Hotspots, complementarity or representativeness? Designing optimal small-scale reserves for biodiversity conservation

Abstract

Reserve networks are a major tool of ecological management aiming at biodiversity conservation. Maximizing the number of species conserved with the minimum land sacrifice is a primary requirement in reserve design. In this study, we examine the efficiency of five different scenarios to conserve: (i) the biodiversity of one target group and (ii) the overall biodiversity of an area. The study was conducted in Dadia Reserve, in northern Greece. Six groups of species were selected to represent its biodiversity: woody plants, orchids, Orthoptera, aquatic and terrestrial herpetofauna, and small terrestrial birds. The scenarios examined represent different conservation approaches to select network sites. For each approach, the starting point was one of the above six groups of species, considered as the target group. In scenario A, which reflects the hotspot approach, the sites richest in species are selected. Scenario B selects the sites most complementary in terms of species richness. The next two scenarios use the principle of environmental representativeness, expressed in terms of habitat (scenario C) or vegetation (scenario D). Under scenario E, sites forming the network are selected at random. The rank of scenarios in terms of preserving the species of the target group was always B > A > C > D > E, irrespective of the group considered as target group. Their rank, when preservation of the total biodiversity was the issue, was B, A > C, D > E.

Full text

Hotspots, complementarity or representativeness? designing
optimal small-scale reserves for biodiversity conservation
Vassiliki Kati
a,*
, Pierre Devillers
b
, Marc Dufr^
ene
c
, Anastasios Legakis
d
,
Despina Vokou
e
, Philippe Lebrun
f
a
Department of Environmental and Natural Resources Management, University of Ioannina, Seferi 2, 30100 Agrinio, Greece
b
Institut Royal des Sciences Naturelles de Belgique, Section de Biologie de la Conservation, rue Vautier 29, B-1000 Bruxelles, Belgium
c
Minist
ere de la R
egion Wallonne, Centre de Recherche de la Nature, des For^
ets et du Bois, Avenue Mar
echal Juin, 23, 5030 Gembloux, Belgium
d
Department of Biology, University of Athens, Zoological Museum, Panepistimioupolis, 15784 Athens, Greece
e
Department of Ecology, Aristotle University of Thessaloniki, School of Biology, UPB 119 54124 Thessaloniki, Greece
f
Universit
ecatholique de Louvain, Unit
ed’

Ecologie et de Biog
eographie, Centre de Recherche sur la Biodiversit
e,
Place Croix du Sud, 5, 1348 Louvain-la-Neuve, Belgium
Received 23 July 2003; received in revised form 4 February 2004; accepted 26 March 2004
Abstract
Reserve networks are a major tool of ecological management aiming at biodiversity conservation. Maximizing the number of
species conserved with the minimum land sacrifice is a primary requirement in reserve design. In this study, we examine the efficiency
of five different scenarios to conserve: (i) the biodiversity of one target group and (ii) the overall biodiversity of an area. The study
was conducted in Dadia Reserve, in northern Greece. Six groups of species were selected to represent its biodiversity: woody plants,
orchids, Orthoptera, aquatic and terrestrial herpetofauna, and small terrestrial birds. The scenarios examined represent different
conservation approaches to select network sites. For each approach, the starting point was one of the above six groups of species,
considered as the target group. In scenario A, which reflects the hotspot approach, the sites richest in species are selected. Scenario B
selects the sites most complementary in terms of species richness. The next two scenarios use the principle of environmental rep-
resentativeness, expressed in terms of habitat (scenario C) or vegetation (scenario D). Under scenario E, sites forming the network
are selected at random. The rank of scenarios in terms of preserving the species of the target group was always B > A > C > D > E,
irrespective of the group considered as target group. Their rank, when preservation of the total biodiversity was the issue, was B,
A > C, D > E.
Ó2004 Elsevier Ltd. All rights reserved.
Keywords: Reserve design; Ecological networking; Biodiversity; Conservation; Complementarity
1. Introduction
Reserves have a major role as a tool for preserving
biodiversity (Margules and Pressey, 2000). In designing
reserve, one of the objectives is to maximize the number
of species conserved with the minimum land sacrifice, or
else satisfy the ‘‘minimal reserve set’’ requirement (re-
view by Cabeza and Moilanen, 2001). Reserves are still
set up in response to political and economic interests
rather than scientific criteria. Nevertheless, a multitude
of methods and scientific approaches have been devel-
oped to facilitate optimal reserve design; they are pri-
marily based on hotspot identification and on
complementary and representative networking.
Biodiversity hotspots are areas with a large number
of species or with large numbers of rare, threatened or
endemic species. Because of these features, they are
considered of high conservation priority (Margules and
Usher, 1981; Prendergast et al., 1993; Muyers et al.,
2000; Rodriguez and Young, 2000). Designing reserve
networks on the basis of the richest-in-species areas re-
flects the traditional practice.
*
Corresponding author. Tel.: +30-26510-60949; fax: +30-26510-
29477.
