A Pattern-Classification Ingredient Approach for Spring Regional Rainstorm Forecasting in Zhejiang
Mao Chengyan
Li Haowen
Fang Lu
Tong Jianping
Zhu Jun
Li Wenjuan
Zhang Chao
Abstract：Based on hourly precipitation observations from 75 national meteorological stations in Zhe-jiang Province and ECMWF ERA5 reanalysis data(0.25°×0.25° horizontal resolution)from 2010 to 2022,this study develops a pattern-classification-based ingredient approach for spring regional rainstorm forecasting.Three dominant synoptic regimes were identified:stationary front rainband,warm-sector shear line,and trough-cold front patterns.For each regime,key environmental predictors were selected,and the random forest algorithm was applied to identify critical ingredients and construct a composite in-gredient index,establishing regime-specific rainstorm forecast models.The models were evaluated through historical hindcast and real-time verification during spring 2023-2024,and compared against ECMWF,GRAPES_GFS,and Zhejiang Provincial OCF(Objectively Corrected Forecast)ensemble products.Results demonstrate that the"pattern-classification first,ingredient-selection second"strategy significantly improves rainstorm hit rates while reducing false alarm ratio and miss rate.In real-time fore-casting during 2023-2024,the models achieved a 95.7%hit rate,with substantially lower miss rates and significantly higher Threat Scores than the reference products.
Keywords：regional rainstormspattern-classification-based ingredient methodrandom forestforecast verification
Publication Date：2026-05-30
Online Publishing Date：2026-09-12(First online date of this platform, not the publication date of the document)
Pages：8( 9-16 )