Data sets obtained from the quantitative analysis of 43 volatile organic compounds in air samples acquired every hour in 50 different sampling days (covering the different seasons) during the years 2005 and 2006 in a suburban area of the NW Spain have been investigated by different chemometric methods including matrix augmentation principal component analysis (MA-PCA) and parallel factor analysis (PARAFAC). The application of these two chemometric methods allowed the estimation of the chemical profiles of the main pollution sources operating over the investigated location and also unravelling their main temporal patterns. Using PCA, it was possible to identify four main different sources related to different VOC families (aromatic, aliphatic, halogenated and biogenic). However, the temporal emission patterns could not be properly resolved when the data of each individual sampling day were considered separately. When VOC emissions during the same week and for the whole year (50 sampling days) were simultaneously analysed by a new matrix augmentation PCA (MA-PCA) strategy and by PARAFAC, a better recovery and description of the daily and hourly temporal trends were achieved. In fact, the combination of MA-PCA with the score rearrangement followed by the first singular value decomposition allowed the extraction of daily cyclical emission trends that were subjacent in the original non-trilinear data matrix by means of a very simple methodology facilitating the interpretation of these complex data sets.

D. Pérez-Rial is very grateful to the Xunta de Galicia for the concession of a PhD Grant during the development of this work and to the Universidade da Coruña for a concession of a fellowship to develop a research stay in the Department of Environmental Chemistry of the IIQAB-CSIC (Barcelona).

Peer reviewed