Dataset used in "Soil Moisture Forecasting integrating Physical-based model and Deep Learning". (1) 1-24.tar is training/test data (after preprocessing) over 24 sub-regions in China. (2) GFS* is 3-day forecast of Global Forecast System (GFS) over 2015-2017 and 2018 years. (3) DEM* and LC* is DEM and land cover in EASE 9km grids. (4) auxiliary.json is utility data (e.g., land mask for sub-task). (5) valid_data.tar contains 2018 year of SoMo.ml, ERA5-Land, SMOS L3, LPRM-AMSR2, which were used to triple collocation analysis in our study. The CMA in-situ datasets only could be available from us after certain permission in CMA.

Codes of "Soil Moisture Forecasting integrating Physical-based model and Deep Learning", Journal of Hydrometeorology are open source in https://github.com/leelew/HybridHydro.