Climate Informed Non-stationary Modeling of Extreme Precipitation in China

Recent years have witnessed climate change characterized by increasingly frequent extreme precipitation events, and the assumption of stationarity in traditional frequency analyses is gradually being questioned. In terms of the current research status in China, there is a lack of thorough investigations on the linkage between extreme precipitation and climate change. This paper aims to determine the dominant climate indices as well as the corresponding significant time scales and periods affecting extreme precipitation over China for dynamic assessments of the upcoming rainstorm risk. Correlations between 15 climate indices and precipitation extremes, as well as the correlations among climate indices, are fully explored to identify potential predictors for non-stationary modeling. Then, 21 non-stationary generalized extreme value (GEV) models are constructed, and the optimal covariates as well as their lag times with extreme precipitation at 769 stations are ascertained in a Bayesian framework. Finally, a complete predictive process is developed, and the national rainstorm risk under non-stationary conditions is assessed. The results indicate that precipitation extremes remain stationary only at 74 stations (less than 10%). WPI is dominant in modeling the variability in precipitation extremes for nearly 22% of the total stations, ranking first among all the climate indices. The predominant time scale affecting extreme precipitation at the majority of stations is 3 months. Ignoring the non-stationarity of extreme precipitation inevitably leads to misperceptions of rainstorm risks, and the spatial distribution of the maximum case of the design rainstorms under non-stationary conditions differs remarkably from that under stationary conditions. Our findings have important implications for the in-depth understanding of the real drivers of extreme precipitation non-stationary and enable advanced predictions of rainstorm risks for mitigating subsequent disasters.