Forewarning extreme precipitation events using scaling behaviors

Extreme precipitation events significantly impact ecological systems, agricultural production, and human societies across various temporal and spatial scales. However, due to their complex physical mechanisms and multi-scale interactions, the accurate simulation and forecasting of extreme precipitation remain challenging. To improve the ability to simulate and predict such events, it is essential to gain a deeper understanding of their fine-scale temporal structure and the mechanisms driving their evolution. In this study, with the observed hourly precipitation data in Eastern China, we investigate the temporal clustering behavior of hourly extreme precipitation and explore the underlying mechanisms through the scaling properties of hourly precipitation. By analyzing the probability distributions and conditional statistics of return intervals and durations, we find that, unlike daily extreme precipitation, hourly extremes tend to occur in clusters; and the time series exhibiting similar scaling properties of precipitation are able to effectively reproduce these clusters. Based on these findings, a method for forewarning extreme precipitation based on their inherent scaling behavior is suggested and it has shown comparable forewarn capabilities when compared to dynamical forecast models. We believe this study can not only deepen the understanding of the mechanisms underlying hourly extreme precipitation but also provide valuable insights for improving the simulation and early warning capabilities for regional hourly extreme precipitation.