Qualitative modeling of groundwater resources using Artificial Neural Network and Gray Wolf Optimizer algorithm (Case Study: Kabudarahang Plain, Hamedan Province, Iran)

Abstract

Background: Climate change and declining rainfall have significantly reduced access to surface water in Iran, particularly in Hamedan Province. This has led to increased reliance on groundwater resources, consequently altering their quality for various uses. Therefore, this study was conducted to develope a qualitative model for assess of groundwater resources in Kabudarahang Plain, Hamedan Province, using an Artificial Neural Network (ANN) and the Gray Wolf Optimizer (GWO) algorithm. Methods: Qualitative data of groundwater resources in Kabudarahang Plain were collected and analyzed over a decade (2012-2022). ANN modeling was employed to predict groundwater quality changes. Additionally, the GWO algorithm was integrated to enhance prediction accuracy. The model utilized three output or dependent variables (TDS, EC, and pH) and six input or independent variables (calcium, magnesium, chloride, sulfate, sodium, and turbidity). Results: : The ANN model demonstrated that over 99% of water quality variations can be attributed to the six input variables. Moreover, the GWO algorithm effectively reduced average prediction errors from 0.0015 to 0.0008. Conclusion: The ANN algorithm exhibited high prediction accuracy, low prediction error, and model optimality, which were further enhanced by the GWO algorithm. This suggests that while the ANN model successfully predicted groundwater quality changes in the study area, the GWO algorithm refined the predictions, and improving the model's overall performance. Considering the complementary nature and effectiveness of ANN and GWO algorithms for prediction, their application to predict qualitative changes in groundwater resources in other regions is recommended.

Keywords

• groundwater quality
• intelligent data modeling
• artificial neural network
• gray wolf algorithm
• kabudarahang plain