Developing predictive models - Conservation and Sustainable Use - page 158

to which we need to characterise this variation depends on the problem we are
addressing. In general, the simpler the model, the better, so structure should only
be introduced if it is fundamental to the dynamics of the system (Section 5.4.1).
For example, it is usually only necessary to include age and sex structure in biolog-
ical models of population dynamics if the harvest is strongly biased towards a par-
ticular age or sex. Individual-based models are computationally very intensive and
are not often needed in this field. Spatially explicit models , on the other hand, are
useful because hunting pressure is often spatially heterogeneous, and the results of
spatially explicit models can vary quite substantially from non-spatial models.
Continuous time models are usually used when an analytical solution is required,
because they can be solved using calculus. However, they are more difficult to
conceptualise and to implement in a simulation model. For this reason, we tend
to use discrete time models—these also often make more sense in bio-economic
systems when time is naturally divided. This is usually by season (open vs. closed
hunting seasons, agricultural seasons, fruiting or birth periods) or by year.
Deterministic models are easier to parameterise and interpret, and in many cases
they give an adequate representation of the system. Including random variation
(such as weather-related variation in survival) is more realistic, and can make a big
difference if the population is small or the variation is large (for example, catastrophic
winter mortality can cause extinction in small populations). Another strength of
stochastic models is that they can be used for risk assessment, while deterministic
models only give the most likely outcome based on a fixed set of parameter values.
Finally, we can create bio-economic models, that attempt to model the sustain-
ability of the system as a whole, or we can model components of the system separ-
ately. For example, we might want to find out what the sustainable level of offtake
is that a given population size can provide, using a population dynamics model.
Or we might wish to use a cost-benefit model to find out how likely it is that a
hunter will decide to break the rules governing hunting, based on his perceptions
of the likely punishment and the profitability of poaching. The trade-off between
complexity and predictive power comes in here again, and in general it is impor-
tant first to determine the key questions , and then build a model appropriate to
answering these questions.
5.2.1 Off-the-shelf packages
There are a number of packages available that enable you to enter parameter values
into a pre-coded population model, and so save yourself the time and effort of
building your own model (see Resources section for websites). The strengths of off-
the-shelf packages are that they are quick and reliable and give a higher level of
analysis than is possible for a first-time modeller. If you are not confident in the
principles of population dynamics, then when building your own model you may
well make fundamental mistakes in parameter estimation and equation specification
that you will not spot. Some packages, like RAMAS, allow you to build quite
sophisticated spatial models, combining GIS data with population models, which
is something that is very difficult to do well without substantial modelling expertise.