E-mail address: vkati@cc.uoi.gr (V. Kati).
0006-3207/$ - see front matter Ó2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.biocon.2004.03.020
Biological Conservation 120 (2004) 471–480
www.elsevier.com/locate/biocon
BIOLOGICAL
CONSERVATION

Complementarity, a term invented by Vane-Wright
et al. (1991), is considered a key-principle in reserve
designing (Pressey et al., 1993; Margules and Pressey,
2000). Its application in site selection ensures that as
much as possible new attributes will be added to an
existing reserve system. These attributes can be species,
endemic species (Kirkpatrick, 1983) or landscape units
(Pressey and Nicholls, 1989). A great number of itera-
tive algorithms have been proposed to provide minimum
sets of complementary sites that can maintain biodi-
versity at its maximum (see review by Csuti et al., 1997;
Pressey et al., 1997).
Representativeness is an old and widely used princi-
ple for reserve selection (Margules and Usher, 1981;
Franklin, 1993). The aim in applying it is to ensure that
all environmental variation is well represented in the
selected reserve network (Faith and Walker, 1996).
Standard typologies of habitats or vegetation types (see
Devillers and Devillers-Terschuren, 1996; Pienkowski
et al., 1996; Stoms et al., 1998) can be used to represent
the diversity of the environment.
The aim of this study is to examine the efficiency of
five different conservation approaches to preserve the
biodiversity (in terms of six groups of species) of a
Mediterranean-type area. To this end, we conducted a
study in Dadia Reserve, in northern Greece, an area
protected because of its high ornithological value. The
groups selected to represent its biodiversity were the
woody plants (tree and shrub species), the orchids,
the Orthoptera, the aquatic herpetofauna (amphibians
and freshwater turtles), the terrestrial herpetofauna
(lizards and terrestrial tortoises), and the small terres-
trial birds (Passeriformes, Columbiformes, Coracifor-
mes and Piciformes). For each of the five conservation
approaches examined, the starting point is one of the
above six groups of species. This is considered as the
target group, for the sake of which the reserve network
is primarily designed. The question we ask is how
much biodiversity of each one of the other groups and
of the total biodiversity of the area is preserved in the
networks, constructed following the different conser-
vation approaches. In scenario A, which reflects the
hotspot approach, the sites richest in species are se-
lected. Scenario B selects the sites most complementary
in terms of species richness. The next two scenarios use
the principle of environmental representativeness, ex-
pressed in terms of habitat (scenario C) or vegetation
(scenario D). Under scenario E, sites forming the re-
serve network are selected at random. The five ap-
proaches are evaluated according to their efficiency to
preserve non-target group and the total biodiversity of
the area. Our ultimate goal is to provide guidelines as
to the best practice for designing local, small-scale re-
serve networks, under different regimes of data, budget,
and time availability, particularly for the Mediterra-
nean region.
2. Methods
2.1. Study area
The study area is situated in northeastern Greece
(longitude 26°000to 26°190, latitude 40°590to 41°150). It
covers 430 km2, of which 424.6 km2belong to the Re-
serve of ‘‘Dadia-Lefkimmi-Soufli Forest’’, abbreviated
as Dadia Reserve. After the C
ORINEORINE
typology (Devillers
and Devillers-Terschuren, 1996), nine main vegetation
types occur in the sampled area, further divided to
nineteen sub-types (Table 1). A map of the main vege-
tation types of Dadia Reserve is presented in Kati et al.
(2004a).
2.2. Dataset
We used a dataset of 194 species (Kati et al., 2004b):
46 woody plant species, 19 orchid species, 39 Orthoptera
species, 10 species of aquatic herpetofauna, 10 species of
terrestrial herpetofauna, and 70 species of small terres-
trial birds. Species were sampled within 33 sites in the
study area, ranging in size from 0.02 km2up to 0.2 km2,
during the years 1998–1999. Sampling methodologies
are described in Kati et al. (2004b). None of the species
sampled was endemic. For Orthoptera, terrestrial her-
petofauna species and birds, data were semi-quantita-
tive, whereas for woody plant species, orchids and
aquatic herpetofauna species, data were of the type
presence-absence.
2.3. Data analysis
The diversity of sites was estimated in terms of species
richness (S), weighted species richness (WS), and Shan-
nonindex (H0). The weighted species richness (WS) is the
sum of the vulnerability indices of all the species of the
site. The vulnerability index of a species, taking values
0–35, estimates its status in the European Union (EU).
It is calculated on the basis of its distribution, on both a
coarse (in grids of 500 km 500 km) and a fine scale (in
grids of 10 km 10 km), of its population size (or its
habitat rarity), and of its population trend (for full de-
scription, see Bezzel, 1980). Vulnerability indices for
orchids are published in Devillers et al. (1991), for Or-
thoptera in Kati et al. (2004a), and for herpetofauna and
birds in Kati (2001).
