A Strategy for High-Resolution Ensemble Prediction Part I: Definition of Representative Members and Global-Model Experiments