3. Results and discussion
3.1. Effect of temperature on growth rates
The growth rates, l, were plotted versus temperature
(Figs. 2–4). No large difference of behaviour atsuboptimal temperature values between the strains of
a same species was observed, although the strains
were selected to be representative of different food
contexts. Therefore all the strains of a same species
were analysed together with the same statistical
model.
The residual error variance was analysed graphically
(data not shown). This variance was composed
of uncertainty (lack of accuracy due to the measurements
errors, the discrepancy between the laboratories, etc.) and of biological strain variability. However,
our study did not lead to a characterisation of
each compound of this variability: our purpose wasto suggest a modelling approach to simulate the
temperature effect on the behaviour of bacterial
species on food, whatever the strain or the labo-ratory. Particularly with E. coli (Fig. 2) and B.
cereus (Fig. 3), the residual error was bigger when
temperature values were above the optimal temperature
than for the other temperature levels. This
phenomenon could be explained by biological
considerations such as high sensibility of the strain
to these mild temperature treatments. Consequently,
to calculate the interval limit (see below), the supraoptimal
temperature levels were not taken into account.