It is easy to underestimate the power of natural
dispersal to re-supply isolated sites because the
process is so difficult to observe. We can illustrate
this by stepping outside the metapopulation litera-
ture and considering the regular supply of migrant
moth species to the remote Norfolk Island, 676 km
from the nearest land mass (Holloway 1996). A
survey over a 12 year period suggested that the res-
ident Macrolepidoptera total 56 species, and
migrants as many as 38 species. In addition, some
of the resident species are also reinforced by
the occasional arrival of extrinsic individuals. The
rapid recolonization of the Krakatau islands, some
40 km from the mainland, by so many species of
plants and animals, provides another example of
the ability of many species to cross substantial
distances. There are many such demonstrations to
be found in the biogeographical literature, but the
Norfolk Island data provide a powerful illustration
because quantifying the extreme, the tail of the
distribution, is generally difficult.
In contrast, there are many species of relatively
limited powers of dispersal, which like the
checkerspot butterfly may show distance limitation
to patch recolonization on a scale of a few kilo-
metres. For them to survive in increasingly
fragmented habitats may require active interven-
tion by people, to enable dispersal of viable
propagules between isolates (Primack and Miao
1992; Whittaker and Jones 1994a,b). Most metapop-
ulation simulation models assume no distance
effects, and thus may perform poorly in simulating
species that exhibit a strong gradient in arrival rates
within a particular set of isolates (Dawson 1994;
Fahrig and Merriam 1994).
The importance of continual immigration into
so-called ‘sink habitats’ is such that in some popu-
lations the majority of individuals occur in sink
habitats (Pulliam 1996). Furthermore, a habitat may
contain a fairly high density of individuals but be
incapable of sustaining the population in the
absence of immigration. Density can, therefore, be a
poor indicator of habitat quality. Thus, while some
suitable habitats will be unoccupied because of
their isolation from source populations, other
‘unsuitable’ habitats will be occupied because of
their proximity to source populations. As pointed
out earlier, if these effects operate, snapshot inci-
dence functions may be misleading.Deterministic extinction and colonization within metapopulations
Extinction models tend to be based on stochastic
variation, in some cases emphasizing demographic
stochasticity, sometimes taking genetic stochasticity
and feedback into account, and more recently incor-
porating environmental stochasticity. However,
Thomas (1994) has argued that, in practice, most
extinctions are deterministic, and that they can
be attributed directly to hunting by humans,
introductions of species, and loss of habitat. In
many cases documented in the literature, virtually
the entire habitat was lost or modified, causing
100% mortality. Stochastic extinction from surviv-
ing habitat fragments is minor by comparison.
According to this view, stochastic events are super-
imposed as decoration on an underlying determin-
istic trend. To put this in a metapopulation context,
if the reason for patches ‘winking out’ is habitat
changes rather than—for sake of argument—a
period of unfavourable weather, then the local
habitat is likely to remain unsuitable after extinc-
tion and so will be unavailable for recolonization.
In cases, fairly subtle changes in habitat may pro-
duce a species extinction that might be mistakenly
classified as ‘stochastic’. Thomas cites examples
from work on British butterflies, whereby even
changes from short grass to slightly longer grass
can result in changes in the butterfly species found
in the habitat (cf. Harrison 1994).262 ISLAND THEORY AND CONSERVATION
Table 10.1Problems that have been identified with classic
metapopulation models
- Occurrence of core–satellite relations
- The rescue effect is not dealt with adequately
- Distance effects and varying dispersal abilities
- The occurrence of ‘sink’ habitats
- Metapopulations are very difficult to replicate, tests are
problematic - Much extinction is deterministic