Identifying priority areas for bioclimatic representation under climate change: a case study for Proteaceae in the Cape Floristic Region, South Africa

www.elsevier.com/locate/biocon

Biological Conservation 125 (2005) 1–9

BIOLOGICAL

CONSERVATION

Identifying priority areas for bioclimatic representation underclimate change: a case study for Proteaceae in the Cape

Floristic Region, South Africa

Christopher R. Pyke a,*, Sandy J. Andelman a, Guy Midgley b

a National Center for Ecological Analysis and Synthesis, 735 State Street, Suite 300, Santa Barbara, CA 93101, USAb National Botanical Institute, Kirstenbosch Private X7, Claremont 7735, South Africa

Received 14 April 2004

Abstract

Biological reserves are established to protect natural resources and represent the diversity of environments found within a region.

Unfortunately, many systems of protected areas do not proportionally capture the range of environmental conditions occupied by

species and communities. Combinations of habitat loss and climate change may exacerbate these representational biases, and result

in future distributions of environmental conditions that bare little resemblance to historic patterns. New protected areas need to be

established to correct existing biases, and create conservation networks that remain representative despite climate change, habitat

loss, and changes in species distributions. We demonstrate a new method to identify and prioritize habitat based on its value for

improving bioclimatic representation. We assessed representation provided by existing protected areas for 301 Proteaceae species

under historic and projected 2050 climate across the Cape Floristic Region in South Africa. The existing reserve system has relatively

modest biases with respect to current species distributions and climate. However, if the system is not supplemented, protected areas

in 2050 will capture an increasingly skewed sample of climatic conditions occupied by Proteaceae. These biases can be repaired

through the systematic establishment of new protected areas, and many of the most valuable areas coincide with high priority eco-

system components and irreplaceable elements identified in the Cape Action for People and the Environmental conservation plan.

Protecting these areas achieves nearly the best possible improvement in climatic representation while also meeting biodiversity rep-

resentation goals.

� 2004 Elsevier Ltd. All rights reserved.

Keywords: Climate change; Biodiversity; Systematic conservation planning; Reserve design and selection; Bioclimatic representation

1. Introduction

Reserve networks are a cornerstone of biodiversity

conservation strategies. Effective reserve networks mustrepresent the full range of biodiversity within the region

of interest (Margules and Pressey, 2001). Ideally, habitat

in reserves should also represent the same breadth and

diversity of environmental conditions found across the

ranges of target species and communities (Noss, 2001).

0006-3207/$ - see front matter � 2004 Elsevier Ltd. All rights reserved.

doi:10.1016/j.biocon.2004.08.004

* Corresponding author. Tel.: +1 703 549 0611.

E-mail address: [email protected] (C.R. Pyke).

However, world-wide, existing reserve systems provide

a biased sample of both biodiversity (Margules and

Pressey, 2001; Andelman and Willig, 2003; Rodrigues

et al., 2004) and environmental conditions (Scott et al.,2001; Rouget et al., 2003a,b), resulting in the over-repre-

sentation of some elements and no protection for others.

Climate change poses a major threat to species persis-

tence, and it challenges the effectiveness of reserve net-

works as a conservation strategy. Once designated,

reserves are fixed in space. Yet, relatively small changes

in climate can lead to shifts in the distribution of suit-

able habitat and environmental conditions for a species

2 C.R. Pyke et al. / Biological Conservation 125 (2005) 1–9

or community, and jeopardize the role of reserves as safe

havens for biodiversity. The implications of climate

change for species persistence within reserves may be

confounded by habitat loss outside reserves (Pyke,

2004; Pyke, 2005). Existing representational biases in re-

serve networks make it difficult to predict the outcomesof habitat loss-climate change interactions, and future

distributions of environmental conditions available to

species and communities may bear little resemblance

to current or historic patterns.