We define target group the group of species for which
we are designing the reserve network and as non-target
groups all other groups. For every target group, we
constructed a similarity matrix of the samples, where its
members were found, using two coefficients widely used
in ecological studies – Sørensen and the Steinhaus
assymetrical coefficients of similarity. Sorensen coeffi-
cient behaves very well for binary data and Steinhaus is
its equivalent for semi-quantitative data (Legendre and
472 V. Kati et al. / Biological Conservation 120 (2004) 471–480

Legendre, 1998). Sampling units were then ordinated
across axes using the Principal Coordinate Analysis with
corrected eigenvalues (Dist.P.Co.A.) (Legendre and
Anderson, 1998). For the above analyses the R3 soft-
ware package (Legendre and Vaudor, 1991) was used.
The coordinates of Dist.P.Co.A were used as inputs into
k-means clustering (Legendre and Legendre, 1998) using
the
FASTCLUSFASTCLUS
procedure (S.A.S., 1985). Wherever the
Table 1
Description of sites samples in terms of their size and of the class to which they belong, according to the C
ORINEORINE
typology classification (the first two
digits of the C
ORINEORINE
codes correspond to the main habitat types), and results of the clustering procedure
Habitat types Corine code Description of
habitat types
Site area
(ha)
Number of cluster
Orchids Orthoptera Aquatic
herpetofauna
Terrestrial
herpetofauna
Birds
Broad-leaved
forests
41.1B 41.19311 Beech wood 20 3 75–5
2037635
41.76 Oakwoods
(Quercus
frainetto/cerris)
2026–35
2026–35
41.733 Oakwoods
(Quercus
pubescens)
20–7–36
20–3–26
Oakwoods with
bush
undergrowth
20–1–36
2053–26
Mixed forests 43.7 Mixed pine-oak
woods
2027–35
2027335
Coniferous
forests
42.661(C) Pinewoods
(Pinus nigra)
357534
42.85 A Pinewoods
(Pinus brutia)
Pinewoods with
bush
undergrowth
2057–13
557–24
1557–34
Riverine
vegetation
44.514 Riparian
vegetation
(Alnus glutinosa)
20–7415
20–7415
44.615 Riparian vegeta-
tion (Populus sp.)
20–7111
20–7111
Sclerophyllous
scrub
32.313 High maquis
(Arbutus sp.)
1517––6
1517–16
32.161 Deciduous oak
mattoral
2023–26
32.21A4 Bushes (Phyllirea
latifolia)
10–4–23
32.32 Low ericaceous
maquis
10–4–27
10–4–27
Humid
grassland
37.4 Humid
grasslands
1041236
341233
Dry grasslands 34.53 Xeric grasslands 10 – 5323
5–5–23
34.2 Heavy-metal
grasslands
2–5–13
Rural mosaics 84.4 Rural mosaics 20 – 2112
20–2112
Crops 82.11 Field crops 20 – 2118
20–2118
Total 20 19 518 5 7638
Numbers given under each group of species represent the cluster within which the respective sites were grouped.
Dashes indicate absence of the group in the respective sites.
V. Kati et al. / Biological Conservation 120 (2004) 471–480 473

number of species included in the matrix was low (ma-
trix for Orthoptera in shady sites and matrix for ter-
restrial herpetofauna), we used the less robust Ward’s
minimum variance method (Legendre and Legendre,
1998). The outputs were hierarchical dendrograms; the
distinct clusters that were produced represent the dif-
ferent habitat types of the target group.
2.4. Reserve selection procedure
Information concerning the construction of the five
conservation scenarios, upon which choice of the opti-
mal reserve network for the target group is based, is
given in Table 2. Scenarios A and B are species-based
and provide pragmatic solutions (specific sites selected).
The other scenarios provide a number of possible solu-
tions satisfying specific requirements. All calculations
were carried out with the help of the S.A.S. package.
2.4.1. Scenario A: the hotspot approach
For every target group i, we selected the richest sites
until all species of the group were represented. They
formed the reserve network A for the group i. When two
sites had the same species richness (S), we opted for the
site with the highest weighted species richness (WS), and
in case of further equality, we chose the site with the
highest Shannon’sindex (H0). Once the specific network
A for group iwas established, we calculated the number
of species of the non-target groups included in it.
2.4.2. Scenario B: the principle of complementarity
For every target group i, we ran an optimal selection
algorithm in order to identify the minimum set of sites
that preserve the maximum number of species of the
group. By definition, scenario B produces an optimal
network for any given number of sites, 1 to k, where kis
the number of sites required to protect all species of the
group. This algorithm combines randomly any number
of sites for 20,000 times, calculates the number of species
included in each combination and detects the site com-
bination with the maximum number of species of the
target group. Sometimes, the algorithm provides many
solutions for the same number of sites. In this case, we
opted for the set with the maximum cumulative weigh-
ted species richness. Once the specific network B for
group iand for a given number of sites was established,
we calculated the number of species of the non-target
groups included in it.
