Hydrogeochemical characterization and water quality evaluation associated with toxic elements using indexing approaches, multivariate analysis, and artificial neural networks in Morang, Tunisia

Abstract In arid regions like the Mornag region of Tunisia, groundwater is a critical resource for agriculture, drinking water, and ecosystem sustainability. However, the degradation of groundwater quality due to high levels of potentially toxic elements (PTEs) poses significant risks to human health and the environment. Addressing this issue is crucial for ensuring water security and protecting public health in such vulnerable regions. This study aims to assess groundwater quality, identify contamination sources, and evaluate the associated health risks using an integrated approach combining geochemical modeling, machine learning (ML) techniques, and water quality indices. The methodology involved geochemical modeling, detection of recharge sources, and the development of an optimized model using artificial neural networks (ANN) to forecast integrated weight water quality indices (IWQI), carcinogenic risk (CR), hazard risk (HI), and heavy metal pollution index, metal index, and contamination degree (HPI, MI, and C d ) related to PTEs. The findings revealed that the aquifer system contains groundwater of paleo-meteoric origin, influenced by cation exchange, mixing, and water–rock interactions. The Mornag groundwaters were oversaturated with calcite, dolomite, magnesite, and aragonite minerals, while undersaturated with halite, epsomite, gypsum, and anhydrite minerals. Based on IWQI values, 36.84% of water samples fell within the “no restriction” class, indicating excellent water quality, while 26.31% were classified as “low restriction,” suitable for irrigation without threatening soil fertility or plant production. However, 13.15% of samples fell within the “moderate restriction” class, and 23.7% exhibited high to severe restrictions, indicating significant deterioration in irrigation water quality. The HPI indicator showed no significant contamination from PTEs such as Cd, Cr, F, Fe, As, and Ni. Most samples (47.36%) were classified as “very pure” based on MI values, with 42.10% as “pure,” and only 5.26% slightly affected by PTEs. However, two samples from the western part of the Mornag region showed elevated MI values, suggesting potential geogenic or anthropogenic contamination sources. Chromium was identified as a significant cancer risk for both adults and children in many water samples. The use of Monte Carlo simulation proved effective in predicting dermal carcinogenic risks. Additionally, ANN models demonstrated high accuracy in predicting all indices, with R 2 values ranging from 0.90 to 0.98 during training and 0.86 to 0.94 during testing. These findings provide valuable insights for policymakers and stakeholders to implement effective groundwater management and mitigation strategies in arid regions.