9503295 Whitaker The primary objective of this project is to develop improved methods for ensemble weather prediction. Numerical experiments with idealized models will be performed to examine the utility of different strategies for ensemble prediction. the reduced dimension of the models will allow the principal investigator and his colleagues to perform experiments with much larger ensemble sizes than would be feasible with complex, operational forecast models. Archived operational ensemble predictions produced by the National Meteorological Center will be examined to test diagnostic techniques and methods for predicting forecast skill suggested by the idealized model experiments. The research will focus on three critical topics: (1) the method of constructing the initial conditions for ensemble runs; (2) determining the optimal number of ensemble members required for useful estimates of the most probable forecast state and (3) how systematic errors in the model affect the efficacy of the ensemble prediction. This research has an enormous potential for contributing to the improvement of operational medium-range weather forecasts. ***