多源融合降水实况分析产品在山东一次强降水过程中的质量评估

Quality Assessment of Multi-Source Merged Precipitation Live Analysis Product in a Heavy Precipitation Process in Shandong

摘要: 利用山东省地面观测小时降水数据，针对2021年发生在山东地区的一次强降水过程，从降水时空特征、降水分量级检验等多个方面，评估分析了1 km逐小时降水实时多源融合实况分析产品的数据质量。结果表明：1) 融合降水产品质量较高，能较好地反映降水落区与降水变化趋势，整个降水过程的平均绝对误差为0.21 mm，均方根误差为1.04 mm，相对偏差为3.12%，相关系数达0.97。2) 在降水发生初期和末期，融合降水产品与观测降水的相对偏差较大、相关性较差，在降水较强的时段降水产品与观测降水更接近。3) 融合降水产品在弱降水量级的技巧评分和命中率最好，但对弱降水存在明显的高估；对于大值降水，降水产品的命中率较低，并存在低估大值降水的现象。

Abstract: Based on the hourly precipitation data of ground observation in Shandong Province, the 1 km hourly precipitation real-time multi-source merged live analysis product is evaluated in a heavy precipitation process in Shandong in 2021 from the aspects of precipitation spatial-temporal characteristics and precipitation component level inspection. The results show that: 1) The quality of merged precipitation product is high, and it can better reflect the precipitation area and precipitation change trend. The mean absolute error of the whole precipitation process is 0.21 mm, the root mean square error is 1.04 mm, the relative deviation is 3.12% and the correlation coefficient is 0.97. 2) In the initial stage and final phase of precipitation, the relative deviation of the merged precipitation product and the observed precipitation is relatively large, and the correlation is relatively poor, while precipitation product is closer to the observed precipitation in the period of strong precipitation. 3) The skill score and hit rate of the merged precipitation product are the best in the weak precipitation level, but there is an obvious overestimation of the weak precipitation; for the large-value precipitation, the hit rate of the precipitation product is low, and there is a phenomenon of underestimating the large-value precipitation.