Prioritizing the World's Islands for vertebrate-eradication programmes

Abstract

In the last 400 years, more species have become extinct on small islands than on continents. Yet, scant attention has hitherto been paid to prioritizing island restorations. Nevertheless, considerable conservation effort is now devoted to removing a major cause of these extinctions – invasive alien vertebrates. Because modern techniques allow the clearance of invasive vertebrates from quite large islands (up to 1000 km2), many islands are candidates for restoration. A robust strategy for allocating available funds is urgently needed. It requires, for each candidate island, an objective estimation of conservation gain and a method for predicting its financial cost. Our earlier work showed that a good first-pass estimate of vertebrate eradication costs can be made using just island area and target species. Costs increase with island area, while rodents are more expensive per unit area than ungulates. Here, we develop a method for assessing the conservation benefit of a proposed eradication and apply the method to threatened birds, but not other taxa. The method, combining information on how threatened a species is, on the impact of alien vertebrates on that species and on the islands on which the species occurs, allows us to present a means of determining which islands yield the greatest conservation benefit per unit of expenditure on vertebrate eradication. In general, although greater overall benefit would accrue to birds from eradication of invasive vertebrates on larger islands, benefit per unit of expenditure is the highest on relatively small islands, and we identify those that should be priority targets for future eradications. Crucially, this quantitative assessment provides considerable efficiency gains over more opportunistic targeting of islands. The method could be adapted to prioritize islands on a regional or national basis, or with different conservation gains in mind.

Full text

Costing eradications of alien mammals from islands
T. L. F. Martins
1,2
, M. de L. Brooke
3
, G. M. Hilton
2
, S. Farnsworth
3
, J. Gould
3
& D. J. Pain
2
1 Centre for Ecology and Conservation, University of Exeter in Cornwall, Tremough, Penryn, UK
2 Royal Society for the Protection of Birds, Sandy, Bedfordshire, UK
3 Department of Zoology, University of Cambridge, Cambridge, UK
Keywords
rats; cats; goats; restoration.
Correspondence
M. de L. Brooke, Department of Zoology,
University of Cambridge, Downing Street,
Cambridge CB2 3EJ, UK.
Tel: 01223 336610;
Fax: 01223 336676
Email: m.brooke@zoo.cam.ac.uk
Received 15 November 2005; accepted
2 June 2006
doi:10.1111/j.1469-1795.2006.00058.x
Abstract
The ability to estimate costs of alien species eradications is essential for a rigorous
assessment of priorities for island restoration. Using a global data file from
41 islands, mostly gleaned from the ‘grey’ literature, we show that the cost of
vertebrate eradications can be satisfactorily predicted if island area and species to be
eradicated are known. About 72% of the variation in cost can be explained by island
area, whereas, for a given area, rodent eradications are 1.7–3.0 times more expensive
than ungulate eradications. Costs per hectare decrease with island size. Restricting
the analysis to roughly half the data set, the relatively homogeneous half concerned
with New Zealand islands, we identify two further influences on cost: date of
eradication and distance to the main airport (an indicator of remoteness). For a
given area, costs have declined over time but increase with island remoteness. This
information therefore provides conservation planners with a robust, if preliminary,
estimate of the cost of any proposed eradication programme.
Introduction
Despite the expected economic benefits of biodiversity con-
servation (Balmford et al., 2002), current conservation
resources fall well short of those needed to prevent major
extinctions (Balmford et al., 2003). Estimates suggest that
effective conservation outwith reserves across the world
might cost US$ 290 billion year
1
(1996 prices) whereas the
establishment and maintenance of an ecologically represen-
tative global network of protected areas would cost
US$ 27.5 billion year
1
, as compared with a current expen-
diture on reserves of US$ 6 billion year
1
(James, Gaston &
Balmford, 1999). In the face of this shortfall, an important
strand of conservation biology has investigated means of
optimizing the selection of protected areas (review in Cabeza
& Moilanen, 2001). However, no similar attention has
hitherto been paid to the issue of prioritizing island restora-
tions even though, in the last 400 years, more species have
become extinct on small islands than on continents (Manne,
Brooks & Pimm, 1999). Nevertheless, a significant propor-
tion of conservation effort is now devoted to controlling or
removing the major cause of these extinctions from islands
(Johnson & Stattersfield, 1990): invasive alien vertebrate
species. This effort has been catalysed inter alia by the
development of anti-coagulant toxins and effective bait
delivery systems, which now allow islands of up to c. 100
and 300 km
2
to be cleared of rats Rattus spp. and cats Felis
catus, respectively (Cooper et al., 1995; NZ-DOC, 2003;
Nogales et al., 2004), which are among the most damaging
and widespread of alien vertebrate taxa. Such advances
mean that a robust strategy for allocating the available
funds is needed because a very large number of islands are
apparently urgent candidates for restoration. For example,
there are several hundred islands across the world where
globally threatened bird species occur alongside harmful
alien vertebrates (own data).
