Application of Receptor Models to PM10 and PM2.5 Samples.

The human activities in their various aspects cause a change in the natural air quality. This change results more marked in very populated areas and in that with high industrialization. Some pollutants emitted are typical of a particular activity. Each source of pollution is identified by its profile in the composition of the emissions in the environment. Multivariate receptor models can be used in order to apportion pollutants to the different sources assessing the contribution of each source to the total pollution. This paper deals with the analysis, by Absolute Principal Component Scores (APCS) model, of PM10 and PM2.5 data collected during several projects (Comune di Bari project, SITECOS project, "P r o g r a m m a T r i e n n a l e p e r l a T u t e l a d e l l 'A m b i e n t e d e l l a R e g i o n e P u g l i a " p r o j e c t , e t c . ) c a r r i e d o u t i n Puglia Region during 2005-2006. In this paper a preliminary application of APCS model to a matrix of data collected in Bari during October 2005 is shown: it has been possible to identify five sources. Moreover the APCS model has been compared with two other multivariate receptor models: TTFA and NMF with different sparseness levels. The error, calculated by Frobenius norm, on the reconstructed concentration matrix is less when the APCS model is used.