Current conservation strategies do not incorporate

analytical techniques to repair these biases. New tools

are required to create reserve networks that protect spe-

cies, communities, and ecosystems and the diversity ofenvironmental conditions they require for survival. Here

we present a novel heuristic method for identifying areas

with value for improving representation of bioclimatic

conditions for multiple species in the context of regional

climate change. We then apply a variant of the method

to evaluate the adequacy of bioclimatic representation

within existing or proposed protected area networks.

We focus on species in the family Proteaceae, in theCape Floristic Region (CFR) of South Africa (Fig. 1),

a global hotspot for biodiversity (Cowling et al., 1996;

Myers et al., 2000). Many Proteaceae species are rare

(124 species) and half are listed as endangered or vulner-

able by the World Conservation Union (Hilton-Taylor,

1996; IUCN, 2000; WCMC, 2002). Moreover, many of

these species are associated with the Fynbos Biome,

whose areal extent is expected to contract with climatechange (Midgley et al., 2002). Proteaceae in the CFR

Fig. 1. Location of study area in South Africa. The inset illustrates the distri

study area.

are protected by a relatively extensive system of nature

reserves; however, the distribution of protected habitat

is geographically biased (Rouget et al., 2003b). Habitat

associated with steep slopes and high elevations is over

protected, at the expense of low-relief, coastal areas

(Rouget et al., 2003b).

2. Methods

2.1. Present and future distributions of climate and

Proteaceae

Data on historic environmental conditions and cur-rent species distributions were available across the

CFR for a grid of 1 · 1 min cells (approximately

1.85 · 1.55 km at this latitude). The species distributions

were obtained from the Protea Atlas Project (PAP),

(www.protea.worldonline.co.za/default.htm), compris-

ing more than 250,000 records of presence and absence

for 340 species at more than 60,000 geo-referenced

locations.Future climate projections are based on Schulze and

Perks (1999), according to the 2050 projections for the

region from the Global Change Model HadCM2n.

Midgley et al. (2002) downscaled coarse-scale (3� · 3�)projections of climatic conditions for 2050 to regional

1 · 1 min grids. Downscaled change vectors were ap-

plied to historic climate grids (Schulze, 1997) to yield

map-based predictions of future distributions of climateacross the CFR.

bution of protected areas, degraded lands, and intact habitat across the

C.R. Pyke et al. / Biological Conservation 125 (2005) 1–9 3

Expected future species distributions of Proteaceae

species in 2050 were modeled using bioclimatic envelope

models that combined Generalized Additive Models

(GAMs) of environmental suitability for each species

based on five environmental parameters critical for plant

physiological function (details in Midgley et al. (2002))with constraints on species dispersal distances and rates

(primarily based on dispersal agent). Receiver Operator

Curve (ROC) thresholds were used to limit future distri-

butions and to create predictions of species distributions

for 2050. The result is a set of binary range maps with

1 · 1 min cells. Based on these models, bioclimatic con-

ditions suitable for 301 species are expected to be avail-

able in 2050. Bioclimatic conditions currently occupiedby 39 species in 2000 will no longer exist at this scale

of analysis in 2050.

2.2. Uncertainty in climatic and biogeographic projections

The climatic and biogeographic models used for this

analysis are subject to substantial uncertainties. These

uncertainties result from assumptions about parameters,mechanisms, and socio-economic conditions that are dif-

ficult to estimate or validate. However, climate modeling

from multiple international groups indicates a broad

consensus that this region will experience warmer condi-

tions given current trends. The magnitude and spatial

distribution of these changes vary by model and socio-

economic scenario, but the complexity of our coupled

analysis limits us to using a single projection from onemodel for biogeographic analysis. Perhaps more impor-

tant is uncertainty in modeling species responses to cli-

mate change. Bioclimatic envelope models assume that

historic relationships between climate and biogeography

can, at least in part, be used to predict future distributions

(Pearson et al., 2002; Thuller, 2003). This assumption will

not always hold true (Davis et al., 1998). Despite these

concerns, such models describe fundamental ecologicalrelationships, and are valuable for understanding biogeo-

graphic patterns for many species at regional and global

scales (Pearson and Dawson, 2003). Given these uncer-

tainties, the results presented here should be viewed as

a preliminary solution to one plausible realization of a

complex system.

