(Lomolino 1984a). When introduced into islands,
Blarinabrought about drastic declines of Microtus
densities, restricting it to its optimal habitat and,
in at least one case, causing its extinction. As the
vole is the better disperser, the two species exhibit
a negative distributional relationship across the
islands studied. Lomolino’s studies thus demon-
strate the significance both of recurrent arrivals and
losses and of ecological controls such as predation
in structuring the mammalian communities of rela-
tively poorly isolated islands. As the islands in
question undergo seasonal variation in their immi-
gration filters, such that at times they are more akin
to habitat islands than real islands, care must be
taken in transferring such findings to other systems
with different scales of isolation. The reliability of
incidence functions is explored further in Box 5.2.
Linking island assembly patterns to habitat factors
Haila and Järvinen (1983) and Järvinen and Haila
(1984) examined the bird communities of the Åland
islands, off Finland in the Baltic. There are thou-
sands of islands in this group, ranging from only a
few square metres to the main island of 970 km^2.
Isolation of the islands tends to be slight, the main
island being only 30 km from Sweden, and 70 km,
via a steppingstone chain, from Finland. Twenty
species of land birds found in the same latitudes in
Finland and/or Sweden do not occur on the main
island. These form three groups:
●species for which habitat is available on Åland,
and for which absence is believed to be due to lack
of over-water dispersal (e.g. green woodpecker,
marsh tit, and nuthatch)
●species for which historical explanations can be
offered (e.g. collared turtle dove and tree sparrow
have recently expanded their mainland distribu-
tions but have yet to reach the island)
●species lacking because of absence of suitable
habitat (e.g. great grey shrike).
They found that species numbers and density of
the bird communities tended to be similar where
habitats were of highly similar structure, and
tended to differ when habitats were not directly
comparable. They also noted that small, uninhab-
ited islands in the archipelago lack those species of
birds that are associated either directly or indirectly
with humans. Their overall conclusion was that
habitat factors provided explanations for the
assembly of the bird communities of the Åland
islands.
Graves and Gotelli (1983) used a null modelling
approach in analysing the avifaunas of former
land-bridge islands: Coiba, San José, Rey, Aruba,
Margarita, Trinidad, and Tobago. They found in
general that the avifaunas of these islands could be
viewed as a random subset of the mainland
‘habitat’ pool when judged at the family level. They
used families because they felt unable to assign
species to guilds, and because species within a
family are usually ecologically and morpho-
logically similar. While acknowledging that this is
not ideal—families do not necessarily represent
units of interspecific competition—they contend
that non-randomness of island avifaunas might
reasonably be expected to be detectable at the
family level. Using a simple model assuming all
source-pool species to be equiprobable colonists,
they found that out of 40 families, 3 are unusually
common on land-bridge islands: pigeons
(Columbidae), flycatchers (Tyrannidae), and
American warblers (Parulidae), and only one, the
puffbirds (Bucconidae), was present less often than
expected. Thus, although in absolute terms many
species and families are absent from the islands, the
proportional representation of most families is con-
sistent with mainland source pools. Comparison of
the whole pool of species with the habitat pool
showed that, as expected, the habitat pool is a supe-
rior predictor of species richness in each family.
This implies that it is important when predicting
species losses in fragmented landscapes to take
account of patterns of habitat change that may
unfold within fragments after isolation. Examination
of the mainland range of the species in the pools
showed that those species with widespread main-
land ranges are disproportionately common on
islands. This indicates either that they have
persisted better since sea levels rose to produce the
present degree of island isolation, or that122 COMMUNITY ASSEMBLY AND DYNAMICS