What question did this study set out to answer?

March 25, 2026Open Access

Machine learning modelling of a nonlinear environmental index with sensitivity analysis for groundwater assessment

Key Points

The study aims to assess groundwater quality in the Marvdasht aquifer using the Groundwater Quality Index (GWQI) and machine learning approaches.
Groundwater quality parameters were measured including pH, EC, TDS, Na⁺, and others.
GWQI was calculated based on WHO standards and machine learning models (ANN, SVM, RF) were applied.
A spatial map of groundwater quality was generated using ArcGIS.
Sensitivity analysis identified TDS and EC as the most critical parameters affecting GWQI.
The groundwater quality was classified as hard to very hard and unsuitable for drinking in some areas.
ANN model achieved the highest accuracy with R2 = 0.999 and RMSE = 1.02.
Sodium-chloride indicated significant water type evolution through rock-water interaction.
Variations in TDS and EC were highlighted as key factors in groundwater quality.

Abstract

The aim of this study was to evaluate groundwater quality the Marvdasht aquifer using the Groundwater Quality Index (GWQI), which was determined using a conventional method and also predicted using three machine learning algorithms: Artificial Neural Network (ANN), Support Vector Machine (SVM) and Random Forest (RF). For this purpose, groundwater quality parameters (pH, EC, TDS, TH, Na⁺, Ca2⁺, Mg2⁺, Cl⁻, HCO₃⁻, SO₄2⁻, and K⁺) were measured. The GWQI was calculated based on WHO drinking-water standards, and a spatial map was generated in ArcGIS. Results showed that TDS, EC, Na⁺, Cl⁻, and Mg2⁺ were the most influential parameters controlling groundwater quality. The groundwater quality was classified as hard to very hard and not suitable for drinking purposes in the southern and southeastern areas of the study site, while the northern area showed good quality due to limestone formations and recharge from a dam. Results also revealed that Na–Cl was the dominant water type, which indicates the groundwater evolution through rock–water interaction, dissolution, and evaporation. TDS mean variation index of 27.8 and EC with mean variation index of 15.7 were selected as the most sensitive quality parameters in GWQI calculation. Among the applied models, the ANN provided the highest accuracy (R2 = 0.999; RMSE = 1.02), followed by SVM and RF for predicting GWQI from groundwater quality parameters. It was concluded that the integration of GWQI, ANN, and GIS effectively captured spatial variations and provided a reliable framework for groundwater monitoring in arid regions.

Bookmark

View Full Paper