Characterizing spatiotemporal dynamics of hourly extreme precipitation in China with an adaptive 3D clustering approach

Study region: China, spanning the East Asian monsoon region and its northwestern inland transition zones, where extreme precipitation events (EPEs) show strong spatial heterogeneity and pose substantial hydrological-disaster risks. Study focus: This study investigates the spatiotemporal evolution characteristics and propagation patterns of EPEs across China during 1985–2024. The central objective is to move beyond station-based, two-dimensional, or daily-scale descriptions and reconstruct complete precipitation-system trajectories in a continuous three-dimensional space. To achieve this, an adaptive 3D spatiotemporal clustering framework based on 3D-DBSCAN was developed and applied to ERA5 hourly precipitation data. Extreme precipitation grids were mapped into a standardized longitude-latitude-time space, while k-distance-based parameter optimization and physical constraint filtering were used to reduce subjective parameter selection and avoid artificial fragmentation of fast-moving or intermittent events. Precipitation-weighted centroids were then used to quantify event movement distance and direction, and to classify typical propagation patterns. New hydrological insights for the region: The framework identified independent EPEs and revealed five dominant propagation patterns. EPEs in China show a clear northwestward movement preference, whereas southward events are infrequent and spatially limited. Northwestward events, especially ultra-long-distance events, exhibit the greatest hydrological extremeness, with longer durations, larger affected areas, and higher cumulative magnitudes. These trajectories indicate a regional “westward shift and northward leap” of extreme-rainfall activity, linked to the East Asian summer monsoon, western Pacific subtropical high, and mid-latitude westerlies.