Contrasting oxygen and carbon dioxide contributions to respiratory quotient variability in fresh and fresh-cut produce under modified atmospheres

Respiratory quotient (RQ) integrates carbon dioxide production and oxygen consumption, but its ratio structure can amplify variability when oxygen consumption is low. This study quantified treatment-dependent RQ dynamics and compared the contributions of oxygen and carbon dioxide to RQ variability in fresh and fresh-cut produce under modified atmospheres. Time-series data for lettuce, spinach, baby plum tomato, and mixed salad exposed to four gas treatments and three cutting classes were analysed using generalized estimating equations (GEE). An empirical second-order response surface supported Monte Carlo simulation, Tornado analysis, partial rank correlation coefficients (PRCC), and blockwise semi-partial R2 analysis. GEE identified significant coefficients for Air, spinach, baby plum tomato, selected gas × cutting interactions, and selected time-dependent terms. The Mix C × fine-cut interaction showed the largest positive interaction coefficient (β = 0.4378, p = 0.042). In the baseline-filtered analysis, the CO₂ Tornado span exceeded the O₂ span (3.258 versus 0.733), with similar results after winsorization (3.367 versus 0.743). Under restriction to the higher-RO2 subset, however, the ranking reversed (O₂: 1.136; CO₂: 0.797). Winsorized PRCC showed a nominally statistically significant but extremely weak association for CO₂ (− 0.0708, p = 0.0148) and no meaningful association for O₂ (0.0006, p = 0.9824). Removal of the O₂-associated predictor block produced a larger reduction in model fit than removal of the CO₂-associated block (0.0677 versus 0.0377). These contrasting rankings demonstrate that atmospheric influence on RQ is both metric- and response-domain dependent. RQ is therefore best interpreted as a treatment-dependent trajectory alongside RO₂ and RCO₂, rather than as a standalone biochemical marker of fermentation. • GEE identified atmosphere × cutting effects on respiratory quotient dynamics. • CO₂ dominated Tornado spans in baseline and winsorized analyses. • Strict response-domain restriction reversed the ranking toward O₂. • PRCC showed only a weak conditional CO₂ association with RQ. • RQ sensitivity depended on both analytical metric and response domain.