A high-performance Coupled Human And Natural Systems (CHANS) modelling framework for flood risk assessment and emergency management

Urban flood risk has surged in recent years due to unsustainable urban development, changes of hydrological processes and frequent occurrence of extreme weather events. Addressing this challenge requires capturing the dynamic interactions between human and natural systems. This study presents an innovative Coupled Human And Natural Systems (CHANS) modelling framework which integrates high-performance hydrodynamic and agent-based models to simulate real-time flood-human interactions at high spatial resolution. The framework is enhanced with a reinforcement learning (RL) module to support AI-guided flood risk management, including optimal resource allocation during emergencies.

Applied to the 2015 Desmond flood in the Eden Catchment (UK) and urban flooding in Can Tho City (Vietnam), the CHANS framework demonstrates its capacity to replicate household-level responses and assess flood mitigation strategies, such as early warnings, sandbag distributions, temporary flood defence and mobile pump deployments. Results show that early warnings combined with temporary defences reduced inundation by 30% in Carlisle, saving up to £30 million. RL-guided mobile pump strategies in Can Tho outperformed traditional methods, improving flood mitigation efficiency by up to 4× during post-flooding events.

By incorporating human behaviour, decision-making, and AI optimisation, the CHANS framework provides a robust tool for enhancing flood risk management strategies, contributing to more resilient and adaptive disaster response planning.