Identification of Redundant Air Quality Monitoring Stations using Robust Principal Component Analysis

Abstract

Air quality monitoring stations are essentials for monitoring air pollutants and, therefore, are essential to protect the public health and the environment from the adverse effects of air pollution. Two or more stations may monitor the same pollutant behavior. In this scenario, the equipment must be reallocated to provide a better use of public resources and to enlarge the monitored area. The identification of redundant stations can be carried out by the application of principal component analysis (PCA) as a grouping technique. The principal component analysis is a set of linear combinations of the original variables constructed to explain the variance–covariance structure of the data. It is well known that outliers affect the covariance structure of the variables. Since the components are computed by using the covariance or the correlation matrix, the outliers also affect the properties of the components. This article proposes a grouping methodology that applies robust PCA to identify air quality monitoring stations that present similar behavior for any pollutant or meteorological measure. To illustrate the usefulness of the proposed methodology, the robust PCA is applied to the management of the automatic air quality monitoring network of the Greater Vitoria Region in Brazil that consists of 8 stations. It was found that four components could explain 84% of the total variability, and it is possible to create a group composed of at least two stations in each one of the components. Therefore, the redundant stations can be installed in a new site to expand the monitored area.