2.3. Bioclimatic assessment and priorities for repairing

representation

Data about the geographic distribution of species, cli-

mate, and habitat were used to address three research

goals: (1) characterize bioclimatic representation within

protected areas for Proteaceae species under 2000 and

2050 climates, (2) identify areas outside of existing pro-

tected areas with high value for improving bioclimaticrepresentation for multiple species, and (3) evaluate

the potential of existing conservation plans, if imple-

mented, to improve representation. We used a four-step

procedure:

2.3.1. Bioclimatic representation: Evaluate the

representation of mean annual precipitation (MAP)

across the range of each species for habitat both inside and

outside of biological reserves under 2000 and 2050

climates

We define a bioclimatic representation goal such that

reserves provide a proportional sample of the distribu-

tion of environmental conditions across the range of

each target species. Representation is bioclimatic, be-

cause it is a product of the geographic distributions of

both species and climate. There is considerable uncer-tainty associated with predicting future species distribu-

tions. Thus reserve-siting methods that set objectives

based on direct measures of species representation, such

as a given number of occurrences or populations (Camm

et al., 2002) or a fraction of the total species range

(Rodrigues et al., 2004), entail some risk that reserves

will be designated in places where target species will

not persist. Rather than focusing on the precise loca-tions where species are expected to occur in 2050, our

objective is to identify places valuable for the conserva-

tion of the range of bioclimatic conditions currently

occupied by each species, despite changes in climate

and large-scale habitat loss. In other words, we are inter-

ested in maintaining a biophysical ‘‘stage’’ for each spe-

cies to act on, even though the ‘‘stage’’ may be

shrinking, as well as moving across the landscape. Weillustrate this approach using current and future MAP.

An analysis of MAP cannot hope to capture the full

range of environmental variability important to so many

different species (Faith and Walker, 1996; Midgley et al.,

2003), but it can serve to illustrate methods and provide

one important metric of ecological impacts.

2.3.2. Bioclimatic representation Index (RI): Summarize

the performance of the existing reserve network for each

species using a new metric called the bioclimatic

representation index (RI)

The RI for any species is the difference between the

species� region-wide average MAP in 2000 and the re-

serves-only average MAP in 2050. Region-wide averages

are only the most basic model of complex intra-range

environmental variability, and more comprehensivetreatments could identify many additional dimensions

of environmental variation (see Faith and Walker,

1996; Pyke and Fischer, in press). Assuming a worst case

scenario, in which all habitat outside protected areas

may be lost, we calculated the average bioclimatic RI

for all species using Eq. (1), where HA (all habitat) is

the region-wide average across all habitat under 2000

climate and HP (protected habitat) is the average forhabitat in existing protected areas under 2050 climate

for n species.

4 C.R. Pyke et al. / Biological Conservation 125 (2005) 1–9

RI ¼

Pn

1

HA � HP

n: ð1Þ

An RI value of 0.0 indicates that habitat within exist-

ing biological reserves has exactly the same mean for the

environmental variable as for the species range as awhole. Index values <0.0 indicate that, in the future, re-

serves will be wetter than current region-wide condi-

tions. Index values >0.0 indicate that protected areas

in the future will be drier than current region-wide con-

ditions. Thus, the magnitude of the departure from 0.0

provides an estimate of the degree of representational

bias and the sign indicates the direction. The index can

be calculated either for current conditions or for pro-jected future ranges of species or anticipated protected

areas. It is possible for unusual distributions of habitat

to confound the RI index; however, with so many differ-

ent possible combinations of habitat, species ranges, and

climate, this is unlikely.

