Abstract:

From July 27 to August 2, 2023, an extreme heavy rainfall event occurred in North China (referred to as the “23·7” event), and significant discrepancies are found among different numerical weather prediction models. In this study, three global models and seven regional models are selected to systematically evaluate the forecast performance for this heavy rainfall event under complex terrain conditions, based on a categorical analysis of 3-hourly accumulated precipitation, combined with precipitation verification and surface wind field analysis. The results show that the global models perform relatively well in predicting precipitation of ≥0.1 mm and ≥1 mm, but with significant overforecasting, and exhibit obvious underprediction for precipitation of ≥10 mm. The regional models exhibit superior overall forecasting performance to global models, especially for precipitation of ≥10 mm, for which their forecasting skill is significantly higher. Model resolution is found to exert a considerable influence on the prediction of rainfall at different intensities. Better performance is achieved by high-resolution models compared with low-resolution models, particularly for precipitation of ≥5 mm, though exceptions do exist. With regard to the fitting of precipitation frequency distribution against precipitation intensity, regional models are demonstrated to be overall distinctly superior to global models. Analysis based on the Froude number (Fr) reveals the synergistic mechanism between terrain and wind fields. Overestimation of wind speed or underestimation of terrain height in models leads to the misrepresentation of flow-around as orographic updraft, inducing biases in the predicted precipitation location. In contrast, the characteristics of orographic lifting or flow around can be better preserved in regions with reasonable local Fr values.

Key words: extreme precipitation, forecast evaluation, topography, wind field characteristics

摘要：

2023年7月27日—8月2日，华北地区发生特大暴雨（简称“23·7”过程），不同数值预报模式预报差异显著。本研究选取3个全球模式、7个区域模式，对逐3 h降水量开展分级研究，结合降水检验与地面风场分析，系统评估全球与区域模式在复杂地形条件下对本次暴雨的预报性能。结果表明：全球模式对≥0.1 mm、≥1 mm量级的降水预报较准，但空报明显，对≥10 mm量级的降水有明显漏报；区域模式总体预报能力优于全球模式，尤其对≥10 mm量级的降水预报能力显著高于全球模式。模式分辨率对不同量级降水预报能力具有一定影响，高分辨率模式比低分辨率模式具有更好的预报性能，尤其是对≥5 mm量级的降水预报效果更佳，但也存在例外。在降水频次随降水强度分布的拟合方面，区域模式性能整体明显高于全球模式。基于弗劳德数Fr（Froude number）分析揭示地形-风场协同作用机制，模式风速偏大或低估地形高度会将绕流误报为越山气流，导致降水落区预报偏差，而在局部Fr值合理区域能较好维持越山或绕流特征。

关键词: 极端降水, 评估检验, 地形, 风场特征