Nonstationarity-based evaluation of flood frequency and flood risk in the Huai River basin, China

摘要

Using monthly streamflow data from 9 gauging stations covering a period of 1960-2014 across the Huai River basin, China, the Pettitt method and GAMLSS model were used to quantify flood frequency and related implications for flood hazards. Results indicated that: (1) no significant change points were observed in flood flows at the Huangchuan, Hengpaitou and Bengbu stations. However, significant hydrological alterations were detected for flood flows at other 6 stations with change points during similar to 2000; (2) flood flows at the Bantai, Jiangjiaji and Hengpaitou stations were nonstationary but stationary flood flows were detected at other 6 stations. The Weibull distribution was the appropriate probability distribution describing flood flows across the Huai River basin and the lognormal distribution is the second best distribution; (3) no significant difference was detected for 10- and 20-year flood frequencies using the Pearson III type distribution function under the stationarity assumption. However, significant and increasing differences were expected for 30-, 50- and 100-year floods. It can be concluded that the assumption of stationarity, if the flood flows are actually nonstationary, can produce biased flood frequencies using Pearson III type distribution, and this result provides a critical reference for flood frequency analysis under the nonstationarity and/or stationarity assumption; (4) after 30 years, the flood magnification factor at Hengpaitou and Bengbu stations increased from 1 to 1.12 and from 1 to 1.06, respectively, implying higher flood risks at the Hengpaitou station. In addition, after 30 years, the return period of 100 year flood decreased from 100 years to less 70 years at the Hengpaitou station and from 100 years to nearly 78 years, implying higher frequency of floods with return periods of 100 years. Moreover, significant relations were detected between annual peak flood flow and flood-affected crop areas, implying that annual peak flood