ANN-Based Predictive Modeling of Aquifer Dynamics in Quetta Valley

Abstract

Monitoring and understanding groundwater level (GWL) dynamics in the Quetta Valley have gained significant attention due to the increasing reliance on groundwater resources. Accurate forecasting of GWL is critical for effective resource management, mitigating risks such as overexploitation, water quality degradation, and land subsidence. This study employs an Artificial Neural Network (ANN) model to predict GWL fluctuations using data collected from 14 monitoring stations over a 40-year period (1980–2020), encompassing a wide range of GWL changes. The dataset was divided into two subsets: 77.23% for training and 22.76% for testing the model. Model performance was assessed quantitatively using root mean square error (RMSE), coefficient of determination (R²), and mean absolute error (MAE). Additionally, qualitative evaluations were conducted through time-series line plots and scatter plots, offering insights into the model's predictive accuracy. The findings provide a deeper understanding of GWL dynamics in the Quetta Valley and demonstrate the potential of ANN models as effective tools for sustainable groundwater management.