Geoscience Reference
In-Depth Information
drawbacks, which are often only apparent when they are contrasted with one
another directly. However, there will undoubtedly be strong size structure along
any particular food chain, as is evidenced by themass-mass relationships. If this
is our interest then the size-class-based food webs will be likely to better capture
this due to the less severe species averaging effects, although the exact THof any
one node is often more difficult to define with precision.
Similar arguments explain the effects associated with the different groupings
on other response variables, in particular the unimodal patterns uncovered in
out-degree (vulnerability) and prey range for the size-class grouping. An
argument can be made that the hump shape in both cases could be the result
of sampling, as intermediate predator size classes were better sampled (see
Figure A1
), whereas small and large predators were rare. This meant that the
range of the diet of the large and small individuals was less well resolved, and
hence the extremes were less likely to appear, which resulted in ranges appear-
ing narrower. The negative part of the out-degree slope is partly due to a lack
of predators large enough to feed on the larger size classes in the environment
(prey growing into a size refugia), similar to the explanation of the negative
out-degree relationship suggested for taxonomic food webs (
Digel et al., 2011
).
However, the initial upwards part of the slope can to some degree be a
consequence of sampling. That is, improved sampling of the diets of predators
in the intermediate size classes entailed a greater number of individuals in prey
size classes being discovered (due to the scaling between predator and prey
size), and rarer size classes, understandably, had fewer links either to or from
other size classes (see
Figure A2
for all study sites and
Figure A3
for an
example (Broadstone Stream) of yield-effort curves per size class). Indeed,
linear multiple regression models considering both mass and sample size
suggested that the only significant predictor for out-degree and prey range
was the sample size of a size class (
Table A3
). There are, however, also
plausible real phenomena supporting the lower vulnerability seen for smaller
size classes. Prey can be too small to be handled efficiently (
Brose et al., 2008;
Petchey et al., 2008
) and some of the smallest predators (e.g. tanypod midges)
partly feed on non-animal food when small and become more carnivorous as
they grow (
Woodward et al., 2005b
). The multiple linear regression did indi-
cate an effect of sample size for in-degree (generality) as well, but weaker than,
and interacting with, the effect of predator mass (
Table A3
).
C. Dynamic Implications—Parameterisation of Dynamic
Food Web Models
Dynamic models of food webs taking body sizes into account are mainly of
three different types. One modelling approach is species oriented in the
sense that it assigns one and the same size to all individuals within a species