We also used Moran�s I, an index of spatial autocor-

relation (Goodchild, 1986), implemented in the ArcGIS

Workstation�s GRID module (ESRI, 2002), to evaluatechanges in the spatial structure of MAP between 2000

and 2050. The Moran�s I index ranges from �1 (check-

erboard, uncorrelated) to +1 (smooth, clustered, highly

correlated). MAP naturally would be expected to show

strong patterns of spatial autocorrelation across a spe-

cies range (Moran�s I index values approaching 1.0)

(Koenig, 1999); however, habitat loss can disrupt spatial

patterns and change regional correlation structure(Wimberly et al., 2000).

2.3.3. Geographic weightings: Apply the RI as a weighting

to prioritize areas where the addition of new protected

areas might improve the distribution of bioclimatic

representation in reserves

We split the future (2050) range of each species,

identifying the portion of the range projected to be un-der-represented, and weighting cells within that area

according to the absolute value of the RI. This was

implemented in ArcGIS Geographic Information Sys-

tem using Arc Macro Language (AML) scripts to per-

form map algebra calculations on 1� · 1� raster grids

(limited by the spatial resolution of the climate data).

AML scripts to perform this analysis are available from

the corresponding author. If the species� MAP RI wasgreater than 0.0 (a surplus of dry areas, deficit of above

average precipitation areas), unprotected habitat areas

with 2050 MAP values higher than the species� region-wide 2050 mean were selected and assigned the absolute

value of the RI. If the RI was smaller than 0.0 (a surplus

of wet areas, a deficit of relatively low precipitation

areas), unprotected areas with MAP values lower than

the species� region-wide mean were selected and assignedthe absolute value of the representation ratio. This re-

sulted in a stack of 301 species grids with portions of

the range above or below the region-wide average for

each species attributed with the absolute value of their

representation ratio. This relatively simple method only

breaks the range into two bins, high and low. A more

sophisticated implementation might further subdividethe range and capture more dimensions of environmen-

tal variation, but this would substantially increase com-

plexity, computational effort, and data quality

requirements (see Pyke and Fischer, in press).

2.3.4. Aggregate representation index deficit (ARID)

map: Combine the RI weighted grids for each species to

create a composite map indicating high value conservation

areas

We created a composite grid, describing the ARID

for MAP, by summing the 301 weighted species. The

magnitude of the ARID grid corresponds to a compos-

ite representation deficit for all species occurring at a gi-

ven pixel. Low values indicate that environmental

conditions are already represented in protected areas.

High values indicate that environmental conditions arerelatively under-represented. The ARID map provides

a practical heuristic that directs attention toward areas

that could improve the representation of future biocli-

matic conditions for many species.

2.4. Evaluate existing and proposed reserves

This bioclimatic assessment procedure also can beused to evaluate the representational value of existing

and proposed protected areas. We demonstrate this for

the implementation stages of the Cape Action for People

and the Environment (CAPE) conservation plan (Cowl-

ing et al., 2003; Gelderblom et al., 2003). The CAPE

plan identifies seven implementation stages ultimately

leading to a reserve system covering 52% of the 87,892

km2 CFR. We calculated the value of these implementa-tion stages for bioclimatic representation, both individ-

ually and in aggregate by calculating average ARID

scores for each stage, and the cumulative average RI

score (Eq. (1)) at the end of each stage in the growing

reserve system. This provided information about the

representation value of specific stages, as well as the to-

tal improvement in representation achieved by imple-

menting the entire reserve network.

