R E L A X AT I O N A N D T U R N O V E R — T H E E V I D E N C E 269

Table 10.2 Classes of edge-related changes triggered by the process of forest fragmentation, as informed by the
Minimum Critical Size of Ecosystem project. The first-order effects may lead to second-order and, in turn, third-order
knock-on effects (modified from: Lovejoy et al. 1986)

Class Description of change

Abiotic Temperature
Relative humidity
Penetration of light
Exposure to wind
Biological
First-order Elevated tree mortality (standing dead trees)
Treefalls on windward margin
Leaf-fall
Increased plant growth near margins
Depressed bird populations near margins
Crowding effects on refugee birds
Second-order Increased insect populations (e.g. light-loving butterflies)
Third-order Disturbance of forest interior butterflies, but increased population of light-loving species
Enhanced survival of insectivorous species at increased densities (e.g. tamarins)*
*
Does not apply to birds (Stouffer and Bierregaard 1995).

islands, thereby making such sites less suitable for, be heterogeneous (i.e. structured). Some species
or removing altogether, particular species tend to be highly stable in their distribution across
(McGuinness 1984; Simberloff and Levin 1985; habitat islands. Most turnover involves ‘ephemerals’,
Dunstan and Fox 1996). species marginal to the habitat, or successional
change (Chapters 4–6). The metapopulation studies
examined above provide further insights into
10.7 Relaxation and turnover—the turnover. They suggest a tendency for large
evidence populations to persist, while smaller satellite popu-
lations may come and go, without jeopardizing the
If habitat islands behave according to the expecta- metapopulation. Nonetheless, species do disappear
tions of the EMIB, they should be supersaturated from particular habitat islands, and from entire
with species immediately after system fragmenta- landscapes, and in cases the loss is a global one.
tion. Subsequently, species numbers in the isolates Earlier, we noted that many species losses are
should ‘relax’ to a lower level (Fig. 10.5). Immi- attributable to deterministic causes, meaning hunt-
gration and extinction should continue to occur, ing, habitat changes, and the like. The relaxation
both during the relaxation period and subse- effect, in contrast, is based on stochastic processes.
quently, when the island has found its new, lower How strong is this island relaxation effect?
equilibrium richness level. The time taken for relax- One of the best-known examples of relaxation on
ation to occur is called the ‘lag time’ and the antici- ecological timescales is that of birds lost from Barro
pated eventual species loss is termed the extinction Colorado island, in Panama. It is not an ideal study
debt (Ewers and Didham 2005). If these anticipated in the present context in that, unlike most habitat
effects are strong, even high-priority species—the islands, it is actually now surrounded by water. The
very species that you wish to conserve—may in island was formerly a hilltop in an area of continu-
time disappear from a reserve. As we have noted, ous terrestrial habitat, but it became a 15.7 km2
however, the empirical evidence from a wide range island of lowland forest when the central section of
of real and habitat islands is that turnover tends to the Panama Canal Zone was flooded to make Lake
270 I S L A N D T H E O RY A N D C O N S E R VAT I O N

Gatun in 1914. Of about 208 bird species estimated to been found to be vulnerable to fragmentation
have been breeding on Barro Colorado island elsewhere (below). In a simple sense, Barro Colorado
immediately following isolation in the 1920s and island provides evidence of relaxation, but the
1930s, 45 had gone by 1970 (Wilson and Willis stochastic signal is seemingly much smaller than
1975). Of these, some 13 have been attributed to that produced by changing habitat and food-web
‘relaxation’. relationships, and it might therefore be asked
The other losses could be attributed to particular ‘which is the signal and which the noise in the
forms of ecological change. At the time of the analy- system?’ (cf. Sauer 1969; Lynch and Johnson 1974;
sis, much of the forest was less than a century old, Simberloff 1976).
following abandonment of farming activity before That species numbers and composition may
1914. Many of the birds lost were typical of second change as a consequence of fragmentation is not in
growth or forest-edge, suggesting a probable dispute. However, it is remarkably difficult to find
successional mechanism as forest regeneration good quantitative data for the island relaxation
reduced the availability of these more open habi- effect for systems of habitat islands (Simberloff and
tats. Some were ground nesters, which may have Levin 1985; and see review in Shafer 1990). Most
been eliminated by their terrestrial mammalian studies, including that for Barro Colorado island,
predators. The latter became abundant because of lack precise information about the species composi-
the disappearance of top carnivores (such as the tion before fragmentation (e.g. Soulé et al. 1988).
puma) with large area requirements (Diamond The Barro Colorado island study is also fairly typi-
1984). This effect, of increasing numbers of smaller cal in the losses recorded being mostly ‘determinis-
omnivores and predators in the absence of large tic’ in nature. The predictions of species loss
ones, has been termed mesopredator release (Soulé through ‘relaxation’ derive from the EMIB, but the-
et al. 1988), and it has now been documented in oretically relaxation may also affect metapopula-
several other systems (Fig. 10.7; Crooks and Soulé tions. Hanski et al. (1996) speculate that the pace of
1999; Laurance 2002). Some of the birds lost were fragmentation in many regions has been so rapid
members of the guild of ant followers, which have that ‘scores of rare and endangered species may
already be “living dead”, committed to extinction
because extinction is the equilibrium toward which
their metapopulations are moving in the present
fragmented landscapes’ (p. 527). However, their
findings are heavily dependent on models of the
data of a single species of butterfly, the Glanville
+ fritillary (Melitaea cinxia). The effect identified
– appeared to be far from instantaneous, and might
Fragment age turn out in practice to be a relatively weak effect in
Fragment – (time since
area relation to some of the other processes discussed in
isolation)
this chapter.
A classic illustration both of genuine reasons for
+ – – concern and of the uncertainty involved in projec-
tions is provided by the tropical moist forests of the
Atlantic seaboard of Brazil, which have been
reduced over the past few centuries to only about
7% of their estimated former cover (Ribon et al.
Figure 10.7 Model of the combined effects of trophic cascades and 2003). Based on the ‘rules’ derived from island
island biogeographical processes on top predators (for example,
coyote), mesopredators (domestic cat) and prey (scrub-breeding birds)
theory, some 50% of species are expected to go
in a fragmented system. Direction of the interaction is indicated with extinct. To date, no extinctions have been docu-
a plus or minus. (From Crooks and Soulé 1999.) mented with certainty (Whitmore and Sayer 1992;
 R E L A X AT I O N A N D T U R N O V E R — T H E E V I D E N C E 271

Ribon et al. 2003; Grelle et al. 2005), although many One important caveat to this study is, however,
species are Red-listed as vulnerable, endangered or that the historical data for the Kakamega area were
critically endangered based on range, or population derived from general collecting/observations from
estimates (Ribon et al. 2003). Brooks and Balmford the Kakamega area, with values assigned to the
(1996) compare losses of birds projected using a fragments on the basis of the assumption of
species–area model with those listed by the IUCN z  0.15 (slope of SAR) for their pre-isolation states.
as ‘threatened’ and find congruence, i.e. they argue Hence, the figures derived for relaxation times
that the predictions are basically correct, but that should be seen as still of the nature of first approx-
there is a substantial lag between the habitat imations. Moreover, the relaxation times per frag-
loss/fragmentation process and global extinction ment do not tell us overall species losses from the
of the species. There is, moreover, evidence of system. In practice, the data reviewed by Brooks et al.
local extirpation within the existing range for (1999) suggest that some loses have occurred, with
bird species. Thus, within the Viçosa region 10 of an estimated 73 Kakamega species not having
(a 120 km2 area in southeastern Brazil) over the been recorded from these forests in recent years. Of
last 70 years, it appears that at least 28 bird species those species counted as lost at least one is thought
have become locally extinct, with 43 ‘critically to have been trapped to extinction for the caged-bird
endangered’, and 25 ‘vulnerable’: in total 61% of trade; three others were at their eastern range limits
the original avifauna has been significantly reduced within the study area, so their loss from this area
in incidence (Ribon et al. 2003). Nectarivorous might be related to other processes and may not
species were least at risk, next came omnivores and be indicative of their risk of global extinction. In
carnivores, with frugivores and insectivores in practice, a single study system, of five fragments,
greatest jeopardy. based on relatively poorly specified empirical data,
Assuming relaxation to be underway in these is a rather limited basis for building predictive mod-
fragmented systems, the next problem is to esti- els, and there are surprisingly few other such studies
mate the lag time between fragmentation and even- with which to compare the values obtained.
tual species losses. Brooks et al. (1999) attempt this Box 10.2 appraises the use of species–area models
for tropical forest birds in an area of fragmented in predicting regional species extinctions as a func-
upland forest in Kenya. They base their analyses on tion of regional habitat loss, arguing that although
island theory, assuming z values (slope of the SAR) there is a crude relationship between habitat loss
pre-isolation of 0.15, and post-isolation of 0.25 (i.e. and species loss, there is no logical connection
for the new ISAR), and that the rate of decline in between such models and island theory. It should
species is approximately exponential. This postu- also be recognized that even where species–area
lates that fragments will lose approximately half analyses are linked phenomenologically to island
the species that will eventually be lost every x theory (i.e. they are being applied to systems of
years: this they termed the ‘half-life’. The difficulty fragments using parameters derived from other
is finding systems with ‘before’ and ‘after’ data. such systems, which also vary over similar ranges
The system they chose for the test is the Kakamega of area and isolation), they provide only a coarse,
forest, where forest fragmentation occurred over stochastic simulation of the impacts of fragmenta-
60 years ago (but where the time frame is not per- tion. The actual drivers of extinction may be
fectly specified). Their analyses of five fragments structured, and in some instances avoidable by
suggests a half-life varying in relation to fragment appropriate management. For instance, hunting
area from 23 (100 ha) to 80 (8600 ha) years, and of and collection of birds and mammals may each be
approximately 50 years for a 1000 ha fragment. As important drivers of the decline of particular
might be anticipated from island theory, relaxation species (Peres 2001; Ribon et al. 2003). With respect
times were suggested to be scale dependent in rela- to the Atlantic forest system, conservation action
tion to both range in area (‘grain size’) and degree has likely already been and will continue to be
of isolation (cf. Box 10.2). crucial to staving off the final extinction of highly
272 I S L A N D T H E O RY A N D C O N S E R VAT I O N

