Ambient aerosol composition by infrared spectroscopy and partial least squares in the chemical speciation network: Multilevel modeling for elemental carbon

Abstract

Fourier transform infrared spectroscopy (FT-IR) has been used to predict elemental carbon (EC) on polytetrafluoroethylene (PTFE) filter samples from the United States Environmental Protection Agency's Chemical Speciation Network (CSN). This study provides a proof-of-principle demonstration of using multilevel modeling to determine thermal/optical reflectance (TOR) equivalent EC (a.k.a., FT-IR EC) on PTFE samples collected in the CSN. Initially, spectra from nine geographically disperse sites were pooled and calibrated directly to collocated TOR EC measurements. The FT-IR EC quantified in test samples was deemed substandard when judged against an earlier study, e.g., R 2 D 0.760 and median absolute deviation (MAD) D 26.7%. Upon scrutinizing each sample's absolute prediction error and squared Mahalanobis distance, Elizabeth, NJ predictions were found to exhibit atypical systematic errors, motivating the development of a multilevel classification and calibration procedure. Atypical Elizabeth spectra were distinguished from the (typical) CSN spectra by training a partial least-square discriminant analysis. Predicting EC using calibrations dedicated to either atypical or typical samples produced a satisfactory improvement in overall performance (R 2 D 0.886, MAD D 19.8%). Analysis of the atypical FT-IR spectra and select TOR thermal fractions suggested that Elizabeth samples contained elevated levels of diesel particulate matter 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 was predominately determined by aliphatic C-H, C D C aromatic, and functional groups associated with oxidation. This study provides preliminary confirmation that FT-IR EC may be accurately determined from source-oriented calibrations under a combined classification and calibration methodology. EDITOR Paul Ziemann

Atmospheric particulate matter characterization by Fourier Transform Infrared spectroscopy: a review of statistical calibration strategies for carbonaceous aerosol quantification in US measurement networks

“…This finding suggests that variables eliminated for being uninformative in the calibration set samples may lead to undesirable oversimplification of a model that may be used with samples with potentially different composition, though this hypothesis has yet to be tested with calibrations developed with ambient measurements as reference, where the extent of extrapolation may not be so severe as with calibrations developed with laboratory standards. Weakley et al (2016Weakley et al ( , 2018a applied BMCUVE to second derivative or spline corrected 30 spectra in the CSN network. Improved MDL but otherwise similar performance metrics to the raw (full wavenumber) calibration model was obtained using the reduced model for TOR OC (performance described in Section 3.2.1), though the individual contributions of baseline correction and wavenumber selection to improvement in MDL was not investigated.…”

“…However, wavenumbers remaining after uninformative ones were eliminated (Section 3.3.2) differed when using baseline corrected and raw spectra 5 -even while the two maintained similar prediction performance. Weakley et al (2016) and Weakley et al (2018a) used the Savitzky-Golay method and spline correction method for TOR OC and EC, respectively, in the 2013 CSN network, but did not systematically investigate the isolated effect of baseline correction on predictions without additional processing. A formal comparison between the derivative method against raw and spline-corrected spectra have not been performed, but this is an area warranting further investigation.…”

“…While the nonanalyte regions of the spectra are implicitly assumed, the flexibility of the local splines combined with an iterative method for readjusting the non-analyte region effectively reduced the number of tuning parameters from four (in the global polynomial approach) to one. The spline baseline method was used for TOR EC prediction in 2013 CSN (Weakley et al, 2018a). Second derivative baseline correction method was applied to 2013 CSN TOR OC calibration (Weakley et al, 2016).…”

“…were attributed to C-C and C=C stretching in skeletal ring structures of aromatic or graphitic carbon Weakley et al, 2018a). While this absorption band corresponds to lattice vibrations in graphitic carbon (Tuinstra and Koenig, 1970) and commonly used in…”

“…For TOR analysis, VIP scores have been used to interpret wavenumber importance (Dillner and Takahama, 2015a, b;Weakley et al, 2016Weakley et al, , 2018a. VIP scores can also be used as a filtering method (Section 3.3.2) for wavenumber selection (e.g.,…”