3. Results

We confirmed previous analyses of broad environ-

mental types within the CFR indicating a modest repre-

sentational bias in existing reserves under current

climate (Table 1) (Rouget et al., 2003b). Average MAPwithin existing reserves is 9% greater than for current

habitat region-wide (Fig. 2(a)). The overall correlation

Table 1

Summary statistics for MAP across the Cape Floristic Region study

area

2000 2050

All habitat Reserves All habitat Reserves

MAP (mm/year)

Mean 466 633 367 498

SD 291 368 242 302

Minimum 68 82 39 64

Maximum 3345 3345 2784 2784

C.R. Pyke et al. / Biological Conservation 125 (2005) 1–9 5

between MAP inside reserves and region-wide is strong

(r2 = 0.97, df = 299, p < 0.001). Existing reserves do a

nearly perfect job capturing current region-wide maxi-

mum MAP values (r2 = 0.99, df = 299, p < 0.001), but

biases toward higher elevation areas mean that region-

wide minimum MAPs are occasionally poorly repre-

sented (r2 = 0.94, df = 299, p < 0.001) (Fig. 2(b) and

(c)). This is reflected in an overall 10% difference in min-imum MAP and no net difference in maximum MAP.

Future patterns of bioclimatic representation may

not be as robust. A comparison of all habitat under cur-

rent climate with only habitat in existing reserves, with

2050 MAP and 2050 Proteaceae distributions, shows a

positive, but weaker relationship than exists today

(r2 = 0.75, df = 299, p < 0.001) (Fig. 2(a)). Representa-

tion for region-wide extremes declines more dramati-

Res

erve

s(2

000)

ME

AN

500

1000

1500

2000

Res

erve

s(2

000)

MIN

200

400

600

800

(a)

(d) (e)

(b)

All habitat (2000) MEAN All habi

Res

erve

s(2

050)

ME

AN

500 1000 1500 2000

050

010

0015

00

Res

erve

s(2

050)

MIN

200 400

020

040

060

080

010

00

Fig. 2. Correlations for individual species (n = 301) between MAP across regi

climate (y-axis) with historic ranges: (a) Average MAP, (b) minimum MAP a

MAP, (e) minimum MAP and (f) maximum MAP. Lines are 1-to-1 correlat

cally (Fig. 2(e) and (f)). Projections for 2050 suggest

that existing reserves will capture a lower proportion

of region-wide minimums (r2 = 0.48, df = 299, p <

0.001) and maximums (r2 = 0.76, df = 299, p < 0.001)

than under current climate and species distributions.

These changes in aggregate MAP representationwill also be accompanied by changes in the spatial

pattern of MAP. MAP within existing protected areas

tends to provide precipitation patterns with lower spa-

tial autocorrelation than is found region-wide

(r2 = 0.87, df = 299, p < 0.001) (Fig. 3a). For current

(2000) climate, the average Moran�s I statistic for

MAP inside reserves is 0.38, compared to 0.42 for re-

gion-wide habitat. This indicates a reduction in spatialcorrelation if habitat outside of reserves is lost. The

average Moran I statistic for all species in reserves un-

der 2050 climate drops slightly to 0.35. However, this

small change in average conditions masks a substan-

tial decline in species-by-species correlations

(r2 = 0.21, df = 299, p < 0.001) (Fig. 3b). This suggests

that many individual species will experience substan-

tial changes in the spatial pattern of MAP availableacross their future ranges.

RI values varied widely among Proteaceae species.

The range of values approximately followed a normal

distribution with a mean of –87 mm/year (SD = 149) .

This pattern suggests a systematic change in

eser

ves

(200

0)M

AX

500

1000

2000

3000

(f)

(c)

tat (2000) MIN All habitat (2000) MAX

600 800

Res

erve

s(2

050)

MA

X

500 1000 2000 3000

050

010

0015

0020

0025

00

on-wide habitat (x-axis) and reserves-only habitat under historic (2000)

nd (c) maximum MAP: with projected 2050 species ranges, (d) average

ion lines, and they are not fitted to the data.

