The life cycle of mesoscale convective systems (MCSs) in eastern China is yet to be fully understood, mainly due to the lack of observations of high spatio-temporal resolution and objective methods. Here, we quantitatively analyze the properties of warm-season (from April to September of 2016) MCSs during their lifetimes using the Himawari-8 geostationary satellite, combined with ground-based radars and gauge measurements. Generally, the occurrence of satellite derived MCSs has a noon peak over the land and an early morning peak over the ocean, which is several hours earlier than the precipitation peak. The developing and dissipative stages are significantly longer as total durations of MCSs increase. Aided by three-dimensional radar mosaics, we find the fraction of convective cores over northern China is much lower when compared with those in central United States, indicating that the precipitation produced by broad stratiform clouds may be more important for northern China. When there exists a large amount of stratiform precipitation, it releases a large amount of latent heat and promotes the large-scale circulations, which favors the maintenance of MCSs. These findings provide quantitative results about the life cycle of warm-season MCSs in eastern China based on multiple data sources and large numbers of samples.

This study demonstrates the capability of a cloud model in simulating a real-world tornado using observed radiosonde data that define a homogeneous background. A reasonable simulation of a tornado event in Beijing, China, on 21 July 2012 is obtained. The simulation reveals the evolution of a descending reflectivity core (DRC) that has commonalities with radar observations, which retracts upward right before tornadogenesis. Tornadogenesis can be divided into three steps: the downward development of mesocyclone vortex, the upward development of tornado vortex, and the eventual downward development of condensation funnel cloud. This bottom-up development provides a numerical evidence for the growing support for a bottom-up, rapid tornadogenesis process as revealed by the state-of-the-art mobile X-band phase-array radar observations. The evolution of the simulated tornado features two replacement processes of three near-surface vortices coupled with the same midlevel updraft. The first replacement occurs during the intensification of the tornado before its maturity. The second replacement occurs during the tornado’s demise, when the connection between the midlevel mesocyclone and the near-surface vortex is cut off by a strong downdraft. This work shows the potential of idealized tornado simulations and three-dimensional illustrations in investigating the spiral nature and evolution of tornadoes.

In this study, a machine learning-based quality control algorithm for heavy rainfall was developed by integrating automatic weather station observations with remote sensing data, minute-level data, and metadata. Based on heavy rainfall samples from 1 June 2022 to 31 December 2024, the performances of four gradient boosting models—eXtreme Gradient Boosting (XGBoost), Light Gradient Boosting Machine (LightGBM), Categorical Boosting (CatBoost), and Gradient Boosted Regression Trees (GBRT)—significantly outperformed precipitation-threshold-based conventional methods, including regional extreme value checks, temporal consistency checks, and others. Specifically, the XGBoost in particular achieves an increase in precision by 0.110 and recall by 0.162. This translates to a substantial reduction in both false alarms (higher precision) and missed detections (higher recall) of anomalous heavy rainfall events, thereby significantly enhancing the reliability of the quality-controlled data. The radar composite reflectivity, satellite cloud-top temperature, and minute-level precipitation were identified as dominant contributors to model predictions. The integration of multi-sensor observations effectively addressed limitations inherent in conventional threshold-based approaches. Through SHapley Additive exPlanations (SHAP)-based interpretability analysis, the model’s decision logic was shown to align with meteorological physical principles. Characteristic patterns such as combinations of low radar reflectivity and elevated cloud-top temperatures were flagged as anomalous rainfall events, typically corresponding to manual operational errors. Moreover, the model identified anomalous minute-level precipitation extremes to be critical signals for detecting instrument malfunctions, data encoding and transmission errors. The physical consistency of the model’s reasoning enhances its trustworthiness and supports its potential for operational implementation in heavy rainfall quality control.