These models have acceptable adjusted R2 values for being a simple
quadratic model using limited (three replications) data (Fig. 6). If additional
data were available, the adjusted R2 values would improve with a
resulting improvement in the accuracy of the models. The fit would also
improve if there was less variation in the initial bacterial populations.
The initial microbial populations in this study differ because the pork
chops were of different post-mortem ages when we received them
from the retail store, varying from 3, 7, and 14 days post-mortem. If
the meat was obtained from the production source during fabrication,
there would likely be less variation in the initial populations. Despite
this variation, the models still had a good fit and adjusted R2 values.
However, these models are limited to vacuum-packaged pork chops
stored at 4 °C because those were the conditions tested. These models
also only estimate microbial populations, and not sensory shelf life,
which could be different from the microbial spoilage threshold. Despite
these limitations, these models show proof of concept and prove that
shelf life models can be developed from an instrument measurement,
such as dissolved CO2 concentrations. It also proves that nondestructive,
rapid, and relatively easy to use instruments, such as those
that measure dissolved CO2, can be used to estimate remaining shelf life
by an approximate microbial population. These models also provide an
indication of a “real-time” measurement of shelf life, which is an improvement
upon the three-day, “historical” data delivered currently by traditional
plate count methods to determine shelf life.