0198566123.pdf

authors or reviewers to be uninteresting. One
example of no relationship is provided by Dunn
and Loehle (1988) who examined a series of upland
and lowland forest isolates in south-eastern
Wisconsin and for each data set reported slopes not
differing statistically from zero. Their isolates were
not real islands, nor was there an enormous range
of areas in their data set. Factors other than size
of isolate, such as disturbance and forest-edge
effects, were argued to be influences on plant
species number. Their result was consistent with
the indirect nature of the control exerted by area,
but also with the obvious point that meaningful
depiction of species–area effects requires the
selection of a wide range of island areas (Connor
and McCoy 1979).
Power’s (1972) study of bird and plant species
on the California islands incorporated 16 islands or
island groups with a broad range (0.5–347 km^2 ) of
areas. The results were presented in a path dia-
gram showing the relationships among variables
as indicated by stepwise multiple regression
(Fig. 4.5). Although area was an important deter-
minant of plant species number, it was a relatively
poor predictor of bird species numbers. The role of
climate in mediating plant species richness was
indicated by the role of latitude. Power suggested
that near-coastal islands with richer and struc-
turally more complex floras tended to support a
larger avifauna. The analysis thus indicates the
importance of hierarchical relationships between
plants and birds in determining island species

numbers, a line of argument which may also be

developed in a successional context (Bush and

Whittaker 1991).

Distance and species numbers

The depauperate nature of isolated archipelagos

has long been recognized (Chapter 3). Several

studies have shown that isolation explains a signif-

icant amount of the variation in species number

once area has been accounted for in stepwise mul-

tiple regression (e.g. Fig. 4.5; Power 1972; Simpson

1974). Problems arise from the practical difficulties

of measuring the isolation of a target island

because of the existence of other islands (the con-

figuration of which varies through time: Chapter 2),

which may serve as networks for species move-

ments across oceans, and because of the multi-

collinearity often found with other environmental

variables (e.g. Kalmar and Currie 2006). Moreover,

the effective isolation of islands and habitat islands

is not determined solely by the distance from a

designated source pool. Rather, for real islands it

varies in relation to wind and ocean currents

(Spencer-Smith et al. 1988; Cook and Crisp 2005),

and for habitat islands it also depends on the char-

acteristics of the landscape matrix in which they

are embedded (Watson et al. 2005).

Several authors have made use of the power

function model as indirect evidence for the pres-

ence or absence of equilibrium, and in particular

cases, it may be possible to construct highly plausible

ISLAND SPECIES NUMBERS AND ISARS: WHAT HAVE WE LEARNT? 91

Number of

plant species

Isolation

Unknowns

Latitude

Area

Unknowns Number ofbird species

0.579

0.249

0.171

0.668

0.142

0.190

Figure 4.5Path diagram showing relationships among variables in explanation of species numbers of plants and birds on the California islands
(modified from Power 1972). The coefficient associated with each path is the proportion of the variation in the variable at the end of the path
explained by the independent variable at the beginning of the path, while holding constant the variation accounted for by other contributing
variables.