Unraveling spatiotemporal distribution of extreme precipitation in the southern Tibetan Plateau: Synergistic effects between atmospheric circulation and topography

This is the authors' abstract. We don't add key points for this paper.

The Yarlung Zangbo River Basin. Amid global climate change, the intensity of extreme precipitation across the Tibetan Plateau has increased. However, accurately forecasting the spatiotemporal patterns of extreme precipitation in high-altitude basins with complex terrain remains challenging. This study selects the Yarlung Zangbo River Basin in the southern Tibetan Plateau as a case study and analyzes the spatiotemporal trends of extreme precipitation intensity from 1961 to 2022. A convolutional neural network–long short-term memory model incorporating dynamic atmospheric circulation indices and static topographic characteristics is developed to predict monthly spatiotemporal variations in extreme precipitation intensity. The results indicate that the monthly maximum 1-day precipitation (Rx1day) and the monthly maximum 5-day precipitation (Rx5day) extreme precipitation intensity indices exhibit overall non-significant increasing trends across the basin, although significant upward trends are observed in the central and eastern regions. Regarding model performance, the average Nash–Sutcliffe efficiency for spatiotemporal predictions of Rx1day and Rx5day are 0.62 and 0.67, respectively, while the corresponding Pearson correlation coefficients reach 0.78 and 0.81, demonstrating satisfactory predictive accuracy. Simulation results reveal that atmospheric circulation indices combined with high-resolution topography significantly improve the model’s predictive accuracy, increasing the average Nash–Sutcliffe efficiency for Rx5day by over 6 %. High-resolution topographic data enhance the model’s ability to capture spatial features, thereby improving prediction accuracy. This study establishes an innovative framework for predicting extreme precipitation in high-altitude basins with complex terrain, offering important implications for regional disaster prevention/mitigation and water resource management. • Analyzes spatiotemporal trends of extreme precipitation in the Yarlung Zangbo River Basin. • Develops a deep learning model for spatiotemporal prediction of extreme precipitation. • Circulation indices and topographic data improve the accuracy of the model.