Predictive modeling of fluoride and nitrate health risks using artificial neural networks

Abstract

The health risk associated with high levels of fluoride (F⁻) and nitrate (NO₃⁻) in groundwater is making groundwater unsafe for drinking globally. This concern has justified a continuous study and health risk assessment of groundwater. High levels of F⁻ and NO₃⁻ have been reported in some parts of the Bole District of Ghana. However, their health risks are unknown. Hence, the study assesses the health risks of F⁻ and NO₃⁻ using indexical proxies and machine learning techniques. Hydrochemical analysis reveals that while most parameters meet WHO standards, elevated fluoride, nitrate, and total dissolved solids (TDS) levels are present in certain communities. Fluoride risk classification shows that 83.3 % of the area faces very low risk, but 3.3 % is exposed to extreme fluoride risk. Health risk assessments using hazard quotient (HQ) indicate that adults are more vulnerable to fluoride exposure, with 53.3 % of samples exceeding safe levels, while 40 % of children are at risk from nitrate contamination. An artificial neural network (ANN) model was used to predict health risks associated with fluoride and nitrate exposure. The model performed well for fluoride predictions, achieving R² values above 0.97 for both training and testing datasets. However, its performance was less reliable for nitrate predictions, particularly for children, where the testing R² dropped to 0.60. Sensitivity analysis identified SO₄²⁻, Cl⁻, Na⁺, and Mg²⁺ as key ions influencing health risk predictions. These results highlight the need for stricter regulation of groundwater sources in high-risk areas and the promotion of low-cost water treatment technologies to mitigate contamination.