Severe precipitation ranks among the most impactful meteorological disasters, exerting profound social and economic consequences. Forecasting heavy rainfall on an hourly scale, driven by complex multi-scale thermodynamic processes, remains a formidable challenge in numerical weather prediction (NWP). This study introduces a deep-learning-based fusion rain network (FRNet) designed to improve intense rainfall forecasting by effectively integrating the strengths of multi-scale model rain features. The model’s performance in 3-h rainfall forecasts within a 24-h lead time over North China was comprehensively validated. Statistical assessments, based on independent tests conducted for 2021 and 2022, demonstrated FRNet’s robust generalization capability. The results of the evaluations reveal that the deep learning model significantly enhances forecast accuracy across various precipitation thresholds—ranging from light to torrential rain—across all lead times within the 24-h period, when compared with the original model forecasts and the traditional multi-model similarity ensemble correction method. Notably, improvements become more pronounced with increasing precipitation intensity. Case studies of typical intense rainfall events have illustrated that FRNet effectively synthesizes the strengths of global models in predicting rainband location and movement, as well as mesoscale numerical models in forecasting intensity and fine-scale structures. This integration results in a more comprehensive and precise adjustment of precipitation forecasts, offering superior guidance for intense rainfall events compared to any single numerical model. Structure–amplitude–location (SAL) verification shows that FRNet significantly adjusts the structure and location of heavy precipitation but shows overestimation at the same time. Ablation experiments confirm that the performance gains of FRNet are primarily driven by the inclusion of high-resolution global and mesoscale model forecasts, complemented by geographic and temporal information. This study underscores the potential of deep learning techniques to integrate more multi-scale physical and observational features from different sources, paving the way for future advancements in intense rainfall forecasting.

Evaluation constitutes a crucial component in the creation and application of numerical models. As these models have escalated in complexity in recent years, there has been an increasing demand for more accurate forecasting, thereby catalyzing substantial advancements in model verification and evaluation techniques. This paper begins with a review of recent domestic and international progress in model evaluation and verification, and subsequently delves into the significant contributions made by Chinese researchers in evaluating the dynamical frameworks of next-generation numerical models, the spatiotemporal variability of the East Asian monsoon, and the assessment of weather and climate conditions around the Qingzang plateau. Given the identification of precipitation as a key process, the paper presents a comprehensive review of evaluation efforts pertaining to diurnal variations and process evolution in precipitation. The paper concludes by outlining potential future research outlooks, including the integration of multi-source observation data, the development of evaluation techniques specific to certain models and application contexts, and the synergy between model evaluation and machine learning technology.

Using CMPAS precipitation product and ERA5 reanalysis dataset, spatial characteristics of rainbelts are quantified based on the method of object-based diagnostic evaluation (MODE). The widescale rainbelts are selected and further sorted into zonal and meridional types based on different orientations. The rainbelts in the Yangtze-Huaihe river valley have been classified according to their spatial distributions. Precipitation characteristics and associated atmospheric structure have been analyzed. Comparing spatiotemporal characteristics of different rainbelts, it is found that the widescale rainbelts tend to have a southwest-northeast orientation, and the diurnal variation of precipitation amount (frequency) shows two peaks, i.e., one is in the early-morning (midnight), and the other is in the late afternoon. Precipitation intensity peaks from night to early morning. The zonal rainbelts are southwest-northeast oriented, advancing from south to north during the warm season. Precipitation is concentrated from mid-June to mid-July, corresponding to the Meiyu period. The zonal rainbelts are featured by high frequency of heavy precipitation and broad area. The diurnal variations of precipitation amount, frequency and intensity of the zonal rainbelts are basically the same as those of widescale rainbelts. The meridional rainbelts occur more frequently during the monsoon break period in August, featured by frequent precipitation over relatively small area. The diurnal variation of precipitation amount (frequency) in the meridional rainbelts peaks in the afternoon, and precipitation intensity in the late-afternoon is greater than that at night. Further comparison of the atmospheric structure associated with the two different types reveals that both rainbelts have notable baroclinic structures with warm anomalies in the mid-to-upper levels and cold anomalies in the lower levels. The anomalous circulation of wind field is characterized by an anticyclone in the upper layer and a cyclone in the lower layer. However, the zonal warm anomaly has a large temperature gradient on the north and south sides; while the the meridional warm anomaly has the gradient on the east and west sides. The zonal type is located in the center of the high-level anticyclone. The high-level divergence is conducive to the generation of upward movement, and the low-level southwesterly wind transports abundant water vapor, which is conducive to the formation of large-scale precipitation. The meridional type is located at the back of the high-level anticyclone. Southeasterly and northeasterly wind anomalies prevail in lower levels, leading to meridional spatial distribution of precipitation. Local convergence results in relatively scattered centers of heavy precipitation.

Evaluation constitutes a crucial component in the creation and application of numerical models. As these models have escalated in complexity in recent years, there has been an increasing demand for more accurate forecasting, thereby catalyzing substantial advancements in model verification and evaluation techniques. This paper begins with a review of recent domestic and international progress in model evaluation and verification, and subsequently delves into the significant contributions made by Chinese researchers in evaluating the dynamical frameworks of next-gene-ration numerical models, the spatiotemporal variability of the East Asian monsoon, and the assessment of weather and climate conditions around the Qinghai–Xizang Plateau (QXP). Given the identification of precipitation as a key process, the paper presents a comprehensive review of evaluation efforts pertaining to diurnal variations and process evolution in precipitation. The paper concludes by outlining potential future research outlooks, including the integration of multi-source observation data, the development of evaluation techniques specific to certain models and application contexts, and the synergy between model evaluation and machine learning technology.