Use of advanced receptor modelling for analysis of an intensive 5-week aerosol sampling campaign

This study investigates the application of the Multilinear Engine (ME) to a highly time-resolved aerosol concentration data set. Aerosol concentration, ambient gas and meteorological data were collected from 20 August to 25 September 2003 in Toronto, Ont., Canada. A suite of instruments was incorporated into the study, which produced a diverse air quality data set. The instrumentation included an aerosol mass spectrometer (AMS) that was used to analyse for non-refractory submicron particles and an aethelometer used to collect black carbon (BC) measurements, both at a time resolution of 15 min. A particles-into-liquid sampler coupled with an ion chromatrograph (PILS–IC) collected 15-min data on water-soluble inorganic aerosols during a portion of the total sampling period. The ME was applied to the AMS and aethelometer data, along with meteorological and time-of-day information. Five sources were identified: secondary sulphate characterized by high SO42− concentration and winds from the southwest; secondary nitrate characterized by high NO3− concentration and a significant diurnal trend with a peak in concentration during the morning; fresh organic (1) associated with BC and elevated concentration in the morning suggesting that this source was attributable to local vehicle traffic; fresh organic (2) which was not associated with a strong diurnal or wind direction trend; and aged organic aerosol which was associated with m/z=44m/z=44, showed a multiday pattern of elevated concentration and was not strongly associated with any particular wind direction. This study also investigated the minimum number of days of 15-min time-resolved data which were required by the ME to produce meaningful source identification. It was determined that 6 days of data was the minimum required to produce source profile and temporal trends which were comparable to the ME results using the full sampling period.