Quality control of semi-continuous mobility size-fractionated particle number concentration data

Abstract

Abstract Fine and ultrafine particles have been postulated to play an important role in the association between ambient particulate matters and adverse health effects. As part of the EPA Supersite Program, the Southern California Particle Center & Supersite has conducted a series of monitoring campaigns that contribute to a better understanding of the sources, chemical composition and physical state of ambient aerosols. The Scanning Mobility Particle Sizer (SMPS) was deployed to semi-continuously measure mobility size-fractioned particle number concentrations. As part of the quality control efforts, we developed a two-stage graphic and statistical procedure to label and identify potentially discordant observations. The first stage considered the entire size-fractionated data by date-time as a whole to plot total concentration (TC) vs. coefficient of variation (CV), both in log scale. TC represents the magnitude of overall concentration for a size distribution; while CV represents the relative variability. This plot was used to partition all size distributions into four to five distinct regions. In each region, a generalized extreme studentized deviate (ESD) and a modified Z -score procedure were applied to identify potential discordant outliers. We have found that the majority of particle size distributions are concentrated within a ‘normal’ region, with TC ranging from 10 2 to 10 5 cm −3 and CV varying between 20% and 200%. Size distributions that are contaminated with discordant outliers are displayed distinctly from the ‘normal’ region and form four to five clusters in the Log TC–Log CV plot. The pattern of clusters in the plot is consistent among the four sampling sites in this study, suggesting the robustness of this technique. The generalized ESD and modified Z -score effectively identify discordant outliers and reveal that the pattern of clustering outliers are consistent within each distinct region. It has, thus, been concluded that the new approach is a useful quality control tool to identify potential discordant outliers in SMPS data.