Analysis of the nonstationarity characteristics and future trends of flood extremes in the Dongting Lake Basin

Abstract

Study region: Dongting Lake Basin of China. Study focus: This study aimed to analyze the nonstationary characteristics and future trends of flood extremes in the Dongting Lake Basin. The spatiotemporal variations in the flood extremes that occurred over the past 60 years were explored using the trend-free prewhitening-MK (TFPW-MK) method, the nonstationary characteristics of flood extremes were detected using the Pettitt test, and the nonstationarity analysis was performed using the GAMLSS model. Furthermore, a combination of TFPW-MK and the Hurst exponent was employed to predict the future trends in flood extremes. New hydrological insights for the region: (1) Evident variations were observed in the flood extremes from most hydrometric stations. Among the 10 stations with a decreasing extreme trend, seven exhibited evident decreases. Such decreases probably result from the impact of water conservancy projects. Among the remaining six stations with an increasing extreme trend, two exhibited evident increases. Such increases probably result from the effect of highly extreme precipitation. (2) Flood extremes from most (9 of 16) stations showed nonstationarity. The lognormal distribution was the optimal distribution of extreme values for nonstationary stations, whereas the Gamma distribution was the optimal distribution of extreme values for stationary stations. Despite the preferable fitting efficiency of the GAMLSS model, its simulation performance for nonstationary stations that showed significant trends must be improved. (3) The flood extremes of the 14 stations will maintain existing trends in the future. The Xiangtan Station exhibited anti-persistence and might show a decreasing trend in the future, whereas Shimen Station did not show persistence.

Keywords

• Flood extreme
• Nonstationarity
• Future trend
• GAMLSS model
• Dongting Lake Basin