The prioritization of invasive alien species eradications
on islands requires, for each candidate island, a system for
objective estimation of the conservation gain and an intern-
ally consistent method of predicting its financial cost. Using
a global data file on vertebrate eradications, we address the
latter issue. We ask which variables, among a number of
plausible candidates such as island area, isolation, topogra-
phy, project date and taxa targeted, actually influence the
cost of an eradication project. We hypothesize that larger,
more isolated, more rugged islands will be more costly, that
costs per unit area may decline over time as efficiency
improves, and that smaller species such as rats may be more
costly to eradicate than larger species such as goats. The
results of this analysis therefore allow the first-pass estima-
tion of the likely costs of projects still in the planning stage.
Methods
Our global data set comprises information on 41 invasive
vertebrate eradication projects. Twenty of these (49%) were
carried out on New Zealand offshore islands, nine (22%)
either in the UK (1) or in the UK’s Overseas Territories (8),
four (10%) in the Seychelles, three (7%) on Australia’s
Animal Conservation 9(2006) 439–444 c2006 The Authors. Journal compilation c2006 The Zoological Society of London 439
Animal Conservation. Print ISSN 1367-9430

offshore islands, two (5%) in Mauritius and one each (2%)
on USA islands, Indonesian islands and Antigua. Because
New Zealand provided about half the data, we analysed
both the entire data set and, separately, the New Zealand
data set to investigate whether the latter yielded insights
obscured by between-country noise.
The data comprise (1) information on the eradication
itself, namely species eradicated (rat or others), year of
eradication, cost in US$ and status of the eradication
(successful or unsuccessful two years after the project), and
(2) island topographical and geographical data, specifically
area (km
2
), distance to the nearest main airport (km),
maximum altitude (m) and ‘ruggedness’ (see below).
On the grounds that goods and personnel are readily
moved between international airports but that such move-
ment often becomes more difficult and expensive when by
sea or when to local airports, we took the distance of each
island from the nearest main airport as a measure of
isolation, termed ‘remoteness’. This distance was measured
from The Times Atlas of the World (1999), with main
airports being those marked in the atlas by an airport
symbol enclosed in a circle.
Insofar as the steepness of the terrain may influence costs,
our ‘ruggedness’ variable is intended to capture this, inde-
pendent of both area and max altitude (which we also
include as separate variables). Because altitude tends to
increase with area, we take the standardized residuals from
the regression of log
10
max altitude on log
10
area (log
10
alti-
tude=1.67+0.434 log
10
area) as our ruggedness variable.
Costs were converted from the local currency to US$ at
prevailing exchange rates and adjusted to 2003 prices. A
summary of data used is given in Table 1.
The 41 projects targeted species in various combinations
(Table 1): there were 29 rodent, four ungulate, two cat,
one rabbit Oryctolagus cuniculus, one brushtail possum
Trichosurus vulpecula and four (rodent+other) eradica-
tions. Because we were interested in the impact, if any,
of species targeted on cost, it was necessary to categorize
these projects before analysis. A priori, we have three
main taxon categories: rodents, cats and ungulates. The
methods deployed to eradicate these three groups are
markedly different (although, even within taxon groups,
there is substantial variation in methods in a database
as wide-ranging as ours). We therefore aim to estimate
costs separately for each of these groups, regardless of
whether they are statistically distinguishable. This then
leaves the question of how to treat the single eradications
of rabbit and possum, and those islands from which rodents
plus other species were removed. Kapiti Island was cleared
of brushtail possums using trapping, shooting and hunting
with dogs (Brown & Sherley, 2002), a method that is broadly
similar to that used in most cat eradications. We there-
fore include Kapiti with the cat eradications. Round Island
was cleared of rabbits using a brodifacoum-baiting opera-
tion, similar to most rat eradications, and is therefore
pooled with those.
