Flood risk curve development with probabilistic rainfall modelling and large ensemble climate simulation data: a case study for the Yodo River basin

Abstract

A flood risk curve is the relation between annual maximum economic damage due to floods and its exceedance probability, which provides useful information for quantitative flood risk assessment. This study proposed to examine the applicability of d4PDF, a large ensemble climate projection dataset, to develop a probabilistic flood risk curve for the Yodo River basin (8,240 km²), Japan. The d4PDF is a climate dataset under historical and 4 K increase conditions with tens of ensembles and provide a physically-based and reliable estimation of ensemble flood risk curves and their future changes. We identified that d4PDF rainfall data has bias for the spatial variability of rainfall probably due to coarse spatial resolution, while not for basin-averaged rainfall. This typical type of bias was removed by incorporating basin-averaged rainfall of d4PDF and observed spatial pattern of rainfall into analytically-based probabilistic rainfall modelling. Derived ensemble flood risk curves provided a histogram of T-year flood damage. The histogram had a long tail and showed that T-year flood damage may be larger than its deterministic estimate located at the median. Estimated ensemble flood risk curves at present/future climates showed a clear increase of flood risk and its uncertainty at 4 K increase scenario.