The small but growing number of communities
in which interaction strength has been estimated
suggests that strong skew in interaction strength
may be a general feature of communities, with
many links between species having negligible ef-
fects on the dynamics of either party, and a few
links having very strong impacts (reviewed by
Wootton and Emmerson 2005, Fig. 8.7). This has
important implications for community stability, as
skew in interaction strength has been shown to
confer stability on trophic networks in theory
(McCannet al. 1998; Emmerson and Yearsley 2004)
and in simulation studies of empirical food webs
(Emmerson and Raffaelli 2004).
8.5.2 Empirical evidence
Simulation-based approaches to exploring food web
stability have generally measured stability as resis-
tance to small perturbations imposed on species ran-
domly. A critical question for applied ecology is how
interaction strength of a species covaries with its
vulnerability to real perturbations such as fishing,
and how those perturbations ripple through the
web. Soberingly, analysis of a diverse Caribbean
reef food web found that, while combinations of
strong interactions capable of generating trophic cas-
cades were quite rare, human impacts fell dispropor-
tionately on species involved in such interaction
combinations, primarily because large- bodied spe-
cies are both strong interactors and disproportion-
ately targets of human impacts (Bascompteet al.
2005). Thus, in contrast to the stabilizing effects of
many weak interactions predicted by theory in di-
verse communities such as reefs, the targeted har-
vesting of large predators can have important
cascading impacts because fishing imposes strong,
persistent and non-random perturbations.
There is reason to expect that the results of Bas-
compteet al. (2005) may be common in that strong
interactors are typically the larger species in a com-
munity (although see Sala and Graham 2002 for an
exception), and at least for direct harvesting such as
fishing, large species are also those targeted prefer-
entially. Thus, it is likely that there will often be
positive covariance between a species’ interaction
strength and its vulnerability to fishing. Although
these conclusions are preliminary, this suggests
that the preponderance of weak interactions in
many food webs is unlikely to protect them from
the specific sorts of impacts imposed by fishing.8.6 Alternate stable states and regime shifts in marine ecosystems
8.6.1 Conceptual background
An ecological phenomenon of growing concern in
understanding marine ecosystem dynamics and
their implications for society is the phenomenon
of regime shifts. Regime shifts can be defined as
relatively rapid transitions between distinct and
relatively long-lasting, semi-stable states of a sys-
tem (Knowlton 2004; Steele 2004). The potential
existence of alternate stable states has been a sub-
ject of keen interest and controversy in ecology for
decades (Lewontin 1969; Sutherland 1974). In re-
cent years, regime shifts have gained new promi-
nence in the context of conservation and
management (Schefferet al. 2001). The term has
been used in two ways. The first definition is phe-
nomenological, and refers to multi-year periods of
relative stability in time series of observational
data that are separated by abrupt shifts to intervals
that fluctuate around a different mean. The sec-
ond, stricter definition involves the dynamics of
the system and refers to the existence of two or
more semi-stable states, or alternate attractors, in
an ecosystem. Regime shifts in the phenomenolog-
ical sense have been described in a number of
marine ecosystems, and appear often to track de-
cadal-scale climate variation (e.g. Overlandet al.
2006).
Recent syntheses (Collieet al. 2004) have defined
three types of regime shifts, which are actually
points along a continuum (Fig. 8.8a): (1) a smooth
regime shift, defined by a quasi-linear relationship
between a forcing variable and a response variable,
(2) an abrupt regime shift, in which the relationship
is non-linear and (3) a discontinuous regime shift,
in which the relationship is not only non-linear but
the trajectory of the response variable differs when
the forcing variable is declining compared with
when it is increasing. The latter situation results in
two possible states of the response variable at a
given value of the forcing variable and is also110 APPLICATIONS