Considering the four islands from which rodents plus
other species were removed, the removal of rats and cats on
Tuhua was a carefully planned combined operation, invol-
ving secondary poisoning of cats which ate and were
themselves then poisoned by rats that had already ingested
brodifacoum (Nogales et al., 2004). On Pitcairn, secondary
poison was a major component of the eradication, coupled
with some trapping and hunting (Nogales et al., 2004). On
Inner Chetwode and Stanley, the eradication of wekas
Gallirallus australis and rabbits, respectively, was achieved
via their consumption of brodifacoum, intended for rats. In
the light of these details, we do not consider that these
operations provide sufficient data to examine the very
important question of the extent to which the costs of
simultaneous eradications of more than one species are
additive. Instead, we treat all four eradications as rodent
eradications in the analysis. None of them generates notable
outliers from the overall rodent cost-estimation function.
Hence, in the worldwide analysis, we have four ungulate
eradications, three ‘cat’ eradications (i.e. including the
eradication of possums from Kapiti Island) and 34 ‘rodent’
eradications (including the eradication of rabbits from
Round Island). For New Zealand, we have 16 rodent
eradications (including Tuhua, Stanley and Inner Chet-
wode), three ungulate eradications and one possum eradica-
tion (Kapiti).
General linear models (GLMs, Minitab 13.1) were used
to examine the variables that predict costs of eradication.
Full models containing all potential explanatory variables
were developed, with stepwise deletion of least significant
variables, until a minimum adequate model was obtained in
which all variables were significant (Po0.05). Because of the
small number of observations (n=41 islands) relative to
explanatory variables (n=7), we did not attempt to test for
non-linear effects or interactions among variables.
In GLMs, success was treated as a binary categorical
variable and taxon as a three-level categorical variable (see
above). The remaining explanatory variables were modelled
as covariates. The GLM is robust to departures from
normality of explanatory variables, but nevertheless we
normalized the heavily right-skewed distributions of island
area and remoteness by log
10
transformation. Year of
eradication,ruggedness (see above) and max altitude were
untransformed. The response variable, cost of eradication,
was log
10
transformed to remove a strong right-skew. All
statistical tests were two-tailed.
Results
Our a priori expectation was that island area would have a
very strong influence on costs. This expectation was met
(Fig. 1), and the linear regression of cost on area gives log
10
cost of eradication (US$)= 4.27 (SE 0.069)+0.770
(SE 0.076) log
10
island area (km
2
)[F
1,39
=102, Po0.0001,
R
2
(adjusted)=71.7%].
Because slope of the regression is significantly less than 1,
costs per unit area decline with increasing island area.
Although such a cost–area relationship is intuitively ob-
vious, its influence is so overwhelming that clarifying its
slope is of greater value in providing an accurate estimate of
Animal Conservation 9(2006) 439–444 c2006 The Authors. Journal compilation c2006 The Zoological Society of London440
Costing island eradications T. L. F. Martins et al.

Table 1 The 41 eradication projects analysed in this study
Island Country Year Area (km
2
) Airport (km)
Successful
(1= yes,
0=no) Cost Rodent Cat Ungulate Other Source
Ascension UK Overseas Territory 2003 88.00 1081 1 815 661 0 1 0 0 Royal Society for the Protection of Birds
Bird Seychelles 1996 1.01 270 1 5169 1 0 0 0 www.islandconservation.org/islanderad.html
Bottom Falklands 2001 0.08 1375 1 3201 1 0 0 0 R. Ingham (pers. comm.)
Breaksea New Zealand 1990 1.70 531 1 48 796 1 0 0 0 Pestlink, NZ Department of Conservation database
Campbell New Zealand 2003 113.00 1050 1 1 249726 1 0 0 0 Pestlink, NZ Department of Conservation database
Chetwode
(inner and outer)
New Zealand 1996 2.78 75 1 43 778 1 0 0 Weka Pestlink, NZ Department of Conservation database
Curieuse Seychelles 2000 3.00 45 1 67 290 1 0 0 0 J. Millett (pers. comm.)
Cuvier New Zealand 1993 1.70 105 1 16 968 1 0 0 0 Pestlink, NZ Department of Conservation database
Denis Seychelles 2000 1.40 80 1 56 994 1 0 0 0 J. Millett (pers. comm.)
