FloodDepth-ML-HWMs

This study applies ML models across three large U.S. coastal watersheds during Hurricanes Michael, Ida, and Ian. The models were developed under three scenarios, differing in the flood observational data used for training and validation: Scenario 1: Trained only on stream gauges. Scenario 2: Trained on both stream gauges and HWMs. Scenario 3: Trained on both stream gauges and HWMs while incorporating HWM uncertainty via a proposed loss function. Model transferability was tested across all scenarios.