2.4.3. Scenario C: habitat representativeness
The third scenario introduces a rather novel ap-
proach: the principle of habitat representativeness (Sa-
etersdal and Birks, 1993). By this, it is meant that the
final reserve network will include all habitat types of the
target group. In the current study, these habitats are
defined after the distinct clusters produced by the clus-
tering procedure. In practical terms, for each target
group i, we ran an algorithm that picked randomly
(20,000 permutations) a given number of sites (n¼1to
the number of clusters of each target group) under the
condition that these sites belong to different clusters. For
each network of nsites, the algorithm produced 20,000
outputs calculating the average, minimum and maxi-
mum number of species included. For each target group
i, the procedure ended when the final reserve network
consisted of as many sites as the number of clusters of
the target group.
2.4.4. Scenario D: vegetation representativeness
Application of this approach ensures that a reserve
network will include the full spectrum of the environ-
mental variability of the area, as expressed by the dif-
ferent vegetation types occurring in it. We used the
C
ORINEORINE
typology system to express the environmental
Table 2
Methods used for the construction of conservation networks and outputs of the scenarios examined
Scenario/criteria
applied
Sites selected Type of algorithm used Output
Target group Non-target groups
A/hotspot approach nSites with the greatest number
of species (S)
Non applicable Network A Species number conserved in network A
B/complementarity nComplementary sites with the
greatest cumulative number of
species (Snetwork)
Optimal algorithm Network B Species number conserved in network B
C/habitat
representativeness
nRandom sites from those
belonging to different clusters
Random selection
algorithm respecting
typology
Network C Mean, minimum, maximum number of
species conserved in it
D/vegetation
representativeness
nRandom sites from those
belonging to different C
ORINEORINE
types
Random selection
algorithm respecting
typology
Network D Mean, minimum, maximum number of
species conserved in it
E/random choice nSites randomly chosen Random selection
algorithm
Network E Mean, minimum, maximum number of
species conserved in it
474 V. Kati et al. / Biological Conservation 120 (2004) 471–480

variability; this is represented by nine main habitat types
(corresponding to the first two digits of the coding sys-
tem, Table 1). For each target group i, we ran an al-
gorithm that picked randomly (20,000 permutations) a
given number of sites (n¼1–9), under the condition
that these sites belong to different C
ORINEORINE
habitat types.
For each network on n sites, the algorithm produced
20,000 outputs calculating the average, minimum and
maximum number of species included. The procedure
ended when a network of n¼9 sites was formed.
2.4.5. Scenario E: random choice
Random selection of sites is still a very common
practice in designing reserve systems. This approach
provides a measure of comparison for the efficiency of
the previous four conservation scenarios. For every
target group i, we ran a random selection algorithm
(20,000 permutations). The algorithm produced 20,000
outputs calculating the average, minimum and maxi-
mum number of species included in the network of n
sites (n¼1–33). The procedure ended when a network of
n¼33 sites was formed.
2.4.6. Scenario comparison
We assessed the efficiency of the reserve networks
developed under each of the five scenarios to conserve:
(a) any target group i, and (b) the biodiversity overall.
For every target group, and for a number of sites 1 to
k, we used species gain (g) as a measure to compare the
randomly produced network E with the other four net-
works. We define species gain gjas the difference SjSE,
where Sjis the percentage of species of the target group
conserved within each one of networks A to D, and SEis
the percentage conserved in the random network (Table
3). The parameter gis scale-independent and, therefore,
permits comparisons regardless of the number of sites
forming each particular network. Using one-way AN-
OVA and Dunnet T3 post hoc test (SPSS), we further
examined whether the mean values of species gain gj
differed significantly (p<0:05) among the four
networks.
The total biodiversity (BD6) is equal to the sum of
species of all groups studied. We define biodiversity gain
(gBD
i) as the difference of the percentage of the total
biodiversity conserved within any network (A, B, C, and
D), designed after a target group i, from that conserved
in network E (Table 3). By use of the parameter gBD,we
can compare the efficiency of each approach, regardless
of the number of sites selected to form the network. To
do so, we applied the same procedure as above (for each
target group) using this time biodiversity gain (instead
of species gain). We also designed the optimal biodi-
versity network (G) by running the optimal selection
algorithm, as described under scenario B, and we cal-
culated the maximum biodiversity gain (GBD)asthe
difference of the percentage of the total biodiversity
conserved within the optimal network from that in the
random network (Table 3).
We examined whether the mean values of the biodi-
versity gain gBD
j(j refers to networks A–D) of the net-
works having nsites differed significantly (p<0, 05)
from each other and from the maximum biodiversity
gain (GBD). To do so, we used one-way ANOVA and
Tukey and Dunnet T3 post hoc test (homogeneity of
variance satisfied or not, respectively) (SPSS).
3. Results
The clustering procedure gave 5 habitat types for
orchids, 7 for Orthoptera, 6 for aquatic herpetofauna, 3
for terrestrial herpetofauna and 8 for birds (Table 1).
