272 ISLAND THEORY AND CONSERVATION
Box 10.2 Does island theory provide a basis for the use of species–area regressions to
predict extinction threat?
In the text, we discuss a number of studies that
have used species–area models to predict the
eventual species losses following fragmentation
(the so-called ‘extinction debt’). A growing body
of literature has applied this approach at varying,
sometimes very coarse, scales (e.g. Brooks et al.
2002), to forecast species losses in the tropics as a
function of forest loss. Recent work has attem-
pted both to assess extinction threat and validate
the application of species–area models by
comparing at a regional scale the forecasts of
species–area models to lists of threatened species.
Where congruence is found, the analyses are
regarded as mutually supporting: island effects are
working, systems are ‘relaxing’ to lower species
number, and given time (the ‘lag effect’), will re-
equilibrate to a lower species richness. The
application of species–area models is generally
taken to be supported theoretically by work on
island species–area relationships, from which the
slope parameter z is derived.
Returning to the Atlantic forest system discussed
in the text, the study by Grelle et al. (2005)
assesses the sensitivity of species–area based
predictions of extinction to the z parameter (slope
of the relationship) selected. Their paper provides
an analysis of data for Rio de Janeiro State, in
which present and historical coverage of forest are
estimated at 19% and 100%, respectively.
Following other recent studies, they first predicted
extinctions using a species–area model with a z
value of 0.25, and compared the predicted species
losses with the number of species identified by an
expert workshop as threatened with extinction.
Their analysis suggested that the species–area
model overestimated extinction threat for each of
mammals, birds, reptiles and amphibians. When
they repeated the analysis for endemic species
only, they found that the species–area model
successfully predicted threatened endemic
mammals when using a z of 0.25, and threatened
endemic birds when using a z ranging between
0.1–0.19.
We have detailed this study as it is one of
several in the literature (e.g. Brooks and Balmford1996; Brooks et al. 2002; Thomas et al. 2004)
that use species–area regression as a means of
forecasting species threatened by, or committed
to, extinction. However, the way in which the
species–area models are used in this and other
studies is conceptually decoupled from the island
theory from which it seemingly derives. In
Chapter 4 we cite MacArthur and Wilson (1967)
for the observation that ISAR z values may vary
between 0.2 and 0.35, and note that the range in
values may indeed be much greater (Williamson
1988). Therefore, first, a z of 0.25 is a largely
arbitrary ‘middle’ value to take. Second, and more
crucially, this z value has been derived from
analyses of true isolates: it tells us approximately
how many species are held in each of a series of
isolates/islands of different size. Note that the
total number of species held across the series of
isolates is not derivable from the z parameter of
the isolates: as the overall richness of the system
depends on the degree of compositional overlap
(nestedness) among the isolates. Yet, in recent
studies such as these, the z value is applied not to
separate fragments, but to the entire region. That
is, the authors assume that the whole system (in
this case Rio de Janeiro State) is acting as a single
fragment, which has lost 81% of its area and
which is therefore travelling along a trajectory
towards its lower equilibrium point. However, in
practice, we are talking about an area in which
forest habitat persists in very large numbers of
patches, scattered across that region in varying
configurations. This loss of habitat is certainly
bound to vastly reduce the number of individuals
of forest species across that region, but whether
that results in particular species falling below their
minimum viable population in all isolates is not
knowable simply from the regional figure for
habitat loss (see e.g. Loehle and Li 1996).
There is thus no link of scientific logic that can
carry us from empirical demonstrations of z values
for islands and isolates, via MacArthur and
Wilson’s EMIB, and the related theoretical idea of
species relaxation (Diamond 1975b), through to
the assumption of a particular shape and slope of