keep the highest diversity of butterfly species by
means of a number of small reserves, each targeted
to provide particular key habitats (Ewers and
Didham 2005). They should not, however, be too
small, lest they consist principally of ‘sink’ popula-
tions. However, trophic level and body size are not
the only considerations. We have also to think
about how species use their habitat space. For
instance, birds of prey may have large territories
incorporating both good and poor habitat.
Conservation of large birds of prey in Scotland,
such as the golden eagle (Aquila chrysaetos) and the
osprey (Pandion haliaetus), has required a range of
measures taken to discourage and prevent killing of
the birds within a mixed-use landscape. It has also
required intensive but highly localized efforts to
protect nesting sites from egg collectors. The
‘reserves’ required are tiny compared to the ranges
of the individuals.
Although, in general, most studies advocate
relatively large reserves at the expense of larger
numbers, in cases it may be better to have a set of
several small reserves. First, they may incorporate
more different habitats, i.e. they may capture beta
diversity across a landscape. Secondly, competition
may, in theory, lead to the exclusion of species of
similar niches from any given reserve and so it may
be good to have several reserves so that different
sets of species may ‘win’ in different reserves.
Thirdly, there is an epidemiological risk inherent in
having ‘all your eggs in one basket’, and, fourthly,
particular species may rely upon small islands. For
instance, small estuarine islands in the Florida Keys
area appear to provide important breeding sites for
several species of waterbirds (Erwin et al. 1995).
In organizing our thinking about the implications
of the SLOSS debate, we would do well to consider
two aspects of spatial scale: study system grain and
extent (cf. Whittaker et al. 2001). Grainrefers to the
size of the sampling unit in spatial analyses. In this
context, grain size is variable: we refer in fact to the
range of areas of the habitat islands. Log–log plots
of species richness and area (e.g. Fig. 4.2) often
indicate simple linear trends of converging richness
between fragmented and unfragmented areas as
sample area increases. However, recent theoretical
discussion has reopened the question of whether
this is the right approach to fitting trend lines to
such data (e.g. Lomolino 2000c; Tjørve 2003; Turner
and Tjørve 2005), and in particular whether more
complex forms of relationship should be tested for
at the ‘small island’ end of the relationship.
Empirically, it appears that superior model fits can
be derived for many ‘island’ data sets using a form
of break-point regression, consistent with the exis-
tence of a so-called small-island effect. In practice,
this means that there is no systematic increase in
richness until a critical area threshold is crossed,
producing a two-phase relationship shifting from
flat to increasing (Fig. 4.6; Lomolino and Weiser
2001; Triantis et al. 2006). The implication of this
finding is that the slope and intercept of ISARS fit-
ted using log–log models may be more sensitive to
the range in area of isolates sampled than generally
realized. This in turn may impact on our assess-
ment of the relative richness gain from having
larger-sized reserves: although it should again be
noted that simply being able to predict the richness
of each isolate does not reveal the overall richness
of the system.
Turning to the study system extent, if we imagine
two networks of a dozen habitat islands, each net-
work identical in terms of the starting richness of
each island, the conservation implications of increas-
ing distance separating each of the habitat islands
work in perhaps two key ways. First, the further
apart they are, the lower the probability of their
species populations mutually reinforcing one
another, thereby preventing extinction from the net-
work. Second, the further apart they are, the more
likely they are to sample different habitats, and if
they are really far apart they may even sample dif-
ferent biomes or different biogeographical source
pools. Thus a widely scattered system of reserves
may capture more differentiation diversity (sensu
Whittaker 1977): but whether it can retain it in the
long run may depend on the extent to which each
reserve can sustain its initial species complement in
the long term. Hence, the answer to whether it is bet-
ter to have one large or several small reserves may be
sensitive to the grain and extent of the system. Of
course, the SLOSS ‘principles’ were advocated most
strongly for circumstances in which the biogeogra-
phy of an area is poorly known. When working at aTHE ‘SINGLE LARGE OF SEVERAL SMALL’ (SLOSS) DEBATE 267