Environmental Engineering Reference
In-Depth Information
So instead, we need to assess the relative contribution of a range of hypothesised
factors to changes in our target variable. For example, we may wish to see how dif-
ferences in both habitat and current hunting effort are linked to population abun-
dance, and how these effects interact with the type of weapon hunters use
(Box 4.4). As Box 4.4 shows, data limitations are such that we cannot always dis-
entangle confounding factors without collecting more or different data. But if we
start out with a strong set of hypotheses about the processes underlying our obser-
vations, we are less likely to be misled by our answers—even if the answer is that we
don't know.
This study shows just how difficult it is to disentangle confounding effects on
animal abundance. Each species shows different responses to hunting, distance
from the village and habitat type—for example, porcupines are associated with
rivers and logging roads, but their abundance is not affected by hunting. Many of
the primates are not found near the village, even when hunting pressure and habi-
tat type are taken into account, but squirrels are more abundant near the village.
Overall, however, animal abundance is strongly negatively related to current hunt-
ing pressure.
The whole way of thinking about statistical modelling in biology has been revolu-
tionised over the last few years (e.g. Hilborn and Mangel 1997; Burnham and
Anderson 2002; Hobbs and Hilborn 2006). This is partly due to the wide avail-
ability of statistical computing packages that can deal easily with multivariate
analyses, partly due to the rising use of Bayesian statistics, and partly due to the
increasing recognition that if science is to influence policy, it needs to present
results in a way that quantifies uncertainty in a policy-relevant way. This approach
to modelling involves:
●
Developing
competing statistical models
which reflect alternative explana-
tions of the processes involved in changes in the variable of interest. The
Box 4.4
Disentangling habitat and hunting as determinants of mammal
abundance in tropical forests.
Many studies of tropical forest mammals have shown that animal densities
increase with distance from human settlement, and this is often attributed to a
gradient in hunting pressure (e.g. Peres and Nascimento 2006). However, habitat
is also likely to change with distance from a village, for example from agriculture
to secondary to primary forest. In a study in Equatorial Guinea, Janna Rist and
colleagues tried to disentangle the effects of current and past hunting pressure,
habitat type and distance from the village as determinants of the abundance of a
range of mammal species (see table below).
