© 2006 Nature Publishing Group
Prioritizing global conservation efforts
Kerrie A. Wilson1, Marissa F. McBride1, Michael Bode1& Hugh P. Possingham1
One of the most pressing issues facing the global conservation
community is how to distribute limited resources between
regions identified as priorities for biodiversity conservation1–3.
Approaches such as biodiversity hotspots4, endemic bird areas5
and ecoregions6are used by international organizations to priori-
tize conservation efforts globally7. Although identifying priority
regions is an important first step in solving this problem, it does
not indicate how limited resources should be allocated between
resources between regions identified as priorities for conserva-
tion—the ‘conservation resource allocation problem’. Stochastic
dynamic programming is used to find the optimal schedule of
resource allocation for small problems but is intractable for large
easy-to-use and easy-to-interpret heuristics that closely approxi-
mate the optimal solution. We also show the importance of both
correctly formulating the problem and using information on how
investment returns change through time. Our conservation
resource allocation approach can be applied at any spatial scale.
We demonstrate the approach with an example of optimal
resource allocation among five priority regions in Wallacea and
Sundaland, the transition zone between Asia and Australasia.
Conservation organizations allocate resources to areas that have
been identified as priorities for conservation investment3,7. These
priority regions are identified using information on relative bio-
diversity values, past or present threats to these values, and current
levels of protection9. Species richness, or endemic species richness, is
typically used to estimate the biodiversity value of a region10. The
relative cost of conservation in different regions is ignored in the
identification of priority regions despite evidence that its inclusion
improves the cost-effectiveness of conservation prioritization11–15.
Some international organizations rank these regions in terms of
their priority for funding, but the approaches used to derive these
rankings are not solutions of a properly formulated problem4,6,16. If
the objective is to maximize the total number of species conserved,
then this objective is unlikely to be achieved if regions are prioritized
only on the basis of species richness. This is because regions that are
highly threatened but marginally less species-rich may lose many
species before being considered for conservation investment. Like-
wise, if the relative cost of investing in different regions is not taken
into account, resources may be directed to expensive regions when
the same amount of resources might have conserved more species if
invested in regions with lower land-acquisition and management
costs. The efficient allocation of conservation resources will be
achieved only if the problem includes data on biodiversity, threat
and cost, and is rigorously formulated.
Allocation of conservation resources, like any problem in decision
theory, requires a broad goal, a specific objective, a set of constraints,
asetofpossible actionsthat formastrategy,andanunderstanding of
the system dynamics provided by equations that link the actions
and constraints to the objective1(see Methods). Here, the goal is to
maximize biodiversity conservation through the creation of reserves,
given ongoing habitat destruction and the constraint of a fixed
each year—depends on endemic species richness, forest conversion
rates (and the uncertainty associated with these rates), land cost and
initial conditions (area of land currently reserved, converted or
otherwise; see Table 1 and Supplementary Table). This decision
theoreticformulation provides anexplicit and transparent statement
resource allocation: biodiversity values, threats, costs, investment
returns and data uncertainty.
We find an optimal resource allocation schedule using stochastic
dynamic programming (SDP)8,17. The SDP algorithm finds the
optimal allocation decision each year given the current state of the
with more than a few regions is computationally intractable, so we
heuristics that we propose as approximations are ‘maximize short-
term gain’ and ‘minimize short-term loss’ (see Methods). We
compare the performance of these heuristics to priority setting
approaches based on simple rankings, using a case study from
To illustrate our conservation resource allocation approach, we
first compare the allocation of resources between two regions,
Table 1 | Biodiversity, threat and cost data for the five priority regions
Priority region Area
(km2) in 1997
(km2) in 2003
No. of endemic bird
ActualRank ActualRank ActualRank
Southern peninsular Malaysia
1The Ecology Centre, Schools of Integrative Biology and Physical Sciences, The University of Queensland, Brisbane, Queensland 4072, Australia.
Vol 440|16 March 2006|doi:10.1038/nature04366
© 2006 Nature Publishing Group
Borneo and Sumatra, using the parameters in Table 1. We find that
the optimal schedule is to allocate all resources to Sumatra for over a
decade. Once all species occurring in Sumatra are conserved, invest-
ment is scheduled for Borneo. For this initial case study, the heuristic
that minimizes short-term loss most closely approximates the
optimal solution (Fig. 1). However, if there is uncertainty regarding
our ability to invest in a region for the whole planning period, for
gains is likely to result in the greatest number of species conserved.
