Analysis of functional groups in atmospheric aerosols by infrared spectroscopy: ElnetPLS model for statistical selection of relevant absorption bands for OC predictions

Organic carbon (OC) is a major component of atmospheric particulate matter (PM). Typically OC concentrations are measured using thermal methods such as thermal-optical reflectance (TOR) from samples collected on quartz filters. However, TOR measurements are destructive and expensive. We estimate TOR OC concentrations using Fourier transform infrared (FT-IR) spectra of ambient samples collected on Teflon filter. We have developed a sparse statistical calibration model (ElnetPLS), which excludes unnecessary wavenumbers from infrared spectra during the model building process, permitting the identification and evaluation of the most relevant vibrational modes of molecules in complex aerosol mixtures associated with reported TOR OC concentrations. The sparsest ElnetPLS model has similar model performances of the full (2784) wavenumber models while requiring only ten wavenumbers associated with carbonyl groups