ARS | Publication request: Using multiple calibration sets to improve the quantitative accuracy of partial least squares (PLS) regression on open-path fourier transform infrared (OP/FT-IR) spectra of ammonia over wide concentration ranges

Submitted to: Applied Spectroscopy
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: April 22, 2011
Publication Date: July 8, 2011
Citation: Shao, L., Liu, B., Griffiths, P.R., Leytem, A.B. 2011. Using multiple calibration sets to improve the quantitative accuracy of partial least squares (PLS) regression on open-path fourier transform infrared (OP/FT-IR) spectra of ammonia over wide concentration ranges. Applied Spectroscopy. 65(7):820-824.

Interpretive Summary: A technique of using multiple calibration sets in partial least squares regression (PLS) was proposed to improve the quantitative determination of ammonia from open-path Fourier transform infrared spectra. The spectra were measured near animal farms, and the path-integrated concentration of ammonia fluctuated from nearly zero to a high of about 1000 ppm-m. The PLS regression using a single calibration set over such a large concentration range had decreased quantitative accuracy due to the nonlinear relationship between concentration and absorbance. In the PLS regression with multiple calibration sets, each calibration set accounted for a smaller concentration range, which significantly decreased the serious nonlinearity problem in PLS regression with a single calibration set. It was found that it was possible to build the multiple calibration sets easily and efficiently without extra measurements. The PLS regression with multiple calibration sets is more flexible and accurate than in the case when a single calibration set is used to process the spectra measured around animal farms where the concentration of ammonia varies significantly. The relative error was reduced from about 6% to below 2%, and the best results were obtained with 4 calibration sets.

Technical Abstract: A technique of using multiple calibration sets in partial least squares regression (PLS) was proposed to improve the quantitative determination of ammonia from open-path Fourier transform infrared spectra. The spectra were measured near animal farms, and the path-integrated concentration of ammonia fluctuated from nearly zero to a high of about 1000 ppm-m. The PLS regression using a single calibration set over such a large concentration range had decreased quantitative accuracy due to the nonlinear relationship between concentration and absorbance. In the PLS regression with multiple calibration sets, each calibration set accounted for a smaller concentration range, which significantly decreased the serious nonlinearity problem in PLS regression with a single calibration set. The relative error was reduced from about 6% to below 2%, and the best results were obtained with 4 calibration sets, each covering one quarter of the entire concentration range. It was also found it was possible to build the multiple calibration sets easily and efficiently without extra measurements.