Alr#*krk fnbimmwnr Vol. 21. No. 9. pp. 18774889. 1987. om-fMI/ti7 53.00+0.00 Printed in GUI Britain. Pcr#mmn Journals L:d. DETERMINATION OF SOURCES OF ATMOSPHERIC AEROSOL IN COPENHAGEN BASED ON RECEPTOR MODELS DORTE KRONBORG Statistical Research Unit, University of Copenhagen, Blegdamsvej 3, DK-2200 Copenhagen N, Denmark and FINN PALMGREN JENSEN, KRISTIAN KEIDING* and NIELS Z. HEIDAM National Agency of Environmental Protection, Air Pollution Laboratory, Ris# National Laboratory, DK- 4000 Roskilde, Denmark (Firsr received 22 Nouember 1985 and receiued for publication 6 March 1987) Abstract-A statistical analysis is presented of atmospheric aerosol element composition data collected at five monitoring stations in Copenhagen in 1983. The objective is to identify soucccs of air pollution, to estimate the variation of total suspended particles (TSP) for the different sources and to calculate the annual average source contribution to TSP. The basis for analysis is the chemical mass balance equations. Factor analysis methods are used to identify major sources of pollution and the contribution from each of the sources to the variation in TSP is estimated by an additional multiple regression. Finally, the influence of serial correlations between daily clement contributions is discussed and it is stressed that the result of the analysis is not seriously affected by these correlations. I. INTRODUCTlON This paper deals with the identification and quantitai- ive determination of sources of total suspended par- ticulate matter (TSP), and presents a statistical analysis of aerosol samples collected in Copenhagen during 1983. The approach is receptor oriented. Receptor models have been used widely and successfully in resolving aerosol element composition data. A review of the developments in this field is given by Gordon (1980). Henry et al. (1984) have given a review of statistical techniques used in analysing receptor models. The fundamental idea of receptor models is to apportion the ambient element concentrations at the receptor site to the different sources which contribute to the mass of total suspended particles at the receptor site. The source emission is assumed to be character- ized by its elemental composition and it is furthermore assumed that under transport from source to receptor the source signatures are preserved to such a degree that identification of sources from the element com- position at the receptor can be based on knowledge of the source element composition. In practice a number of sources with chemically almost similar aerosols are present around the receptor, and at the receptor site the source immission profile then represents the entire type of sources. l Present address: Environmental Engineering Laboratory, Department of Civil Enginccrin~ University of A&bore Sohngaardsholmsvcj 57, DK-9000 Aalborg, Denmark. In the receptor based model for the aerosol element composition the concentration of each element is a sum of the contributions from a number p of major sources each characterized by its source immission profiles plus a contribution from some unspecified sources such that xil = j$, OijBjr + ujf> (1) where xi, is the concentration of element i on day t, fljt is the total amount of particles from source j on day t, oijdenotes the mass fraction ofelement i in source jand Ui, is the portion of the concentration of the element not accounted for by the p sources. In matrix terms, the content of n elements may be written x, = @,+u,. (2) The domain of variation of the source immission profilesisOdoijdl,i=l ,..., n,j=l,..., pand i$, aij < 1. j = 1, . . . , p. By letting V, denote the amount of total suspended particles (TSP) on day t not amounted for by the p sources, the relation between the mass of TSP. xTsp, and the source contributions can be expressed as The main objective of the present study was to identify and to quantify sources of particles on the basis of the above relations.