Optimization of the water quality monitoring network in a basin with intensive agriculture using artificial intelligence algorithms

Abstract

This research applies artificial intelligence algorithms for optimizing the water quality monitoring network in a representative basin with intensive agricultural and livestock activities. This study used the water quality database provided by the National Water Commission (CONAGUA). Bi-monthly monitoring was registered from 2013 to 2020 for 23 water quality parameters in 23 sampling locations in tributaries and the mainstream river. Therefore, it was necessary to apply principal component analysis to reduce the dimensionality of the data and thus identify the parameters that contribute most to the variation in the water quality. This artificial intelligence algorithm promoted the ease of clustering sampling sites with similar water quality characteristics by reducing the number of variables involved in the database. The reduction highlighted nutrients (TN and TP), parameters related to dissolved organic matter (NH3-N and TOC), and pathogens such as fecal coliforms. The similarity of sampling sites was determined through hierarchical clustering using the Euclidean distance as a measure of dissimilarity and the Ward method as a grouping method. As a result, nine clusters were obtained for the rainy and dry seasons, reducing approximately 50% of the sampling sites and generating an optimized network of 11 sampling sites.