Su et al. (2026) Spatiotemporal Evolution and Attribution Analysis of Groundwater Storage in Heilongjiang Province Based on Interpretable Machine Learning
⚠️ Warning: This summary was generated from the abstract only, as the full text was not available.
Identification
- Journal: Water
- Year: 2026
- Date: 2026-11-11
- Authors: Yanmin Su, Huili Gong, Xin Liu, Jingcheng Tian, Xinyong Zhang, Beibei Chen, Chaofan Zhou, Lin Zhu, Jiajun Zhu, Lin Guo
- DOI: 10.3390/w18222816
Research Groups
Not specified in the provided text.
Short Summary
This study derived Groundwater Storage Anomaly (GWSA) in Heilongjiang Province from 2002–2025 using GRACE data and the water balance equation, identifying its spatiotemporal evolution and drivers. It found an overall provincial GWSA decline of -1.32 mm/yr, with significant regional variations, primarily driven by paddy-rice cultivation.
Objective
- To derive the Groundwater Storage Anomaly (GWSA) in Heilongjiang Province from 2002–2025 and characterize its spatiotemporal evolution.
- To explore the nonlinear associations between stage-specific GWSA trends and various hydro-climatic and anthropogenic factors using an interpretable machine learning model.
Study Configuration
- Spatial Scale: Heilongjiang Province, China, with specific focus on Songnen Plain and Sanjiang Plain.
- Temporal Scale: 2002–2025 for GWSA derivation; analysis divided into four stages (P1–P4) based on change points in September 2014, June 2018, and November 2023.
Methodology and Data
- Models used:
- Water balance equation (for Groundwater Storage Anomaly derivation)
- Binary Segmentation (Binseg) algorithm (for identifying structural change points)
- XGBoost model (for exploring nonlinear associations)
- SHAP (SHapley Additive exPlanations) (for interpreting XGBoost model)
- Data sources:
- GRACE satellite data
- Hydro-climatic factors (not explicitly detailed, but implied as part of the eight factors)
- Anthropogenic factors (e.g., paddy-rice variables)
Main Results
- The Binary Segmentation algorithm identified three change points in the GWSA evolution: September 2014, June 2018, and November 2023, dividing the period into four distinct stages (P1–P4).
- Over the full period (2002–2025), the provincial GWSA showed an overall declining trend of -1.32 mm/yr.
- Significant spatial heterogeneity was observed: the Songnen Plain exhibited a slight recovery (+1.70 mm/yr), while the Sanjiang Plain showed a marked declining trend (-4.61 mm/yr).
- Paddy-rice variables were identified as the leading correlates influencing GWSA trends in stages P1, P3, and over the entire study period. Increased paddy rice cultivation and faster paddy expansion were consistently associated with more negative GWSA trends, with the strongest impact observed in the Sanjiang Plain.
- Temperature was the primary correlate in the weaker P2 model, though this was not interpreted as a direct warming effect.
Contributions
- Provides uncertainty-qualified evidence crucial for stage-sensitive groundwater monitoring strategies.
- Offers valuable insights for managing irrigation-related groundwater stress, particularly in high-latitude agricultural regions.
- Integrates GRACE satellite data with advanced statistical (Binseg) and machine learning (XGBoost, SHAP) techniques to provide a comprehensive and interpretable analysis of GWSA drivers.
Funding
Not specified in the provided text.
Citation
@article{Su2026Spatiotemporal,
author = {Su, Yanmin and Gong, Huili and Liu, Xin and Tian, Jingcheng and Zhang, Xinyong and Chen, Beibei and Zhou, Chaofan and Zhu, Lin and Zhu, Jiajun and Guo, Lin},
title = {Spatiotemporal Evolution and Attribution Analysis of Groundwater Storage in Heilongjiang Province Based on Interpretable Machine Learning},
journal = {Water},
year = {2026},
doi = {10.3390/w18222816},
url = {https://doi.org/10.3390/w18222816}
}
Original Source: https://doi.org/10.3390/w18222816