Enhancing groundwater quality prediction through ensemble machine learning techniques.

Area of Science:

• Environmental Science
• Hydrogeology
• Data Science

Background:

• Groundwater quality assessment traditionally relies on costly and time-consuming field sampling and laboratory analysis.
• Accurate prediction of groundwater quality is crucial for sustainable water resource management, especially in unconfined aquifers.

Purpose of the Study:

• To introduce a novel machine learning (ML) framework for predicting groundwater quality index (GWQI).
• To develop and evaluate an innovative stacking ensemble learning model for enhanced GWQI prediction accuracy.
• To create a spatially explicit map of GWQI for an unconfined aquifer in northern Iran.

Main Methods:

• Utilized 250 groundwater samples to evaluate the groundwater quality index (GWQI).
• Employed machine learning classifiers including AdaBoost (ADA), quadratic discriminant analysis (QDA), and stacking ensemble learning (SEL).
• Introduced a novel Quadratic-Ada-Stacking Ensemble Learning (QA-SEL) model and validated its performance using ROC curves and statistical indicators.

Main Results:

• All tested ML algorithms demonstrated high accuracy in GWQI prediction.
• The novel QA-SEL model achieved superior performance with an overall accuracy of 0.95, precision of 0.95, recall of 0.96, and ROC of 0.96.
• A validated GWQI map was generated using the QA-SEL model and GIS, showing predicted GWQI classes across the study area.

Conclusions:

• The QA-SEL model provides an economically efficient and highly accurate method for predicting groundwater quality.
• This ML-based framework offers a replicable solution for groundwater quality assessment in other plain areas.
• The study highlights the potential of advanced ML techniques to support sustainable groundwater resource management.