Abstract Anomalous precipitation, with unexpected intensity and/or spatiotemporal structures, makes flood risk management in the East Asian monsoon regions challenging, where interacting circulation systems generate highly variable precipitation events. We develop an event‐level quantitative framework based on a novel lightweight Trans‐Unet model and apply it to the Hanjiang River Basin as proof‐of‐concept using a 59‐year, 1‐km daily precipitation data set. The framework reveals four patterns: West‐type (37.0% of events), Middle‐type (25.3%), Southeast‐type (21.7%), and East‐type (16.0%). West‐type events show bimodal seasonality in July and September, with July peaks linked to ∼20% fewer September events, suggesting potential implications for adaptive reservoir regulation. A probabilistic analysis further quantifies the influence of synoptic drivers on anomalies, for example, a westward extension of the western North Pacific Subtropical High, reaching as far as Taiwan, increases West‐type frequency by ∼23%, while a northward‐displaced, intensified East Asian jet at 200 hPa over northern China enhances occurrence by ∼72%.

Read original article

Read original article