Non-stationary frequency analysis of annual extreme rainfall volume and intensity using Archimedean copulas: A case study in eastern China

Abstract

Owing to hydrological change and human activities, increasing attention is being paid to the evolution of dependence structure of hydrological characteristics, especially for non-stationary multivariate frequency analysis. In this study, we performed a non-stationary frequency analysis of annual extreme rainfall of four study regions in eastern coastal China. Two indexes separately accounting for rainfall volume and intensity were derived from daily precipitation data recorded by four pairs of gauges. In particular, based on the compatibility of data sets examined by the copula equality test, index series sharing similar dependence structures were merged to enrich data bases for copula modelling. A moving time window of 30 years was applied to the index series for non-stationarity analysis. An investigation of dependence evolution was conducted by detecting trends and change-points in the correlation coefficient series. Time-dependent Archimedean copulas and Generalized Extreme Value (GEV) distribution were employed to model the joint and marginal distributions, respectively. Kendall’s return period was employed to compute the design values of the indexes. Results showed an intensifying tendency of extreme rainfall volume and/or intensity in three of the study regions and uneven spatial and temporal distribution of precipitation. Finally, the occurrence risk of annual extreme rainfall corresponding to different joint return periods (JRPs) was analyzed based on the observed index series. The study highlights incorporating non-stationarity in multivariate hydrological analysis and merging compatible data sets to improve copula inference.