PDFID: A high-resolution flood inundation dataset with a long time series

Abstract

Study regionPoyang Lake and East Dongting Lake, China. Study focusOn the basis of data on the extent of the inundation of water bodies, the water level, and DEM data, this study constructed an hourly flood inundation dataset (PDFID) for the East Dongting Lake and Poyang Lake regions with a 10 m resolution and a coverage area of >5000 km2 from 2012 to 2022 by using the RFim model and the U-net model. Validation of the results on Sentinel-1 SAR images and Sentinel-2 MSI optical images showed that the overall accuracy of the dataset ranged from 88% to 97%, with a kappa coefficient ranging from 0.601 to 0.932. The accuracy for the East Dongting Lake region ranged from 88% to 97%, with a kappa coefficient of >0.67, and for the Poyang Lake region, the accuracy ranged from 89% to 97%, with a kappa coefficient of >0.6. New hydrological insights for the regionFlood inundation datasets with a high spatiotemporal resolution and long time series enable us to analyze and comprehend floods’ processes, patterns, and dynamics, facilitating the enhanced development and management of flood prevention and mitigation strategies. However, they have yet to be made available. Thus, this study can provide a fundamental data source for analytical mining and predicting the flooding processes in basins.