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 1996; Brooks et al. 2002; Thomas et al. 2004)
have used species–area models to predict the that use species–area regression as a means of
eventual species losses following fragmentation forecasting species threatened by, or committed
(the so-called ‘extinction debt’). A growing body to, extinction. However, the way in which the
of literature has applied this approach at varying, species–area models are used in this and other
sometimes very coarse, scales (e.g. Brooks et al. studies is conceptually decoupled from the island
2002), to forecast species losses in the tropics as a theory from which it seemingly derives. In
function of forest loss. Recent work has attem- Chapter 4 we cite MacArthur and Wilson (1967)
pted both to assess extinction threat and validate for the observation that ISAR z values may vary
the application of species–area models by between 0.2 and 0.35, and note that the range in
comparing at a regional scale the forecasts of values may indeed be much greater (Williamson
species–area models to lists of threatened species. 1988). Therefore, first, a z of 0.25 is a largely
Where congruence is found, the analyses are arbitrary ‘middle’ value to take. Second, and more
regarded as mutually supporting: island effects are crucially, this z value has been derived from
working, systems are ‘relaxing’ to lower species analyses of true isolates: it tells us approximately
number, and given time (the ‘lag effect’), will re- how many species are held in each of a series of
equilibrate to a lower species richness. The isolates/islands of different size. Note that the
application of species–area models is generally total number of species held across the series of
taken to be supported theoretically by work on isolates is not derivable from the z parameter of
island species–area relationships, from which the the isolates: as the overall richness of the system
slope parameter z is derived. depends on the degree of compositional overlap
Returning to the Atlantic forest system discussed (nestedness) among the isolates. Yet, in recent
in the text, the study by Grelle et al. (2005) studies such as these, the z value is applied not to
assesses the sensitivity of species–area based separate fragments, but to the entire region. That
predictions of extinction to the z parameter (slope is, the authors assume that the whole system (in
of the relationship) selected. Their paper provides this case Rio de Janeiro State) is acting as a single
an analysis of data for Rio de Janeiro State, in fragment, which has lost 81% of its area and
which present and historical coverage of forest are which is therefore travelling along a trajectory
estimated at 19% and 100%, respectively. towards its lower equilibrium point. However, in
Following other recent studies, they first predicted practice, we are talking about an area in which
extinctions using a species–area model with a z forest habitat persists in very large numbers of
value of 0.25, and compared the predicted species patches, scattered across that region in varying
losses with the number of species identified by an configurations. This loss of habitat is certainly
expert workshop as threatened with extinction. bound to vastly reduce the number of individuals
Their analysis suggested that the species–area of forest species across that region, but whether
model overestimated extinction threat for each of that results in particular species falling below their
mammals, birds, reptiles and amphibians. When minimum viable population in all isolates is not
they repeated the analysis for endemic species knowable simply from the regional figure for
only, they found that the species–area model habitat loss (see e.g. Loehle and Li 1996).
successfully predicted threatened endemic There is thus no link of scientific logic that can
mammals when using a z of 0.25, and threatened carry us from empirical demonstrations of z values
endemic birds when using a z ranging between for islands and isolates, via MacArthur and
0.1–0.19. Wilson’s EMIB, and the related theoretical idea of
We have detailed this study as it is one of species relaxation (Diamond 1975b), through to
several in the literature (e.g. Brooks and Balmford the assumption of a particular shape and slope of
 R E L AT I O N A N D T U R N O V E R — T H E E V I D E N C E 273

relationship between the percentage of forest loss attention to be given to species judged most at
on a regional basis and regional species losses. In risk, and necessarily involve an assessment of what
short, the use of species–area regressions in this factors threaten particular species. This sort of
way appears to be currently without a theoretical information would seem to be of inherent value to
justification, and thus represents merely an conservation agencies. Moreover, the argument
exercise in curve-fitting. that we should use species–area modelling in the
Taken as such, what do we learn from such absence of adequate data for other regions or
studies? The Grelle et al. (2005) study illustrates other taxa, arguably breaks down with the
that, in practice, there is no reason to favour a z realization that the success of the study in fitting
value of 0.25 over other slope parameters in species–area models to independent data on
making what amounts to a back-of-the-envelope species threats varied between taxa, with
sketch of how regional habitat loss may relate to extinction risks of reptiles and amphibians not
loss of species regionally. In practice, we would successfully predicted in any of their models. This
be better to use species-level assessments is one area within conservation biogeography
(whether underpinned by rigorous survey data or where current theory appears inadequate and in
expert assessments) than to rely on such urgent need of attention (as Whittaker et al.
species–area fits. Such assessments enable 2005).

threatened species (Birdlife EBA factsheet 75, Means and Simberloff 1987; Brown and Opler
www.birdlife.org/datazone/, visited July 2004). 1990). In the Neotropical gallery forests, Kellman
Given these varying forms of human intervention, it (1996) reports that there is no evidence for floral
may be impossible to resolve the question of whether impoverishment with distance from the peninsu-
the Atlantic forests system provides support for, or a lar base, possibly because the frugivores that
refutation of, the theory of species relaxation. disperse the seeds of the plants are capable of
The assumption that fragmentation will necessarily adjusting their ranging behaviour to accommo-
lead to ongoing species losses through relaxation date the long, thin peninsular configuration, or
has been challenged by Kellman (1996). He argues indeed to move through the sparsely wooded
that reference to past environmental change and savanna matrix surrounding the peninsulas. These
species responses to it shows that many species are peninsulas have a long history, and thus appear to
more flexible in their ecological requirements than provide evidence for adequate gene flow of plant
generally recognized. Kellman contends that many populations to the peninsular tips over long peri-
plant communities are actually undersaturated and ods, i.e. there has been plenty of time for ‘lag
that systems of fragments may be able to sustain effects’ to unfold, but at least for plants, there is no
increased species densities for significant periods of evidence of species impoverishment. However, it is
time (although not necessarily indefinitely). He unclear how good an analogue these peninsulas
bases his case in large part on studies from provide for the rapid conversion of forested areas to
Neotropical gallery forests, which constitute nar- fragmented systems.
row peninsulars of forest extending through a There is a long-term fragmentation project which
matrix of savanna ecosystems. According to equi- sheds some light on these questions, and in which
librium theory, the tips of the peninsulas should be data were collected before fragmentation: the
impoverished because of their relative isolation. Minimum Critical Size of Ecosystem project. It is
This extension of island theory is termed the penin- located near Manaus, in the Brazilian Amazon, and
sula effect. It has been tested at a coarse scale for sites were turned into habitat islands, isolated from
Florida and for Baja California, and has generally intact forests by just 70 to 650 m of pasture, between
been found to be a fairly weak effect (Brown 1987; 1980 and 1984. While some patches remained
274 I S L A N D T H E O RY A N D C O N S E R VAT I O N

