Atmospheric Aerosol Size Distribution at Sumas Mountain, Vancouver, BC.

Atmospheric size distributions provide fundamental information for studying atmospheric particle physics. To study and monitor the air quality in the Lower Fraser Valley, Vancouver, BC, Pacific 2001 field study was held in Vancouver during August 13th to September 1st 2001. As part of the Pacific 2001 study, aerosol size distributions were measured at Eagle Ridge, just east of Abbotsford, BC. Atmospheric particles were sampled from a 3/8-inch stainless steel tubing with an inverted U-shape inlet. The sample line located at 2 meters above the top of a trailer that has an elevation of 300 meters above sea level. Particles coming from the sample line were sized using a TSI 3071 Differential Mobility Analysis (DMA) and then counted by a TSI 3010 Condensation Nucleus Counter (CNC). The system measured atmospheric particles with diameters from 9.3 to 604 nm. Each five-minute scan consisted of 30 size bins increasing exponentially in size. All the data was corrected for CNC counting efficiency and multiple charging effects to represent a more realistic description of the size distribution. During the field study, 19 days of data for both sunny and rainy days were obtained. By comparing the measured total concentration from 10 nm and 6 nm onwards, identified local nucleation event was observed in 4 days. Clear observation of particle transport from other remote places to the monitoring site was observed in 13 days during the study. Nucleation mode particles were usually observed between 10 a.m. to noon in all sunny days. These nucleation mode particles are thought to due to nocturnal inversion. As that happen, pollutants that were accumulated overnight in the valley were transported up to the site, with the combination of the sunlight, trigger the formation of new particles. The formation of these nucleation mode particles was usually followed by a short particle growth period, which may last from two to five hours, except for one day, which this growth period extends to midnight. Owing to the high variability of the size distribution, to fit these data using multi-LogNormal distributions may result instability problem. To accomplish this, we have explored a new method to analyse the data. Applying the Principal Component Analysis (PCA) to the data may provide more stable distribution for describing the data. Furthermore, the PCA also provide a mean for reducing a massive data set to a manageable data set.