An agent-based modeling method to determine the number of flash flood warnings for improving the warning response levels

The effectiveness of flash flood warnings depends on people’s responses to the warnings. Although the impacts of the number of warnings ( NOW ) on the warning response levels have been explored by statistical methods, a simulation method is required to determine the NOW for improving the warning response levels. An agent-based model (ABM) has been proposed to quantify the relationship between the warning response levels and the NOW, and then a robust decision making has been used to determine the robust NOW due to the deep uncertainties of ABM parameters. Our proposed method was applied to Liulin Town, China, which experienced a serious flash flood disaster in 2021. The results show that the robust NOW gradually increases from six to eight times and stabilizes at eight times as the lead time of the flash flood warnings increases from 120 min to 300 min. The robust NOW is the highest for a warning frequency changed from low to high, followed by a constant warning frequency, and finally a warning frequency changed from high to low. The applicable scopes of the robust NOW are also figured out in terms of parameter ranges. The proposed method provides valuable insights into the determination of NOW for improving the warning response levels and the effectiveness of flash flood warnings.