isolated, the land surrounding others was tain time frame. However, it has yet to be satisfac-
abandoned during the study and became infilled torily demonstrated that the stochastic ‘island’
by forest regrowth dominated by Cecropia. relaxation effect is generally a strong one, and it
Stouffer and Bierregaard (1996) examined the use seems likely that there will be a degree of scale
of these forest fragments by understorey humming- dependency in relation both to area and isolation:
birds. The three species found before fragmentation factors that warrant further theoretical and empiri-
persisted at similar or higher numbers in the frag- cal attention. Recent work suggests that there may
ments through the 9 years of the study. Use of the be taxon dependent thresholds of habitat loss,
fragments did not differ between 1 and 10 ha frag- above which few species losses occur, but beyond
ments. Furthermore, it mattered little whether the which losses accelerate (Ewers and Didham 2005).
surrounding matrix included cattle pasture, aban- Given the significance of the assumptions of relax-
doned pasture, or Cecropia-dominated second ation to the debate on species extinction rates, it is
growth. This particular guild of birds thus appears surprising that more attention has not been given to
to be able to persist in a matrix of fragments, sec- demonstrations of these effects (see Simberloff and
ondary growth, and large forest patches. Levin 1985; Simberloff and Martin 1991; Whitmore
By contrast, the guild of insectivores, which and Sayer 1992; Brooks et al. 1999; Laurance 2002).
dominate the understorey bird community, was
much more responsive to fragmentation: both abun-
10.8 Succession in fragmented
dance and richness declined dramatically (Stouffer
landscapes
and Bierregaard 1995). Three species of obligate
army-ant followers disappeared within 2 years. The processes of land-use change which create frag-
Mixed-species flocks drawn from 13 species disag- mented systems typically initiate successional
gregated over a similar timescale, although three of changes in the habitat island remnants. This was
the species persisted in the fragments. The 10 ha apparent in the Barro Colorado island study.
fragments were less affected than the 1 ha frag- Another example is provided by Weaver and
ments, and in time the species flocks reassembled in Kellman’s (1981) study of newly created woodlots
those 10 ha fragments that were surrounded by in southern Ontario, Canada. The species losses that
Cecropia regrowth. The communities in these recon- occurred did not take the form of stochastic
necting fragments converged compositionally on turnover, with different species ‘winning’ in differ-
pre-isolation communities, while communities in ent woodlots. Instead, the ‘relaxation’ involved the
the smaller or more isolated fragments continued to successional loss of a particular subset of species.
diverge. Particular terrestrial insectivores (e.g. With appropriate management, the observed losses
Sclerurus leafscrapers and various antbirds) failed could be avoided. If occurring at all in the absence
to return to any of the fragments during the study. of disturbance and succession, turnover appeared to
To sum up: (1) the impact of fragmentation on birds be very slow-paced among the vascular flora. In this
is strongly ecologically structured, (2) different system, the EMIB effects were clearly weak, and
‘guilds’ respond in divergent fashion, and (3) the subordinate to other ‘normal’ ecological processes.
types of species most vulnerable to fragmentation In practice, nature-reserve management often
are fairly predictable, with insectivores that partici- pays considerable attention to ecological succession,
pate in mixed-species flocks being the most vulner- particularly in small reserves. Failure to do so often
able guild (e.g. Lovejoy et al. 1986; Stouffer and leads to the loss of desirable habitats. This applies to
Bierregaard 1995, 1996; Christiansen and Pitter 1997; many of the lowland heath reserves of southern
Ribon et al. 2003). England, which have long been anthropogenically
We may thus conclude that habitat fragmentation maintained. Without appropriate management,
is associated with compositional change and most areas suffer woodland encroachment. The
species losses. Theory predicts significant losses concepts of minimum dynamic areas and patch
where fragmentation is extreme, but over an uncer- dynamics (Pickett and Thompson 1978) are thus
 T H E I M P L I C AT I O N S O F N E S T E D N E S S 275

relevant to management plans. Simplified repre- contribute a simple answer to the SLOSS question,
sentations of woodland dynamics view a stand of as a strong degree of nestedness implies that most
trees as going through phases of youth, building, species could be represented by conserving the
maturity, and senescence, each of which may sup- richest (largest) patch. A low degree of nestedness,
port differing suites of interacting species. Reserves on the other hand, would mean that particular
should therefore be large enough (or managed) to habitat patches sample distinct species sets and that
ensure that they contain enough habitat patches at an array of reserves of differing size and internal
different stages of patch life-cycles to support a full richness may be required to maximize regional
array of niches. A related and often contentious diversity (cf. Kellman 1996). Knowledge of nested
issue is how fire regimes are managed. In fire-prone subset structure might therefore provide a basis for
regions there is often a cyclical pattern of post-burn predicting the ultimate community composition of
succession and fuel accumulation, leading to an a fragmented landscape, particularly if it is possible
increased likelihood of fire, repeating the cycle. The to attribute patterns to particular causes (Worthen
maintenance of a particular fire regime and patch 1996, Fischer and Lindenmayer 2005).
mosaic structure may be of crucial relevance to Blake’s (1991) study of bird communities in iso-
species diversity in a reserve system, but is often lated woodlots in east-central Illinois shows how
poorly understood (Short and Turner 1994; Milberg nestedness calculations can provide useful insights.
and Lamont 1995). Fire is also politically contentious He demonstrated a significant degree of nested-
because of the threat it poses to people and property. ness, particularly among birds requiring forest inte-
Much active management of nature reserves is rior habitat for breeding and among species
thus about keeping a mosaic of different succes- wintering in the tropics. In contrast, species breed-
sional stages, and not allowing the whole of the ing in forest-edge habitat showed more variable
reserve to march through the same successional distribution patterns. Blake’s findings resonate
stage simultaneously. The relevance of such succes- with those reported from São Paulo by Patterson
sional dynamics to species changes in reserve sys- (1990), who found significant nestedness amongst
tems appears to have received much less attention sedentary bird species, but that the full data set (i.e.
in the theoretical conservation science literature also including transient species) was non-nested.
than its significance warrants (but see Pickett et al. These results indicate that some species, often those
1992). Moreover, continuing changes in the habitats of most conservation concern, will be lacking from
of the matrix can also be extremely important to the any number of small patches, but can be found in
fate of species populations in the reserves them- the larger, richer patches.
selves (Stouffer and Bierregaard 1995, 1996; Gascon Tellería and Santos (1995) have demonstrated
et al. 1999; Daily et al. 2003). the apparent importance of nestedness of habitat.
They studied the winter use of 31 forest patches
(0.1–350 ha) in central Spain by the guild of pari-
10.9 The implications of nestedness
forms (Parus and Aegithalos, Regulus, Sitta, and
As we established in Chapter 5, many island and Certhia—tits, goldcrests, nuthatches, and treecreep-
habitat island data sets exhibit nestedness. That is, ers). They found that birds with similar habitat
when organized into a series of increasing species preferences tend to disappear simultaneously with
richness, there is a significant tendency for the reduction in forest size, thereby producing a nested
species present in species-poor (small) patches to be pattern of species distribution.
found also in successively richer (larger) patches. Simberloff and Martin (1991) have argued that
Although nestedness can be driven by selective col- establishing nestedness across a whole system is no
onization or by habitat nestedness, it appears that longer particularly exciting, but that some benefit
differential extinction plays a major role in produc- can come from examining discrepancies from nest-
ing nested structure in many habitat island data edness. Cutler (1991) developed an index, U, which
sets (Wright et al. 1998). A nestedness analysis can was designed to account for both unexpected
276 I S L A N D T H E O RY A N D C O N S E R VAT I O N

presences and unexpected absences. He used it to decision-making—input in which the contribution

identify what he termed holes and outliers. Holes of the nestedness statistics is fairly limited.
are where widespread species are absent from Moreover, there are many different indices for
otherwise rich faunas, and outliers are where assessing nestedness, and much debate over
uncommon species occur in depauperate faunas. which is the best to use for particular purposes
He applied his index to the data for boreal (Rodríguez-Gironés and Santamaría 2006). Cook
mammals and birds of the North American Great (1995) undertook a comparative study of 38 insular
Basin.He found that for mammals most of the systems using 6 different indices of nestedness,
departures from perfect nestedness were due to ordered both by area and richness. The analyses
holes, whereas for birds most departures were due ordered by richness returned higher nestedness
to outliers. Cutler speculates that this difference scores than those ordered by area in between 35
could be a function of the superior dispersability of and 37 cases (depending on the metric used).
birds, allowing them to generate greater numbers Cook’s analyses show that the broad patterns of
of outliers by recolonization events. ‘Supertramp’ nestedness are robust, but that there is a danger
species (Chapter 5) might also appear as outliers in of over-interpreting the details of a nestedness
analyses of this sort. The preponderance of holes in analysis.
the mammal data supports the importance of Formal consideration of compositional structure
extinction as a driving force in this system, with the was, for too long, sidelined in the SLOSS debate
holes representing departures from the generally and related literature. The attention now being
predictable sequence of species losses that have given to species compositional patterns across sys-
occurred during the Holocene. tems of habitat islands is thus welcome (Fischer
Such analyses might appear to provide impor- and Lindenmayer 2005). A nestedness index can
tant information for conservation. However, provide one compositional descriptor and can per-
Simberloff and Martin (1991) rightly caution that, to haps aid identification of risk-prone species; but
a large degree, extinction across these mountain- should not be given primacy in conservation plan-
top habitat islands has been inferred rather than ning. The discovery of nestedness at a particular
demonstrated to have happened. The simple statis- point in time does not necessarily provide clear
tics can, moreover, mask more complex and impor- insights as to the probability of maintaining the
tant underlying details. In illustration, they show same sets of species (or any particular species) over
that the same nestedness score can be generated for time (Simberloff and Martin 1991). The isolates may
species with widely differing underlying distribu- be subject to turnover and/or species attrition in
tions. In the Maddalena archipelago (Sardinia), two new ways dictated by the changing biogeographic
species have a nestedness score of zero (by the circumstances of the landscape in which the frag-
Wilcoxon statistic). One, the rock dove (Columbia ments occur. As Worthen (1996, p. 419) put it, nest-
livia), is found across the entire range of island sizes, edness is not a ‘magic bullet’, ‘ . . . no single index
and is absent from only 2 of 16 islands. The second, should be expected to distil the informational con-
the little ringed plover (Charadrius dubius), has been tent of an entire community, let alone predict how it
observed only once in the archipelago, breeding will react to habitat reduction or fragmentation’.
only in one year. As they point out, there may be dif-
ferent explanations for accordance or deviation
10.10 Edge effects
from nestedness in particular species. The explana-
tion of such patterns requires basic information on The boundary zone, or ecotone, between two habi-
species’ presence–absence, abundances, minimum tats, being occupied by a mix of the two sets of
area requirements, habitat use at different sites, and species, is often richer in species per unit area than
temporal regularity in occupying differing types of either of the abutting core habitat types. In the
patch. In their view, given such data, it becomes reserve context, attention has focused on the
possible to offer sensible input into conservation reserve edge, typically where a woodland reserve is
 EDGE EFFECTS 277