When we modify the problem to include a random probability
of investment ceasing, the optimal allocation schedule more
closely reflects the heuristic that maximizes short-term gain. The
heuristic that maximizes short-term gain allocates funding to both
regions simultaneously, in proportion to the marginal returns from
investment (Fig. 2).
each approach using different combinations of relative threat and
relative endemic species richness for two hypothetical regions. Both
heuristics perform well, but the heuristic that minimizes short-term
loss is most similar to the optimal SDPsolution and outperforms the
other heuristic for most parameter sets (Fig. 3). The heuristic that
maximizes short-term gain performs best when the threat levels of
the two regions are similar, and performs poorly when the regions
have very different threat levels but similar endemic species richness
(Fig. 3a). The heuristic that minimizes short-term loss performs
slightly worse than the optimal solution when both the relative level
of threatand the endemic species richness of the tworegions are very
similar (Fig. 3b). If the annual budget is increased, land parcels are
reserved at a faster rate. This mitigates forest conversion and,
consequently, the difference between the approaches in the number
of species conserved is reduced. When the relative cost of land
acquisition in each region is varied, the results are similar to those
in Fig. 3 once the axes are adjusted to a species gain per dollar basis.
We next evaluate the performance of the heuristic algorithms,
which we have shown to be close to optimal, for five priority regions
from Southeast Asia. We compare the results of the algorithms with
rankings based on endemic species richness, threat and cost (Table 1
and Fig. 4). This case study is used to illustrate our resource
allocation approach, which can be applied to any number of priority
regions for which a schedule for resource allocation is required. The
approach can also be applied at any spatial scale, from global level
problems to those at a local level. Our decision theory approach
recommends initially investingall resourcesin Sulawesiand noother
place until all the species occurring in Sulawesi are conserved. Only
after this should investment proceed in Sumatra, Borneo and Java.
After investment in Sulawesi ceases, the heuristic that maximizes
short-term gain recommends roughly equal investment in Sumatra,
Borneo and Java, and investment is scheduled last for Malaysia
By contrast, ranking the five regions based on individual criteria
does not provide an obvious schedule for resource allocation
(Table 1). The three ranking criteria suggest that Sulawesi is the
highest priority for conservation investment and Malaysia is the
lowest. On the basis of only endemic bird richness, the rankings
would recommend that investment should occur first in Sulawesi,
second in Borneo, third in Java, fourth in Sumatra, and last in
Malaysia. Although these simple rankings are not widely different
from the results of our decision theory approach, there are discrep-
ancies, which would occur more frequently for problems involving
of funds to allocate between the regions (nor whether funds should
be allocated totally to a particular region or distributed between
be directed towards Sulawesi until its species are conserved or,
alternatively, that investmentsshould be inproportion to the relative
number of endemic species occurring in these regions. It is also not
clear how these criteria should be combined to provide an allocation
schedule that will maximize the protection of biodiversity. For
example, if priority is determined only by endemic species richness,
ignoring cost and threat, then Borneo’s priority is overestimated.
Similar confusion can arise if we prioritize only on threat or cost.
We have formulated the conservation resource allocation problem
identifying management actions, acknowledging constraints and
incorporating uncertainty. Our problem formulation has five main
within the priority regions and, consequently, the particular parcels
where resources should be allocated are not identified. Second, our
economic model is very simple. Third, we have not accounted for
the temporally heterogeneous nature of land availability for reser-
biodiversity and assumed that numbers of endemic species reserved
Figure 1 | Proportion of endemic species reserved through time in Borneo
and Sumatra. The average proportion is calculated by four different
resource allocation approaches: SDP, maximizing short-term gain,
minimizing short-term loss, and random allocation between regions. The
annual budget is US$1 million and assumes no pre-existing reserves. The
minimize short-term loss heuristic most closely approximates the optimal
greatest number of endemic species reserved at early time steps.
Figure 2 | Proportion of the total area of Borneo and Sumatra reserved
through time. The average proportion of the total area reserved is
calculated by three different resource allocation approaches. a, SDP. b, The
heuristic that maximizes short-term gain. c, The heuristic that minimizes
short-term loss. The annual budget is US$1 million and assumes no
superior, approaches allocate sequentially between regions.
NATURE|Vol 440|16 March 2006
© 2006 Nature Publishing Group
follows a species–area relationship. Last, we have assumed that the
amount of resources invested in a region is directly proportional to
the probability of species persisting. These assumptions can be
relaxed within the framework that we present.
There is a need for computationally feasible and understandable
algorithms that can deliver near-optimal solutions for large con-
explore were developed to solve a properly formulated problem,
perform surprisingly well relative to the optimal SDP solution, and
are superior to simple ranking approaches. Minimizing short-term
loss most closely approximates the optimal allocation schedule and
maximizing short-term gain is close to optimal despite ignoring
threat, although it underperforms if threat levels are very different.
