Bai et al. (2025) Effect of natural and human factors on hydrological connectivity in the arid region: Application of water-ecological network and XGBoost model

Identification

• Journal: Journal of Hydrology Regional Studies
• Year: 2025
• Date: 2025-11-25
• Authors: Yufan Bai, Fan Yang, Wei Deng, Hao Zhang
• DOI: 10.1016/j.ejrh.2025.102972

Research Groups

• College of Geography and Resources, Sichuan Normal University, Chengdu, China
• Institute of Mountain Hazards and Environment, Chinese Academy of Sciences, Chengdu, China

Short Summary

This study developed a node-corridor water-ecological network in Altay Prefecture, China, to quantify structural and functional hydrological connectivity and used an XGBoost model to analyze the nonlinear impacts and critical thresholds of natural and human factors on connectivity in an arid region.

Objective

• To quantify structural and functional connectivity of water-ecological networks in arid regions using a node-corridor approach.
• To investigate the direction, intensity, and threshold effects of natural and human factors on hydrological connectivity.

Study Configuration

• Spatial Scale: Regional scale (Altay Prefecture, Xinjiang, China, approximately 118,000 square kilometers), with a uniform spatial resolution of 1000 meters.
• Temporal Scale: Analysis based on data from recent years, including annual means and inter-monthly variations (April-October).

Methodology and Data

• Models used: XGBoost (eXtreme Gradient Boosting), Graph theory parameters, Network indices, Cost distance model. (Linear Regression and Support Vector Regression were used as benchmark models).
• Data sources:
  • Digital Elevation Model (DEM): Shuttle Radar Topography Mission (SRTM) dataset.
  • Normalised Difference Vegetation Index (NDVI): NASA (MOD13Q1 NDVI).
  • Meteorological data (monthly potential evapotranspiration, monthly precipitation, atmospheric humidity index, soil moisture): National Tibetan Plateau Science Data Center.
  • Snowmelt datasets: National Glacial Permafrost Desert Science Data Center.
  • Land use data: National Earth System Science Data Center.
  • Population and Gross Domestic Product (GDP) data: Resource and Environmental Science Data Platform.

Main Results

• High structural connectivity (index exceeding 0.48) was found south of Beitun City and in the eastern and southeastern Ulungu Lake agricultural area, while Gobi and desert areas exhibited extremely low values (index less than 0.16).
• Functional connectivity ranged from 0.17 to 1.0, being strongest in snowmelt-recharged mountainous areas (0.79–1.0, peaking in May) and generally lower in plain areas.
• Comprehensive water-ecological network connectivity was highest in the northern mountainous areas and agricultural production areas east of Ulungu Lake (values ranging from 0.56 to 0.70).
• Among natural factors, NDVI significantly promoted connectivity (r = 0.727), whereas elevation (r = –0.514), mean temperature (r = –0.564), and annual evapotranspiration (r = –0.266) inhibited it (all p < 0.001).
• Human factors exhibited critical thresholds: when GDP density exceeded 400,000 CNY/km² or population density surpassed 10 persons/km², high-intensity human activities reversed the marginal benefit of ecological connectivity.
• The optimized XGBoost model achieved an R² of 0.727 on the test set, demonstrating superior performance compared to linear regression (R² = 0.604) and support vector regression (R² = 0.700).

Contributions

• Innovatively quantified arid-zone water transport loss using cost paths within a constructed water–ecological network, providing a new perspective and quantitative tool for understanding regional hydrology.
• Employed an underutilized node-corridor approach at a grid scale to assess structural and functional connectivity in arid regions.
• Advanced functional connectivity evaluation by integrating network indices and cost distance, quantifying total material expenditure through path resistance values beyond conventional minimum moisture thresholds.
• Quantified the nonlinear effects and critical threshold responses of both natural and human factors on hydrological connectivity using the XGBoost model.
• Provided scientific references and decision-making support for the restoration and stabilization of water-ecological networks in arid regions, contributing to UN Sustainable Development Goals.

Funding

• The Third Xinjiang Scientific Expedition (Grant No. 2021xjkk070204)

Citation

@article{Bai2025Effect,
  author = {Bai, Yufan and Yang, Fan and Deng, Wei and Zhang, Hao},
  title = {Effect of natural and human factors on hydrological connectivity in the arid region: Application of water-ecological network and XGBoost model},
  journal = {Journal of Hydrology Regional Studies},
  year = {2025},
  doi = {10.1016/j.ejrh.2025.102972},
  url = {https://doi.org/10.1016/j.ejrh.2025.102972}
}