Moran index (all habitat 2000)

Mor

an in

dex

(res

erve

s-on

ly 2

000)

0.0 0.2 0.4 0.6 0.8

-0.2

0.0

0.2

0.4

0.6

0.8

Moran index (all habitat 2000)M

oran

inde

x (r

eser

ves-

only

205

0)0.0 0.2 0.4 0.6 0.8

-0.2

0.0

0.2

0.4

0.6

0.8Increasing spatial autocorrelation

Decreasing spatial autocorrelation

(a) (b)

Fig. 3. Correlations in Moran�s I spatial autocorrelation index for: (a) all region-wide habitat under 2000 climate and reserves-only habitat under

2000 climate, and (b) all region-wide habitat under 2000 climate and reserves-only habitat under projected 2050 climate and species ranges. Moran�s Ivalues range from �1.0 (contrasting, checkerboard landscapes) to +1.0 (smooth, clustered landscapes). Random landscapes have a Moran�s I valueof 0.0.

6 C.R. Pyke et al. / Biological Conservation 125 (2005) 1–9

environmental representation between 2000 and 2050. A

qualitative assessment of the ARIDmap suggests at least

three areas with high value for multiple species (Fig. 4).

Predictably, these correspond to gaps in the existing re-

serve system, where habitat outside reserves is heavily de-

graded and is poorly represented in protected areas.

Implementation of the CAPE conservation planwould lead to significant improvements in the represen-

tation of environmental conditions. Areas selected for

implementation in Stages 1 (average RI: 148, maximum

RI: 9625) and 2 (average RI = 111, maximum RI: 3587)

have the greatest absolute value for bioclimatic repre-

Fig. 4. ARID scores for MAP across 301 Proteaceae species (see E

sentation (Fig. 5). However, relatively high value areas

are available in every implementation stage. The combi-

nation of relatively high value and large area mean that

the greatest improvements in representation occur with

the implementation of Stages 0 through 2 (Fig. 6).

Stages 3 through 6 yield only a 4.5% improvement in

average RI with the addition of 27,713 km2 (+67%)more habitat. Implementation of Stage 6 yields nearly

the best possible average RI for this region of �56

mm/year. The residual reflects region-wide representa-

tional differences that cannot be repaired with remaining

habitat under future climate conditions.

q. (1)). Circles highlight concentrations of high-value habitat.

0 1 2 3 4 5 6

050

100

150

Ave

rage

AR

ID (

mm

/yr)

CAPE implementation stage

0 1 2 3 4 5 6

020

0040

0060

0080

0010

000

Max

imum

AR

ID (

mm

/yr)

CAPE implementation stage

Fig. 5. ARID scores associated with CAPE conservation plan implementation stages: (a) average ARID value by stage and (b) maximum ARID

value by stage.

Total reserve area (km2)

Ave

rage

RI (

mm

/yr)

10000 20000 30000 40000 50000 60000 70000

-110

-100

-90

-80

-70

-60

1

0

2 3 4 56

Fig. 6. Improvement in average RI for all 301 Proteaceae species with increasing reserve area. The points are labeled by their CAPE plan

implementation stage.

C.R. Pyke et al. / Biological Conservation 125 (2005) 1–9 7

4. Discussion

The CFR�s relatively extensive reserve system pro-vides reasonably good environmental representation to-

day. However, its modest bioclimatic biases are likely to

be exacerbated by climate change and species responses.

For most Proteaceae species, this means that environ-

mental representation within protected areas will be-

come more biased in the future. Repairing these biasesnow, while opportunities still exist for conservation of

suitable habitat and environmental conditions, will cre-

ate a more robust biophysical foundation for population

8 C.R. Pyke et al. / Biological Conservation 125 (2005) 1–9

and community processes and increase the likelihood

that the region�s ecosystems respond more predictably

to climate change (Pyke, 2004).

The methods we present are general, but they are lim-

ited by the assumptions and uncertainties of the under-

lying data. The approach is sensitive to uncertainty inclimate change projections and in modeled species re-

sponses, as well as to the spatial grain of climate data,

the extent of the study area, and the choice of climatic

and habitat baselines. Despite these caveats, improving

proportional environmental representation in reserve

networks will almost always enhance biodiversity con-

servation strategies. The most effective strategy will re-

duce biases with respect to existing climate, while alsoseeking opportunities to safeguard conditions that may

become rare in the future. Opportunities to improve bio-

climatic representation can only diminish over time with

increasing habitat loss.

