Sun et al. (2026) Climate-hydrology-topography-anthropogenic factors jointly drive the evolution of vegetation coverage in semi-arid regions: A downscaling approach based on random forest and nonlinear residual correction

Identification

• Journal: Environmental Impact Assessment Review
• Year: 2026
• Date: 2026-04-08
• Authors: Jiaxin Sun, Tiejun Song, Xiaosi Su, Weihong Dong, Hang Lyu, Yuyu Wan, Xiaofang Shen
• DOI: 10.1016/j.eiar.2026.108457

Research Groups

• Key Laboratory of Groundwater Resources and Environment (Jilin University), Ministry of Education, Jilin, Changchun 130021, China
• Jilin Provincial Key Laboratory of Water Resources and Water Environment, Jilin University, Jilin, Changchun 130021, China
• Institute of Water Resources and Environment, Jilin University, Jilin, Changchun 130021, China

Short Summary

This study developed a synergistic downscaling approach combining random forest and nonlinear residual correction to analyze the spatiotemporal dynamics of annual mean Normalized Difference Vegetation Index (NDVI) in the Songnen Plain from 1985 to 2022, revealing a significant upward trend in NDVI and identifying key driving factors and their interactions, including the crucial role of groundwater depth.

Objective

• To develop a synergistic downscaling approach for generating continuous, high-spatial-resolution Normalized Difference Vegetation Index (NDVI) datasets in semi-arid regions, addressing limitations of previous methods.
• To analyze the spatiotemporal dynamics of annual mean NDVI in the Songnen Plain, China, from 1985 to 2022.
• To identify and quantify the driving factors of vegetation dynamics, including climate, hydrology, topography, and anthropogenic factors, with a particular focus on the regulatory role of groundwater.

Study Configuration

• Spatial Scale: Songnen Plain, Jilin Province, China.
• Temporal Scale: 1985 to 2022.

Methodology and Data

• Models used: Random forest (for downscaling), nonlinear residual correction (for downscaling), Geographical detector (for analyzing driving factors).
• Data sources: Normalized Difference Vegetation Index (NDVI) data (implied satellite-derived for downscaling), environmental feature variables including climate (e.g., precipitation), topography, anthropogenic factors (e.g., land-use change, population density), and groundwater depth.

Main Results

• The developed synergistic downscaling approach achieved a high fitting accuracy (R² = 0.90), demonstrating superior performance compared to conventional methods.
• The regional annual mean NDVI exhibited a significant upward trend over the study period, with a growth rate of 0.001 per year.
• The primary driving factors for NDVI varied across different sub-regions:
  • In the Low Plain, Land Use and Land Cover Change (LUCC) had the strongest explanatory power (multi-year average = 0.22).
  • In the Piedmont Plain, precipitation was the most influential factor (multi-year average = 0.17).
  • In the High Plain, population density showed the highest explanatory power (multi-year average = 0.14).
• Although groundwater depth alone exhibited low explanatory power, its interactions with climate, topography, and anthropogenic factors were notably significant in influencing vegetation dynamics.

Contributions

• Developed a novel and highly accurate synergistic downscaling approach combining random forest and nonlinear residual correction, providing a robust method for generating high-spatial-resolution NDVI data.
• Generated a continuous, high-spatial-resolution NDVI dataset for semi-arid regions, specifically addressing the data gap around the year 2000.
• Enhanced the understanding of vegetation dynamics in semi-arid regions by comprehensively analyzing the joint influence of climate, hydrology, topography, and anthropogenic factors.
• Highlighted the often-overlooked but significant interactive role of groundwater depth in regulating vegetation cover in semi-arid environments.
• Provided a solid data foundation and scientific guidance for regional ecological management and restoration efforts.

Funding

Not specified in the provided text.

Citation

@article{Sun2026Climatehydrologytopographyanthropogenic,
  author = {Sun, Jiaxin and Song, Tiejun and Su, Xiaosi and Dong, Weihong and Lyu, Hang and Wan, Yuyu and Shen, Xiaofang},
  title = {Climate-hydrology-topography-anthropogenic factors jointly drive the evolution of vegetation coverage in semi-arid regions: A downscaling approach based on random forest and nonlinear residual correction},
  journal = {Environmental Impact Assessment Review},
  year = {2026},
  doi = {10.1016/j.eiar.2026.108457},
  url = {https://doi.org/10.1016/j.eiar.2026.108457}
}