Deep Learning Reveals Circulation-Driven Intensification of Extreme Precipitation in the Lower Yangtze River Basin

In recent decades, the lower Yangtze River Basin has experienced a notable rise in both the frequency and intensity of extreme precipitation events (EPEs). In this study, we employ a convolutional neural network (CNN) to identify circulation patterns associated with EPEs, referred to as extreme precipitation circulation patterns (EPCPs). The CNN effectively captures key atmospheric features, including a southward-shifted upper-level westerly jet stream, a deepened East Asian trough, and a southwestward extension of the western North Pacific subtropical high (WNPSH). These circulation changes promote increased water vapor and vertical ascent in the region, providing favorable conditions for EPEs. Over recent decades, EPCP occurrence has increased, and precipitation on EPCP days has intensified. These trends pertain to frequency and EPCP-day moisture transport. Moisture budget analysis reveals that the intensification of EPEs on EPCP days is primarily driven by strengthening horizontal dynamical moisture advection on EPCP days, which is closely linked to the intensification of the WNPSH and a southward weakening of the upper-level westerly jet. Our findings highlight the critical role of dynamically driven moisture transport in enhancing EPEs over the lower Yangtze River Basin in a warming climate.