Multi-model Integration Precipitation Forecasting Based on TCH Theory: Recent Advances and Challenges

The rapid development of numerical weather prediction (NWP) models offers new opportunities for improving quantitative precipitation forecasting, while raising challenges in objectively integrating multi-model forecasts. This study presents recent advances in an operational multi-model integration precipitation forecasting method based on the generalized Three-Cornered Hat (TCH) theory.Seven NWP models routinely operated at the National Meteorological Center of the China Meteorological Administration are considered, including ECMWF, GERMAN, NCEP, GRAPES_3KM, BEIJING_MR, GUANGZHOU_MR, and SHANGHAI_MR. The method applies TCH theory to estimate the relative error characteristics of precipitation forecasts from different models. A Bayesian framework is then used to derive objective, model-dependent weighting coefficients, enabling short-range multi-model integration forecasts.The integration performance is evaluated using Threat Score (TS) metrics for 2025. Results show that the TCH-based integration consistently outperforms the single ECMWF model across all precipitation categories. The 24-hour heavy rainfall TS reaches 0.2357, a 48% improvement, while the TS for extreme rainfall events reaches 0.1354, a 141% improvement relative to ECMWF.The multi-model integration products have been operationally implemented at the National Meteorological Center, providing critical support during high-impact weather events, highlighting both recent advances and remaining challenges in operational multi-model precipitation forecasting.