Improved methods for performing multivariate analysis and deriving background spectra in atmospheric open-path FT-IR monitoring

Open-path Fourier transform infrared spectrometry (OP/FT-IR) may improve the temporal and spatial resolution in air pollutant measurements compared to conventional sampling methods. However, a successful OP/FT-IR operation requires an experienced analyst to resolve chemical interference as well as to derive a suitable background spectrum. The present study aims at developing a systematic method of handling the OP/FT-IR derived spectra for the measurement of photochemical oxidants and volatile organic compounds (VOCs) in urban areas. A classical least-squares (CLS) method, the most frequently used regression method in OP/FT-IR, is modified to constrain all the analyzed chemical species concentrations within a physically reasonable range. This new CLS method, named constrained CLS, may save the effort of predetermining the chemical species to be analyzed. A new background spectrum generation method is also introduced to more efficiently handle chemical interferences. Finally, CLS is shown to be prone to propagating errors in the case that a few data points contain a significant amount of error. The L1-norm minimization method reduces this error propagation to considerably increase the stability compared to CLS. The presently developed analysis software based on these approaches is compared with the other conventional CLS method using an artificially made single-beam spectrum as well as a field single-beam spectrum.