Flood risk analysis integrating of Bayesian-based time-varying model and expected annual damage considering non-stationarity and uncertainty in the coastal city

Abstract

Flood disaster is more serious in coastal cities due to the combined impact of rainfall and tides. Accurate assessment of coastal flood risk is essential for planning effective and targeted adaptation under changing environment. The objective of the study is to propose an integrated framework for future flood risk assessment by Bayesian-based time-varying model and expected annual damage (EAD) in the coastal city. To decrease the uncertainty of non-stationary frequency, Bayesian Model Averaging (BMA) and time-varying parameter distribution (TVPD) models were employed to establish the non-stationary distributions of rainfall and tides, and copula function was adopted to determine the joint and co-occurrence probability. Subsequently, to reflect flooding probability and inundation damage simultaneously, the EAD was applied to quantify flood risk by copula function and hydrodynamic model. The variation and uncertainty of flood risk were also investigated under changing environment. Taking Haidian Island in Hainan Province, China as a case study, the results show that the non-stationary distributions of rainfall and tides can be appropriately derived based on BMA and TVPD models. The joint and co-occurrence probabilities increase significantly under non-stationary scenario with the average rates of 33.22 % and 64.82 %, respectively. Moreover, the EAD will be underestimated by 20.56 %–69.84 % in 2030–2060 year without considering non-stationarity, and the uncertainty of EAD rises with the increase of the design year. The approach and result of our study can help decision makers evaluate the future flood risk in the coastal city, and provide support for sustainable flood management to adapt the climate change.