Inclusion of the variability of model parameters on shelf-life estimations for low and intermediate moisture vegetables

Shelf-life is the time period during which products retain market-acceptable quality while meeting legal and safety requirements. Deterministic models yield single value estimations of shelf-life typically based on average or worst-case values for input parameters. In deterministic calculations, considering the input parameter variability can be challenging. In this study, a Monte Carlo procedure and the G.A.B. model for moisture sorption isotherms were used to predict shelf-life frequency distributions for intermediate moisture (IM) tomato slices, and low moisture (LM) onion flakes and sliced green beans. End of shelf-life for IM tomato slices (initial a_w = 0.8) was assumed to occur for a 10% moisture loss, and when a_w changed from 0.25 to 0.4 for LM onion flakes and LM sliced green beans. The estimated shelf-life for tomato slices, LM onion flakes, and LM sliced green beans based on the deterministic approach was 243, 86, and 79 days, respectively. The Monte Carlo procedures yielded shelf-life frequency distributions with values ranging 181-366, 76-95, and 71-90 days, respectively. Products would fail before the deterministic shelf-life value with an unacceptably high probability of 51.6, 48.6, and 53.0%, respectively. If 5% is an acceptable probability that the actual shelf-life is shorter than specified, the estimated values would be 211, 81, and 73 days, respectively. X_m and K were the most influential G.A.B parameters on the shelf-life of the three products. The package area, product amount, and water vapor transmission rate were high contributors and had the expected effect on shelf-life as demonstrated by deterministic estimations.