Quantitative volatilomics identifies robust multi-class markers for hazelnut shelf-life monitoring in a multi-year industrial GC×GC study.

The volatile fraction of hazelnuts undergoes complex changes during industrial storage, especially those driven by lipid auto-oxidation. Hazelnut quality monitoring in the food industry mainly relies on standard fat oxidation indexes (e.g., free acidity, peroxide value, p-anisidine index) or on single-compound indicators such as hexanal, which fail in capturing the multi-pathway nature of deterioration. This study investigated the dynamics of 43 quantified volatile compounds spanning six chemical classes (aldehydes, alcohols, ketones, volatile fatty acids, lactones, and terpenes) in 428 industrial samples collected across three consecutive harvesting years (2020-2022), two geographical origins (Turkey and Italy), and two storage conditions (vacuum-packed and ambient atmosphere). Samples were analysed at seven time points over 24 months using headspace solid-phase microextraction comprehensive two-dimensional gas chromatography coupled with mass spectrometric and flame ionization detection (HS-SPME-GC × GC-MS/FID). The harvesting year emerged as a dominant source of baseline compositional variability, highlighting the necessity of multi-year designs for reliable marker discovery. Storage conditions showed a time-dependent effect, with vacuum packaging providing significant protection beyond 9-12 months. PCA and Spearman's correlation analysis consistently revealed that shelf-life deterioration was not governed by a single compound or pathway, but by the interrelated, sequential involvement of multiple chemical classes. Hexanal exhibited variable and storage-dependent network connectivity, further questioning its reliability as a standalone quality marker. These findings demonstrate that an expanded multi-class volatile panel is necessary to capture the full complexity of hazelnut shelf-life deterioration, providing broader mechanistic coverage, earlier detection of degradation, and more reliable quality assessment in real industrial conditions.