A SEARCH TECHNIQUE FOR CHOOSING OPTIMUM PARAMETERS IN NUMERICAL WEATHER PREDICTION

The steepest descent(or ascent)search is employed for finding optimum diffusion coefficients in T42L9G model,with a view to improving the model's computational stability or prediction accuracy.The method of the steepest descent search is first described,in which the golden section search is chosen as the fundamental one-dimensional search used in the multi-dimentional steepest descent search,and then the optimization of the diffusion coefficients is described.