In the CFR, areas with high value for bioclimatic rep-

resentation are strongly associated with areas identified

as high priorities for biodiversity conservation. Our re-

sults indicate that implementing Stages 1 and 2 of theCAPE conservation plan will substantially improve bio-

climatic representation within reserves. These stages fo-

cus on areas with high value for bioclimatic

representation, and also contain underrepresented eco-

logical conditions (e.g., ecotones and edaphic gradients)

and irreplaceable distributions of Proteaceae. The per-

formance of representation deficit scores like ARID

might be improved with the addition of measures ofother environmental variables such as soils or geomor-

phology to provide a more fine-grained index of envi-

ronmental diversity.

Conserving biodiversity in a dynamic world requires

new strategies that go beyond static reserve siting meth-

ods (e.g., Meir et al., 2004) based on species representa-

tion objectives to focus on maintaining suitable

conditions for long-term species persistence, despitechanges in climate and large-scale habitat loss. We have

demonstrated that it is possible to identify areas where

conservation action can mitigate existing representa-

tional biases in environmental conditions and increase

opportunities for species to adapt to future conditions,

without sacrificing current biodiversity representation

goals.

Acknowledgements

This work was supported by The Nature Conser-

vancy�s David H. Smith Fellows Program (CRP), the

National Center for Ecological Analysis and Synthesis

at the University of California, Santa Barbara (NSF

grant DEB-0072909), by NSF grant DEB-0074676(SJA), the Climate Change Research Program, Center

for Applied Biodiversity Science, Conservation Inter-

national, and Thomas Lacher from the Caesar Kle-

berg Chair in Wildlife Ecology at Texas A&M

University. We thank L. Hannah, R. Cowling, and

two anonymous reviewers for comments on the

manuscript.

References

Andelman, S.J., Willig, M.R., 2003. Present patterns and future

prospects for biodiversity in the Western Hemisphere. Ecology

Letters 6, 818–824.

Camm, J.D., Norman, S.K., Polasky, S., Solow, A.R., 2002. Nature

reserve selection to maximize expected species coverage. Opera-

tions Research 50, 946–955.

Cowling, R.M., Rundel, P.W., Lamont, B.B., Arroyo, M.K., Arian-

outsou, M., 1996. Plant diversity in Mediterranean-climate regions.

Trends in Ecol and Evol 11, 362–366.

Cowling, R.M., Pressey, R.L., Rouget, M., Lowbard, A.T., 2003. A

conservation plan for a global biodiversity hotspot – the Cape

Floristic Region, South Africa. Biological Conservation 112, 191–

216.

Davis, A.J., Jenkinson, L.S., Lawton, J.H., Shorrocks, B., Woods, S.,

1998. Making mistakes when predicting shifts in species range in

response to global warming. Nature 391, 783–786.

ESRI, 2002. ArcGIS. ESRI, Redlands, CA.

Faith, D.P., Walker, P.A., 1996. Environmental diversity: on the

best-possible use of surrogate data for assessing the relative

biodiversity of sets of areas. Biodiversity and Conservation 5,

399–415.

Gelderblom, C.M. et al., 2003. Turning strategy into action: imple-

menting a conservation action plan in the Cape Floristic Region.

Biological Conservation 112, 291–297.

Goodchild, M.F., 1986. Spatial Autocorrelation Catmog 47. Geo

Books, Norwich.

Hilton-Taylor, C., 1996. Red Data List of Southern African Plants.

National Botanical Institute, Pretoria, South Africa.

IUCN, 2000. IUCN Red List of Threatened Species. IUCN, Gland,

Switzerland.

Koenig, W.D., 1999. Spatial autocorrelation of ecological phenomena.

TREE 14, 22–26.

Margules, C.R., Pressey, R.L., 2000. Systematic conservation plan-

ning. Nature 405, 243–253.

