Implement Technology of Optimal Threat Score Correction Method for Numerical Model Precipitation Forecast

This study focuses on enhancing the accuracy of 12-h cumulative precipitation forecasts from the model derived from the European Centre for Medium-Range Weather Forecasts using Fujian, Jiangxi, Zhejiang, and Shanghai as the research areas. It introduces three technologies either cited or initiated by Fujian to implement the OTS (Optimal Threat Score) correction method and compares them with similar technologies. Results are summarized in the following points: (1) A 3-year quasi-symmetric sliding window sampling method was used to collect training samples for calculating the OTS correction threshold. This approach proved to be more effective than collecting samples from the same season over the previous 3 years. (2) The magnitude of the OTS correction threshold F₁ (i.e., elimination threshold) is closely related to the 2-m temperature. As the temperature rises, the elimination threshold first increases and then decreases. Grouping modeling based on the maximum 2-m temperature forecast yielded an elimination threshold under different temperature conditions. This method simultaneously reduced the false alarm ratio and missing ratio for light rain, improving the equitable threat score for light rain by 5.0%–8.2%. (3) The study compared two schemes for interpolating model precipitation forecasts before applying the OTS correction method. The first scheme employed the inverse distance weighted interpolation method, which outperformed the second scheme using the nearest neighbor interpolation method.