous. The model is constructed by combining a computationally fast clear sky RT model (e.g., AIRS Radiative Transfer Algorithm, or SARTA) with an adding-doubling model. The former provides the optical thickness of the background atmosphere to initialize the adding-doubling process and the latter is used to solve the radiative transfer involving multiple scattering. Different from other RT models, in which the bulk optical properties of a scattering medium such as a cloud or aerosol layer are computed directly from the single-scattering properties of the particles in the medium, the present model employs a pre-computed database (i.e., a look-up table) and an interpolation scheme to obtain these properties. This simplification improves the computational efficiency by more than two orders of magnitude in comparison with the well-known discrete ordinate radiative transfer model (DISORT). Comparison of results with DISORT show that the root-mean-square differences of the brightness temperatures computed from our fast model and DISORT are less than 0.3K under most circumstances.