Identification of Spectral Regions for Quantification of Red Wine Tannins with Fourier Transform Mid-Infrared Spectroscopy

Jacob Skibsted Jensen, Max Egebo, Anne S. Meyer

Research output: Contribution to journalJournal articleResearchpeer-review

Abstract

Accomplishment of fast tannin measurements is receiving increased interest as tannins are important for the mouthfeel and color properties of red wines. Fourier transform mid-infrared spectroscopy allows fast measurement of different wine components, but quantification of tannins is difficult due to interferences from spectral responses of other wine components. Four different variable selection tools were investigated for the identification of the most important spectral regions which would allow quantification of tannins from the spectra using partial least-squares regression. The study included the development of a new variable selection tool, iterative backward elimination of changeable size intervals PLS. The spectral regions identified by the different variable selection methods were not identical, but all included two regions (1485−1425 and 1060−995 cm−1), which therefore were concluded to be particularly important for tannin quantification. The spectral regions identified from the variable selection methods were used to develop calibration models. All four variable selection methods identified regions that allowed an improved quantitative prediction of tannins (RMSEP = 69−79 mg of CE/L; r = 0.93−0.94) as compared to a calibration model developed using all variables (RMSEP = 115 mg of CE/L; r = 0.87). Only minor differences in the performance of the variable selection methods were observed.
Original languageEnglish
JournalJournal of Agricultural and Food Chemistry
Volume56
Issue number10
Pages (from-to)3493-3499
ISSN0021-8561
DOIs
Publication statusPublished - 2008

Cite this

@article{0912213e5a714737af66ddd182f527f9,
title = "Identification of Spectral Regions for Quantification of Red Wine Tannins with Fourier Transform Mid-Infrared Spectroscopy",
abstract = "Accomplishment of fast tannin measurements is receiving increased interest as tannins are important for the mouthfeel and color properties of red wines. Fourier transform mid-infrared spectroscopy allows fast measurement of different wine components, but quantification of tannins is difficult due to interferences from spectral responses of other wine components. Four different variable selection tools were investigated for the identification of the most important spectral regions which would allow quantification of tannins from the spectra using partial least-squares regression. The study included the development of a new variable selection tool, iterative backward elimination of changeable size intervals PLS. The spectral regions identified by the different variable selection methods were not identical, but all included two regions (1485−1425 and 1060−995 cm−1), which therefore were concluded to be particularly important for tannin quantification. The spectral regions identified from the variable selection methods were used to develop calibration models. All four variable selection methods identified regions that allowed an improved quantitative prediction of tannins (RMSEP = 69−79 mg of CE/L; r = 0.93−0.94) as compared to a calibration model developed using all variables (RMSEP = 115 mg of CE/L; r = 0.87). Only minor differences in the performance of the variable selection methods were observed.",
author = "Jensen, {Jacob Skibsted} and Max Egebo and Meyer, {Anne S.}",
year = "2008",
doi = "10.1021/jf703573f",
language = "English",
volume = "56",
pages = "3493--3499",
journal = "Journal of Agricultural and Food Chemistry",
issn = "0021-8561",
publisher = "American Chemical Society",
number = "10",

}

Identification of Spectral Regions for Quantification of Red Wine Tannins with Fourier Transform Mid-Infrared Spectroscopy. / Jensen, Jacob Skibsted; Egebo, Max; Meyer, Anne S.

In: Journal of Agricultural and Food Chemistry, Vol. 56, No. 10, 2008, p. 3493-3499.

Research output: Contribution to journalJournal articleResearchpeer-review

TY - JOUR

T1 - Identification of Spectral Regions for Quantification of Red Wine Tannins with Fourier Transform Mid-Infrared Spectroscopy

AU - Jensen, Jacob Skibsted

AU - Egebo, Max

AU - Meyer, Anne S.

PY - 2008

Y1 - 2008

N2 - Accomplishment of fast tannin measurements is receiving increased interest as tannins are important for the mouthfeel and color properties of red wines. Fourier transform mid-infrared spectroscopy allows fast measurement of different wine components, but quantification of tannins is difficult due to interferences from spectral responses of other wine components. Four different variable selection tools were investigated for the identification of the most important spectral regions which would allow quantification of tannins from the spectra using partial least-squares regression. The study included the development of a new variable selection tool, iterative backward elimination of changeable size intervals PLS. The spectral regions identified by the different variable selection methods were not identical, but all included two regions (1485−1425 and 1060−995 cm−1), which therefore were concluded to be particularly important for tannin quantification. The spectral regions identified from the variable selection methods were used to develop calibration models. All four variable selection methods identified regions that allowed an improved quantitative prediction of tannins (RMSEP = 69−79 mg of CE/L; r = 0.93−0.94) as compared to a calibration model developed using all variables (RMSEP = 115 mg of CE/L; r = 0.87). Only minor differences in the performance of the variable selection methods were observed.

AB - Accomplishment of fast tannin measurements is receiving increased interest as tannins are important for the mouthfeel and color properties of red wines. Fourier transform mid-infrared spectroscopy allows fast measurement of different wine components, but quantification of tannins is difficult due to interferences from spectral responses of other wine components. Four different variable selection tools were investigated for the identification of the most important spectral regions which would allow quantification of tannins from the spectra using partial least-squares regression. The study included the development of a new variable selection tool, iterative backward elimination of changeable size intervals PLS. The spectral regions identified by the different variable selection methods were not identical, but all included two regions (1485−1425 and 1060−995 cm−1), which therefore were concluded to be particularly important for tannin quantification. The spectral regions identified from the variable selection methods were used to develop calibration models. All four variable selection methods identified regions that allowed an improved quantitative prediction of tannins (RMSEP = 69−79 mg of CE/L; r = 0.93−0.94) as compared to a calibration model developed using all variables (RMSEP = 115 mg of CE/L; r = 0.87). Only minor differences in the performance of the variable selection methods were observed.

U2 - 10.1021/jf703573f

DO - 10.1021/jf703573f

M3 - Journal article

VL - 56

SP - 3493

EP - 3499

JO - Journal of Agricultural and Food Chemistry

JF - Journal of Agricultural and Food Chemistry

SN - 0021-8561

IS - 10

ER -