Application and evaluation of a multi-model dynamic-weighting ensemble precipitation forecast method in Yunnan

To address the systematic biases in numerical model precipitation forecasts over the complex terrain of Yunnan, this study develops and evaluates a forecast method combining the Frequency Matching Method (FMM) with a multi-model dynamic-weighting ensemble approach. Independent sample tests for 2024 show that the ensemble forecast effectively overcomes the systematic biases of individual models by accurately simulating the precipitation amplitude. Compared to individual models, it achieves higher rain/no-rain accuracy, optimal Threat Scores (TS) and Bias scores across all precipitation levels, and a significantly lower false alarm ratio. Case studies further confirm its advantages in forecasting the location and intensity of heavy precipitation. This integrated method provides an effective pathway for enhancing the forecast skill for precipitation, particularly for critical heavy rainfall events, in the complex terrain of Yunnan.