3.1. Optimal reserve design for the conservation of one
target group
Results in Table 4 show the species gain (g)ineachof
the non-randomly constructed networks for the special
case of knumber of sites (given by scenario B), where k
is equal to 9 for the woody plants, 4 for orchids, 6 for
Orthoptera, 4 for aquatic herpetofauna, 2 for terrestrial
herpetofauna, and 8 for birds.
Species gain (g) differed significantly among networks
(F¼70:732, p<0:01; one way ANOVA). Dunnet T3
post hoc tests (p<0:05) showed that the hierarchy of
networks was always the same regardless of the target
group. Networks were ranked as follows B(7%) >
A(17%) > C(5%) > D. Numbers in parentheses show
Table 3
Parameters used for the comparison of networks’ efficiency; see text for explanation of symbols
Target of the network Parameter Parameter calculation
One group of species kMinimum number of sites needed to conserve all species of the group
Species percentage (S) Number of species/total number of species
Species gain (g)gj¼SjSE,(j¼A, B, C, D)
Biodiversity Total biodiversity (BD6)BD
6¼(species of all groups/194)*100
Biodiversity gain (gBD
j)gBD
j¼ðBD6
jBD6
EÞ,(j¼A, B, C, D)
Ideal biodiversity gain (GBD)GBD ¼ðBD6
optimal network BD6
EÞ100
V. Kati et al. / Biological Conservation 120 (2004) 471–480 475

how much more species on average are preserved in each
network compared to the next best alternative. Network
D, the worst alternative, was compared with the ran-
domly constructed network E; species gain values dif-
fered significantly from 0 (t¼4:250, 0 <0:01; ttest).
Network D preserved on average 2% more species than
network E.
3.2. Optimal reserve design for the conservation of
biodiversity
Table 5 ranks the different networks after their effi-
ciency to conserve the biodiversity of the area overall.
Biodiversity gain differed significantly (F¼211:757,
p<0:05) among networks designed after all six groups
combined; the rank of networks was the following:
G > B, A > C, D (post hoc tests, p<0:05). Biodiversity-
gain values differed in all cases from zero (ttest:
tA¼7;358, tB¼12:583, tC¼3:955, tD¼3:591,
tG¼22:761, p<0:05, for networks A, B, C, D and G,
respectively). This proves that each of the non-randomly
constructed networks is more efficient in conserving
biodiversity than the random one (E). However, the
above hierarchy was not always the same when net-
works were designed after only one target group (Table
5). For instance, scenario A did not differ significantly
from scenarios C and D, when the target group was the
aquatic herpetofauna, while scenario B did not differ
from scenarios C and D, when birds were the target
group. Besides, there were no significant differences
among scenarios in the case of terrestrial herpetofauna;
this is due to the fact that the optimal network for this
group consists of only two sites (k¼2), thus making
the sample size too small to demonstrate significant
relationships.
4. Discussion
4.1. Scenario efficiency
The success of a reserve network to conserve the bi-
ological diversity of an area depends on the quantity and
quality of biological data used to design it. Our results
show that, when conservation efforts target one group,
networks made in a complementary way (under scenario
B) are much more efficient than those based on the
richness hotspot approach (scenario A); this finding is in
agreement with results of a previous study (Williams
et al., 1996). When species richness data are missing but
there is information on the habitat preference of the
target group within a region (e.g., the Mediterranean), it
Table 5
Scenarios’ rank after their efficiency to conserve biodiversity for a network of 1, 2, 3 up to knumber of sites
Target group Network efficiency kFstatistic (p<0:05) Post hoc test
Woody plants G >B>A>D 9 155.810 Tukey
Orchids G > BA>C D > E 4 37.153 Tukey
Orthoptera G > BA>C D > E 6 95.296 Tukey
Aquatic herpetofaua G >BA>CD>E 4 25.900 Dunnett T3
Terrestrial herpetofauna G B A C D > E 2 17.235 Dunnett T3
Birds G >AB>D C > E 8 60.147 Dunnett T3
Biodiversity G > BA>C D > E – 211.757 Dunnett T3
Networks underlined do not differ significantly among them.
Table 4
Networks’ efficiency for the conservation of one target group in terms of species gain (g)
Target group Network of ksites g
ABCD
Woody plants 9 30 41 – 0
Orchids 4 52 68 19 1
Orthoptera 6 27 32 11 5
Aquatic herpetofauna 4 37 57 19 5
Terrestrial herpetofauna 2 36 56 2 0
Birds 8 10 24 8 3
Average 32 46.3 11.8 2.3
Given are the corresponding values (percentages) for every target group, under every scenario and for the minimum surface reserve network of k
sites (for definition of gand k, see Table 3).
