Fluorescence fingerprinting to predict multiple chromatic and phenolic parameters for a diversified set of red wines

Absorbance-transmittance and fluorescence excitation-emission matrix (A-TEEM) spectroscopy was utilised to study red wines (n = 113) undergoing maturation or bottle ageing. Chromatic and phenolic parameters including wine colour density, hue, monomeric pigments, various forms of polymeric pigments, tannin concentration using precipitation assays, SO₂ resistant pigments, total anthocyanins, and total phenolics were modelled. Machine learning models using extreme gradient boosting (XGB) regression were developed for EEM data and fused A-TEEM data based on reference chemistry, achieving high R² and low RMSE values of Train and cross-validation (CV) models across all parameters. XGBR Test models based on the A-TEEM dataset showed enhanced prediction accuracy for certain parameters than using EEMs alone. Overall, this study provided insights into effectively determining colour and phenolic profile of wines with a rapid fluorescence spectroscopy approach, offering a valuable tool that could be used to facilitate research, wine quality control, and decision-making in wineries.