A fuzzy logic-based approach for groundwater vulnerability assessment.

Area of Science:

• Environmental Science
• Hydrogeology
• Artificial Intelligence

Background:

• Groundwater vulnerability assessment is crucial for resource protection.
• The traditional DRASTIC model relies on expert opinion, introducing uncertainty.
• Existing methods struggle to accurately predict groundwater vulnerability.

Purpose of the Study:

• To develop an improved groundwater vulnerability assessment model using Mamdani fuzzy logic (MFL) and data mining.
• To address the uncertainty issues inherent in the DRASTIC model.
• To evaluate the proposed MFL model's performance in the Qorveh-Dehgolan plain (QDP) and Ardabil plain aquifers.

Main Methods:

• Calculated DRASTIC index for Ardabil and QDP aquifers.
• Developed two scenarios of the Mamdani fuzzy logic model: one with seven parameters and another with four.
• Utilized Heidke skill score (HSS) and total accuracy (TA) to validate model performance.

Main Results:

• The DRASTIC model's nitrate concentration-based results showed limitations in verification.
• The MFL model achieved higher TA and HSS values compared to the DRASTIC method.
• The first MFL scenario (seven parameters) yielded TA of 0.75 and HSS of 0.51 for Ardabil, and TA of 0.45 and HSS of 0.33 for QDP.

Conclusions:

• The Mamdani fuzzy logic model, especially with fewer parameters, provides a more reliable and practical approach to groundwater vulnerability assessment.
• The integration of fuzzy logic and data mining effectively handles uncertainty in vulnerability assessments.
• The proposed model demonstrates superior performance over traditional methods for aquifer protection.