Under extreme uncertainty, maximizing short-term gain is the most
risk-averse approach as it provides a buffer against an uncertain
We recommend that conservation organizations maximize short-
termgain, unlesstheregionsofconcernhavesimilarendemic species
richness and very different levels of threat. In such circumstances,
resource allocation should minimize short-term loss. We argue that
conservation investments should be evaluated as any investment is
evaluated: that is, with a clearly defined objective and an assessment
of how well the returns from the investment meet this objective.
Responsible conservation organizations and international agencies
should consider embracing a decision theoretic approach when
scheduling the allocation of conservation resources.
The first step in formulating the conservation resource allocation problem is to
define a quantifiable objective. Our objective is to maximize the number of
endemic species remaining across all regions when habitat conversion ceases
because there is no unreserved or unconverted land (see Supplementary
Methods ‘Problem formulation’). We assume that, for each region, the number
of endemic species conserved per unit area is a monotonically decreasing
function ofthe area reserved. Therefore, ourconservation returns ineach region
area curve19–21(see Supplementary Methods (ii)). In principle, any relationship
between area and endemic species conserved could be used. The next step in
formulating the problem is determining what actions are possible, in this case
thus constrained bya fixed annual budget. Althoughwe recognize that funds for
conservation can be directed towards many kinds of activity (for example,
restoration programs, the purchase of forestry concessions and species recovery
programs), we focus on the acquisition of land for reservation. We assume that
each land parcel can be classified as reserved, available for reservation or
converted (anthropogenically altered and assumed no longer suitable habitat
for the endemic species of the region). Threat is modelled by assuming that a
constant proportion of available parcels in each region are converted each year.
To incorporate the uncertainty associated with parcel loss22–24(see Supplemen-
tary Methods (iii)), conversion is represented as a stochastic process with a
binomial distribution. We estimate the cost of reservation in each region using
statistical models12,25(see Supplementary Methods (iv)).
We use SDP and two myopic heuristics to determine how many parcels to
reserve in each region each year. We compare these results to a random
acquisition process. Once theoptimal solutionis obtained, we forward-simulate
the resulting acquisition schedule 10,000 times for each parameter set to
calculate the expected number of species conserved. The solutions found
using SDP are optimal in the face of uncertainty18,26–28. Owing to an exponen-
tially increasing state space, however, ‘the curse of dimensionality’ limits SDP to
problems with few regions. The maximize short-term gain heuristic selects
parcels for reservation that result in the greatest increase in the number of
endemic species conserved. This heuristic ignores threat and is myopic, con-
sidering only the short-term future when selecting the next parcel to reserve and
heuristic is also myopic: it selects parcels that will minimize the expected loss of
species from the system in the next time step (see Supplementary Methods (i)).
Received 27 September; accepted 24 October 2005.
1. Possingham, H. P., Andelman, S. J., Noon, B. R., Trombulak, S. & Pulliam, H. R.
in Research Priorities for Conservation Biology (eds Orians, G. & Soule ´, M.)
225– -244 (Island Press, Washington DC, 2001).
Pimm, S. L. et al. Can we defy nature’s end? Science 293, 2207– -2208 (2001).
Dalton, R. Biodiversity cash aimed at hotspots. Nature 406, 818 (2000).
Myers, N., Mittermeier, R. A., Mittermeier, C. G., da Fonseca, G. A. B. & Kent, J.
Biodiversity hotspots for conservation priorities. Nature 403, 853– -858 (2000).
Stattersfield, A. J., Crosby, M. J., Long, A. J. & Wege, D. C. Endemic Bird Areas
of the World: Priorities for Biodiversity Conservation (Birdlife International,
Cambridge, UK, 1998).
Olson, D. M. & Dinerstein, E. The global 200: priority ecoregions for global
conservation. Ann. MO Bot. Gard. 89, 199– -224 (2002).
Halpern, B. S. et al. Gaps and mismatches between global conservation
priorities and spending. Conserv. Biol. (in the press).
Bellman, R. Dynamic Programming (Princeton Univ. Press, Princeton, NJ, 1957).
Wilson, K. A. et al. Measuring and incorporating vulnerability into conservation
planning. Environ. Manag. 35, 527– -543 (2005).
10. Orme, C. D. L. et al. Global hotspots of species richness are not congruent with
endemism or threat. Nature 436, 1016– -1019 (2005).
