Flood risk analysis of reservoirs based on full-series ARIMA model under climate change

Abstract

• A novel full-series ARIMA stochastic model has been built. • This model can reflect the statistical characteristics of past and future floods. • The BMA method was utilized to analyze the uncertainty of flood frequency curves. • More comprehensive and reliable flood risks considering the weight of each GCM have been obtained. To assess the flood risks of reservoirs in a changing environment, this study considers the uncertainty of future floods and argues that the future flood projected by Global Climate Models (GCMs) is only one of the possibilities. This possibility contains some main statistical characteristics of future floods. The autoregressive integrated moving average (ARIMA) model was used to extract such statistical characteristics of the future flood series. Moreover, the initial values of the model were assumed to obey the probability distribution derived from historical floods. Hence, a novel full-series ARIMA stochastic model which can simultaneously reflect the statistical characteristics of past and future floods has been built to simulate the possible future extreme floods. Synthetic flood series under different future climate scenarios were randomly generated by the proposed model. Then the flood risks exceeding the design flood and check flood can be obtained. The Bayesian model averaging (BMA) method was used to weight the flood risks projected by each GCM to better incorporate the uncertainty of GCMs. The Yangfanggou Reservoir in Yalong River basin of China was selected as a case study. The future flood risks under three Representative Concentration Pathways (RCPs) were assessed. The results highlight that future flood risks have an increasing trend. Under future RCP2.6, RCP4.5 and RCP8.5 scenarios, the risk rates of exceeding design flood are in the range of 0.63%–0.74%, 1.31%–4.58% and 2.63%–5.23%, respectively. The risk rates of exceeding check flood are in the range of 0.20%–0.25%, 0.41%–2.14% and 0.97%–2.48%, respectively. This study provides a comprehensive evaluation of flood risk and could be useful when planning to fortify existing structures and update design standards.