Zhang et al. (2026) Impact of spatial scale on the sensitivity of the water supply-demand balance to driving factors

Identification

• Journal: Ecological Indicators
• Year: 2026
• Date: 2026-03-14
• Authors: Lei Zhang, Ning Li, Van beng
• DOI: 10.1016/j.ecolind.2026.114744

Research Groups

• Jingjinji Spatial Intelligent Perception Collaborative Innovation Center, Hebei University of Engineering, Handan 056038, China
• Department of Economics and Finance, Jiangsu Normal University, Zhenjiang 221116, China

Short Summary

This study develops an integrated water footprint accounting framework to diagnose water stress and its drivers across grid (~1 km²), municipal, and sub-basin scales in the Yellow River Basin from 2000 to 2024. It reveals a widening upper-to-lower reach divergence in water stress, driven by coupled socio-hydrological interactions that are often masked by aggregated analyses, providing scale-differentiated management recommendations.

Objective

• To develop and validate a transferable multi-scale water footprint accounting framework (blue, green, grey water) that exposes scale-dependent supply-demand mismatches in the Yellow River Basin.
• To identify the process scales at which supply-side versus demand-side drivers operate using Random Forest importance ranking and Geodetector factor interaction analysis.
• To translate these findings into concrete, sector-specific, and scale-differentiated intervention guidance for water managers.

Study Configuration

• Spatial Scale: Yellow River Basin (approximately 795,000 km²), analyzed at three levels: grid (~1 km² cells), municipal (prefecture-level cities), and sub-basin (natural hydrological divisions).
• Temporal Scale: 25-year period from 2000 to 2024.

Methodology and Data

• Models used:
  • Integrated water footprint accounting framework (blue, green, grey water components).
  • GIS-based supply-demand mapping.
  • Calibrated water balance approach (for Water Yield estimation, validated with Kling-Gupta Efficiency (KGE) and relative volume error (RVE)).
  • Random Forest (RF) ensemble learning model (for driver attribution and importance ranking).
  • Geodetector model (for spatial heterogeneity and factor interaction analysis).
  • Ordinary Least Squares (OLS) regression with Moran eigenvector spatial filtering (for marginal influence and residual analysis).
• Data sources:
  • Remotely sensed/Reanalysis: ERA5-Land precipitation (downscaled to ~1 km² using CHELSA V2.1), GLEAM V3.7 (actual and potential evapotranspiration), China Land Use/Cover Dataset (CLUD) at 30 m resolution (resampled to 1 km²), National Irrigation District Database, SoilGrids 2.0, NASADEM elevation data, NPP-VIIRS nighttime lights, WorldPop population datasets, MODIS and ESA CCI products (NDVI, soil moisture, land surface temperature).
  • Observation/Institutional: Yellow River Water Resources Bulletin (streamflow, water allocation), basin management reports, provincial water bulletins (sectoral water withdrawal), pollution census data (Third National Pollution Source Census 2017–2019), China City Statistical Yearbook (socioeconomic indicators).

Main Results

• The basin-wide Composite Water Stress Index (WSI) increased from 0.46 (2000–2005) to 0.63 (2019–2024), crossing the high-stress threshold (WSI > 0.6) in 2016.
• Significant spatial heterogeneity was observed, with the lower reach irrigation district reaching severe stress levels (WSI > 0.80) by 2019–2024, while the upper reach maintained low stress (0.18–0.27).
• The blue water footprint averaged 39.0 ± 4.8 km³ per year, with irrigated agriculture accounting for 58.3%. Total blue water demand increased by 19% from 2000 to 2024, driven by industrial water intensity (+31%) and urban domestic consumption (+44%).
• The grey water footprint averaged 16.2 ± 2.1 km³ per year, primarily from agricultural non-point source pollution (52%), and grew faster than the blue water footprint in the first half of the study period.
• Random Forest identified irrigated area fraction (IAF) and annual precipitation (PRE) as the most important predictors of WSI spatial variability. Supply-side drivers explained 41% of WSI variance, while demand-side drivers accounted for 53%.
• Geodetector interaction analysis revealed that coupled factor pairs, not single factors, govern severe stress zones. The strongest nonlinear enhancement was between IAF and groundwater dependency index (GWD) (joint q-value = 0.83), and a bilinear enhancement between population density (POP) and industrial water intensity (IWI) (q = 0.74).
• Aggregation bias was significant: municipal-level aggregation understated peak stress by approximately 23% compared to grid-level estimates, confirming that water stress severity is scale-contingent.
• OLS residual analysis showed systematically under-predicted water stress in lower-reach irrigation districts, likely due to unmeasured groundwater depletion and inter-provincial water transfers.

Contributions

• Developed a comprehensive multi-scale (grid, municipal, sub-basin) blue-green-grey water footprint accounting framework for the Yellow River Basin over 25 years.
• Quantitatively demonstrated how water stress patterns and their underlying drivers vary across spatial scales, exposing critical governance blind spots.
• Integrated 25 years of environmental and socioeconomic data to provide actionable, scale-differentiated water management recommendations.
• Showcased the importance of coupled socio-hydrological drivers (e.g., irrigated area and groundwater dependency) in governing severe deficit zones, advocating for integrated rather than sector-isolated governance.
• Validated that aggregating fitted grid surfaces to administrative and hydrological units yields more reliable cross-scale inference than conventional rescaling.

Funding

• Science Research Project of Hebei Education Department (Project No. ZD2022092).

Citation

@article{Zhang2026Impact,
  author = {Zhang, Lei and Li, Ning and beng, Van},
  title = {Impact of spatial scale on the sensitivity of the water supply-demand balance to driving factors},
  journal = {Ecological Indicators},
  year = {2026},
  doi = {10.1016/j.ecolind.2026.114744},
  url = {https://doi.org/10.1016/j.ecolind.2026.114744}
}