Understanding the evolutionary processes and causes of groundwater drought using an interpretable machine learning...

Area of Science:

• Hydrology
• Climate Science
• Machine Learning

Background:

• Groundwater drought assessment is challenging due to limited direct observation.
• Understanding groundwater drought evolution is crucial for water resource management.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting groundwater drought.
• To identify key factors influencing groundwater drought using SHAP analysis.
• To project future groundwater drought trends under climate change scenarios.

Main Methods:

• Employed machine learning models, including XGBoost optimized by the Sparrow Search Algorithm (SSA).
• Utilized Shapley Additive Explanation (SHAP) for model interpretability and feature importance analysis.
• Evaluated eight models for groundwater drought prediction in the West Liao River Plain (WLRP).

Main Results:

• The SSA-optimized XGBoost model demonstrated high performance (AUC: 0.922, F1-score: 0.84).
• Standardized Precipitation Evapotranspiration Index (SPEI) at 12- and 24-month scales were identified as key predictors.
• Long-term meteorological drought, over-extraction, and urbanization were found to exacerbate groundwater drought.

Conclusions:

• Machine learning and SHAP analysis provide a robust framework for understanding groundwater drought.
• Long-term meteorological drought significantly impacts groundwater drought, with interactions from other factors.
• Future climate change (SSP5-8.5) is projected to increase the frequency, extent, and severity of groundwater drought.