Physics-embedded Fourier Neural Network for Spatiotemporal Dynamics Modeling (Flood Forecasting Benchmark Dataset)

[Flood Forecasting Benchmark Dataset] A significant flood event occurred in the Pakistan study area from August 18 to August 31, 2022, resulting in substantial increases in flood coverage. Additionally, a catastrophic flood struck Beira, Mozambique, from March 14 to March 20, 2019, due to heavy rainfall. Consequently, we establish a flood forecasting benchmark dataset by numerically simulating the Pakistan flood over 14 days (covering 85,616.5 km2) and the Mozambique flood over 6 days (covering 6,190.9 km2). The dataset features a spatial resolution of 480m × 480m and a temporal resolution of 30 seconds for training, validation, and testing. [Data Records] The dataset comprises three folders: Pakistan Flood folder, Mozambique Flood folder, and Relevant Data folder. Pakistan Flood folder includes data on the 2022 Pakistan flood and Mozambique Flood folder includes the 2019 Mozambique flood, both with a spatial resolution of 480 m and a temporal resolution of 30 seconds. The Relevant Data folder includes DEM, land use and land cover, and rainfall data for the two flood events.