Identification of non-conventional groundwater resources by means of machine learning in the Aconcagua basin, Chile

Abstract

Study region: Our study region is the Aconcagua basin, central Chile. The catchment, home to over 500,000 people, currently experiences a multiyear drought that threatens water security. In this context, there is an impending need to explore unconventional groundwater resources, such as peripheral hard rock and deep aquifers. Study focus: The focus of this study is the application of machine learning techniques to identify areas of potentially untapped groundwater resources in the middle and upper reaches of the Aconcagua basin. New hydrological insights for the region: Machine learning classifiers accurately depict those areas known for their high groundwater potential, including the San Felipe, Putaendo, Panquehue, Catemu and Llay-Llay sectors. The Mesozoic sedimentary sequences and intrusive units of the coastal range, together with the Neogene units of the Andes range, are correctly identified as areas of very low hydrogeological interest. A major novelty of this study is the delineation of several new areas of potentially high groundwater prospect, namely, the geological domain associated with the Pocuro fault system, the Teatinos-Volcan-River fault system, and the Las Chilcas volcano-sedimentary sequence, in the Chacabuco Range. These findings are in line with the results of complementary studies, thus suggesting that machine learning applications may successfully underpin groundwater exploration across other groundwater basins in the coastal Andean region.

Keywords

• Groundwater potential mapping
• Machine learning
• Groundwater exploration
• Artificial intelligence
• Aconcagua basin