Assessment of water quality and potential health risks through probabilistic approaches is fundamental in determining the uses of water and prevent health problems. This study aimed to investigate the presence of toxic/non-toxic elements, evaluate water quality and assess health risks in a arid region in Iran. Eighteen elements were detected in the drinking water and followed the order of concentration Na > Ca > Mg > Sr > K > Si > Fe > Li > Al > P > Zn > Mo > Cr > Ba > Se > V > Nb > As, with Mg most frequently exceeding the regulatory limits. Based on deterministic and probabilistic approaches, the water quality index (WQI) showed that 4.3% of the areas had unsuitable conditions (most influential parameter is Li), while the heavy metal evaluation index (HEI) was in mild classification. The geochemical origin of elements was studied. Li being predominantly correlated with Fe and Mg and a suitable model of Li occurrence was predicted using artificial neural network. The hazard quotient for all elements, except for Li, was less than 1, and cancer risk for Cr was less than 10(-4). The hazard index for children, teens and adults was 3.4, 2.5 and 2.1, respectively, while it was only 0.38, 0.28, 0.23 in the blank water sample. The results of Sobol's sensitivity analysis showed that Li and Cr concentration in waters and interaction with body weight had the greatest effect on non-carcinogenic and carcinogenic rates. [GRAPHICS] .