Multi-dimensional radiative transfer computations using a scalable parallel implementation of the discrete-ordinates method

A scalable parallel implementation of the discrete-ordinates method for the solution of the radiative transfer equation is discussed. The approach presented uses a parallel architecture to solve the radiative transfer equation in three-dimensional Cartesian geometries. Two numerical experiments are presented. The first examines the effect of grid size and number of processors on computational time for a given problem, whereas the second experiment examines the effect of quadrature order on computation times. For both experiments, the parallel algorithm is used to compute upwelling brightness temperatures from a precipitating cloud. Using the parallel model, brightness temperature fields for three-dimensional precipitating atmospheres can be computed much more quickly than by a serial model.<
>