基于BMA-QM的福建区域降水预报订正及暴雨检验研究

Precipitation Forecast Correction and Rain-storm Test in Fujian Province Based on Bayesian Model Averaging-Quantile Map-ping Method

摘要: 以全球交互式大集合预报系统提供的六种模式模拟的福建省及其周边区域2019年至2020年的4月至6月的日降水数据结果作为降水预报资料，以国家气象信息中心多源融合降水格点产品作为降水观测资料。对降水预报资料进行贝叶斯模型平均(简称BMA)方法订正后，对BMA方法订正结果使用分位数映射法(简称QM)进行再订正。结果表明：1) BMA方法订正后，提升模式在晴雨和小雨的预报技巧，但在中雨和大雨中未体现出优势。2) BMA-QM方法订正后，保持晴雨处的预报技巧之余，提升了小雨处的预报技巧，尤其提高了大雨处的预报技巧。3) BMA-QM方法订正之下，暴雨TS (0.163)和暴雨ETS (0.147)的提升率分别为98.30%、108.16%。BMA-QM方法订正可以提升模式对强降水的预报技巧，这将对做出更为准确的暴雨预警有着重要的意义。

Abstract: The 24-hour cumulative precipitation data from April to June in Fujian Province and its surround-ing areas from 2019 to 2020 simulated by six models provided by TIGGE are used as precipitation forecast data. The grid precipitation observation data are provided by CMPA. After the precipitation forecast data were revised by Bayesian model averaging, the results of BMA were revised using quantile-mapping. The results show as follows: 1) After BMA revision, the forecasting skills of mod-els in clear-rainy and light rain have been improved, but there is no improvement in moderate rain and heavy rain. 2) After BMA-QM revision, the forecasting skill of light rain, especially heavy rain, are improved while maintaining the forecasting skill of clear-rainy. 3) After BMA-QM revision, Thereat Score (0.163) and Equitable Threat Score (0.147) in rainstorm increased by 98.30% and 108.16% respectively. The correction of BMA-QM can improve the prediction skills of heavy precip-itation, which is of great significance to make more accurate rainstorm warning.