10th International Aerosol Conference September 2 - September 7, 2018 America's Center Convention Complex St. Louis, Missouri, USA
Abstract View
Fourier-transform Infrared Determination of Organic and Elemental Carbon: Anomalous Samples and What They Tell Us about Composition and Sources
ANN DILLNER, Andrew Weakley, Bruno Debus, Satoshi Takahama, University of California, Davis
Abstract Number: 1359 Working Group: Carbonaceous Aerosol
Abstract Several studies have demonstrated that thermal optical reflectance (TOR) OC and EC in fine aerosol can be predicted from Fourier transform infrared (FT-IR) spectra of aerosol samples collected on polytetrafluoroethylene (PTFE) filters. Calibrations for OC and EC are developed using partial least squares regression and provide equivalent OC and EC mass with accuracy and precision comparable to TOR. For most samples, one calibration model developed from samples from multiple sites is needed to predict OC and EC across sites and seasons. Here, we discuss cases in which multi-level modeling is needed for high quality predictions. For example, in the Chemical Speciation Network (CSN), which provides speciated particulate matter concentrations at urban sites, samples from one site with (presumably) distinctive EC sources are not well predicted by the calibration developed for all samples. To accurately predict EC on these anomalous samples, “ordinary” EC is distinguished from more “anomalous” EC using the sample’s FT-IR spectrum and a classification algorithm (partial least squares discriminant analysis). Samples collected from the Elizabeth, NJ sampler, located near a major tollway, often contained anomalous EC. Further analysis of FT-IR spectra and TOR fractions suggested that Elizabeth samples likely contained elevated levels of EC from fresh diesel exhaust as evidenced by the use of organic nitrogen functional groups for prediction, very low average OC/EC, and minimal charring during TOR speciation. FT-IR EC from the other eight sites in this study was predominately determined by aliphatic C-H, C=C aromatic, and functional groups associated with oxidation. Two calibration models were developed, one for ordinary and one for anomalous spectra. Classified spectra were allocated to their associated calibration resulting in a significant reduction in prediction error (30% to 21%; p<0.05). This study provides insights into appropriate tools for accounting for significantly different chemical compositions at different sites and into the character of carbonaceous aerosol at heavily diesel impact sites as compared to more typical urban environments. Additional examples from the Interagency Monitoring to Protect Visual Environments (IMPROVE) network will be presented.