Abstract
This study assessed the discriminative potential of leg meat from White Kołuda geese fed two diets using volatile compounds and fatty acids as predictors, applying calibration, internal cross-validation, and external cross-validation over four years of data. Key discriminant VOCs in meat were identified, together with discriminant FAs, lipid parameters, and amino acids as precursors of meat VOCs. Spectral models based on volatile compounds showed similar performance for calibration and internal validation, but external cross-validation accuracies were nearly 30 percentage points lower, highlighting the importance of realistic validation. A more complex SRN-based classification algorithm did not improve performance. The most reliable evaluation used an external cross-validation strategy based on the “one-year vs. three-year” approach. Fatty acid predictors expressed as peak-area percentages yielded the best results, with external cross-validation accuracies of 98.1%, outperforming volatile compound predictors. These fatty acid-based models provide robust, reliable classification of goose leg meat according to diet.

This publication has 26 references indexed in Scilit: