Dynamic Assessment of Flood Risk Under Varied Precipitation Scenarios

This is the authors' abstract. We don't add key points for this paper.

Floods rank among the most common yet damaging disasters worldwide, resulting in large-scale infrastructure damage, human casualties, and persistent economic and environmental harm. Flood risk assessment is vital for life-loss mitigation and socio-economic sustainability. This work establishes a comprehensive flood risk index set, including geology, topography, environment, meteorology, hydrology, demography, economy, flood-prevention infrastructure, government rescue power, medical service, disaster shelters, and public disaster risk reduction awareness, to improve the assessment rationality. Moreover, a novel combination method of TFARetNet, GTCW, AHP, and CRITIC is proposed to perform dynamic flood risk assessment under varied precipitation scenarios to improve the assessment accuracy. Poyang Lake region (19630.5 km2, including 279 towns), an area chronically affected by recurrent flooding disasters, is selected as a benchmark. The testing Accuracy, Recall, F1-score, and AUC values attain 92.39%, 96.29%, 0.9270, and 0.9238, respectively, and the risk result is validated by 126 flood events from 2010 to 2024 collected from official news reports. The risk of flooding fluctuates considerably based on rainfall amount, intensity, frequency, and duration. High risk regions are mainly distributed in the lake riparian zones and fluvial corridors. These regions feature low-lying topography, dense water networks, limited floodwater storage capacity, developed economy, concentrated high-value assets, or thriving agricultural and aquaculture industries, and deficiencies in flood defenses and rescue capacities regarding extreme precipitation episodes. In terms of the assessed risk, some risk mitigation measures are recommended, and these recommendations offer promising solutions for the government to mitigate and regulate flood risks.