Apple Drying Process Lethality Prediction Using Response Surface Methodology

• Higher temperature and airflow increased lethality, thicker bed reduced lethality. • Water activity reduction was a practical proxy for apple drying process lethality. • Polynomial response surface methodology was applied to predict process lethality. • Model predicted Salmonella inactivation at a w 0.60 with high accuracy and low error. • External validation confirmed strong agreement of predictions with observations. This study evaluated the effects of drying temperature, bed depth, and airflow on Salmonella lethality achieved during hot-air drying of inoculated apple cubes to a fixed water activity level, using a three-factor, three-level Box-Behnken design. Apples were dried under 15 conditions, and Salmonella inactivation was estimated at a product water activity (a w ) of 0.60, a level representative of a relatively higher-a w dried apple product. Across all trials, Salmonella inactivation correlated linearly (R 2 = 0.91–0.97) with a w reduction, allowing a w to be used as a practical proxy for drying process lethality estimation. Estimated Salmonella inactivation at a w 0.60 ranged from 2.25 ± 0.11 to 4.97 ± 0.21 log CFU/4 cubes. A refined response surface methodology (RSM) model captured the significant linear effects of temperature, airflow, and bed depth, and a marginal interaction between airflow and bed depth. Higher temperature and airflow enhanced lethality, while a deeper bed reduced lethality. However, the positive effect of airflow diminished at greater bed depths. Model validation under three randomly selected drying conditions showed strong agreement between model predictions and experimental observations, with a low root mean square error of prediction (0.371 log CFU/4 cubes), an accuracy factor of 1.101, and a bias factor of 1.073 (7.3% overprediction). These results demonstrate that water activity reduction can be used as a reliable indicator of microbial lethality in hot-air apple drying and that the developed RSM model provided a robust tool to predict Salmonella inactivation across practical processing conditions.