Improvements to the discrete ordinates method for fast atmospheric radiative transfer model development

The success of a variety of fast methods for radiative transfer in the atmosphere (i.e. correlated-k, radiative perturbation, principal component) relies on the accurate collection of pre-calculated radiative parameters. This large amount of data is obtained with the aid of either well establish deterministic (discrete ordinates, spherical harmonics, zone method) or statistical (Monte Carlo) methods resulting in critical computational burdens. In this paper, improvements on spatial accuracy, convergence rate and computational times are introduced to the Discrete Ordinates Method in 1D, 2D and 3D domains. The introduction of flux limiters in the spatial discretization, pre-conditioners matrices and parallelization of the Newton-Krylov (GMRES) solution of the final system of equations allows for these improvements, respectively. A test case in an atmospheric profile including ozone and scattering characteristic shows the applicability of the improvements on the principal component method for fast radiative transfer calculations, resulting in a considerable reduction of the computational burden.