AAAR 34th Annual Conference
October 12 - October 16, 2015
Hyatt Regency
Minneapolis, Minnesota, USA
Abstract View
Source Apportionment of PM2.5 in St. Louis Using Chemical Speciation Network Data
LI DU, Jay Turner, Washington University in St. Louis
Abstract Number: 599 Working Group: Source Apportionment
Abstract The USEPA Chemical Speciation Network (CSN) was implemented in 1999 to provide insights into fine particulate matter (PM2.5) chemical composition in urban areas and to support air quality standard compliance and health related studies. In this presentation, positive matrix factorization implemented by EPA PMF v5.0 was utilized to identify and apportion local and regional source contributions to PM2.5 mass at five CSN/CSN-protocol sites in metropolitan St. Louis with temporal coverage varying from 2 to 15 years. Regional sources collectively accounted for ~72-91% of the observed PM2.5 mass with secondary sulfate as the major contributor followed by biomass burning and secondary nitrate. A factor characterized by high calcium loading was resolved which exhibited characteristics of a regional-scale source instead of point sources as was concluded in previous studies for St. Louis. Impacts from local point sources such as a steel mill and a brass works were resolved in certain single site analyses. Source apportionment analysis using data pooled from multiple sites exhibited both advantages and disadvantages in resolving sources and quantifying their impacts at individual sites. For example, a traffic factor at the suburban sites was better defined upon pooling with data from an urban site. In contrast, the ability to identify point sources at a given site was diminished by the multisite analysis. Multisite analyses led to source contribution estimates that are biased from those derived by single-site analyses, especially for local point sources. In addition to the conventional methods of determining the optimal solutions, perturbation of the input uncertainty matrix was also applied and shown to provide information on the stability of the solutions. Modeling uncertainties were evaluated by bootstrap, displacement and bootstrap-displacement methods as implemented in the EPA PMF tool. The sensitivity of the outcomes from these uncertainty estimation methods to the structure of the solutions was explored.