Mapping the vulnerability to groundwater contamination using machine learning: Mayarí watershed, Cuba

e414

Abstract

The Mayarí watershed, located in the eastern region of Cuba, has been declared of national interest by the National Council for Watershed Basins; therefore, studies aimed at preserving its natural resources are of special importance.The objective of this article is to evaluate the intrinsic vulnerability of groundwater to contamination in the Mayarí watershed due to the impact of the vertical migration of potential contaminants discharged on its surface. The k-means machine learning algorithm was employed using the variables of topographic slope, soil attenuation index, density of tectonic faults, and geological formations. The spatiotemporal variation of vegetation and soil salinity in the basin was analyzed according to the Normalized Difference Vegetation Index (NDVI) and the Soil Salinity Index (NDSI), using Landsat satellite images from the years 2000 and 2023. An intrinsic vulnerability map of groundwater to contamination from the infiltration of potential surface contaminants was obtained at a scale of 1:100,000, which distinguishes areas of low, moderate, high, and extreme vulnerability. The vegetation index increased while soil salinity decreased during the analyzed period, which positively impacts the protection of water resources. The presented procedure is part of an early warning system designed to prevent the potential deterioration of groundwater quality in the Mayarí watershed. It can be applied to other basins across the national territory and at more detailed work scales.