Novel approach for predicting groundwater storage loss using machine learning.

Area of Science:

• Environmental science
• Hydrology
• Data science

Background:

• Comprehensive national estimates of groundwater storage loss (GSL) are crucial for effective natural resource management.
• Data-scarce regions face significant challenges in monitoring groundwater resources, with Iran's major aquifers critically depleted.
• Existing methods for GSL estimation can be time-consuming and costly, particularly in areas with limited observational data.

Purpose of the Study:

• To introduce and evaluate a novel approach using Artificial Intelligence (AI) and machine learning (ML) for estimating GSL.
• To identify key water budget variables that are most influential in predicting GSL.
• To develop an efficient and reliable method for GSL calculation suitable for data-scarce regions.

Main Methods:

• Utilized easily accessible water budget variables including aquifer area, storage coefficient, groundwater use, return flow, discharge, and recharge.
• Applied various AI and ML techniques, including Harris Hawks Optimization Adaptive Neuro-Fuzzy Inference System (HHO-ANFIS) and Least-Squares Support Vector Machine (LS-SVM).
• Assessed the correlation between 11 investigated variables and GSL, selecting the most significant ones for model development.

Main Results:

• Identified agricultural water consumption, aquifer area, river infiltration, and artificial drainage as highly correlated variables with GSL (0.84, 0.79, 0.70, and 0.69, respectively).
• Selected 9 out of 11 investigated variables for the final GSL prediction model.
• Demonstrated the efficiency of ML methods in discriminating input variables for reliable GSL estimation, with HHO-ANFIS showing the highest predictive accuracy.

Conclusions:

• AI and ML methods offer a significant advancement in estimating GSL, providing time and cost savings.
• The proposed methodology is particularly well-suited for data-scarce regions with high uncertainty and limited groundwater level observations.
• The study highlights the potential of AI/ML to improve groundwater resource management through accurate and efficient GSL estimations.