surrounded by a non-woodland matrix. For some (a)

species groups, possession of a large expanse of 1.0 specie
s
Edge
such edge habitat can increase the overall species 0.8
cies

Proportion
number found on a reserve, an apparently benefi- re spe
0.6 Co
cial effect (Kellman 1996). But, the argument goes, if Mammals
the additional species supported by the edge habi- 0.4

tat are those of the matrix, then they do not depend 0.2 Edge-co
re specie
s
on the reserve and should be discounted in evalu- 0
ating the merits of the reserve system. 0 500 1000 1500 2000
In illustration, Humphreys and Kitchener (1982)
(b)
classified vertebrate species into those dependent 1.0 ecies
Edge sp
on reserves and those which persist in disturbed
0.8
areas in the intervening agricultural landscape of cies

Proportion
re spe
0.6 Co
the Western Australian wheat belt (Fig. 10.8). They Passerine
found that the small habitat islands, which were 0.4 birds
mostly edge habitat, were disproportionately rich 0.2 Edge-co
re specie
in ubiquitous species. Those species restricted to s
0
woodlots typically required a larger area (cf. 0 500 1000 1500 2000
Hinsley et al. 1994). This pattern was found sepa-
(c)
rately for lizards, passerine birds, and mammals. 1.0 s
By analysing all species within a taxon together, specie
Edge
0.8
small reserves came out in a better light than they ies
Proportion

pec
warranted. Similar results are reported by Watson 0.6 res
Co Lizards
et al. (2004a,b) in a study of birds in littoral forests 0.4
and surrounding habitats in south-eastern Madag-
0.2 Edge-core species
ascar. Core forest locations were found to be richer
than edge or matrix habitats, with some 68% of the 0
0 500 1000 1500 2000
forest dependent species found to be edge-sensitive.
Frugivorous species and canopy insectivores Area (ha)

(including six endemic vanga species) were gener- Figure 10.8 The proportion of the core species (those dependent
ally edge-sensitive, while in contrast, sallying on the reserves) and the edge species (those which can also survive
insectivores were edge-preferring. At least part of outside the reserves in disturbed habitat) which were present in a
the edge-sensitivity recorded was attributable to series of 21 nature reserves in the Western Australian wheatbelt in
changes in vegetation structure at the remnant edges. relation to the number of each present in a reserve of 2000 ha. The
curves with negative slopes represent the difference between the
Consistent with this edge sensitivity, forest curves for edge and core species, and represent the relative excess of
dependent species were generally lacking from edge species over core species compared with their distribution at
fragments of less than 10 ha, thereby demonstrat- 2000 ha. Data for (a) mammals, (b) passerine birds, and (c) lizards.
ing a small-island effect (Chapter 4), masked in (Redrawn from Humphreys and Kitchener 1982, Fig. 1.)
analyses of all bird species richness (Fig. 10.9).
Hansson (1998) used a rather different approach in In particular circumstances, ecotones may have
his studies of edge effects in ancient woodlands in negative implications for core woodland species
Sweden, but again found differences in bird com- (Wilcove et al. 1986), in that although they may sup-
munity composition that appeared to relate to habi- port additional species or larger populations of
tat structure at the edges. He analysed the degree of non-woodland core species, these populations may
nestedness, finding that whereas census data for have negative interactions with deep-woodland
birds from the centres of the patches were nested, species. Studies in North America have suggested
the samples from the edge habitat were non-nested. that the nesting success of songbirds is lower near
278 I S L A N D T H E O RY A N D C O N S E R VAT I O N

70 were found to be higher 500 m into the forest than

60
30 m and 100 m from the edge. In contrast to
R2 = 0.88 Burkey’s egg predation data, Delgado et al. (2005)
Species Richness

50
report higher predation pressure under closed
40 R2 = 0.92 canopy than within edge habitats in their study of
30 artificial nest predation by ship rat (Rattus rattus) in
20 R2 = 0.78
laurel forest on Tenerife. This illustrates that the
relationship between a reserve and its surrounding
10
matrix is not subject to easy generalization. There
0 are species that share both zones, and just as there
are matrix species which may impact negatively
0.1 1 10 100 1000
upon core reserve species, there may also be reserve
Area (Log, ha)
species (dependent on it for breeding and cover)
Figure 10.9 Bird species richness–area relationships for littoral which exploit resources in the matrix. The concepts
forests, southeastern Madagascar. Closed circles (and bold, solid line): of source and sink populations may again be useful
all bird species; open circles: forest dependent bird species, fitted by in this context, as the maintenance of maximal
linear regression (solid line), and by break-point regression (broken
diversity across a landscape depends on sufficient
line), demonstrating a small-island effect. (From Watson et al. 2004b.)
source or core habitat for species of each major
habitat. The identification of edge effects provides a
the forest edges than in the interior because of part of such an analysis.
higher densities of nest predators—e.g. blue jay
(Cyanocitta cristata), weasel (Mustela erminea), and
racoon (Procyon lotor)—around forest edges. 10.11 Landscape effects, isolation, and
Wilcove et al. (1986) estimated that as a result, corridors
reserves of less than 100 ha will not support viable
The benefits of wildlife corridors
populations of forest songbirds. A few studies of
birds and invertebrates have suggested that some The premise of much of the literature discussed in
species prefer edge habitats for breeding even this chapter is that habitat connectivity is beneficial
though mortality rates at edges may be higher, a to long-term survival, as it enables gene flow
phenomenon termed an ecological trap (Ewers and within populations and metapopulations. Habitat
Didham 2005). Paton (1994) has undertaken a connectivity might be achieved by having stepping
critical review of studies of nest predation in edge stones, or corridors, of suitable habitat linking
habitats. He concluded that, in general, there does larger reserves together. In practice, habitat corri-
appear to be evidence for a depression of breeding dors act as differential filters, enabling the move-
success, due either to enhanced predation or ment of some species but being of little value to
through parasitism. However, these forms of edge others. As Spellerberg and Gaywood (1993) point
effect usually operate only within about 50 m of an out, we are not merely concerned with forest penin-
edge, a narrower belt than some have claimed (e.g. sulas or hedgerows. All sorts of linear landscape
Terborgh 1992). Nonetheless, such an effect implies features, such as rivers, roads, and railways, may
that the effective reserve area may be less than that act as conduits for the movement of particular
described by the perimeter of the isolate. species. Equally, they may represent barriers or
Edge effects may be taxon or even system hazards for others (Reijnen et al. 1996). Harris (1984)
specific. For instance, Burkey (1993) undertook an points out that landscapes with great topographic
experimental study of egg and seed predation with relief channel their kinetic energy via dendritic trib-
distance from the edge of a patch of Belizian rain utaries, main channels, and distributaries, and that
forest. Egg predation rates were higher in a 100 m these in turn influence the landscape template in
edge zone, but, conversely, seed predation rates characteristic ways. The stream-order concept of
 L A N D S C A P E E F F E C T S, I S O L AT I O N, A N D C O R R I D O R S 279