Double Falklands 2001 0.09 1375 1 370 1 0 0 0 R. Ingham (pers. comm.)
Double Islands
(Larger Island)
New Zealand 1989 0.19 96 1 3271 1 0 0 0 Pestlink, NZ Department of Conservation database
Double Islands
(Smaller Island)
New Zealand 1989 0.08 96 1 1919 1 0 0 0 Pestlink, NZ Department of Conservation database
Ducie Pitcairn Islands 1998 0.60 2700 1 32 191 1 0 0 0 Wildlife Management International
Enderby New Zealand 1993 7.10 910 1 10 698 0 0 Cattle 0 Pestlink, NZ Department of Conservation database
Flat Mauritius 1998 2.00 60 1 64 381 1 0 0 0 J. Hartley (pers. comm.)
Fregate Seychelles 2000 2.20 45 1 61 916 1 0 0 0 J. Millett (pers. comm.)
Great Barrier New Zealand 1987 32.30 93 0 32975 0 0 Goat 0 Pestlink, NZ Department of Conservation database
Green Island Antigua 2001 0.43 40 1 16 115 1 0 0 0 www.islandconservation.org/islande rad.html
Hawea New Zealand 1986 0.09 531 1 36 101 1 0 0 0 Pestlink, NZ Department of Conservation database
Kapiti New Zealand 1986 19.70 50 1 149 498 0 0 0 Possum Pestlink, NZ Department of Conservation database
Korapuki New Zealand 1987 0.18 96 1 3858 1 0 0 0 Pestlink, NZ Department of Conservation database
Lord Howe Australia 2001 14.60 700 1 48 125 0 0 Goat 0 Parkes, Macdonald & Leaman (2002), Anon. (2003)
MacQuarie Australia 2000 122.50 1200 1 2 356 350 0 1 0 0 G. Copson (pers. comm.)
Mokohinau New Zealand 1991 1.00 111 1 21621 1 0 0 0 Pestlink, NZ Department of Conservation database
Mou Waho New Zealand 1996 1.40 312 1 8243 1 0 0 0 Pestlink, NZ Department of Conservation database
Oeno Pitcairn Islands 1998 0.60 2400 1 32 191 1 0 0 0 Wildlife Management International
Otata New Zealand 1991 0.22 36 1 8208 1 0 0 0 Pestlink, NZ Department of Conservation database
Outer Falklands 2001 0.20 1375 1 895 1 0 0 0 R. Ingham (pers. comm.)
Palmyra USA 2001 2.29 1300 0 111 007 1 0 0 0 B. Flint (pers. comm.)
Pitcairn Pitcairn Islands 1998 5.00 2500 0 225 334 1 1 0 0 Wildlife Management International
Ramsey UK 2000 2.53 150 1 28 972 1 0 0 0 I. Bullock (pers. comm.)
Raoul New Zealand 1986 29.38 1160 1 551 470 0 0 Goat 0 Pestlink, NZ Department of Conservation database
Red Mercury New Zealand 1992 2.25 110 1 24 126 1 0 0 0 Pestlink, NZ Department of Conservation database
Round Mauritius 1986 1.50 40 1 48 286 0 0 0 Rabbit J. Hartley (pers. comm.)
Rurima New Zealand 1984 0.08 348 1 7366 1 0 0 0 Pestlink, NZ Department of Conservation database
Sandy Lacepede Australia 1986 4.49 1100 1 51 653 1 0 0 0 www.islandconservation.org/islanderad.html
Animal Conservation 9(2006) 439–444 c2006 The Authors. Journal compilation c2006 The Zoological Society of London 441
Costing island eradicationsT. L. F. Martins et al.

costs than is determining the role of other, secondary
variables.
We subsequently investigated the effect of other second-
ary variables on the estimation of costs. A full GLM
including all variables was reduced to give a final minimum
adequate model, which contained log
10
island area
and taxon as significant predictors of cost (Table 2). Para-
meter estimates for the taxon effect indicate that eradication
of rodents might be costlier per unit area than ungulates.
The difference is large: according to this model, rodent
eradications are estimated to be c. 1.7 times more expensive
per unit area than ungulate eradications. Including the
taxon variable in the model has a minor influence on the
estimate of the slope of the area effect.
Success was the first variable to be dropped from the
model; there was no evidence of a difference in costs between
successful and unsuccessful operations. However, our data
set comprises only three unsuccessful eradications (Pitcairn,
Palmyra, Great Barrier), and hence this is a weak test. None
of the remaining variables that were dropped from the
model approached significance.
