Classification of Multi-pollutant Peak Events in Mobile Monitoring Data

摘要

Background: Proximity to major roadways is associated with both chronic and short-term adverse health outcomes. Near-roadway exposures are characterized in part by transient events that are several orders of magnitude above background. We hypothesize that the composition, frequency and duration of these peaks is a characteristic of a roadway and is associated with traffic composition, driving conditions and atmospheric reactivity. METHODS: The University of Washington mobile monitoring platform sampled 16 pollutants in 10 second intervals over 2-week periods in summer and winter 2012 in Baltimore, MD. Multi-pollutant peaks were defined as time periods where a 10 s increase in two or more independently measured pollutants were outlier values (z＞3.5). The peaks were grouped based on the correlation structure using K-means analysis to identify characteristic pollutant profiles of these events. RESULTS: There were 89 summer and 72 winter multi-pollutant peaks, evenly distributed across days but not geographically. Four distinct profiles were identified with cluster analysis. One-third of all peaks were characterized by extreme increases in both BC and NO2. These peaks occurred on roadways with high truck traffic. Another fraction of events was associated with rapid increases in NOx and O3 suggesting the plumes are subject to photochemical processes. In winter, a peak was identified with high increases in NOx and particle-bound PAH; events likely associated with a combustion source. Lastly, in both summer and winter some peaks were associated with increases in only PM10 and PM2.5 likely indicating re-suspension of road dust. CONCLUSIONS: Analyzing extreme events in mobile monitoring data yields unique profiles, attributable to different roadway processes. Some events relate to primary sources, while others reflect on-roadway chemical transformations. The frequency of these peaks and their composition is linked to important roadway and traffic features.