the fluvial geomorphologist may be a means of

understanding species distributions within the
landscape (Fig. 10.10), and of understanding move-
ments of species between disjunct habitat patches
(Fig. 10.11). Studies of features such as hedgerows
have not always found them to be as effective in
connecting up woodlands as was hoped, with some
species moving as well through surrounding fields
and others simply not moving well along the
hedges. However, habitat corridors may be impor-
tant in particular cases.
A useful illustration of how corridors can be ben-
eficial comes from the study by Saunders and
Hobbs (1989) of Carnaby’s cockatoo (Calyptorhyncus
funereus latirostrus) from the Western Australian
wheat belt, an area of 140 000 km2 in the south-west
of the state, 90% of which has been cleared for
agriculture. The Carnaby’s cockatoo is one of
Australia’s largest parrots and was once the most
widely distributed cockatoo in the region. It breeds
in hollows in eucalypt trees and eats seeds and
insect larvae from plants in the sand-plain heath. It
congregates in flocks, both to nest and forage. Figure 10.10 Association of different-sized carnivorous mammal
Individuals may live 17 years and breeding status is species with stream order and typical food particle size in accordance
not attained until at least 4 years of age, each pair with the stream-continuum concept. (Redrawn from Harris 1984,
rearing a single offspring at a time. The widespread Fig. 9.8.).
clearance of the land has removed extensive areas of
the vegetation type in which they feed, replacing it fragmented landscapes. Given the relatively long
with annual crops that are useless to the birds. In life cycle of these birds, and their flocking behav-
recent developments, wide verges of native vegeta- iour, failure can have a sudden and wholesale
tion have been left uncleared along the roads. These expression in the bird’s disappearance. The big
act to channel the cockatoos to other areas where reduction in area is fairly recent, and the cockatoo is
food is available. not yet in equilibrium with the new regime. It is
Cockatoos have not survived in areas of earlier still on a downward path and is viewed as a species
clearances, carried out without these connecting in danger.
strips, as once they have run out of a patch of
acceptable habitat it takes a long time for the flock
The benefits of isolation
to find another patch of native vegetation. The
more patchy the vegetation, the less successful the Although much of the literature is concerned with
birds are at supplying adequate food to rear their maintaining population movements, there are also
nestlings. Furthermore, the narrow road verges of contexts in which isolation of populations of a target
these early clearances result in a higher incidence of species might be desirable (Simberloff et al. 1992),
road deaths. The example illustrates that the degree e.g. where there is a threat from a disease organism.
and nature of connectivity of different landscape Young (1994) evaluated 96 studies of natural
elements—in this case of the breeding and feeding die-offs of large mammal populations, defined as
habitats, and of the vehicular hazards—are critical cases where numbers crashed by 25% of the
to the survival of this species in these newly individuals or more. He noted that populations
280 I S L A N D T H E O RY A N D C O N S E R VAT I O N

0 8

km

Diamond

Peak

ge
Wilderness

a de s R an
C as c
Figure 10.11 A possible spatial and size-frequency distribution of old-growth habitat islands arranged along riparian strips at progressively
greater distances from a current wilderness area in the Willamette National Forest, USA, designed to provide the optimal degree of connectivity
between patches in a system structured by the dendritic pattern of drainage. In this hypothetical system, the most distant islands are generally
larger in order to counter presumed lower dispersal. (Redrawn from Harris 1984, Fig. 10.2.)

subject to large-scale phenomena such as drought movement within a metapopulation to enable the
and severe winters may not be protected from such movement of propagules of the target species but
die-offs by population subdivision. But, where the not of disease. But much will depend on the nature
crashes are produced by disease epidemics, subdi- of the intervening landscape matrix and the extent
vision may be beneficial, and the creation of linking to which it filters target and ‘pest’ species (Ricketts
corridors, and translocation efforts, may be harmful. 2001; Ewers and Didham 2005).
Young noted in his review that most herbivore die- The spread of exotic competitors and predators
offs were due to starvation, but carnivore die-offs and of fire into reserves can also be problematic
were more often attributed to disease. An epidemic (Spellerberg and Gaywood 1993; Lockwood and
need not eliminate all the individuals in a popula- Moulton 1994). It has been suggested that alien
tion for it to have a crucial role in causing the even- plant invasion can be reduced by isolating reserves
tual extinction of a population from a reserve. and surrounding them with simplified cropland
Combining these considerations with meta- oceans (Brothers and Spingarn 1992). Short and
population modelling, Hess (1996) argues that Turner (1994) found that the decline and extinction
there should theoretically be an optimal degree of of certain species of native Australian mammals
 L A N D S C A P E E F F E C T S, I S O L AT I O N, A N D C O R R I D O R S 281

were most parsimoniously explained by reference stantial shifts in the distribution of species, and that
to the roles of introduced predators (foxes and cats) the resulting species migrations will be impeded by
and herbivores (rabbits and stock). This explained the human sequestration of land to agriculture and
the continued persistence of the native species on other purposes. Hence, on these grounds, it would
offshore islands from which the exotic species were seem to be prudent to plan more or less continuous
absent. In these cases, isolation is beneficial habitat corridors that straddle major climatic/
(as Janzen 1983) and many reserves set in ‘seas of elevational gradients where this is feasible (e.g.
altered habitat’ within mainland Australia have Bush 1996, 2002).
proved to be insufficiently isolated to prevent their
invasion by exotic pests and pathogens.
Reserve systems in the landscape
As we have seen, the debate about habitat corridors
Corridors or isolation?
has broadened from fairly simple theoretical begin-
Is it possible to offer guidance about the situations nings to a consideration of how a whole range of
when corridors may be beneficial and when not? differing natural, semi-natural, and artificial
First, as Dawson (1994) notes, all-purpose corridors features are configured within a landscape. For
do not exist, as each species has its own require- instance, countless thousands of birds are killed
ments for habitat, its own ability to move, and its each year through collision with motor vehicles
own behaviour. Corridors, like other landscape and with overhead power cables (Bevanger 1996;
components, therefore act as filters. He points out Reijnen et al. 1996). Research has shown that deaths
that many rare and threatened species are unlikely through collisions with cables can be greatly
to benefit from corridors, because the corridor reduced by consideration of important flight paths
would have to contain their presumably rare habi- during construction, by design features of the
tat, i.e. rare species may require odd corridors. gantries, and by attaching a variety of objects to the
Some corridors are essential in providing links cables to enable birds to sight them (Alonso et al.
between preferred habitats for animals that under- 1994). Paved roads and tracks can play an impor-
take regular seasonal migrations, e.g. some fish, tant role as corridors favouring the introduction of
amphibians, reptiles, and for large mammals in alien plants (Arévalo et al. 2005) or animals
both the seasonal tropics and the Arctic. On the (Delgado et al. 2001) into otherwise well-preserved
other hand, for some populations, corridors may ecosystems, but also creating corridors of suitable
act as ‘sinks’, drawing out individuals from the habitat that can in cases allow endemic species (e.g.
main habitat area but not returning individuals to lizards) to disperse through inhospitable forest
supplement it, in which case they may do more matrix, enabling genetic exchange between other-
harm than good. In other cases, they may be fairly wise isolated populations (Delgado et al. in press).
neutral in their ecological cost–benefit, but perhaps On the other hand, wide roads and, especially,
be quite expensive to purchase and set up if not fenced highways, may act to impede the move-
already existing in a landscape. ments of large terrestrial animals, thus fragmenting
Simberloff et al. (1992) point out that numerous populations and interfering with migration. Such
corridor projects have been planned in the USA, information needs to be integrated into improved
potentially costing millions of dollars, despite the management of whole landscapes (Spellerberg and
lack of data on which species might use the corri- Gaywood 1993). In short, a landscape ecological
dors and to what effect. Once again, the cost–- framework is needed.
benefit equation for corridors may depend on Landscape components include both habitat
properties of the system involved, including scale patches and the matrix in which the patches are
considerations and assessment of the major drivers embedded. Modelling exercises typically favour
of species loss. In the longer term, it has been clumped configurations of reserves, treating the
argued that climate change is likely to drive sub- matrix as a uniform ‘sea’. In practice, the optimal
282 I S L A N D T H E O RY A N D C O N S E R VAT I O N

configuration must take account of the details of

the landscape involved (e.g. Fahrig and Merriam
1994). For instance, there may be greater opportu-
nity for dispersal between two distant reserves
linked by a river and its adjacent riparian corridor,
than between two similar but closer reserves sepa-
rated by a mountain barrier of differing habitat
type (Fig. 10.11). For butterflies, differences in land
use as subtle as switches from one woodland type
to another can significantly influence the passage of
butterflies from one favoured habitat patch to
another (Ricketts 2001). An early attempt to place
island theories into realistic landscape contexts was
provided by Harris (1984). Recognizing that there
were limits to the amount of land that would be put Figure 10.12 Harris’s (1984) proposed system for the management
of forest habitat islands on a long-rotation system, such that different
over to reserve use, Harris advocated a series of sectors are cut in a programmed sequence, but the core old-growth
reserves placed within the landscape as dictated by patch is left uncut. This recognizes the importance of patch dynamics to
geographical features such as river and mountains, habitat diversity and how such management may serve both economic
to encourage population flows between reserves. and conservation goals. (Redrawn from Harris 1984, Fig. 9.5.).
Most reserves would remain in forestry manage-
ment. Within such reserves, there should be an
undisturbed core of forest that is never cut, pro- need of the magpie geese to relocate in response to
viding habitat for species requiring undisturbed, patchiness of rainfall. They identify four possible
old-growth conditions, around which commercial strategies for conservation in the Northern Territory:
operations might continue, following a pattern of
rotational partial felling. This would provide a ● The status quo, with improvements to ensure all
mosaic of patches of differing successional stage, vegetation types are included in the network. This
maximizing habitat diversity, while maintaining would not solve the problem, as it is not just
income (Fig. 10.12). His strategy aimed to satisfy vegetation types but the year-to-year variation in
the requirements of island theories, patch dynamic carrying capacity across the region that matters.
models, and economic realities. The implementa- ● Seeking inclusion of known habitat of important
tion of such a policy requires concerted action from species into the reserve system. For the magpie
a wide variety of agencies and is thus easier to geese, this might mean inclusion of all wetlands.
sketch out than to bring to a realization. However, it would be difficult to decide which
species to concentrate on and it would be impracti-
cal to gazette all the land in question because of
Species that don’t stay put competing claims for use.