We developed a similar model using only data from
eradications conducted in New Zealand. The success
Table 1 Continued
Island Country Year Area (km
2
) Airport (km)
Successful
(1= yes,
0=no) Cost Rodent Cat Ungulate Other Source
Sangalaki Indonesia 2003 0.14 500 1 2800 1 0 0 0 www.islandconservation.org/islanderad.html
Stanley New Zealand 1992 1.00 99 1 17064 1 0 0 Rabbit Pestlink, NZ Department of Conservation database
Tawhitinui New Zealand 1983 0.23 51 1 4225 1 0 0 0 Pestlink, NZ Department of Conservation database
Top Falklands 2001 0.12 1375 1 2974 1 0 0 0 R. Ingham (pers. comm.)
Tuhua New Zealand 2000 12.80 129 1 67 543 1 1 0 0 Pestlink, NZ Department of Conservation database
For each island, the year of the project, area, distance to nearest international airport and cost (expressed as US dollars adjusted to year 2003) are shown. The table also indicates whether or not
the eradication was successful and whether it did (code 1) or did not (code 0) target various categories of vertebrate. All rodents were Rattus spp. and all cats were Felis catus.
100
1000
10 000
100 000
1 000 000
10 000 000
0.01 0.10 1.00 10.00 100.00 1000.00
Island area (km )
Cost of eradication (USD)
rodent
ungulate
cat
possum
rabbit
rodent + other
Figure 1 Cost of island eradications as a function of island area, for
different taxa eradicated. The full data set is plotted.
Table 2 Significant variables in the minimum adequate model of
eradication costs using the full data set
Response variable F(d.f.) PEstimate (SE)
Constant o0.001 4.10 (0.14)
Island area 79.7 (1,37) o0.001 0.85 (0.067)
Taxon 3.77 (2,37) 0.032 –
Ungulate 0.00 (0.00)
Rodent 0.22 (0.15)
Cat 0.20 (0.19)
Deleted variables
Success 0.13 (1,32) 0.72
Remoteness 1.05 (1,33) 0.31
Maximum altitude 1.83 (1,34) 0.19
Ruggedness 0.77 (1,35) 0.39
Year 1.03 (1,36) 0.32
Non-significant variables are listed at the bottom, in the order in which
they were deleted from the full model.
Animal Conservation 9(2006) 439–444 c2006 The Authors. Journal compilation c2006 The Zoological Society of London442
Costing island eradications T. L. F. Martins et al.

variable was not considered, because there was only one
New Zealand failure. We combined the single possum
eradication (see Methods) with the rodent eradications to
create a two-level taxon variable (rat/possum vs. ungulate).
A model with area, taxon, year of eradication and
remoteness as explanatory variables gave a good fit to the
data (Table 3). As in the global model, rodent/possum
eradications were significantly more expensive per unit area
than ungulate eradications, in this case by a factor of three.
In addition, in the New Zealand Model, costs were lower in
more recent eradications (Fig. 2), and higher for more
remote islands (Fig. 3).
Thus, for a hypothetical 10 km
2
island situated 100 km
from an airport, costs would decrease from US$ 251 500 if
rodents were eradicated in 1983 (the earliest date for which
we have data), to US$ 88 600 if the operation were done in
1993, to only US$ 31 200 if it were done in 2003 (the latest
date for which we have data).
Similarly, for a hypothetical 10 km
2
island from which
rodents/possums were eradicated in 2000, the predicted cost
would increase from US$ 9800 if it were 10 km from an
airport, to 42 700 if 100 km from an airport to 185 500 if
1000 km from an airport.
Discussion
Our results indicate that island area is the primary determi-
nant of the cost of an eradication (Fig. 1). Seventy-two per
cent of the variation in cost of an eradication can be
explained by area alone. Detecting this strong effect was
made easier by the fact that the areas of the islands in our
data set span more than four orders of magnitude. Thus,
decision makers considering potential eradication pro-
grammes need only know island area, distance from airport
(at least in the New Zealand region) and species to be
eradicated to make internally consistent and robust first-
pass estimates of the likely cost. Of especial consequence is
whether the programme will target rodents or ungulates, the
former being 1.7–3.0 times more expensive for a given island
area. This reflects substantial differences in the methods
typically used to conduct the work (Courchamp, Chapuis &
Pascal, 2003). Although other variables failed to enter our
model, it seems almost inevitable that these variables do
exert some influence on costs. In particular, more rugged
islands are likely to be more costly than flat islands.
