Stochasticity in population models
Consider a stochastic version of the simple birth–death process
described in Newman and Jagoe [45], such as the one
depicted in Figure 1. A single population of size N increases
because of births and decreases because of deaths due to natural
causes and the eventual effect of environmental toxicants.
Heritage and habitat effects influence both birth and natural
death, with intrinsic rates of both processes (b, m) assumed
as normally distributed around a central characteristic value.
Even when both central values are equal (b 5 m 5 0.2) and
in the absence of toxic-induced mortality (t 5 0), the population
can experience severe fluctuations, nearing extinction if
a temporary uneven compensation of stochastic birth–death
processes occurs (Fig. 1a). A slightly increased average mortality
induced by an environmental toxic can be expected to
result in probabilistic outcomes where the population approaches
local extinction. No-observed-effect level-type responses
and temporarily increasing trends are also possible (t
5 0.05; Fig. 1b). Note that the toxic death rate t assumed in this case amounts to depleting by 5% the population stock in
every generation. This would correspond to a lethal concentration
0.05 (LC05), or increasing the mean natural death rate
m by 20%. The ecological relevance of considering unusually
low LC threshold values has been identified in important ecosystem
components such as fish [46]. Because of chance, small
local populations have a high risk of extinction, even when
the expected equilibrium population size is positive [47]. This
example shows that LCs should be interpreted with caution in
estimating sustainable exposures at the population level even
if, for practical reasons, lethal concentration/effect concentration
(LCs/ECs) values are usually the only data available [48].
The exercise in Figure 1 also prompts an interest in inspecting
the ratio of toxic-induced to natural mortality rates (t/m), or
the ratio of t to the standard deviation of m, as potentially
useful end-point parameters to estimate extinction risk.