476 V. Kati et al. / Biological Conservation 120 (2004) 471–480

is preferable to implement the principle of habitat rep-
resentativeness (scenario C) rather than the principle of
vegetation representativeness (scenario D). Finally, in
cases of emergency or whenever scarcity of resources
does not allow collection of data that will support ap-
plication of any of the first three approaches (A, B, C), it
is better to select sites in such a way as to represent the
environmental variability of the area (scenario D) rather
than selecting them at random (scenario E).
When our aim is to conserve the whole biological
diversity of an area, there is not a unique solution. The
ideal approach is to sample as many biological groups
as possible and design their complementary networks;
in this way all sampled components of local biodiver-
sity will be included. Obviously, the principle of com-
plementarity, as expressed in scenario B, is equally
useful when focusing on one group or on a set of
groups. Though ideal, this approach is rather imprac-
tical in the field of conservation biology as, in general,
we need efficient but not time-consuming solutions
(Meffe and Carrol, 1994). Were we able to identify bi-
ological indicators that could represent the full spec-
trum of an area’s biological diversity (Noss, 1990; Caro
and O’ Doherty, 1999; Soberon et al., 2000; Kati et al.,
2004b), we would have a way to circumvent this
problem.
None of the groups that we studied can adequately
represent the whole biodiversity of the area. Neverthe-
less, our results show that selection of the complemen-
tary network of any target group (scenario B) or of the
richest-in-species sites, in which it occurs (scenario A),
are the best possible approaches. Therefore, whenever
time and resources allow it, we should opt for collection
of data for at least one group of species. If we have the
possibility to collect data for two or more groups, then
the most dissimilar in ecological requirements and spa-
tial needs should be selected to better represent different
facets of biological diversity (Noss, 1990). Birds and
invertebrates could be such a pair in our study, reflecting
very different spatial needs, patterns of distribution and
ecological niches. In absence of any detailed data, ap-
plication of the principle of environmental representa-
tiveness (in our case, expressed in terms of vegetation,
scenario D) is the only available and acceptable choice.
As our results show, networks derived under scenario D
are more efficient in preserving biodiversity than the
random ones.
We should note that results concerning the scenarios’
rank depend on the sampling design and the character of
the study area; a different sampling design or a study of
a less natural area may lead to different results, e.g.,
show a weaker efficiency of scenarios A and E as com-
pared to B, C, and D (Tables 4 and 5). In our study, in
most cases, we represented each C
ORINEORINE
habitat type by
two sites (Table 1). With more sites per C
ORINEORINE
type, it
is probable that the richest-in-species sites, selected un-
der scenario A, will most probably belong to the same
C
ORINEORINE
type. In such a case, each site will add only few
new species to the network. Therefore, the cumulative
species number in network A will increase very slowly
and as a consequence the scenario efficiency will be ra-
ther low. On the contrary, the efficiency of scenarios B,
C and D will remain the same as the optimal selection
algorithm is not affected and because only one site is
selected from each of the classes representing environ-
mental variability. As for network E, with more sites per
habitat type, there would be a greater chance for similar
sites to be selected at random. If so, the average cu-
mulative number of species in the network would be low
and in consequence the same would hold true for the
scenario efficiency.
Though all non-randomly designed networks are
found to preserve more efficiently both a target group
and total biodiversity than random ones, the difference
is not always pronounced. The character of the study
area can explain this: it is a natural, well-diversified area,
with few anthropogenic disturbances, and therefore, the
sites sampled represent natural or semi-natural habitats
of different type. Any selected site is rich enough to
contribute new species in the random network.
Given the above, we can conclude that the networks’
rank for conserving one target group is G > B >
C > D > E, whereas that for the whole biodiversity of the
area is G > B > C, D > E; the exact position of network
A in the above schemes depends on the sampling design
and the character of the study area.
4.2. Combining scenarios
In our study, we evaluate and rank five different ap-
proaches as tools for designing local, small-scale reserve
networks. This, however, does not mean that under any
circumstances, the rank will remain the same, or that
those low in the rank have no value as conservation
tools. In fact, the current trend in conservation biology
is to combine many different approaches so as to meet
several criteria, in order to design operational and ef-
fective reserve systems (Belbin, 1995; Wessel et al., 1999;
Hoctor et al., 2000; Noss et al., 1999, 2002; Rodriguez
and Young, 2000).
Although the richness hotspot approach (scenario A)
was found with limited value at local scale, it has sub-
stantial value at larger scales. Identification of biodi-
versity hotspots globally, in terms of species richness,
endemism, rarity and threat is very important as it can
direct conservation efforts towards such priority areas
(Reid, 1998; Muyers et al., 2000).
The principle of complementarity (scenario B) proves
very useful in ecological networking not only in the
frame of the current, small-scale study but also at larger
scales. For instance, Howard et al. (1998) represented
biodiversity in the tropics with five biological groups
V. Kati et al. / Biological Conservation 120 (2004) 471–480 477

and proved that the complementary network of any
biological group conserves more biodiversity than a
randomly selected network. Lombard (1995) repre-
sented biodiversity in reserves of South Africa with six
vertebrate groups. Targeting every group separately, she
found that more reserves were necessary to be included
in the network for protection to be effective. These
should not be hotspots but complementary to the ex-
isting reserves.