In the Northern Territory of Australia, as in many ● Developing very large reserves which span
other systems across the globe, mobility and mas- extensive slices of the environmental gradient, just
sive population fluctuations in response to a highly as does the present Kakadu National Park, a
variable climate are features of the wildlife. 20 000 km2 reserve. However, this would require
Woinarski et al. (1992) cite as a particular example vast reserves in the semi-arid and arid region, in
the magpie goose (Anseranas semipalmata). They which, at the present time, the largest reserve is of
note that a conventional reserve network consisting 1325 km2.
of discrete national parks (such as they have devel- ●Supplementing the representative reserve net-
oped for plants in the region) will not cater for the work with measures that protect wildlife habitat on
 D O E S C O N S E R VAT I O N B I O L O G Y N E E D I S L A N D T H E O RY ? 283

large areas of unreserved land. They see this as the of other forms of compositional structure. These
only viable answer in the long term, requiring a non-random patterns in how communities assem-
broad educational and political strategy to get there ble, or alternatively how they change in response to
(Woinarski et al. 1992; Price et al. 1995). fragmentation, demand that we reappraise the con-
So, wildlife conservation must in part be con- tribution of ‘island ecology’ within conservation
cerned with the practicalities of conserving outside biology.
reserves. The Northern Territory example may Much of the earlier applied island literature
seem an extreme case when viewed from many seemed to assume that matrix contexts were
parts of the world, yet the argument surely applies relatively insignificant and that stochastic decay
to some degree everywhere. towards equilibrium would restore a new, depau-
perate balance. These were arguably useful simpli-
fying assumptions, but the behaviour of habitat
10.12 Does conservation biology need islands is perhaps more realistically captured
island theory? within the notions of flux of nature described by
Zimmerman and Bierregaard (1986) undertook an Pickett et al. (1992). This framework stresses physi-
analysis of Amazonian frogs which demonstrated cal as well as biotic processes (e.g. succession), the
that knowledge of the autecological requirements role of episodic events (e.g. fires or hurricanes), and
of the species would be critical to planning a the differing degrees of connectivity of populations
reserve system (not that a system would necessar- across ‘filtering’ landscapes. We have also seen that
ily be designed just for frogs). They undertook a the nature of the matrix can be crucial to what goes
prediction by a species–area approach and showed on in the habitat islands within the landscape. How
that it lacked relevance in the light of the available much of this is ‘island biogeography’ is a moot
autecological data. Frog species differ in the habitat point, but insularity remains a key property and
required for breeding: whereas some are strictly issue within these other topics.
streamside breeders, others require landlocked per- The application of equilibrium (EMIB) assump-
manently or periodically flooded terra firma pools. tions to reserve systems has been neatly critiqued
The authors concluded that if such data are avail- by Saunders et al. (1991, p. 23):
able for enough species in the system being consid- It is commonly assumed that at some stage the remnant
ered, then there need be no recourse to abstract will re-equilibrate with the surrounding landscape. It is
principles. If they are not available, it is still prefer- however, questionable whether a new stable equilibrium
able to conduct a rapid survey and collect biogeo- will be reached since the equilibration process is liable to
graphic and ecological data. This is just one of be disrupted by changing fluxes from the surrounding
many studies that have championed empirical over matrix, disturbances, and influx of new invasive species.
theoretical ‘island’ approaches. Does conservation The final equilibrium can be likened to an idealized end
biology need island theories? point that is never likely to be reached, in much the same
fashion as the climatic climax is now conceptualized in
succession theory. Management of remnant areas will
A non-equilibrium world? thus be an adaptive process directed at minimizing poten-
tial future species losses.
Worthen (1996) bemoans the fact that the bias towards
species–area relationships over compositional–area The general premise of much of the applied
relationships continues to be a feature of the litera- island biogeography literature is that (1) the alarm-
ture, including some of the most widely used intro- ing fragmentation of many ‘natural’ systems will
ductory ecology texts. This bias continues despite continue and therefore (2) if much wildlife is to sur-
the demonstration, in numerous studies, of nested- vive, it will have to do so in reserve systems, and
ness, community similarity of habitat fragments, that (3) we have very little time in which to act.
the relevance of successional patterns, and a variety These concerns explain the initial attractiveness of
284 I S L A N D T H E O RY A N D C O N S E R VAT I O N

deriving general principles for reserve systems The problem of biodiversity loss (including
from island theory. Increasingly, the resulting lines population-level decline, genetic erosion and species
of reasoning have been attacked on the grounds: losses) is a depressingly real one and the pace of
first, that many valid generalizations are self-obvi- species extinctions is undoubtedly accelerating
ous platitudes; secondly, that models are often both (Whitmore and Sayer 1992; Turner 1996). Yet, as
too simplistic and effectively untestable; and, Simberloff et al. (1992) observe, ideas derived from
thirdly, that many theoretical generalizations have island theory, and the so-called SLOSS principles
achieved the status of dogma (Doak and Mills (specifically the benefits of corridors), became
1994). Yet, the general principles derived from embedded in policies of bodies such as the IUCN
island theory have influenced conservation policy and World Wildlife Fund without having been val-
and they remain pervasive. Because of this, it is idated. Thus, policy-makers and legislators have in
important to understand both the strengths and, cases acted upon such uncertain scientific ‘princi-
particularly, the weakness of the ‘island’ theories ples’. Once released from the bounds of scientific
discussed in this and the previous chapters. journals, scientific information is often poorly
If our focus is on global losses of species diver- understood. It is therefore important not to offer up
sity, then clearly the loss of—and fragmentation general principles and rules where the evidence for
of—species rich habitats is a key concern. But, the them is equivocal and, in cases, contradictory
stochastic consequences of reduced population (Shrader-Frechette and McCoy 1993).
sizes are not necessarily simple to understand and Nonetheless, we should not simply discard the
model. Some species can persist well in fragmented theoretical frameworks of island theory, metapop-
landscapes providing that the areas of remaining ulation theory, etc. They have their relevance to
habitat are not too small and isolated. Much may particular systems and a wider and important role
depend on the previous history of the region in helping us to understand the sorts of effects that
(extent of initial departure from equilibrium might be operative in real-world contexts, and that
assumptions), the taxa considered, and the scale therefore should be tested for or monitored. As
sensitivity of the systems, and how the habitat Doak and Mills (1994) have put it, we should face
matrix (the land outside the fragments) is managed. up to the problems rather than use them as an
Simple species–area models, particularly as applied excuse for ignoring the theoretical debates: ‘it is
on regional scales, may not in the end provide reli- incumbent on us to teach such complexity to
able estimates of how many species really are ‘com- managers and nonbiologists, rather than attem-
mitted to extinction’, i.e. still around but heading pting to snow them with undefendable over-
for an inevitable nemesis. Some would regard this generalizations’.
assessment as complacent and over-optimistic. For The notion that species disassembly (as well as
instance, Pimm and Askins (1995) argue that con- assembly) is fairly structured and predictable, at
cerns about the precise causal mechanisms of least in terms of functional ecological types or
extinctions, as expressed here, miss the point and guilds, is one that has reappeared in several guises
that, in practice, simple predictions of species losses in this chapter (e.g. from studies of relaxation,
based on species–area relationships are actually experimental fragmentation, incidence functions,
fairly reliable. From their analyses of North American nestedness, and edge effects). In the case of the
forest birds they suggest that if such estimates err, Manaus habitat fragmentation study, similar struc-
they tend to be conservative, i.e. predicting fewer turing was evident both in the initial losses (disas-
extinctions than actually occur. We are not con- sembly) and, in particular circumstances, in the
vinced that their claim can be generalized, as so recovery of bird communities that followed. This
much seems to depend on issues such as system reminds us, first, that habitat change does not have
configuration and the actual drivers of species to be a one-way process, and, secondly, that it
threat and extinction (see discussion of the Atlantic might be possible to target conservation efforts
forest analyses, above). fairly directly on to particular sets of vulnerable
 D O E S C O N S E R VAT I O N B I O L O G Y N E E D I S L A N D T H E O RY ? 285

