Chen et al. (2026) Regional stability, local vulnerability: GeoShapley-based diagnosis and planning translation of water retention in Southwest China

Identification

• Journal: Ecological Indicators
• Year: 2026
• Date: 2026-09-17
• Authors: Yedong Chen, Rongbing Xu, Sihao Gu, Zixuan Li
• DOI: 10.1016/j.ecolind.2026.115401

Research Groups

• College of Landscape Architecture and Horticulture, Southwest Forestry University
• Yunnan Key Laboratory of Landscape Plant Resources Cultivation and Application
• Faculty of Engineering, University of Waterloo
• School of Architecture and Urban Planning, Nanjing University
• Chair of Spatial Information and Modelling, TUD Dresden University of Technology
• Leibniz Institute of Ecological Urban and Regional Development (IOER)

Short Summary

This study examines the water retention (WR) trends in Southwest China from 2000 to 2024 using a combination of raster trend diagnosis, GeoShapley-based explanation, and county translation. The results show that while mean WR suggests regional stability, local heterogeneity in variability, direction, and significance is revealed through coefficient of variation and raster trend diagnostics.

Objective

• To diagnose the water retention trends in Southwest China from 2000 to 2024 using a combination of raster trend diagnosis and GeoShapley-based explanation.
• To identify the leading explanatory variables for the fitted WR Sen slope model.

Study Configuration

• Spatial Scale: County-level analysis units (549 counties) across the five-province study region in Southwest China.
• Temporal Scale: Annual WR sequences from 2000 to 2024.

Methodology and Data

• Models used: InVEST annual water-yield module, version 3.14.2, for estimating annual water yield; Random forest and XGBoost models for explaining the spatial heterogeneity of WR Sen slope.
• Data sources: Precipitation, evapotranspiration forcing, land cover, soil-hydrological properties, terrain, and hydroclimate data.

Main Results

• Mean-state evidence describes a regionally stable but internally uneven WR background.
• Raster trend diagnosis reveals local trajectory differences that are not visible in the mean-state map.
• GeoShapley identifies the explanatory structure of WR trends, with Δ annual precipitation, Δ actual evapotranspiration, and saturated hydraulic conductivity as the leading variables.

Contributions

• This study provides a novel approach to diagnosing water retention trends using raster trend diagnosis and GeoShapley-based explanation.
• The results highlight the importance of considering local heterogeneity in variability, direction, and significance when assessing WR trends.

Funding

• This research was funded by the Yunnan Key Laboratory of Landscape Plant Resources Cultivation and Application (Grant No. 2020YF012).
• The Chair of Spatial Information and Modelling at TUD Dresden University of Technology also contributed to the funding of this project (Grant No. IOER-2022-01).

Citation

@article{Chen2026Regional,
  author = {Chen, Yedong and Xu, Rongbing and Gu, Sihao and Li, Zixuan},
  title = {Regional stability, local vulnerability: GeoShapley-based diagnosis and planning translation of water retention in Southwest China},
  journal = {Ecological Indicators},
  year = {2026},
  doi = {10.1016/j.ecolind.2026.115401},
  url = {https://doi.org/10.1016/j.ecolind.2026.115401}
}