A flash flood sweeps through a city street, carrying debris that can knock people off their feet or trap them against obstacles—yet current evacuation models ignore this danger entirely. FAST will build the first computer model that simulates how people behave during flash floods when waterborne debris is present. Existing tools treat evacuees as simple particles moving through water, ignoring how floating objects like cars, bins, or construction materials alter both the flow and people’s decisions. The project combines three approaches: physics-based models of water and debris movement, virtual reality experiments that place volunteers in simulated flood scenarios to record their real reactions, and agent-based models that use machine learning to predict crowd behaviour from those experiments. If successful, the open-source tool could help emergency planners design evacuation routes and public warnings that account for debris hazards—not just water depth and speed. This matters because debris often causes injuries and slows escape in ways current models cannot predict. The research is applied: it directly targets a gap in flood risk management, with a case study to verify the model against real data.
View original technical description
Flash floods in urban areas are particularly dangerous extreme events, given their tragic impacts on human life and significant social and economic damage. Evacuation of citizens during these events is crucial to mitigate their impact. Effective evacuation strategies from flash floods can be developed through the understanding of the interactions among all the components of the process: the individuals, the flow, the waterborne debris and the urban environment. The impact of waterborne debris on the behaviour of individuals has not been addressed by existing studies and modelling tools. FAST aims at reducing human losses through improved evacuation strategies possible thanks to the innovative combination of physics-based, virtual reality and agent-based modelling approaches. This aim will be achieved thanks to the understanding and modelling of the complex interactions among the flow, the waterborne debris, and evacuees. To this end, FAST will develop the first computational methodology that combines reliable simulations of human behaviour and the physical aspects of flash floods, including waterborne debris, to provide the tools to improve evacuation and emergency response strategies during these events. FAST will exploit the innovative capabilities of virtual reality environments to conduct an experiment that will allow the development of an agent-based model of the evacuees' behaviour in the presence of waterborne debris using data-driven and machine learning methods. This model will be verified using a case study for which data on the flow and human behaviour will be collected. It will be made available as open-source code and combined with a hydrodynamic model capable of tracking waterborne debris to be used as a comprehensive simulation tool. This research will allow the fellow to grow as an independent interdisciplinary researcher thanks to training in novel methodologies in virtual reality and agent-based modelling applied to extreme hydrodynamic events.
Plain English summaries and category classifications on this site are generated by AI and may not perfectly reflect the original research.
Is something wrong? Let us know