FloodSim Sandbox: An Immersive Interactive Simulation Framework for Urban Flood Risk Management

Eveline Gonzalez, Ramteja Sajja, Yusuf Sermet, Ibrahim Demir

FloodSim Sandbox is an AI-augmented, immersive digital twin framework for real-time flood visualization, analysis, and decision support. Developed within Unreal Engine 5, it integrates multi-source geospatial data, physically based fluid simulation, and multimodal AI reasoning to model flood dynamics in a high-fidelity digital twin of Iowa City. The system procedurally generates terrain and infrastructure using CityEngine and OpenStreetMap data, employs the Fluid Flux plugin for hydrodynamic simulation, and incorporates a HAZUS-based damage model calibrated with FEMA flood map data. A multimodal AI subsystem interprets visual and quantitative simulation data to deliver scene-specific risk assessments, mitigation strategies, and explainable insights. Interactive visualization, including responsive human and vehicle game characters, enhances engagement and supports scenario-based exploration of flood behavior. A user study for framework evaluation with environmental professionals confirmed the system’s usability and effectiveness compared to traditional 2D flood information tools. Collectively, FloodSim Sandbox provides capabilities for enhancing flood risk communication, participatory education, and adaptive planning by uniting simulation, visualization, and AI-driven analysis within a single digital environment.