species. In this sense, ‘island’ approaches have pro- important may be the activities of animals as
duced some dividends. It is now a question of dispersers and pollinators of plants (and as preda-
learning what they are, and differentiating between tors), both in contexts of forest maintenance and of
those effects that are weak and those that are strong recolonization of disturbed areas. Different types of
(the world is neither wholly non-equilibrial nor animals disperse different, albeit overlapping,
wholly equilibrial). It seems that we can work out suites of plant species. Studies of cleared areas in
which sorts of species are most threatened and the Neotropics have suggested that bats may be
some of the major reasons why they are threatened, more significant in seeding open habitat than are
and it should therefore be possible to offer some day-flying birds. The efficacy of birds, in particular,
management solutions. Some of these insights have may also be positively influenced by the availabil-
indeed been derived from ‘island’ studies. ity of some woody cover in an otherwise open area
For many species, reserves are essential. Very (Estrada et al. 1993; Gorchov et al. 1993; Guevara
often this is because of specific threats posed by and Laborde 1993). Thus, for instance, cheap plant-
people (or our commensals) and the protection that ings of small clumps of plants may encourage
can be afforded within the reserve. Simberloff speedy return of forest around the inocula.
(1992) comes down in general on the side of large, The Krakatau recolonization data demonstrate
continuous blocks of habitat. He argues that no the significance of birds and bats in transporting
existing theory adequately describes the joint seeds between sites, and thus also supports their
effects of loss of area and fragmentation, but that significance for population interchange between
there are empirical observations, such as those to forest patches in the tropics. Maintenance and
do with predation discussed earlier, to suggest that restoration of forest might thus necessitate protec-
for many species the combined impact is indeed tion for vertebrate species, if not in their own right,
strongly deleterious: ‘Probably any species that has for their interactions with plants. It is not ‘merely’
evolved in large, relatively continuous habitat has biodiversity which is at stake, as in cases there may
traits that are maladaptive in small, isolated frag- be huge economic costs to the loss of pollinator
ments’ (p. 85). Beyond this generalization, the species on which economically useful plants
means, rate, and extent to which extinction is deliv- depend (Pannell 1989; Fujita and Tuttle 1991; Cox
ered, are likely to be idiosyncratic. The theoretical et al. 1992; Ricketts 2004). At the same time, the
basis for calculating likely regional and global Krakatau study demonstrated that particular
extinctions as a result of the rapid phase of habitat groups of plants, those dispersed by non-volant
loss currently under way remains highly uncertain. frugivores and the large-seeded, wind-spread
For a careful review of the evidence and basis for species, are likely to be greatly hampered by patch
calculations see Whitmore and Sayer (1992). isolation (Whittaker and Jones 1994a,b; Whittaker
et al. 1997). These topics have recently seen a resur-
gence of research interest (Ewers and Didham
Ecological hierarchies and fragmented
2005), with studies in both Africa and the
landscapes
Neotropics demonstrating the negative impacts on
As with other branches of island theory, approaches frugivore-mediated plant dispersal of forest frag-
to the debates on reserve configuration that take mentation (e.g. Cordeiro and Howe 2001; Martínez-
no account of hierarchical interactions between Garza and Howe 2003).
organisms of differing trophic levels are incomplete It seems likely that some area effects involve
(Terborgh 1992). For many animal groups, plants threshold responses and that these operate via hier-
(collectively or individually) determine the ability archical links within ecosystems (Soulé et al. 1988;
of particular animal species to occupy a habitat Terborgh 1992). For instance, below a certain size,
island. Successional changes in plant communities an isolated habitat island may be incapable of sup-
may often explain turnover in bird or butterfly porting populations of large predatory vertebrates,
communities (Chapter 5; Thomas 1994). Equally with the consequence that higher densities of their
286 I S L A N D T H E O RY A N D C O N S E R VAT I O N

prey species may be maintained, with potentially There may also be important dynamic features,
significant consequences further down the food such as the spatial and temporal oscillations that
chain. Thus, the densities of medium-sized terres- characterize elephant populations in the Tsavo
trial mammals are between 8 and 20 times greater National Park in Kenya, on a cycle of roughly
on Barro Colorado Island, which lacks top preda- 50 years. Even with very large reserves, migratory
tors, than in a comparable ‘mainland’ site in which animals may ignore park boundaries and, espe-
they occur. Terborgh (1992) suggests that this may cially in lean years, may move into areas outside,
be having important selective impacts on plant where they are most unwelcome and perhaps also
regeneration. On some smaller islets nearby in Lake unsafe. These patterns can provide big manage-
Gatun, not only are the (true) predators absent, but ment problems, even when a very large area is
so too are those smaller mammals that act as seed- enclosed as a reserve (Woinarski et al. 1992). The
predators, the squirrels, peccaries, agoutis, and elephants illustrate another idea: that of the
pacas which have such high densities on Barro keystone species, meaning a species that is so
Colorado island. These small islets became domi- critical to the functional character of an ecosystem
nated in the 70 or so years since their isolation by that its removal would cause a chain of alterations
large-seeded tree species. This is plausibly because (e.g. Soulé 1986; Pimm 1991). Elephants have a key
these plants no longer suffer high rates of attrition at role as grazers. Without them successional changes
the point of germination and establishment, and occur in the vegetation, altering the carrying
have therefore been able to out-compete the smaller- capacity of the system for other species. High
seeded species to a greater extent than usual. densities of elephant, on the other hand, can lead
Predators undoubtedly have a crucial role to play to overgrazing and widespread destruction of
in fragmented landscapes, not just in terms of trees. In such cases, conservation management
predation within the habitat island, but also within naturally focuses around managing the keystone
the matrix. Simberloff (1992) notes that in the species.
Manaus fragmentation study the bat falcon (Falco
rufigularis) has repeatedly been seen to chase birds
Climate change and reserve systems
flying over cleared areas from fragment to fragment.
It may therefore be unsurprising that most under- Range shifts driven by significant climatic change
storey birds will not willingly cross gaps of even have been a feature of the Quaternary period. There
80 m. Similarly, in the USA, the northern spotted is no reason to consider these climatic fluctuations
owl (Strix occidentalis caurina), of metapopulation to be at an end, and there is also mounting concern
theory fame, previously was a species which rarely that humans have initiated a phase of rapid climatic
left closed forest. In the newly fragmented land- warming. Researchers are now focusing on the
scapes, as many as 80% of yearling males die, abilities of plant and animal species to track
apparently from predation by great horned owls their required climatic envelope across the now
(Bubo virginianus) and goshawks (Accipiter gentilis) fragmented landscapes (Huntley et al. 1995). Key
as they disperse over cleared areas. This was one factors are the speed of climatic change, the
consideration that persuaded the interagency dispersal abilities of the target species, and
federal committee responsible for the management the characteristics of the landscapes across which
plan to shift their strategy towards larger, more con- species may have to move, which may range from
tinuous blocks. This brief consideration of predation the optimal to the benign to the hostile. Grappling
helps us recall the point that the particular mecha- with such scenarios requires a variety of research
nisms responsible for failures to sustain populations tools (Bush 1996, 2002; Pearson and Dawson 2003;
or to recolonize habitat islands may thus be poorly Araújo et al. 2005b), amongst which are the tools of
predicted by ‘species–area’ type-approaches, or for island biogeography reviewed in these pages
that matter by metapopulation models. (Boggs and Murphy 1997).
 I S L A N D B I O G E O G R A P H Y TO C O U N T RYS I D E B I O G E O G R A P H Y ? 287

