Presentation | H31M: Advancing Flood Characterization, Modeling, and Communication IV Poster
Poster
H31M-1235: Convolutional Neural Network Surrogate Modeling of Flood Inundation Predictions for the United States Operational Hydrological Forecasting Framework
This study improves how to enhance flood maps quickly and accurately using a simple alternative approach. Traditional flood models are very accurate but slow and need a lot of data. Simpler models like the one used by NOAA (called HAND-FIM) are faster but less accurate. To fix this, a deep learning model (U-Net) is created that uses both NOAA’s fast model and detailed simulation data to make better flood maps. The model combines terrain and water data to predict floods more accurately and works well across many U.S. locations. FIMserv helps use this improved model in real-time, making it easier for forecasters and emergency teams to respond to floods.
