Analysis of S2S Multi-models Forecasting Performance for Persistent Heavy Precipitation Events During May-September over South China

CHI Yanzhen; LUO Guanting; WU Zhengqiu; HE Fen

doi:10.16032/j.issn.1004-4965.2025.039

Abstract：This study develops grid-based objective identification criteria for persistent heavy precipitation events(PHPEs)in South China during May-September using daily precipitation observations from 2 407 stations across China(1961-2015).We then evaluated extended-range forecast skill using a Comprehensive Test Score(CTS)and improved dichotomous metrics,assessing multi-model ensemble means constructed from different proportions of top-ranked members based on the reforecast data from four operational models in the World Meteorological Organization's(WMO)Subseasonal-to-Seasonal(S2S)Prediction Project-the European Centre for Medium-Range Forecasts(ECMWF),China Meteorological Administration(CMA),United Kingdom Met Office(UKMO),and National Centers for Environmental Prediction(NCEP).The results show that South China experiences an average of 4.5 PHPEs per year,with significant interannual variability ranging from a minimum of 2 to a maximum of 8 events.The duration is mainly 3-7 days,and the longest duration can be up to 19 days with an increasing trend.The CTS,threat score(TS),and equitable threat score(ETS)decrease as more ensemble members in descending order are included in the mean.Among individual models,NCEP performs best in terms of CTS,while UKMO shows slightly better performance than ECMWF and NCEP according to TS and ETS.The CMA model exhibits the lowest skill across all metrics.The multi-model ensemble achieves the highest forecast skill when combining the top 50%of ensemble members.Additionally,all-ensemble-mean forecasts show higher scores for heavy precipitation events with longer durations.The results can provide valuable insights for real-time monitoring and subseasonal forecasting of PHPEs.