Harnessing deep learning for fusion-based heavy metal contamination index prediction in groundwater

Abstract

Groundwater contamination by heavy metals presents a major environmental threat with serious implications for public health and resource sustainability. This study proposes a novel deep learning-based data fusion framework to predict heavy metal contamination index in groundwater, focusing specifically on Manganese (Mn), Iron (Fe), Arsenic (As), and Lead (Pb)-elements found to exceed World Health Organization (WHO) permissible limits in the Gultepe-Zarrinabad sub-basin, Zanjan, Iran. Five widely used water contamination indices (e.g., EHCI, HPI, HEI, MI, and CI) were integrated into a unified composite metric using a customized root-based data fusion and normalization approach. This fused index was modeled using a Deep Neural Network (DNN) and benchmarked against traditional machine learning models-Decision Tree (DT), k-Nearest Neighbors (KNN), and Artificial Neural Network (ANN). The DNN model achieved superior predictive accuracy (R² = 0.98), with minimal error (RMSE and MAE = 0.01) and excellent generalization capacity, outperforming all other models. This study marks a successful application of a fusion-based DNN approach for comprehensive groundwater heavy metal assessment, demonstrating its strong potential to support AI-enabled environmental monitoring and sustainable water resource management.

Keywords: Aggregated contamination index; Data fusion; Deep Neural Network; Heavy metals.