Hotspots defined after a certain biological group very
seldom coincide with hotspots defined after another
(Prendergast et al., 1993; Lombard, 1995; Gaston and
Williams, 1996; Howard et al., 1998; Ricketts et al.,
1999). In consequence, establishing networks based on
hotspots for one group does not safeguard conservation
of the whole biodiversity of an area. The greatest ad-
vantage of the complementary reserve networks lies in
their flexibility. Given the pressure of political and so-
cio-economic factors against setting land apart for
conservation purposes, decision-makers often need al-
ternative solutions. The complementary approach sat-
isfies this need up to a point and offers a tool for
managing contradictions.
The complementary approach has the tendency to
place reserves in areas of ecological transition, where
the niches of many species overlap (Lombard, 1995). It
is debated if transition zones can be important for the
long-term persistence of biodiversity or if only the
non-transition zones can maintain viable populations
(Ara
ujo, 2002). In this context, the approach of habitat
representativeness (scenario C) can indicate represen-
tative non-transition zones for conservation. Multivar-
iate analysis is a powerful tool in conservation planning,
since it pinpoints the atypical habitats, which should
not be included in a reserve system (Belbin, 1995).
Scenario B can propose more than one optimal solu-
tions, and scenario C can select the most representative
of them.
The current study evaluated the vegetation-based
approach (scenario D) as the least efficient method for
reserve selection, second only to random choice. Nev-
ertheless, with minimum time and budgetary resources,
it is the only realistic one. It is also favored by the fact
that remote sensing techniques (satellite images, GIS,
gap analysis) have resulted in an increase of the quality
and quantity of information at the level of plant com-
munity. Also, taking into account that till now only a
very small fraction (less than 10%) of the total number
of species have been recorded (Wilson, 1992), we realize
that species-based approaches are often impracticable
(Maddock and Du Plessis, 1999). Therefore, the vege-
tation-orientated approach is a very valuable conserva-
tion tool in poorly known areas and at large
geographical scales (e.g., Keel et al., 1991; Nilsson and
G€
otmark, 1993; Strittholt and Boerner, 1995; Awimbo
et al., 1996).
4.3. Reserve design complexity
Reserve design for the long-term persistence of bio-
diversity constitutes a complex problem involving many
important parameters (Cabeza and Moilanen, 2001).
Our study dealt only partially with the problem of rare
species and did not take into consideration important
aspects related to conservation, such as edge effects,
local ecosystem processes, metapopulation analysis,
and evolution. Neither did it take into account
non-ecological factors. But in practice, conservation
decisions are informed, not dictated by science. Deci-
sion-makers have to design the shape, size, inter-dis-
tance and connectivity of the reserves (Kunin, 1997;
Shafer, 2001; Olson et al., 2002; Parks and Harcourt,
2002) under social, political, and economic constraints.
All these factors, ecological or not, should be taken into
consideration in order to ensure the long-term conser-
vation of biodiversity and maintenance of natural eco-
system processes (Margules and Pressey, 2000; Purvis
and Hector, 2000). Our study contributed to meeting the
‘‘minimal reserve set’’ requirement at a local, small-
scale. It can, therefore, provide a substantial basis for
designing reserve networks at such a scale, offering a
number of alternative best possible approaches, de-
pending on data, time, and budget availability. But as a
final remark, we should note that conservation practices
must be as dynamic as ecosystems are; everything is
changing, including the values that we assign to habitats
and taxa. Decisions concerning reserve networks should
be regularly revisited and modified accordingly to meet
newly emerging values and concomitant needs.
Acknowledgements
The first author expresses her thankfulness to Bod-
ossakis Foundation and to A. Onassis Foundation, for
supporting this research in the frame of Ph.D. scholar-
ships for biodiversity issues.
References
Ara
ujo, M.B., 2002. Biodiversity hotspots and zones of ecological
transition. Conservation Biology 16, 1662–1663.
Awimbo, J.A., Norton, D.A., Overmars, F.B., 1996. An evaluation of
representativeness for nature conservation, Hokitika ecological
district, New Zealand. Biological Conservation 75, 177–186.
Belbin, L., 1995. A multivariate approach to the selection of biological
reserves. Biodiversity and Conservation 4, 956–963.
Bezzel, E., 1980. Die brutv~
ogel bayerns und ihre biotope: versuch de
bewertung ihrer situation als grundlage f^
ur planungs- und schu-
tzma. nahmen. Anzeiger Der Ornithologischen Gesellschaft in
Bayern 19, 133–169.
Cabeza, M., Moilanen, A., 2001. Design of reserve networks and the
persistence of biodiversity. Trends in Ecology and Evolution 16,
242–248.
478 V. Kati et al. / Biological Conservation 120 (2004) 471–480

Caro, T.M., O’ Doherty, G., 1999. On the use of surrogate species in
conservation biology. Conservation Biology 13, 805–814.
