Source apportionment of particulate matter in urban aerosol

Source apportionment of particulate matter in urban aerosol Angeliki Karanasiou Institute of Nuclear Technology and Radiation Protection, Environmental Radioactivity Laboratory, N. C. S. R. Demokritos, Athens, Greece

Background • Athens has significant air pollution problems • Non-attainment of the EU Air Quality Standard for Particulate Matter is frequent • Receptor models applied to aerosol chemical composition data can identify the source types (Hopke, 2003) The fundamental principle of receptor modelling is that mass conservation can be assumed and a mass balance analysis can be used to identify and apportion sources of airborne particulate matter in the atmosphere

Receptor Modelling A mass balance equation can be written to account for all m chemical species in the n samples as contributions from p independent sources Where i = 1, …, n samples, j = 1, …, m species and k =1, …, p sources Sources Known Sources Unknown • Chemical Mass Balance • Principal Component Analysis (Watson et al. , 1990) (Thurston and Spengler, 1985) • Unmix (Henry, 2000) • Positive Matrix Factorization (Paatero and Tapper, 1993; Paatero, 1997)

Positive Matrix Factorization Aerosol Mass Balance source profile, μg/μg X n m = G n p F j p i = 1, …, n samples j = 1, …, m species k =1, …, p sources (user specified) source contribution, μg/m 3 Sij: uncertainty in Xij

Aerosol Sampling • Sampling was conducted on three sites located in Athens urban area • Two aerosol-sampling campaigns at each site were performed during March - December 2002 in Athens, covering the cold and warm period of the year • PM 10 and PM 2. 1 samples were collected simultaneously over a 24 h period (55 samples) • Aerosol components determined: metallic elements, sulphate, black carbon Positive Matrix Factorization, PMF 2 Two separate datasets Fine aerosol specie concentration in PM 2 Coarse aerosol specie concentration in PM 10 -PM 2

Aerosol concentrations

Important steps in PMF Sources Preliminary runs to select the number of factors Variables S/N>2: strong variables Uncertainty C 1: Combined standard uncertainty SR: variable reproducibility aj: deviation from the true value C 2: Sampling uncertainty

Aerosol sources Fine aerosol Coarse aerosol • Road dust • Vehicles • Biomass burning • Marine aerosol • Oil combustion • Road dust • Soil • Marine aerosol

Fine aerosol sources Road dust BC, SO 42 -, crustal metals Vehicles diesel and gasoline exhaust emissions

Fine aerosol sources Marine aerosol BC pulled towards zero, Fkey (Wilson, 1975) Biomass burning mixed source? secondary aerosol Dall’Osto and Harrison, 2006

Fine aerosol sources Oil combustion Fe: fuel oil (V, Ni) Samara et al, 2005, Atm Environ, 39, 6430 -6443

Coarse aerosol sources Soil Scheff and Valiozis, Atm Environ, 24, 203 -211, 1990 Road dust

Coarse aerosol sources Marine aerosol (Wilson, 1975)

Sources – fine and coarse mode Certain sources existed in both coarse and fine fractions Fine fraction includes a ‘tail end’ of the coarse mode Karanasiou et al. , 2007. Atmospheric Environment, 41, 2368– 2381.

Positive Matrix Factorization in PM 10 Marine aerosol Vehicles Unidentified Biomass burning Unidentified Road/Soil dust

Tracers Pearson correlation coefficient between source contributions and variable concentrations üRoad dust: Ca üBiomass burning: K üMarine aerosol: Na üOil combustion: Fe üVehicles: Ni üCoarse aerosol: no significant correlations between source contributions and variable concentrations

Wind direction vs Source contribution Fuel oil Vehicles Marine aerosol Biomass burning Road dust • No directional dependence • Sources are dispersed in Athens basin

Conclusions üPMF resolved the major source types üFive sources were identified in fine aerosol representing Road dust, Vehicles, Biomass burning, Marine aerosol and Oil combustion ü Coarse aerosol sources were Road dust, Soil and Marine aerosol üAerosol sources were not clearly identified in PM 10 data set üNi proved to be a good tracer of vehicle emissions in Athens

Take home messages ü Receptor models can provide valuable information on aerosol sources ü PM 10 monitoring could not provide adequate information on aerosol sources üAerosol organic compounds could yield a resolved factor based on the source of the tracer Acknowledgements Special thanks to Prof. Pentti Paatero for the fruitful discussions and suggestions on PMF