Developing predictive models - Conservation and Sustainable Use - page 182

There has been debate about the usefulness of PVAs as a tool for conservation
(Brook et al . 2000; Coulson et al . 2001), which is relevant to any kind of model-
ling exercise that aims to provide management recommendations. The main issue
with PVAs is whether the extinction probabilities that they predict have any mean-
ing, particularly when the probability of irregular and rare events like very bad
weather, disease epidemics or poaching sprees may be very hard to predict despite
having a major influence on probability of extinction. The evidence from extensive
testing of PVAs in simulated environments suggests that they are useful in ranking
management alternatives, even if they can't necessarily accurately predict the
probability of extinction itself (McCarthy et al . 2003).
The distinguishing feature of a PVA is not the form of the model, but the metric
that is used to assess conservation performance—a PVA by definition needs to
produce an estimate of extinction risk. This makes them a fairly limited subset of the
types of model that are used in assessing the sustainability of use. Partly this is because
extinction is not likely to be a probable near-term outcome if a population is large
enough for sustainable use, and partly because we take a broader view that includes
social and financial aspects of sustainability. Hence, although the literature on PVA
is interesting and relevant, it is a relatively small part of the modelling toolkit.
5.4.2 Behavioural models
Behavioural models can be very useful for predicting the behaviour of both prey
and hunters. Within behavioural ecology, optimal foraging theory provides a
range of useful models for predicting prey choice and spatial patterns of foraging,
which can be applied just as well to humans as animals (e.g. Alvard 1993;
Winterhalder and Smith 2000; Rowcliffe et al . 2004). From the economics side,
human hunting behaviour can be modelled using similar techniques, so hunter
effort varies depending on the availability of prey in a particular location. For
example, Moustakas et al . (2006) produced a model in which both fish and fishers
moved through a grid of cells, with the fish looking for foraging and spawning
grounds, and the fishers moving to where the fish were most abundant. Another
set of behavioural models comes from game theory (Binmore 2007). This pre-
dicts how individuals behave when their optimal decision is dependent on the
decisions made by others, and is most useful when there are only a small number
of individuals involved. For example, Mesterton-Gibbons and Milner-Gulland
(1998) used game theory to explore the incentives to monitor others or to poach in
a community-based wildlife management project.
Recently, there has been a lot of interest in agent-based models , which hold out
the prospect of simulating the strategic decision-making of individual hunters.
These models are a sub-class of individual-based population models (IBMs,
Table 5.2), and are as computationally intensive as all IBMs. They can also be dif-
ficult to programme, although there are a number of software packages available
(see Resources section). The distinction between agent-based models and other
IBMs is that the agents respond to their environment, and can therefore act
strategically. In other individual-based models, the individuals just have to be