昆仑山北坡地形复杂，降水日变化特征独特导致其降水精细化预报难度大，准确率低。ECMWF模式的整体预报性能世界领先，但其对昆仑山北坡复杂地形下降水日变化特征的预报能力尚不明确。本文利用2020—2023年夏季自动气象站降水观测资料，检验评估了ECMWF模式对昆仑山北坡不同区域夏季降水日变化特征的预报性能。结果表明：（1） 模式20:00起报的24 h累计降水预报性能优于08:00，模式对昆仑山北坡海拔大于2000 m区域的降水预报能力优于海拔小于2000 m的区域，模式对西昆仑山北坡降水的捕捉能力优于中昆仑山北坡。（2） 模式降水量与观测降水量的日变化特征差异在17:00至次日02:00最大，模式降水频次明显多于观测值，降水强度明显小于观测值，在观测降水较少（多）时段，模式降水易高（低）估；西（中）昆仑山北坡海拔>2000 m（≤2000 m）区域的模式降水量日变化特征与观测降水量的差异较大。（3） 模式降水在西昆仑山北坡以对流性降水预报为主，在中昆仑山北坡以大尺度降水预报为主；模式降水与观测降水日变化特征的误差主要来自对流性降水预报。研究成果可为提高昆仑山北坡夏季降水预报准确率和ECMWF模式降水预报产品订正提供参考。

The northern slope of the Kunlun Mountains has a complex terrain and unique diurnal variation of precipitation, which lead to low accuracy in refined precipitation forecasting. The ECMWF model has world-leading forecast performance, but its ability to predict the diurnal variation in precipitation in complex terrain is still unclear. This study used precipitation data from automatic meteorological stations during the summer of the 2020-2023 season to verify and evaluate the ECMWF model’s forecast performance for the diurnal variation of summer precipitation at various regions on the northern slope of the Kunlun Mountains. The results show the following. (1) The 24 h cumulative precipitation forecast performance of the ECMWF model initialized at 20:00 was better than that which initialized at 08:00; the model’s precipitation forecast capability for areas above 2000 m in altitude on the northern slope of the Kunlun Mountains was better than that for areas below 2000 m in altitude; the ECMWF model’s ability to capture precipitation in the Western Kunlun Mountains was superior to that in the Central Kunlun Mountains. (2) The ECMWF model’s forecasts of daily precipitation variations exhibited the greatest discrepancies from observed precipitation between 17:00 and 02:00 the following day. The model’s predicted frequency of precipitation events was much higher than the observed frequencies, but the predicted intensity of precipitation was markedly lower the observed. During periods of low observed precipitation, the model was prone to overestimating precipitation. In regions of the western (central) Kunlun Mountains, on the north-facing slopes with elevations above 2000 m (and up to 2000 m), the model’s daily variation in precipitation significantly diverged from that of the observed precipitation. (3) The ECMWF model’s precipitation forecasts were dominated by convective precipitation (CP) in the Western Kunlun Mountains and by large-scale precipitation (LSP) in the Central Kunlun Mountains. Within the ECMWF model, LSP had a better capturing ability for observed precipitation than CP, and the discrepancies between model precipitation forecasts and observed precipitation were more likely to come from CP. These findings provide a reference for improving the accuracy of summer precipitation forecasts on the northern slope of the Kunlun Mountains and for the correction of the ECMWF model’s precipitation forecasts.