Effect of spectral pre-processing methods on the evaluation of the color components of red wines using Fourier-Transform Infrared spectroscopy

A rapid method for measuring the (i) free anthocyanins, (ii) copigmented anthocyanins, (iii) polymeric pigment fractions and (iv) actual red color at 520 nm from the Fourier transform infrared (FT-IR) spectrum was evaluated. Twenty wines, mostly one-year old, with red color ranging from almost zero to 10 absorbance units (AU) at 520 nm were analyzed by reference methods for their levels of the selected color components. Partial least squares (PLS) regression, obtained from several spectral data preprocessing methods, was used for multivariate calibration to relate the chemical variables to the FT-IR spectra. The performance of different models was compared in terms of root mean square error of prediction for cross-validation (RMSECV) and coefficient of determination (RI). The best results were obtained using vector normalization and no-preprocessing of spectral data. FTIR measured the copigmentation of wine with 5.7% RMSECV only.