To meet the demand of high precision real-time precipitation forecast for intelligent operation and control of reservoirs, a precipitation ensemble forecast method determined by multi-dimensional evaluation system was proposed. First, a mesoscale numerical weather forecast model covering microphysical, cumulus, longwave/shortwave radiation and other physical parameters was established. Second, error index, deviation score, accuracy index and more than a dozen evaluation indicators were introduced to form a comprehensive evaluation method for precipitation forecast, which integrated the dimensions of precipitation classification, precipitation and spatio-temporal distribution. Taking a typical precipitation event in Mei River Basin as an example, the ensemble prediction model was verified by combining the evaluation results of multi-dimensional scale precipitation forecast. Results show that the ensemble forecast is much more stable and can reduce the model forecast error by taking the parameterized scheme combinations as ensemble forecast members, which are optimized by multi-dimensional precipitation forecast evaluation system. © 2022 Huazhong University of Science and Technology. All rights reserved.
