Probabilistic approach for assessing water quality, carcinogenic and non-carcinogenic risks of drinking water in a semi-arid region: A correlation and clustering by AI approach

Global industrialization and urbanization have raised concerns about widespread contamination of water resources. This study evaluated the water quality index (WQI) and health risk assessment in a region with a hot semi-arid climate using a probabilistic method. In addition, a correlation matrix and hierarchical cluster analysis (HCA), a machine learning technique, were employed to investigate the origins of the contaminants. Around 1.75 % of electricity conductivity (EC), 7.98 % of nitrate (NO₃), 6.03 % of magnesium (Mg), 4.69 % of sulfate (SO₄), 3.5 % of calcium (Ca), 18.29 % of alkalinity (Alk), 16.53 % of total dissolved solids (TDS), 1.76 % of iron (Fe), and 43.61 % of chromium (Cr) samples exceeded WHO/BIS/EPA standards. Based on the deterministic results, the WQI indicated excellent quality. However, using the Monte Carlo simulation method, 78 % of samples were classified as excellent, 22 % as good. The Cr emerged as the most significant influencing factor on WQI. The origin analysis revealed interconnections among the parameters and clustered the drinking water samples. Probabilistic analysis shows that for age < 2, the 95th percentile hazard quotient (HQ) values for fluoride (F), NO₃, and Cr were 1.38, 2.39, and 2.50, respectively. Carcinogenic risks exceeded the acceptable threshold across all age groups, particularly affecting 94.07 % of adults. Sobol's analysis identifies Cr concentration and its interactions as primary health risk factors across age groups. Overall, although the water quality in the study area was rated as excellent to good, the health risk assessment indicated that Cr, NO₃, and F posed potential risks, particularly for younger age groups.