Categorisation of air quality monitoring stations by evaluation of PM10 variability

Air Quality Monitoring Networks (AQMNs) are composed by a number of stations, which are typically classified as urban, suburban or rural, and background, industrial or traffic, depending on the location and the influence of the immediate surroundings. These categories are not necessarily homogeneous and distinct from one another, regarding the levels of the monitored pollutants. A classification providing groups with these features is of interest for air quality management and research purposes, and therefore, other classification criteria should be explored. In this work, the variations of PM10 concentrations in 43 stations in the AQMN of the Basque Country in the period 2005-2012 have been studied to group them according to common characteristics. The characteristic variations in time are synthesised by the autocorrelation function (ACF), with both daily and hourly data, and by the average diurnal evolution pattern of the normalised concentrations on a seasonal basis (Evol-P). A methodology based on k-means clustering of these features is proposed. Each classification gives a different piece of information that has been phenomenologically related with specific dispersion and emission dynamics. The classification based on Evol-Ps is found to be the most influential one when comparing PM10 levels between groups. A combination of these categorisations provides 5 groups with significantly different levels of PM10, improving the discrimination of the conventional classification. Our results indicate that the time series of the pollutant concentrations contain enough information to provide an objective classification of the monitoring stations in an AQMN. (C) 2015 Elsevier B.V. All rights reserved.