Advances in Data Processing for Open-path Fourier Transform Infrared Spectrometry of Greenhouse Gases

Authors: SHAO, LIMIN - University Of Science And Technology Of China; GRIFFITHS, PETER - University Of Idaho; Leytem, April

Submitted to: Analytical Chemistry
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 9/25/2010
Publication Date: 9/30/2010
Citation: Shao, L., Griffiths, P.R., Leytem, A.B. 2010. Advances in Data Processing for Open-path Fourier Transform Infrared Spectrometry of Greenhouse Gases. Analytical Chemistry. 82(19):8027-8033.

Interpretive Summary: Open-path Fourier transform infrared (OP/FT-IR) spectrometry is an almost ideal technique for measuring the concentration of greenhouse gases as it is reasonably sensitive, non-invasive, fast, capable of determining multiple compounds simultaneously, and only measures those compounds that contribute to global warming. Furthermore, the instrumentation is rugged and relatively easy to handle in the field. In the past decade, the application of OP/FT-IR spectrometry for atmospheric monitoring has seen a significant increase in Asian countries such as China and South Korea. However, on a worldwide scale, and especially in the USA, this technique is still undergoing a slow acceptance, largely because of the difficulty of processing the data when commercial instruments are used. The atmosphere is a complex, multi-component system, the study of which is complicated by uncontrolled or unpredicted factors such as wind, rain, snow and dust. As a result, although a typical OP/FT-IR spectrum contains the information on the path-integrated concentration of greenhouse gases, the analysis of these molecules is hampered by various types of interference in addition to the omnipresent absorption of the beam by water vapor and CO2. Furthermore, the chemometric technique that is usually employed, namely classical least squares (CLS) regression, is far more accurate when all molecules in the region being investigated (including the lines in the spectrum of water vapor) obey Beer’s Law, which is seldom the case with atmospheric spectra where interfering water lines are often very intense and are measured with an instrument line shape function that is broader than the lines. As a result, it is often necessary to only use very short regions of the spectrum between the stronger water lines to perform the analysis. A more general and user-friendly software program for OP/FT-IR spectroscopy that is not based on CLS regression would increase the quality and automation of the data processing, and enable researchers without spectroscopic expertise to use the OP/FT-IR technology for atmospheric monitoring more readily. Today, with a factor-based chemometric approach such as partial least squares (PLS) regression, the effect of Beer’s law nonlinearity can be compensated by the addition of additional factors to the model. The effect of these atmospheric interferences is clearly not handled adequately when the data are processed by CLS. The regions that are used in the PLS calculations are far wider and were selected in part because the strongest lines in the spectra of the analytes were found in these regions. Even though there is significant overlap by rotational lines due to lines in both the water and CO2 spectra, their effect is taken care of much better using the PLS approach and the result of these calculations appears to be very accurate. The result is a software package (available on request) with a straightforward graphical user interface, which provides both the analytical results and all the intermediate data including the processed interferograms, single-beam spectra, and absorbance spectra in popular format for easy access.

Technical Abstract: The automated quantification of three greenhouse gases, ammonia, methane and nitrous oxide, in the vicinity of a large dairy farm by open-path Fourier transform infrared (OP/FT-IR) spectrometry at intervals of 5 minutes is demonstrated. Spectral pretreatment, including the detection and correction of the effect of interrupting the infrared beam is by a moving object, and ways of correcting for the effect, and correction for the nonlinear detector response are applied to the measured interferograms. Two ways of obtaining quantitative data from OP/FT-IR data are described. The first, which is installed in commercial OP/FT-IR spectrometers, is based on classical least squares (CLS) regression and the second is based on partial least squares (PLS) regression. It is shown that CLS regression only gives accurate results if the absorption features of the analytes are located in very short regions where lines due to atmospheric water vapor are absent; of the three analytes examined, only ammonia fell into this category. On the other hand, PLS regression allowed what appeared to be accurate results to be obtained for all three analytes.