Razavi-Termeh et al. (2025) Spatially explicit and interpretable GeoAI models for understanding factors controlling groundwater availability

Identification

• Journal: Journal of Hydrology
• Year: 2025
• Date: 2025-11-26
• Authors: Seyed Vahid Razavi-Termeh, Abolghasem Sadeghi‐Niaraki, Farman Ali, Saied Pirasteh, Mahdieh Shirmohammadi, Soo-Mi Choi
• DOI: 10.1016/j.jhydrol.2025.134683

Research Groups

• Department of Computer Science & Engineering and Convergence Engineering for Intelligent Drone, XR Research Center, Sejong University, Seoul, Republic of Korea
• Department of Computer Science and Engineering, School of Convergence, College of Computing and Informatics, Sungkyunkwan University, Seoul 03063, Republic of Korea
• Institute of Artificial Intelligence, School of Mechanical and Electrical Engineering, Shaoxing University, China
• Department of Geotechnics and Geomatics, Saveetha School of Engineering, Saveetha Institute of Medical and Technical Sciences, Chennai 602105, India
• Key Laboratory of Virtual Geographic Environment (Nanjing Normal University), Ministry of Education, Nanjing, China

Short Summary

This study develops an advanced GeoAI approach, optimizing the CatBoost algorithm with the Fruit Fly Optimization Algorithm (FOA) and using SHAP for interpretability, to accurately predict groundwater-prone areas in semi-arid regions. The model significantly improves prediction accuracy and identifies key environmental factors influencing groundwater availability.

Objective

• To develop a spatially explicit and interpretable GeoAI model for predicting groundwater-prone areas.
• To identify the key environmental factors controlling groundwater availability using model interpretability techniques.

Study Configuration

• Spatial Scale: Kazerun and Kooh-Chenar counties, Iran.
• Temporal Scale: Not explicitly defined for the model application, but based on static environmental factors.

Methodology and Data

• Models used: CatBoost algorithm optimized with the Fruit Fly Optimization Algorithm (FOA). Shapley Additive exPlanations (SHAP) for model interpretability.
• Data sources: 16 environmental factors related to groundwater potential, likely including remote sensing, elevation models, and rainfall data.

Main Results

• The optimized CatBoost-FOA model significantly outperformed the standalone CatBoost model.
• CatBoost-FOA achieved an RMSE of 0.086, MAE of 0.068, and AUC-ROC of 0.838.
• The standalone CatBoost model yielded an RMSE of 0.219, MAE of 0.17, and AUC-ROC of 0.741.
• SHAP analysis identified low elevation, high vegetation index (NDVI), high rainfall, and proximity to rivers as significant factors affecting groundwater availability.
• Lower elevation facilitates surface water accumulation and infiltration, higher NDVI indicates sufficient soil moisture, greater rainfall enhances groundwater recharge, and proximity to rivers improves hydraulic connectivity.

Contributions

• Introduction of an advanced GeoAI approach combining CatBoost with the Fruit Fly Optimization Algorithm (FOA) for enhanced groundwater potential mapping.
• Integration of Shapley Additive exPlanations (SHAP) for spatially explicit and interpretable understanding of factors controlling groundwater availability.
• Provides a more accurate and interpretable method for groundwater potential mapping, aiding water resource planning and management.

Funding

• Not explicitly mentioned in the provided text.

Citation

@article{RazaviTermeh2025Spatially,
  author = {Razavi-Termeh, Seyed Vahid and Sadeghi‐Niaraki, Abolghasem and Ali, Farman and Pirasteh, Saied and Shirmohammadi, Mahdieh and Choi, Soo-Mi},
  title = {Spatially explicit and interpretable GeoAI models for understanding factors controlling groundwater availability},
  journal = {Journal of Hydrology},
  year = {2025},
  doi = {10.1016/j.jhydrol.2025.134683},
  url = {https://doi.org/10.1016/j.jhydrol.2025.134683}
}