“Bias Correction Method” for Regional Correction Experiment of Warm Season Rainstorm in Zhejiang.

 

This study employs the K-means clustering algorithm to partition warm-season precipitation in Zhejiang Province into distinct regions. A frequency matching bias correction method is then applied to enhance the spatial and temporal accuracy of the objective consensus forecasting (OCF) 0.05° × 0.05° model precipitation forecast data for the region. The corrected forecasts are subsequently evaluated against the original model outputs. The key findings include: 

(1) Zhejiang Province is optimally divided into four regions characterized by similar precipitation patterns, which exhibit distinct regional features closely linked to the province’s topography and geomorphology. The correction has the most significant impact in northwestern Zhejiang, while its effects are less pronounced in the northeastern coastal areas. 

(2) Both overall correction and regional correction improve forecast accuracy across various precipitation thresholds. Regional correction consistently outperforms overall correction, with the greatest improvements observed in the accuracy of spring rainstorm forecasts, where the TS score increases from 0.157 to 0.280. 

(3) For both systematic and convective precipitation events, regional correction enhances the accuracy of intensity and spatial predictions for rainfall exceeding the rainstorm threshold. This improvement is particularly evident in systematic precipitation, where regional correction increases the hit rate for rainfall above the rainstorm threshold from 0.336 to 0.519 and enables the prediction of extremely torrential rain events. Read the research paper!