Multiple Models Forecast Evaluation of Three Typhoon-induced Heavy Precipitation Events over Zhejiang Province

HE Mu; YU Zhenshou

doi:10.16032/j.issn.1004-4965.2025.070

Abstract：The heavy precipitation which caused by typhoon may lead to huge societal and economic losses,especially in the coastal regions of southeastern China.Based on the hourly rainfall data from 96 national stations in Zhejiang province,the performance of six regional and three global operational numerical weather prediction models for the typhoon-induced heavy precipitation during the 2021-2023 was evaluated.(1)For 24-hour accumulated precipitation,regional models demonstrate a higher Equitable threat score(ETS)and comprehensive scores in rainfall(greater than 0.1 mm)forecast than global models.However,obvious differences exist in torrential rain forecastg.ECMWF showed the best ETS for the In-fa and Doksuri processes,while CMA-TRAMS9 performed best for the Muifa.(2)Precipitation with an intensity greater than 10 mm·(3h)-1 was the major contributor to accumulated precipitation.The ETS for 3-hour heavy precipitation showed that regional models with resolution of 3 km slightly outperform the other models,and CMA-MESO3 had the best skill during the In-fa and Doksuri.(3)Both global and regional models overforecasted precipitation on eastern coast of Zhejiang during the In-fa and Doksuri.This positive bias was contributed to by both overestimated precipitation frequency and intensity.All models reasonably reproduced the early morning peak of precipitation amount before the landfall of In-fa,but the bias of the peak hour after the landfall of In-fa varied greatly between models.Among them,ECMWF,CMA-MESO3 and CMA-BJ9 were more accurate in predicting the main peak hour after landfall.In contrast,the forecasts of the diurnal variation were generally poor for all models before and after the landfall of Doksuri,with the exception of CMA-GD3 after landfall.The results provide users with the bias features and accuracy for the precipitation forecast among operational models,which may be helpful in the improvement of model forecast skill and weather forecasting services.