Rapid Flash Flood Inundation Mapping for Planning and Emergencies

Over the coming decades, Flash Flooding could claim thousands of lives and cost billions of dollars due to climate change - with $2.2 billion in property damage to the US in 2022 alone. Flash floods are dangerous as they occur with little warning due to how quickly they travel and how rapidly water levels rise. Traditional flood models take weeks to run, making them unsuitable for real-time flood forecasting. Since 2022, US states and municipalities have used our ML-based Flood Predictor, built entirely on Databricks, to get real-time insight into flood dangers to their communities.

To accomplish this, we trained a U-net model on FEMA flood studies and then paired the model with terrain data to generate maps of water depths. The output allows communities like the State of Tennessee to determine the vulnerability of power plants and other assets to flash flooding, creating better and more informed emergency plans - saving lives and curbing disaster in real communities.