A team from North Carolina State University has developed a novel tool that uses satellite imagery and machine learning to map urban flooding, addressing a critical need as climate change intensifies storms. The tool, detailed in the journal Natural Hazards, uses open-source satellite data and machine-learning algorithms to predict flood-prone areas in urban environments.
Traditional methods struggle in urban areas due to challenges like building shadows and complex drainage systems, which complicate water flow prediction. The new model, trained on data from Hurricane Ida's 2021 floods, overcomes these issues by utilizing detailed satellite imagery and a Random Forest algorithm. The model's accuracy was validated against Federal Emergency Management Agency (FEMA) flood zones, revealing that many minimal hazard zones experienced more flooding than the higher-risk 500-year flood zones, likely due to their larger area.
The team aims to refine the tool further, adding features like flood depth mapping, and plans to make the code open-source to aid emergency response planning. This innovation promises to enhance urban flood preparedness and resilience planning significantly.
