Datasets Supporting Soil Moisture Downscaling in the Huai River Basin Using a Spatiotemporal Deep Learning Framework

# Data Availability and Description The datasets used in this study are derived from publicly available remote sensing and environmental data sources. All data processing results used for model training, evaluation, and validation have been archived in a public repository to ensure transparency and reproducibility. ## 1. Source datasets The following open-access datasets were used in this study: **Satellite soil moisture dataset** * **ESA Climate Change Initiative Soil Moisture (ESA CCI SM)** ESA CCI Soil Moisture combined product (COM) was used as the target dataset for downscaling. Data are publicly available at: [https://www.esa-soilmoisture-cci.org/](https://www.esa-soilmoisture-cci.org/) * **SMAP Soil Moisture Product** The SMAP Level-3 passive soil moisture product was used for independent comparison and ETC evaluation. Data are available at: [https://nsidc.org/](https://nsidc.org/) **Reanalysis dataset** * **ERA5 Soil Moisture** ERA5 reanalysis soil moisture from the ECMWF was used as a reference dataset for comparison and validation. Data are available at: [https://cds.climate.copernicus.eu/](https://cds.climate.copernicus.eu/) **Environmental predictor datasets** The environmental predictors used in the downscaling framework include: * Precipitation* Relative humidity* Wind speed* Sunshine duration* Maximum temperature* Minimum temperature* Mean temperature* NDVI (Normalized Difference Vegetation Index)* Terrain slope These datasets were obtained from publicly accessible meteorological, remote sensing, and terrain databases. --- ## 2. Derived datasets and model outputs All processed datasets used for model training, evaluation, and analysis—including: * predictor variable datasets* model comparison results* scale comparison results* ETC evaluation datasets* SMAP resampled datasets* factor selection analysis results have been archived and publicly released in the following repository: **Zenodo repository:**DOI: 10.5281/zenodo.19174285 --- ## 3. Description of archived files The archived datasets include the following files: **Model_compare** Performance comparison results of different downscaling models, including: * CNN-only* LSTM-only* CNN–LSTM Evaluation metrics include R, MSE, MAE, and ubRMSE. **Compare_resolution** Performance comparison results for different spatial downscaling resolutions, including: * original ESA CCI resolution* 5 km downscaled product* 1 km downscaled product **PCC_MI** Datasets used for predictor selection analysis, including: * Pearson Correlation Coefficient (PCC)* Mutual Information (MI) These were used to identify the optimal environmental predictors for soil moisture downscaling. **compare_smap** Datasets used for ETC-based comparison between: * downscaled ESA CCI* SMAP* ERA5 **etc_datasets** SMAP datasets resampled and prepared for ETC (Enhanced Triple Collocation) analysis. **select_train_vars** Datasets containing the environmental predictor variables used for CNN–LSTM model training. --- ## 4. Additional information If additional data or clarification is required, readers are encouraged to contact the corresponding author. Correspondence and requests for materials should be addressed to the corresponding author of this paper.