[Paper Review] Groundwater vulnerability assessment in semi-arid regions using GIS-based DRASTIC models and FUZZY AHP: South Chott Hodna

Lakhdar Seraiche, Mostafa Dougha+4 more|arXiv (Cornell University)|Jan 16, 2026

Groundwater and Isotope Geochemistry0 citations

TL;DR

The paper presents a GIS-based framework that enhances the DRASTIC groundwater vulnerability model with land-use data and weighting via AHP and Fuzzy AHP, validated with nitrate data from 70 wells, achieving AUC up to 0.951.

ABSTRACT

Groundwater vulnerability is a major concern in arid regions worldwide, where population growth and intensive agriculture increase the risks of depletion and contamination. This study proposes a hybrid groundwater vulnerability assessment framework that improves the conventional DRASTIC model by integrating land-use data and applying advanced weighting techniques, namely the Analytical Hierarchy Process (AHP) and its fuzzy logic variant (Fuzzy AHP). This method makes expert-based weighting less subjective, better captures anthropogenic effects, and facilitates adaptation to challenging situations and limited data. Four vulnerability maps were produced using Geographic Information Systems (GIS): DRASTIC, DRASTIC_LU, AHP DRASTIC_LU, and Fuzzy AHP DRASTIC_LU. We used nitrate levels from 70 wells to verify our work. We found that agricultural areas, especially those above the alluvial aquifer, were the most vulnerable. The ROC curve analysis showed that the model improved over time, with the area under the curve (AUC) values of 0.812 for DRASTIC, 0.864 for DRASTIC_LU, 0.875 for AHP DRASTIC_LU, and 0.951 for fuzzy AHP DRASTIC_LU. These results show that fuzzy AHP DRASTIC_LU makes groundwater risk assessments much more. The GIS-based hybrid models offer a scalable and transferable method for mapping vulnerability, but they also provide local and regional water resource managers with useful information.

Motivation & Objective

• Motivate groundwater vulnerability assessment in semi-arid regions with increasing population and agriculture.
• Develop a hybrid framework that integrates land-use data into DRASTIC and applies AHP and Fuzzy AHP for weighting.
• Produce and compare four vulnerability maps: DRASTIC, DRASTIC_LU, AHP DRASTIC_LU, and Fuzzy AHP DRASTIC_LU.
• Validate model performance using nitrate concentrations from 70 wells.

Proposed method

• Use GIS-based processing to construct four vulnerability maps: DRASTIC, DRASTIC_LU, AHP DRASTIC_LU, and Fuzzy AHP DRASTIC_LU.
• Incorporate land-use data into the DRASTIC framework (DRASTIC_LU).
• Apply Analytical Hierarchy Process (AHP) to derive expert-based weights for the DRASTIC_LU model.
• Apply Fuzzy AHP to derive fuzzy expert weights for the DRASTIC_LU model.
• Evaluate models using nitrate data from 70 wells as verification data.
• Assess performance via ROC/AUC metrics (0.812, 0.864, 0.875, 0.951).

Experimental results

Research questions

• RQ1 How does incorporating land-use data affect DRASTIC-based groundwater vulnerability assessments in semi-arid regions?
• RQ2 Do AHP and Fuzzy AHP weighting improve the predictive performance of GIS-based vulnerability models compared to the standard DRASTIC approach?
• RQ3 Which vulnerability map (DRASTIC, DRASTIC_LU, AHP DRASTIC_LU, Fuzzy AHP DRASTIC_LU) best explains observed nitrate distributions across wells?
• RQ4 What is the relative improvement in model performance when advancing from DRASTIC to Fuzzy AHP DRASTIC_LU?

Key findings

• Agricultural areas, especially above the alluvial aquifer, show the highest vulnerability.
• DRASTIC_AUC = 0.812; DRASTIC_LU_AUC = 0.864; AHP DRASTIC_LU_AUC = 0.875; FUZZY AHP DRASTIC_LU_AUC = 0.951.
• The hybrid GIS-based models are scalable and transferable for local and regional water resource management.
• The approach better captures anthropogenic effects and reduces expert subjectivity through weighting methods.