Monitoramento de qualidade de água baseado em análise por componentes principais : uma abordagem local

Abstract

Surface water quality monitoring has gained more attention in recent years due to the increasing degradation of aquatic environments. The data obtained from this monitoring is essential for the management of water resources, whose challenge lies in transforming them into relevant information. Often, the difficulty exists not because of the amount of data, but due to the complex existing interactions in watercourses and to the multivariate nature of the data. In this sense, the objective of this work was to apply the statistical multivariate technique, called Principal Component Analysis (PCA), with a focus on three different applications. The methodologies were proposed to be applied to a monitoring station individually (local approach), allowing a detailed description of the water quality conditions in a specific region. As a case study, secondary data from 12 physical-chemical and microbiological parameters were used, collected over 21 years (1997-2018) in a sampling station (BV137) identified in previous studies as a priority for pollution monitoring and control in the Velhas River. In the first application, the PCA model was able to identify critical samples in terms of pollution, and through score plots, it was possible to associate spatial regions in them with environmental pressure factors characteristic of the study area (discharge of domestic and industrial effluents and mining activities). In the second application, the multivariate control charts for the 𝑇 2 and 𝑄 statistics were able to identify 79.73% and 72.97% of the extreme data, respectively, showing promise in monitoring for the detection of critical water quality conditions. In the third application, analyzes involving PCA and non-parametric tests were performed to assess the temporal data variability. The results indicated that three main principal components were responsible for explaining 62.47% of the total variability in water quality, and that January, July and October were the months that most contributed to explain the variability of the data. The analyzes performed can be used to support decision making in water quality monitoring and management activities providing means to handle and evaluate the effectiveness of pollution control and prevention policies.