Assessing water quality index and health risk using deterministic and probabilistic approaches in Darab County, Iran; A machine learning for fluoride prediction

Abstract

The present study employed deterministic and probabilistic approaches to determine the Water Quality Index (WQI) and assess health risks associated with water consumption in Darab County, Iran. Additionally, pollution levels were predicted using a machine-learning algorithm. The study's findings indicate that certain physicochemical parameters of water in some locations exceeded permissible limits (WHO or EPA), with 79.00 % of total hardness (TH) and 21.74 % of Total dissolved solids (TDS) levels exceeding standard values. The WQI for drinking water was determined to be 94.56 % using the deterministic approach, and 98.4 % of samples included the excellent and good categories according to the WQI classification system using the probabilistic approach. Fluoride (F) exhibited the most substantial impact on WQI values. The Artificial Neural Network (ANN) analysis findings suggest that the pH, nitrate (NO₃), and TDS are the most significant factors affecting the prediction of F concentration in water. Multivariate analysis demonstrated that anthropogenic, especially agriculture and geogenic factors, contributed to the water quality in this area. The health risk assessment (HRA) using deterministic methods revealed that water consumption posed a relatively high risk in certain areas. However, Monte Carlo simulation demonstrated that the 5th and 95th percentiles of Hazard Index (HI) for children, teenagers, and adults were within limits of (0.14-2.38), (0.09-1.29), and (0.10-1.00) respectively, with a certainty level of 70 %, 91 %, and 95 %. Interactive indices revealed that the intake of IR and NO₃-IR in children, BW and F-BW in teenagers, and NO₃ and NO₃-IR in adults significantly impacted health risks. Based on these findings, augmenting water treatment processes, regulating fluoride concentrations, and advocating for sustainable agricultural practices complemented by continuous monitoring is imperative.

Keywords: Health risk; Modelling; PMF model; Sensitivity analysis; Water evaluation.