Although the explanatory power of the regression of cost
on area is very high, the confidence intervals around the cost
of a particular eradication, especially on a small island,
remain rather large as a proportion of predicted cost. Clearly,
local factors that are not captured in this generic analysis may
have an important influence on the costs of a given eradica-
tion. However, the absolute precision of cost prediction is not
the main issue here. These models provide a means by which
several or many potential eradications can be compared in a
consistent manner at the pre-planning stage, particularly
Table 3 Significant variables in the minimum adequate model of
eradication costs using only New Zealand data
Response variable F(d.f.) PEstimate (SE)
Constant 0.032 92.7 (39.3)
Island area 49.4 (1,15) o0.001 0.82 (0.12)
Remoteness 12.1 (1,15) 0.003 0.64 (0.18)
Year of eradication 5.23 (1,15) 0.037 0.045 (0.20)
Taxon
a
11.9 (1,15) 0.004
Ungulate 0.00 (0.00)
Rodent 0.51 (0.15)
Deleted variables
Ruggedness 0.52 (1,12) 0.49
Maximum altitude 0.23 (1,13) 0.64
a
For the purposes of this analysis, we used a two-level taxon variable:
rodent/possum and ungulate.
Non-significant variables are listed at the bottom, in the order in which
they were deleted from the full model.
−2
−1
0
1
2
3
1 1.5 2 2.5 3 3.5
Remoteness (log10 distance to airport)
Residual cost of eradication
Figure 3 Relationship between remoteness of island and costs of
restoring New Zealand islands. Y-values are standardized residuals
from a general linear model, with log
10
cost of eradication as response
variable and log
10
island area, year of eradication and taxon eradicated
as explanatory variables [i.e. the minimum adequate model for New
Zealand eradications (see Table 3) with the remoteness variable
removed].
−3
−2
−1
0
1
2
3
1980 1985 1990 1995 2000 2005
Year of eradication
Residual cost of eradication
Figure 2 Relationship between year of eradication and costs of
restoring New Zealand islands. Y-values are standardized residuals
from a general linear model, with log
10
cost of eradication as response
variable and log
10
island area, taxon eradicated and remoteness of
island (log
10
distance from nearest airport) as explanatory variables
[i.e. the minimum adequate model for New Zealand eradications (see
Table 3) with the year variable removed].
Animal Conservation 9(2006) 439–444 c2006 The Authors. Journal compilation c2006 The Zoological Society of London 443
Costing island eradicationsT. L. F. Martins et al.

when combined with an internally consistent means of
estimating conservation benefits (Brooke et al., in prep.).
The regression coefficient for island area (Fig. 1 and
results) is less than unity, implying that eradication pro-
grammes on larger islands cost less per hectare than those on
smaller islands. Such a relation was demonstrated by Towns
& Broome (2003) for New Zealand eradications. These
authors also considered that there had been efficiency gains,
reducing per hectare costs over time, but they did not
present detailed multivariate statistics to support this claim.
Our analysis now detects such gains. When increased effi-
ciency is coupled with the lower per hectare cost of larger
islands, the number of candidate islands clearly expands for
a given budget. But when assessing the case for eradications
on candidate islands of different size, it is crucial to know
the slope of the regression of cost on island area, which our
analysis now provides.
The lower per hectare cost of programmes on larger
islands should not automatically be considered an argument
for targeting eradication expenditure towards larger islands.
Although, on average, the populations of threatened species
will be larger and potentially more viable on larger islands,
there may be counter-arguments in favour of targeting
several small islands. For example, several small islands
might harbour populations of several different endangered
species, and their smaller size could facilitate quarantine
measures against accidental re-introductions of aliens.
Few of the eradications used in our calculations were
undertaken in developing countries. Even those that were
undertaken in developing countries were carried out by
visiting experts from the developed world. Thus, in contrast
to protecting reserves in developing countries, which may be
cheaper because land and labour costs are low (Balmford
et al., 2003), there is at present no case to be made that the
costs of eradications are significantly affected by the devel-
opment status of countries (although this situation may
certainly change in the future).
The results presented here offer a much-needed tool for
comparing the costs of future eradications among sets of
candidate islands and, in due course, for the assessment of
global priorities for restorative island conservation.
