The Application of an Evolutionary Algorithm to the Optimization of a Mesoscale Meteorological Model

Abstract

It is shown that a simple evolutionary algorithm can optimize a set of mesoscale atmospheric model parameters with respect to agreement between the mesoscale simulation and a limited set of synthetic observations. This is illustrated using the Regional Atmospheric Modeling System (RAMS). A set of 23 RAMS parameters is optimized by minimizing a cost function based on the root-mean-square (rms) error between the RAMS simulation and synthetic data (observations derived from a separate RAMS simulation). It is found that the optimization can be done with relatively modest computer resources; therefore, operational implementation is possible. The overall number of simulations needed to obtain a specific reduction of the cost function is found to depend strongly on the procedure used to perturb the ?child? parameters relative to their ?parents? within the evolutionary algorithm. In addition, the choice of meteorological variables that are included in the rms error and their relative weighting are also found to be important factors in the optimization.