Susceptibility mapping of groundwater salinity using machine learning models.

Area of Science:

• Environmental Science
• Hydrogeology
• Data Science

Background:

• Rising groundwater salinity poses global environmental and health risks.
• Effective resource management and mitigation planning require advanced spatial salinity modeling.

Purpose of the Study:

• To apply machine learning (ML) models for groundwater salinity mapping.
• To identify key predictive factors and evaluate model performance for salinity assessment.

Main Methods:

• Employed six ML models: flexible discriminant analysis (FDA), mixture discriminant analysis (MAD), boosted regression tree (BRT), multivariate adaptive regression spline (MARS), random forest (RF), and support vector machine (SVM).
• Utilized simulated annealing for feature selection and k-fold cross-validation for robust model evaluation.

Main Results:

• The support vector machine (SVM) model demonstrated superior performance compared to other ML models.
• Soil order, groundwater withdrawal, precipitation, land use, and elevation were identified as the most significant factors influencing groundwater salinity.
• High groundwater salinity was predicted in southern, northern, northeastern, and western regions, correlating with Entisols, barelands, and low elevations.

Conclusions:

• Machine learning models, particularly SVM, are effective tools for spatial groundwater salinity mapping.
• Understanding the influence of environmental and anthropogenic factors is crucial for managing groundwater salinity.
• The study provides valuable insights for targeted mitigation strategies in vulnerable areas.