Acknowledgements
Data from New Zealand were kindly supplied (two years
ahead of it being accessible on the web) by Dr Wendy Evans,
project manager for Pestlink (web-based animal pest data-
base) from the Northern Regional Office of the Department
of Conservation, New Zealand. Thanks also to Dr Rod
Hitchmough from the Biodiversity Unit of the Department
of Conservation, New Zealand, who played an important
role in establishing contact between Dr Evans and ourselves.
Ann Amer from HSA Systems Ltd provided altitude data
for some New Zealand islands (www.hsa.co.nz). David
Bryant, Matthew Evans and David Gibbons read and
commented on various versions of this paper. T.L.F.M.
was part-funded by the European Social Fund.
References
Anon. (2003). Feral and pest animal control. Environment
News of Environment Unit, Lord Howe Island Board,
summer 2003.
Balmford, A., Bruner, A., Cooper, P., Costanza, R., Farber,
S., Green, R.E., Jenkins, M., Jefferiss, P., Jessamy, V.,
Madden, J., Munro, K., Myers, N., Naeem, S., Paavola, J.,
Rayment, M., Rosendo, S., Roughgarden, J., Trumper, K.
& Turner, R.K. (2002). Ecology – economic reasons for
conserving wild nature. Science 297, 950–953.
Balmford, A., Gaston, K.J., Blyth, S., James, A. & Kapos, V.
(2003). Global variation in terrestrial conservation costs,
conservation benefits and unmet conservation needs. Proc.
Natl. Acad. Sci. USA 100, 1046–1050.
Brown, K.P. & Sherley, G.H. (2002). The eradication of
possums from Kapiti Island, New Zealand. In Turning the
tide: the eradication of invasive species: 46–52. Veitch, C.R.
& Clout, M.N. (Eds). Occasional paper of the IUCN
Species Survival Commission No. 27. Gland, Switzerland
and Cambridge, UK: IUCN.
Cabeza, M. & Moilanen, A. (2001). Design of reserve net-
works and the persistence of biodiversity. Trends Ecol.
Evol. 16, 242–248.
Cooper, J., Marais, A.v.N., Bloomer, J.P. & Bester, M.N.
(1995). A success story: breeding of burrowing petrels (Pro-
cellariidae) before and after the eradication of feral cats Felis
catus at subantarctic Marion Island. Mar. Orn. 23, 33–37.
Courchamp, F., Chapuis, J.-L. & Pascal, M. (2003). Mammal
invaders on islands: impact, control and control impact.
Biol. Rev. 78, 347–383.
James, A.N., Gaston, K.J. & Balmford, A. (1999). Balancing
the earth’s accounts. Nature 401, 323–324.
Johnson, T.H. & Stattersfield, A.J. (1990). A global review of
island endemic birds. Ibis 132, 167–180.
Manne, L.L., Brooks, T.M. & Pimm, S.L. (1999). Relative
risk of extinction of passerine birds on continents and
islands. Nature 399, 258–261.
Nogales, M., Martin, A., Tershy, B.R., Donlan, C.J., Veitch,
D., Puerta, N., Wood, B. & Alonso, J. (2004). A review of
feral cat eradication on islands. Conserv. Biol. 18, 310–319.
[NZ-DOC] New Zealand Department of Conservation (2003).
Campbell Island rat eradication. Available from http://
www.doc.govt.nz/Conservation/Offshore-Islands/
Campbell-Island-Rat-Eradication.asp (accessed June 2003).
Parkes, J.P., Macdonald, N. & Leaman, G. (2002). An
attempt to eradicate feral goats from Lord Howe Island. In
Turning the tide: the eradication of invasive species:
233–239. Veitch, C.R. & Clout, M.N. (Eds). Occasional
paper of the IUCN Species Survival Commission No. 27.
Gland, Switzerland and Cambridge, UK: IUCN.
The Times Atlas of the World (1999). Comprehensive 10th
edn. London, UK: Times Books.
Towns, D.R. & Broome, K.G. (2003). From small Maria to
massive Campbell: forty years of rat eradications from
New Zealand islands. N. Z. J. Zool. 30, 377–398.
Animal Conservation 9(2006) 439–444 c2006 The Authors. Journal compilation c2006 The Zoological Society of London444
Costing island eradications T. L. F. Martins et al.