10.13 Concluding remarks: from held within a reserve, but also which sets of species.
island biogeography to countryside The number of species held in a reserve (or reserve
biogeography? system) is actually less important than to conserve
those species which cannot survive outside the
As made clear at the outset of this chapter, this remnants (e.g. Newmark 1991).
review was not intended as a summary of all impor- Some recent efforts have been made to move
tant branches of wildlife conservation. However, we beyond an exclusive focus on (forest) fragments
have stressed along the way the importance of aute- towards understanding the role of such habitat
cological, distributional, and other forms of data. islands within mixed-use landscapes. This switch in
The student of conservation biology will also need emphasis comes under varying headers. For
to have a knowledge of such themes and issues as example, Watson et al. (2005) show that the incidence
ex situ conservation efforts in zoos, and other dedi- functions of woodland bird species in three different
cated breeding facilities; germplasm (seed) banks; landscapes in the Canberra area (Australia) differ
species translocation schemes; reserve management significantly, seemingly as a function of differences
practices; CITES and other wildlife legislation; in properties of the landscape matrix within which
poaching and trade in endangered species; strategic the woodlands are embedded. Hence, Watson et al.
conservation policy; the philosophy, politics, and (2005; and see Box 5.2) join others (e.g. Gascon
economics of conservation and other fields besides et al. 1999; Cook et al. 2002, Ewers and Didham 2005)
(e.g. Primack 1993; Shrader-Frechette and McCoy in calling for greater attention to ‘matrix effects’.
1993; Whittaker et al. 2005). Hughes et al. (2002) adopted a slightly different
It must also be recognized that there can be many approach to their study in southern Costa Rica,
aims and purposes for shaping conservation focusing on the extent to which native forest species
management, ranging from the aesthetic, through make use of the surrounding countryside: they
the scientific, to the economic. We may wish to con- found that some 46% of bird species foraged often
serve systems or species which are ‘representative’, kilometres away from extensive areas of native
‘typical’, rare, speciose, nice to look at, of recre- forest. Although they stress that not all species can
ational value, or provide economic return (Ratcliffe be so readily accommodated outside large tracts of
1977; Shrader-Frechette and McCoy 1993). Such native forests, their work supports the importance
multiplicities of purpose require requisite tools; the of developing ‘countryside’ landscapes that are
island theories have their place in the tool kit, but biodiversity-friendly and penetrable by native
they should not always be the first to be reached for. fauna (as Harris 1994). Daily and colleagues (e.g.
Allied to the habitat remnant/reserve strategy, Daily et al. 2001, 2005; Hughes et al. 2002) coin the
wherever possible, conservationists should argue term ‘countryside biogeography’ for this switch in
for greater priority to extensive rather than intensive attention from remnants per se, to the way in which
conservation, i.e. for environmental management remnants function within whole landscapes. This
policies to encourage the survival (or passage) of switch in emphasis is similar to that promoted by
many species outside of the more closely protected Rosenzweig (2003) under the heading ‘reconciliation
reserves systems (cf. Rosenzweig 2003; Daily et al. ecology’. Whether we label this shift in emphasis
2005). In short, to work to prevent habitat islands ‘matrix effects’, ‘countryside biogeography’, or
and reserves becoming more and more like real ‘reconciliation ecology’, the common element is a
islands—except in those biogeographical contexts realization that effective conservation must include
where insuralization is actually beneficial to consideration of what happens outside reserves, as
survival prospects. We have seen in this chapter the way we shape the countryside, whether we farm
that issues such as size, shape, and configuration intensively or extensively, whether we retain
within a landscape are important to reserve suc- hedgerows and trees within mixed landscapes, can
cess, not just in terms of how many species will be have profound implications for regional diversity
288 I S L A N D T H E O RY A N D C O N S E R VAT I O N

and for abundances of wildlife (e.g. see Gascon et al. effects, or to environmental disturbance and change.
1999; Gates and Donald 2000). Stochastic demographic effects may have been over-
Conservation requires pragmatic decision- estimated. Genetic effects are potentially complex,
making. As we continue to fragment landscapes, and vary between taxa. The significance of environ-
island effects may inform such decision-making, mental change or catastrophe is probably crucial to
but should not be oversimplified. There is no single population viability, but can be difficult to model.
message, and no single island effect; indeed, MVPs therefore need to be established separately for
insularity may, in at least a minority of cases, bring different types of species, and management regimes
positive as well as negative effects (Lockwood and must be responsive to changing circumstances.
Moulton 1994). Island effects may be weak or strong. The area required by a MVP is termed the mini-
The implications of insularity vary, depending on mum viable area, and may be estimated from
such factors as the type(s) of organism involved, the knowledge of home range size, although this
type of landscapes involved, the nature of the approach assumes no population exchange with
environmental dynamics, the biogeographical other isolates. Another approach to area (or habitat,
setting, and the nature of human use and involve- or isolation) requirements is to use incidence func-
ment in the system being fragmented. It is unfortu- tions, although the patterns revealed may be con-
nate that the term ‘island biogeography theory’ has founded by multicausality, and may fail to predict
become largely synonymous in a conservation changes that follow within fragmented landscapes.
setting with a limited conception of island ecology, Moreover, recent work suggests that area require-
stressing the inevitability of stochastically driven ments for species presence in fragments may vary
trends to equilibrium. Both pure and applied island significantly even within the same region.
biogeography are richer than this. This richness of Where geographically separated groups are
ideas and information needs to be understood and interconnected by patterns of extinction and recolo-
integrated into teaching for both pure and applied nization they constitute a population of popula-
purposes. Moreover, there is also a need for a tions, or a metapopulation. Here, the idea is that
renewed research effort within this area of conser- particular patches have their own internal popula-
vation biogeography, in the search for improved tion dynamics, but when they crash to local extinc-
scientific guidance. tion they are repopulated from another patch
within the metapopulation. Such a scenario alters
the projections of population viability considerably.
10.14 Summary
In many cases, however, habitat patches appear to
The island analogy can be extended to patches of describe a source–sink rather than a mutual sup-
habitats within continents. The conversion of more- port system. In such circumstances, a large patch
or-less wild habitats to other uses is fragmenting, acts as an effectively permanent population, with
isolating, and reducing ‘wild’ areas across much of smaller sink habitats around it being re-supplied
the globe. The implications of this insularization are from the source population. The sink habitats may
examined in this chapter from the scale of individ- have little relevance to overall persistence-time of
ual populations up to whole landscapes. the metapopulation.
The minimum viable population (MVP) is the A heated debate developed around the SLOSS
smallest number of individuals required to ensure ‘principles’, which represented an attempt to
long-term population persistence. We do not know answer the question of whether it was better to
how big MVPs should be. We do know that MVPs advocate Single Large or Several Small reserves of
vary from species to species, and that the effective the same overall area, based on the assumptions of
population size is generally smaller than the actual the EMIB. Unfortunately, island theory provides no
population size, i.e. not all individuals, or all adults, resolution to this debate. In practice, the optimal
are involved in breeding. Population loss may be reserve configuration for one type of organism,
due to stochastic demographic and/or genetic landscape, or scale of study system, may not be so
 S U M M A RY 289

for another. Physical environmental change in both require a more extensive approach to conservation
fragment and ‘matrix’ habitats has not generally than provided for by any reserve system on its own.
been given enough attention in the theoretical liter- These considerations demand a re-evaluation of
ature. Successional change and altered disturbance the place of island biogeography within conserva-
regimes may have long-term implications for tion biology. The equilibrium island theory turns
species persistence in altered landscapes. The evi- out to provide rather equivocal guidance, and
dence for ‘relaxation’, i.e. stochastically driven appears to have often been invoked as a basis for
decline in species number, is surprisingly thin. predictions of species loss where there is a logical
Species composition and richness are liable to disconnection between theory and prediction.
change in fragmented systems: species are indeed Some of the effects traditionally associated with
lost. However, in relation both to SLOSS and island approaches may actually be fairly weak in
metapopulation scenarios, it is apparent that much general, but other features of insularity may pro-
turnover may well be deterministic in nature and vide powerful and fairly immediate influences on
explicable in relation to specific ecological processes species persistence and system change. The role of
such as succession or meso-predator release. hierarchical ecological relationships, e.g. between
Various lines of data suggest strong structure to the plants and their dispersers, between vegetation
disassembly as well as to the assembly of systems. change and altering habitat space for butterflies,
For instance, many systems and species are between understorey insectivores and their
strongly nested in their distributions, and this can, predators, and so forth, may be crucial to the
it seems, be produced in either colonization- or consequences of fragmentation. Such successional
extinction-dominated systems. and food-web controls may be the most powerful
It appears possible to predict the types of species forms of expression of islandness in the altered
that will be in greatest difficulty in fragmented landscapes, and they typically involve interactions
landscapes, and it is these species that require and exchanges between the fragments and the
primacy in relation to reserve configuration, popu- surrounding matrix. The attention being paid to the
lation monitoring, and management planning. The effects of the matrix on persistence in remnants, and
consequences of fragmentation can be severe, but to the use of matrix habitats by native species,
they are poorly predicted by classical island signals an on-going switch in emphasis from
theories, which assume stochastic demographic island thinking to what some term ‘countryside
effects to dominate in a system of fragments search- biogeography’, i.e. to the importance of incorporat-
ing for a new equilibrium in a sea of altered habi- ing conservation concerns in managing whole land-
tats. The nature of the connectivity of habitat scapes, not just isolated reserves.
patches may be crucial. In some cases, corridors A line must be drawn somewhere in any text, and
and short dispersal distances may be beneficial, in this chapter discussion stops short on alternative
whereas in other circumstances isolated reserves approaches to conservation problems, and on
may actually provide the best chance of survival for themes such as the implication of large-scale cli-
a threatened species. Some species thrive in edge mate change under global warming scenarios.
habitats, others are disadvantaged by increased Neither is any attempt made to draw out guiding
edge effects. The ‘filtering’ properties of the matrix principles for conservation from island studies. The
can be crucial and require careful assessment in grand hopes for simple unifying principles have
conservation planning. Landscape effects may be proven to be as illusory as the end of the rainbow.
complex, with the same linear feature acting as It is nonetheless important to appreciate how these
corridor, barrier, and lethal hazard for different ideas have influenced conservation theory, just as it
organisms. Moreover, some species are mobile, sea- is important to tackle the ecological complexities of
sonally or aseasonally, and their protection may increasing insularity within real landscapes.