Wang et al. (2026) Reconstructing and downscaling terrestrial water storage to Reveal vegetation–water coupling across continuous areas of Western China-Central Asia-Western Asia under climate warming

Identification

• Journal: Journal of Hydrology Regional Studies
• Year: 2026
• Date: 2026-04-09
• Authors: Jianshun Wang, Ping Yue, Qiang Zhang, Ying Wang, Xueyuan Ren
• DOI: 10.1016/j.ejrh.2026.103410

Research Groups

• Lanzhou Institute of Arid Meteorology, CMA, Lanzhou, China
• College of Atmospheric Sciences, Lanzhou University, Lanzhou, China
• Meteorology Bureau of Gansu, Lanzhou, China

Short Summary

This study reconstructed and downscaled terrestrial water storage anomalies (TWSA) to a continuous 0.1° resolution dataset for Western China-Central Asia-Western Asia (2001–2020) using STL and Random Forest, revealing a significant TWSA decline across over 70% of the region and elevation-dependent vegetation-water coupling.

Objective

• To reconstruct a complete time series of terrestrial water storage anomaly (TWSA) data and develop a Random Forest (RF) downscaling model.
• To reveal the spatiotemporal characteristics of downscaled TWSA across different land use/land cover (LULC) types, elevation zones, and irrigation types.
• To clarify the response of vegetation to TWSA changes based on the downscaled data over different LULC, elevation, and irrigation types.

Study Configuration

• Spatial Scale: Western China-Central Asia-Western Asia (WCCAWA) region (43.45°E-126.07°E, 21.14°N-55.44°N), with downscaled data at 0.1° resolution.
• Temporal Scale: Monthly data from 2001 to 2020.

Methodology and Data

• Models used:
  • Seasonal Trend Decomposition LOESS (STL) for reconstructing continuous TWSA time series.
  • Random Forest (RF) algorithm for downscaling TWSA.
  • Sen's Slope analysis and Mann-Kendall test for trend detection and significance.
  • Hybrid regionalization (Varimax Rotated Empirical Orthogonal Functions (REOF) and K-means clustering) for delineating homogeneous hydrological regions.
• Data sources:
  • Satellite Gravity Data: Arithmetic mean of GRACE and GRACE-FO TWSA from Center for Space Research (CSR), German Research Centre for Geosciences (GFZ), and Jet Propulsion Laboratory (JPL).
  • Reanalysis Datasets:
    • ERA5-Land (runoff, evapotranspiration (Evp.), soil water content (SWC), snow water equivalent (SWE)).
    • GLDAS Catchment Land Surface Model (CLSM) Terrestrial Water Storage (TWS).
  • Remote Sensing Datasets:
    • GPM IMERG (precipitation, Prc.) at 0.1° resolution.
    • MODIS Normalized Difference Vegetation Index (NDVI).
    • MODIS Land Surface Temperature (LST).
  • Auxiliary Datasets:
    • Shuttle Radar Topography Mission (SRTM) Digital Elevation Model (DEM) at 90 m resolution (resampled to 0.1°).
    • FAO Global Land Cover Network (GLC-SHARE) Land Use/Land Cover (LULC) data at 1 km resolution (reclassified into 5 categories).

Main Results

• The STL method accurately reconstructed TWSA data gaps (R² = 0.933, RMSE = 16.7 mm, MAE = 12.3 mm for validation period).
• The RF downscaling model achieved high overall accuracy (R² = 0.994, RMSE = 5.2 mm, MAE = 2.42 mm) in reproducing large-scale TWSA variability.
• A significant decreasing trend in TWSA (-0.0109 mm/month, P < 0.05) was observed across over 70% of the WCCAWA region from 2001 to 2020.
• The most pronounced TWSA declines occurred in the Caspian Sea basin and the southern Tibetan Plateau.
• Faster water losses were observed in low-elevation irrigated areas and bareland, indicating intensified groundwater and soil-moisture depletion.
• Forests exhibited the weakest water storage reduction among LULC types.
• Non-irrigated areas showed the steepest depletion rates (-0.0117 mm/month), followed by irrigated (-0.0116 mm/month) and rainfed (-0.00951 mm/month) areas.
• Vegetation at higher elevations demonstrated stronger sensitivity to TWSA variability compared to low and mid elevations.
• Sparse vegetation (Bareland) showed higher overall sensitivity to TWSA, and SIF responded more sensitively than NDVI to TWSA changes.

Contributions

• Developed a robust framework integrating STL decomposition and Random Forest algorithms to reconstruct and downscale GRACE/GRACE-FO TWSA, providing a continuous, high-resolution (0.1°) dataset for the WCCAWA region (2001–2020).
• Systematically analyzed the spatiotemporal variability of TWSA across different LULC types, elevation gradients, and irrigation regimes in a vast, transboundary arid and semi-arid region.
• Quantified the ecohydrological coupling by clarifying the response of vegetation (NDVI and SIF) to TWSA changes, highlighting elevation-dependent sensitivity and the role of irrigation.
• Provided new hydrological insights into water redistribution patterns, contrasting human-dominated drylands with climate-sensitive mountain systems, which is crucial for regional water resource management and ecological restoration.

Funding

• National Natural Science Foundation of China (Project No. U2142208, 42230611, 42542502)
• Foundation of Key Talent Program Projects in Gansu Province (Project No. 2023RCXM37)
• Innovation Development Special Project of China Meteorological Administration (Project No. CXFZ2026J006, CXFZ2026J129)
• Natural Science Foundation of Gansu Province (Project No. 25JRRA1110, 24JRRA723, 25JRRA1113)
• Research Projects of Lanzhou Institute of Arid Meteorology, CMA (Project No. KYS2022BSKY01)

Citation

@article{Wang2026Reconstructing,
  author = {Wang, Jianshun and Yue, Ping and Zhang, Qiang and Wang, Ying and Ren, Xueyuan},
  title = {Reconstructing and downscaling terrestrial water storage to Reveal vegetation–water coupling across continuous areas of Western China-Central Asia-Western Asia under climate warming},
  journal = {Journal of Hydrology Regional Studies},
  year = {2026},
  doi = {10.1016/j.ejrh.2026.103410},
  url = {https://doi.org/10.1016/j.ejrh.2026.103410}
}