Groundwater quality index and AI model for evaluating land use change impact in northern Iran: non-carcinogenic risk

ABSTRACT This study aimed to evaluate the groundwater quality index (GWQI) and artificial intelligence (AI) model to evaluate the impact of land use change in northern Iran and the non-carcinogenic risk in Babol drinking water. Changes in the dependent variable F were analyzed and predicted by the LightGBM algorithm. In this study 228 water samples collected from 38 wells during 2016–2021 were analyzed for various chemical parameters to determine the GWQI. The results from the deterministic approach showed that 74.6% of water sources show good quality with the range of GWQI from 49.91 to 107.38. However, the probabilistic approach showed that only 0.01% of the water samples showed excellent quality, while 98.37% were classified as having good quality. Principal component analysis on parameters of chemicals showed three factors that account for 71.55% of the total variance in the groundwater quality dataset. The land use results showed that in residential areas where the population is dense, a higher concentration of NO3– is observed. The results found that the non-carcinogenic risk caused by F− and NO3– was above the risk limit (HQ>1) for infants in the wet season. The LightGBM algorithm results demonstrated strong predictive accuracy for F−.