Meir, E., Andelman, S.J., Possingham, H.P., 2004. Does conservation

planning matter in a dynamic and uncertain world. Ecology Letters

11, 615–622.

Midgley, G.F., Hannah, L., Millar, D., Rutherford, M.C., Powrie,

L.W., 2002. Assessing the vulernability of species richness to

anthropogenic climate change in a biodiversity hotspot. Global

Ecology & Biogeography 11, 445–451.

Midgley, G.F., Hannah, L., Millar, D., Thuiller, W., Booth, A., 2003.

Developing regional and species-level assessments of climate

change impacts on biodiversity in the Cape Floristic Region.

Biological Conservation 112, 87–93.

Noss, R.F., 2001. Beyond Kyoto: Forest management in a time of

rapid climate change. Conservation Biology 15, 578–590.

Pearson, R.G., Dawson, T.P., 2003. Predicting the impacts of

climate change on the distribution of species: are bioclimatic

envelope models useful?. Global Ecology and Biogeography 12,

361–371.

Pearson, R.G., Dawson, T.P., Berry, P.M., Harrison, P.A., 2002.

SPECIES: a spatial evaluation of climate impact on the envelope of

species. Ecological Modelling 154, 289–300.

Pyke, C.R., 2004. Habitat loss confounds climate change impacts.

Frontiers in Ecology and the Environment 2, 178–182.

C.R. Pyke et al. / Biological Conservation 125 (2005) 1–9 9

Pyke, C.R., 2005. Interactions between habitat loss and climate

change: Implications for fairy shrimp in the Central Valley of

California. Climatic Change 121 (3), 429–441.

Pyke, C.R., Fischer, D.T., 2005. Selection of bioclimatically represen-

tative biological reserve systems under climatic change. Biological

Conservation 121 (3), 429–441.

Rodrigues, A.L., Andelman, S.J., Bakarr, M.I., Boitani, L., Brooks,

T.M., Cowling, R.M., Fishpool, L.D.C., da Fonseca, G., Gaston,

K.J., Hoffmann, M., Long, J., Marquet, P.A., Pilgrim, J.D.,

Pressey, R.L., Schipper, J., Sechrest, W., Stuart, S.N., Underhill,

L.G., Waller, R.W., Watts, M.E.J., Yan, X., 2004. Effectiveness of

the global protected area network in representing species diversity.

Nature 428, 640–643.

Rouget, M., Cowling, R.M., Pressey, R.L., Richardson, D.M.,

2003a. Identifying spatial components of ecological and evolu-

tionary processes for regional conservation planning in the Cape

Floristic Region, South Africa. Diversity and Distributions 9,

191–210.

Rouget, M., Richardson, D.M., Cowling, R.M., 2003b. The current

configuration of protected areas in the Cape Floristic Region,

South Africa – reservation bias and representation of biodiversity

patterns and processes. Biological Conservation 112, 129–145.

Schulze, R.E., 1997. South African Atlas of Agrohydrology and

Climatology. Report TT82/96, Water Resource Commission,

Pretoria, South Africa.

Schulze, R.E., Perks, L.A., 1999. Assessment of the impact of climate.

Final report to the South African Country Studies Climate Change

Programme. School of Bioresources Engineering and Environmen-

tal Hydrology, University of Natal, Pietermaritzburg, South

Africa.

Scott, J. et al., 2001. Nature reserves: Do they capture the full range of

America�s biological diversity?.EcologicalApplications11, 999–1007.

Thuller, W., 2003. BIOMOD – optimizing predictions of species

distributions and projecting potential future shifts under global

change. Global Change Biology 9, 1353–1362.

WCMC, 2002. The 2002 IUCN Plant Red Data Book, World

Conservation Monitoring Centre – IUCN, Gland, Switzerland.

Wimberly, M., Spies, T., Long, C., Whitlock, C., 2000. Simulating

historical variability in the amount of old forests in the Oregon

Coast Range. Conservation Biology 14, 167–180.