PMF5.0 vs. CMB8.2: An inter-comparison study based on the new European SPECIEUROPE database

Abstract

Receptor Models are tools widely adopted in source apportionment studies. We describe here an experiment in which we integrated two different approaches, i.e. Positive Matrix Factorization (PMF) and Chemical Mass Balance (CMB) to apportion a set of PM10 (i.e. Particulate Matter with aerodynamic diameter lower than 10μm) concentration values. The study was performed in the city of Genoa (Italy): a sampling campaign was carried out collecting daily PM10 samples for about two months in an urban background site. PM10 was collected on Quartz fiber filters by a low-volume sampler. A quite complete speciation of PM samples was obtained via Energy Dispersive-X Ray Fluorescence (ED-XRF, for elements), Ionic Chromatography (IC, for major ions and levoglucosan), thermo-optical Analysis (TOT, for organic and elemental carbon). The chemical analyses provided the input database for source apportionment by both PMF and CMB. Source profiles were directly calculated from the input data by PMF while in the CMB runs they were first calculated by averaging the profiles of similar sources collected in the European database SPECIEUROPE. Differences between the two receptor models emerged in particular with PM10 sources linked to very local processes. For this reason, PMF source profiles were adopted in refined CMB runs thus testing a new hybrid approach. Finally, PMF and the “tuned” CMB showed a better agreement even if some discrepancies could not completely been resolved. In this work, we compared the results coming from the last available PMF and CMB versions applied on a set of PM10 samples. Input profiles used in CMB analysis were obtained by averaging the profiles of the new European SPECIEUROPE database. The main differences between PMF and CMB results were linked to very local processes: we obtained the best solution by integrating the two different approaches with the implementation of some output PMF profiles to CMB runs.