Cluster analysis of roadside ultrafine particle size distributions

This study reports the diurnal, seasonal, and annual variation of ultrafine particle size distributions in downtown Toronto. The k-means clustering algorithm was applied to five years of size-resolved data for particles with diameters less than 100 nm. Continuous particle number concentrations were measured 16 m from a major arterial roadway between March 2006 and May 2011 using a Fast Mobility Particle Sizer. Eight particle size distribution (PSD) types were identified. The PSD types exhibited distinct weekday–weekend and diurnal patterns. The relative frequency that each PSD occurred varied with season and wind direction and was correlated with other pollutants. These temporal patterns and correlation helped in elucidating the sources and processes that each of the eight PSD represent. Finally, similar PSD types were observed in residential areas located 6 and 15 km away from the central monitoring site suggesting that these PSD types may be generalizable to other sites. Identification of PSD types was found to be a valuable tool to support the interpretation of PSD data so as to elucidate the sources and processes contributing to ultrafine particle concentrations.