Research on objective correction techniques for precipitation forecasts based on optimal zero-rain thresholds and frequency matching in Ningxia

This study aims to improve the precipitation forecast accuracy of CMA-MESO in Ningxia Province. Using 24-hour precipitation data from national/regional stations, gridded precipitation observations, and CMA-MESO forecasts (2020-2023) as training datasets, optimal zero-rain threshold schemes (annual, monthly, terrain height-based, precipitation area-based, and comprehensive) and frequency matching schemes including the graded linear function (FM1) and non-graded Gamma function (FM2) were designed to carry out combined correction experiments, with operational verification conducted in 2024. The results are as follows. (1) The comprehensive optimal zero-rain threshold scheme performs best, with its rain/no-rain accuracy increased by 4.2% and 2.8% for forecasts initialized at 20:00 BT and 08:00 BT, respectively, followed by the monthly and precipitation area-based schemes, while the annual scheme performs worst. (2) The single-factor optimal zero-rain threshold schemes tend to excessively eliminate a large number of true precipitation samples, which significantly reduce the TS score (≥0.1 mm). In contrast, the comprehensive scheme fully considers the differences in precipitation false alarms and effectively avoids the shortcomings of such single-factor schemes, and its TS scores (≥0.1 mm) remain nearly unchanged after correction. (3) Both FM1 and FM2 exhibit good correction performance for light and moderate rainfall. FM1 performs better for forecasts initialized at 20:00 BT, especially for heavy precipitation, and FM2 is superior for forecasts initialized at 08:00 BT.