Flood Forecasting and Assessment Using Deep Recurrent Neural Networks

Recent years have seen an uptick in the frequency of flood records occurring in the United States, with South Carolina (SC) being particularly hard hit. This study developed various deep recurrent neural networks (DRNNs) such as Vanilla RNN, long short-term memory (LSTM), and Gated Recurrent Unit (GRU) for flood simulation. Precipitation and the USGS gaging data were preprocessed and forced into the DRNNs to predict flood events for Pickens County, South Carolina. The DRNNs are trained and evaluated using hourly datasets, and the outcomes were then compared with the observed data as well as with the National Water Model (NWM) simulations. Analysis suggested that the DRNNs skillfully predicted the shape of flood hydrographs, peak rates, and time to peak, while the NWM largely overestimated flood hydrographs. Among different climatic variables that were forced into the DRNNs, rainfall amount and spatial distribution were the most dominant input variables for flood prediction.