Comprehensive assessment of groundwater quality in Mexico and application of new water classification scheme based on machine learning

Abstract

This study conducted a comprehensive evaluation of groundwater quality at 1,068 monitoring sites across all hydrologicadministrative regions in Mexico. Based on the analysis of 14 physicochemical and microbiological parameters, which include fluorides, fecal coliforms, nitrate-nitrogen, arsenic, cadmium, chromium, mercury, lead, manganese, iron, alkalinity, conductivity, water hardness, and total dissolved solids, it was found that 41% of the sites exhibited good water quality. Additionally, 23% of the sites presented regular water quality, while 36% of the sites showed poor water quality. Sites with good water quality exhibited lower concentrations of major ions (Ca, Mg, Na, K, SO4, Cl, and HCO3) compared to sites with regular and poor water quality. Water nomenclature was also estimated using the VL model based on Support Vector Machines with linear kernel, statistical techniques, and Monte Carlo simulation. This model cl sified 87% of the monitoring sites into four basic water classes: Na HCO3 (47%); Na Cl (18%); Ca HCO3 (17%); and Na SO4 (5%). Furthermore, the t-SNE computational algorithm was applied to reduce the dimensionality of the data and visualize it in a 2D plot; in this context, the data corresponds to the chemical concentrations of major ions and contaminants. This alg rithm obtained a clustering cons stent w th the water nomenclature estimated by the VL model. The contaminant study results revealed that all hydrologic-administrative regions presented at least one physicochemical-microbiological parameter that exceeded the acceptable levels defined by regulations of Mexico. Therefore, the implementation of environmental sanitation strategies is crucial to ensure the availability of high-quality water resources that are safe for human health.