Regional frequency analysis of extreme precipitation based on a nonstationary population index flood method

Anthropogenic climate change has led to nonstationarity in hydrological data and their statistical characteristics. To consider nonstationarity in regional frequency analysis, several nonstationary index flood (NS-IF) methods comprising a time-dependent site-specific scaling factor or nonstationary regional growth curves have been suggested. However, these methods have limitations related to underestimation from using sample statistics as a site-specific scaling factor or considering nonstationarity only in regional parameters. To overcome these drawbacks, this study developed a nonstationary population index flood (NS-PIF) method that considers nonstationarity in the statistical characteristics at each site in a region based on nonstationary generalized extreme value distributions. Monte Carlo simulations were conducted for synthetic regions under various nonstationary conditions to compare the performance of the NS-PIF method with those of existing NS-IF methods. Then the applicability of the NS-PIF method to real-world data was assessed via Monte Carlo simulations of regions with annual maximum rainfall data in South Korea. The results indicated that the NS-PIF method can solve the underestimation problem inherent in existing NS-IF methods. Moreover, the NS-PIF method yielded the best performance and provided more reliable and reasonable quantile estimates considering site-specific trends. In addition, the heterogeneity measure based on L-skewness and L-kurtosis was identified as a suitable test of homogeneity for application of the proposed method.