11. Balmford, A., Gaston, K. J., Rodrigues, A. S. L. & James, A. Integrating costs of
conservation into international priority setting. Conserv. Biol. 14, 1– -9 (2000).
12. Moore, J., Balmford, A., Allnutt, T. & Burgess, N. Integrating costs into
conservation planning across Africa. Biol. Conserv. 117, 343– -350 (2004).
13. Frazee, S. R., Cowling, R. M., Pressey, R. L., Turpie, J. K. & Lindenberg, N.
Estimating the costs of conserving a biodiversity hotspot: a case-study of the
Cape Floristic Region, South Africa. Biol. Conserv. 112, 275– -290 (2003).
Figure3|Comparisonof conservationperformanceof heuristicalgorithms
to the optimal solution. Two hypothetical regions are compared that differ
in relative threat and number of endemic species. The performance of
each heuristic is the percentage of endemic species not reserved relative to
the optimum solution. a, Performance of the heuristic that maximizes
heuristic that maximizes short-term gain for most parameter sets.
Figure 4 | Proportion of five priority regions from Southeast Asia reserved
through time. The average proportion of the total area reserved is
determined using two resource allocation heuristics. a, Heuristic that
annual budget is US$1 million and assumes no pre-existing reserves. In
both cases resources are initially allocated completely to Sulawesi and
investment is directed to southern peninsular Malaysia only after
investment has occurred in Sumatra, Java/Bali and Borneo.
NATURE|Vol 440|16 March 2006
© 2006 Nature Publishing Group Download full-text
14. Possingham, H. & Wilson, K. Biodiversity—turning up the heat on hotspots.
Nature 436, 919– -920 (2005).
15. Odling-Smee, L. Conservation: dollars and sense. Nature 437, 614– -616 (2005).
16. Groves, C. et al. Designing a Geography of Hope. A Practitioner’s Guide for
Ecoregional Conservation Planning (The Nature Conservancy, Washington DC,
17. Clark, C. W. & Mangel, M. Dynamic State Variable Models in Ecology (Oxford
Univ. Press, New York, 2000).
18. Meir, E., Andelman, S. & Possingham, H. P. Does conservation planning matter
in a dynamic and uncertain world? Ecol. Lett. 7, 615– -622 (2004).
19. Rosenzweig, M. Species Diversity in Space and Time (Cambridge Univ. Press,
Cambridge, UK, 1995).
20. Brooks, T. M. et al. Habitat loss and extinction in the hotspots of biodiversity.
Conserv. Biol. 16, 909– -923 (2002).
21. MacArthur, R. H. & Wilson, E. O. The Theory of Island Biogeography (Princeton
Univ. Press, Princeton, NJ, 1967).
22. Achard, F. et al. Determination of deforestation rates of the world’s humid
tropical forests. Science 297, 999– -1002 (2002).
23. Linkie, M., Smith, R. J. & Leader-Williams, N. Mapping and predicting
deforestation patterns in the lowlands of Sumatra. Biodivers. Conserv. 13,
1809– -1818 (2004).
24. Curran, L. M. et al. Lowland forest loss in protected areas in Indonesian Borneo.
Science 303, 1000– -1003 (2004).
25. Balmford, A., Gaston, K. J., Blyth, S., James, A. & Kapos, V. Global variation in
terrestrial conservation costs, conservation benefits, and unmet conservation
needs. Proc. Natl Acad. Sci. USA 100, 1046– -1050 (2003).
26. Possingham, H., Day, J., Goldfinch, M. & Salzborn, F. in Decision Sciences: Tools
for Today, 12th Australian Operations Research Conference (eds Sutton, D.,
Cousins, E. & Pearce, C.) 536– -545 (ASOR, Univ. Adelaide, Adelaide, SA, 1993).
27. Costello, C. & Polasky, S. Dynamic reserve site selection. Res. Energy Econ. 26,
157– -174 (2004).
28. Drechsler, M. Probabilistic approaches to scheduling reserve selection. Biol.
Conserv. 122, 253– -262 (2005).
Supplementary Information is linked to the online version of the paper at
Acknowledgements We thank C. Elkin, T. Martin and E. Game for comments on
the manuscript; and P. Kareiva, S. Polasky, R. L. Pressey, S. Andelman and
B. Murdoch for discussions. The work was supported by The University of
Queensland and grants from the Australian Research Council (to H.P.P,
M. A. McCarthy and R. L. Pressey).
Author Information Reprints and permissions information is available at
npg.nature.com/reprintsandpermissions. The authors declare no competing
financial interests. Correspondence and requests for materials should be
addressed to H.P.P. (email@example.com) or K.A.W.
NATURE|Vol 440|16 March 2006