A high-resolution (1 km) dataset of soil thickness, multi-depth soil moisture, and physically constrained water storage for the Tibetan Plateau (2009–2019)

The Tibetan Plateau, often referred to as the "Asian Water Tower," plays a critical role in regional hydrology and climate feedback. However, current macro-scale soil water storage (SWS) assessments in this topographically complex region are widely hindered by the "fixed-depth" (FD) assumption, which erroneously assigns fictitious water to physically impermeable bedrock layers (i.e., "phantom water"). To overcome this systematic bias, we present a comprehensive, high-resolution (1 km) soil hydrology dataset for the Tibetan Plateau, covering 132 continuous months from 2009 to 2019. By employing a digital soil mapping framework and machine learning-based downscaling, we mapped the actual soil thickness (ST) as a rigid physical boundary. Furthermore, we provided downscaled soil moisture across four vertical depth layers and the final ST-constrained shallow SWS (0–200 cm). This dataset mitigates the systematic overestimation caused by traditional FD methods and provides a physically consistent benchmark for regional hydrological modeling, eco-hydrological assessments, and water resource management. Data Structure and Content This repository contains spatial data in GeoTIFF format, structurally divided into two main components: Soil thickness.tif: The spatial distribution of actual soil thickness over the Tibetan Plateau at 1 km spatial resolution. SWS_on the TP (Folder/Zip): Contains 548 GeoTIFF files representing the high-resolution (1 km) downscaled volumetric soil moisture. The data is stratified into four standard vertical profiles (0–10 cm, 10–40 cm, 40–100 cm, and 100–200 cm) across the 132-month study period. SWS_on the TP (Folder/Zip): Contains 132 GeoTIFF files representing the monthly integrated shallow soil water storage (0–200 cm) physically constrained by the actual soil-bedrock interface. Spatiotemporal Properties Spatial Extent: Tibetan Plateau Spatial Resolution: 1 km (~0.00833 degrees) Temporal Coverage: January 2009 to December 2019 (Monthly) Vertical Profiles: 0–10, 10–40, 40–100, 100–200 cm (for Soil Moisture); 0–200 cm (for SWS) Coordinate Reference System (CRS): Albers_Conic_Equal_Area Data Format: GeoTIFF Usage Notes and Limitations Data users should note that the reconstructed storage space in this dataset is strictly limited to the shallow soil layer (0–200 cm) and does not include deep groundwater aquifers. Furthermore, due to the extreme lack of in-situ observations and intense phase transitions (freeze-thaw cycles) during the cold season, the temporal dynamics in winter months should be interpreted with caution. Code Availability and Processing Scripts To ensure full methodological transparency and facilitate reproducibility, we have included the core R scripts used for geospatial modeling and data processing. These scripts utilize high-performance spatial packages (e.g., explicitly relying on the terra namespace for raster operations) and the ranger package for machine learning. The repository includes: Spatial_Prediction_ST_Tibetan_Plateau.R: The primary machine learning script for predicting the spatial distribution of actual soil thickness using a Random Forest algorithm and environmental covariates. ST_QRF_Uncertainty_Mapping.R: Script for generating pixel-level prediction intervals and relative uncertainty maps for the soil thickness product using Quantile Regression Forests (QRF). SM_Downscaling_Workflow_2015_07.R: A representative monthly workflow (using July 2015 as an example) demonstrating the dynamic spatial downscaling of coarse-resolution soil moisture to 1 km resolution. SWS_Constrained_Integration.R: The core integration script that calculates the pixel-level shallow soil water storage (SWS). It demonstrates the proportional truncation logic that restricts water integration strictly within the predicted soil-bedrock interface.