AAAR 37th Annual Conference October 14 - October 18, 2019 Oregon Convention Center Portland, Oregon, USA
Abstract View
Excitation-Emission Matrix Fluorescence Spectroscopy for Source Apportionment of Combustion Sources: Comparison to Positive Matrix Factorization Results from an Exposure Assessment Panel Study
JAY RUTHERFORD, Timothy Larson, Edmund Seto, Igor Novosselov, Jonathan Posner, University of Washington
Abstract Number: 726 Working Group: Source Apportionment
Abstract Although many consider the link between particulate matter (PM) and negative health effects as a well-established fact, the significance and estimated severity of these health effects have been a topic of recent debate. To address these ongoing questions and to help inform effective mitigation strategies into the future, further study of PM health effects is necessary. The development of low-cost particle counters has enabled highly resolved spatial and personal measurement of PM, but these devices don’t allow for source apportionment.
Here we present a method for identifying PM source using excitation-emission matrix (EEM) fluorescence spectroscopy and machine learning algorithms aimed at enabling low-cost PM source apportionment. Previously we presented an EEM-machine learning method to identify combustion generated PM2.5 from wood smoke, diesel soot, and cigarette smoke from personal exposure monitoring filters collected in a laboratory environment. In the present work, we apply the same EEM-machine learning methodology to identify combustion sources present in PM samples collected during an exposure assessment panel study. We obtained archived field samples from the panel study and the associated positive matrix factorization (PMF) source apportionment results that used elemental analysis by X-ray fluorescence and light absorbing carbon measurements. We show that EEM spectra from cyclohexane extracts of these filter samples can be used to predict the same combustion sources determined by PMF with R2 values up to 0.84. The use of this EEM-machine learning approach may be used to conduct PM exposure studies that include source apportionment of combustion sources at a low-cost.