Abstract
Single particle analysis of an industrially polluted atmosphere in Port Talbot, South Wales, United Kingdom was conducted using Aerosol-Time-of-Flight Mass Spectrometry (ATOFMS). During the four week sampling campaign, a total of 5,162,018 particles were sized in the size range 0.2-1.9 mu m aerodynamic diameter. Of these, 580,798 were successfully ionized generating mass spectra. K-means clustering employed for analysing ATOFMS data utilized 96% of the hit particles to generate 20 clusters. Similar clusters were merged together and 17 clusters were generated from which 7 main particle groups were identified. The particle classes include: K-rich particles (K-CN, K-NO3, K-EC, K-Cl-PO3 and K-HSO4), aged sea salt (Na-NO3), silicate dust (Na-HSiO2), sulphate rich particles (K-HSO4), nitrate rich particles (AlO-NO3), Ca particles (Ca-NO3), carbon-rich particles (Mn-OC, Metallic-EC, EC, EC-NO3 and CC-EC), and aromatic hydrocarbon particles (Arom-CN, Fe-PAH-NO3 and PAH-CN). With the aid of wind sector plots, the K-Cl-PO3 and Na-HSiO2 particle clusters were related to the steelworks blast furnace/sinter plant while Ca-rich particles arose from blast furnace emissions. K-CN, K-EC, Na-HSiO2, K-HSO4, Mn-OC, Arom-CN, Fe-PAH-NO3, and PAH-CN particles were closely linked with emissions from the cokemaking and mills (hot and cold) steelworks sections. The source factors identified by the ATOFMS were compared with those derived from multivariate analysis using Multilinear Engine (ME-2) applied to filter samples analysed off-line. Both methods of source apportionment identified common source factors including those within the steelworks (blast furnace, sinter, cokemaking), as well as marine, traffic and secondary particles, but quantitative attribution of mass is very different. (C) 2014 Elsevier Ltd. All rights reserved.