Csuti, B., Polasky, S., Williams, P.H., Pressey, R.L., Camm, J.D.,
Kershaw, M., Kiester, R.A., Downs, B., Hamilton, R., Huso, M.,
Sahr, K., 1997. A comparison of reserve selection algorithms using
data on terrestrial vertebrates in Oregon. Biological Conservation
80, 83–97.
Devillers, P., Devillers-Terschuren, J., 1996. A Classification of
Palearctic Habitats. Council of Europe, Strasbourg.
Devillers, P., Devillers-Terschuren, J., Ledant, J.-P., 1991. C
ORINEORINE
Biotopes Manual – Habitats of the European Community Data
Specifications – Part2. Office for Official Publications of the
European Community, Luxembourg.
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.
Franklin, J.F., 1993. Preserving biodiversity: species, ecosystems or
landscapes? Ecological Applications 3, 202–205.
Gaston, K.J., Williams, P.H., 1996. Spatial patterns in taxonomic
diversity. In: Gaston, Kevin J. (Ed.), Biodiversity. A Biology of
Numbers and Difference. Blackwell Science, Oxford, pp. 202–229.
Hoctor, T.S., Carr, M.H., Zwick, P.D., 2000. Identifying a linked
reserve system using a regional landscape approach: the florida
ecological network. Conservation Biology 14, 984–1000.
Howard, P.C., Viskanic, P., Davenport, T.R.B., Kigenyi, F.W.,
Baltzer, M., Dickinson, C.J., Lwanga, J.S., Matthews, R.A.,
Balmford, A., 1998. Complementarity and the use of indicator
groups for reserve selection in Uganda. Nature 394, 472–475.
Kati, V., 2001. Methodological approach on assessing and optimizing
the conservation of biodiversity: a case study in dadia reserve
(Greece). Ph.D. Universit
e catholique de Louvain.
Kati, V., Dufr^
ene, M., Legakis, A., Grill, A., Lebrun, Ph., 2004a.
Conservation management for Orthoptera in the Dadia Reserve,
Greece. Biological Conservation 115, 33–44.
Kati, V., Devillers, P., Dufr^
ene, M., Legakis, A., Vokou, D., Lebrun,
Ph., 2004b. Testing the value of six taxonomic groups as biodiver-
sity indicators at local scale. Conservation Biology 18, 1–9.
Keel, S., Gentry, A.H., Spinzi, C., 1991. Using Vegetation Analysis to
Facilitate the Selection of Conservation Sites in Eastern Paraguay,
62–73.
Kirkpatrick, J.B., 1983. An iterative method for establishing priorities
for the selection of nature reserves: an example from Tasmania.
Biological Conservation 25, 127–134.
Kunin, W.E., 1997. Sample shape, spatial scale and species counts:
implications for reserve design. Biological Conservation 82, 369–
377.
Legendre, P., Anderson, M.J., 1998. Program Dist.P.Co.A. Departe-
ment de sciences biologiques, Universit
e de Montreal. Web page.
Available from <http://alize.ere.unmontreal.ca/~casgrain/en/labo/
distpcoa.html>.
Legendre, P., Legendre, L., 1998. Numerical Ecology, second English
edition, Developments in Environmental Modelling 20. Elsevier,
Amsterdam.
Legendre, P., Vaudor, A., 1991. The R Package – Multidimensional
Analysis, Spatial Analysis Departement de Sciences Biologiques.
Universite de Montreal, Montreal.
Lombard, A., 1995. The problem with multi-species conservation: do
hotspots, ideal reserves and existing reserves coincide? South
African Journal of Zoology 30, 145–163.
Maddock, A., Du Plessis, M., 1999. Can species data only be
appropriately used to conserve biodiversity? Biodiversity and
Conservation 8, 603–615.
Margules, C.R., Pressey, R.L., 2000. Systematic conservation plan-
ning. Nature 405, 243–253.
Margules, C.R., Usher, M.B., 1981. Criteria used in assessing wildlife
conservation potential: a review. Biological Conservation 21, 79–
109.
Meffe, G., Carrol, R., 1994. What is conservation biology? In: Meffe,
G., Carrol, R. (Eds.), Principles of Conservation Biology. Sinauer
Associates, Inc., USA, pp. 3–23.
Myers, N., Mittermeier, R.A., Mittermeier, C.G., da Fonseca, A.B.,
Kent, J., 2000. Biodiversity hotspots for conservation priorities.
Nature 403, 853–858.
Nillson, C., Gotmark, F., 1993. Protected areas in Sweden: is natural
variety adequately represented? Conservation Biology 6, 232–241.
Noss, R.F., 1990. Indicators for monitoring biodiversity: a hierarchical
approach. Conservation Biology 4, 355–364.
Noss, R.F., Carroll, C., Vance-Borland, K., Wuerthner, G., 2002. A
multicriteria assessment of the irreplaceability and vulnerability of
sites in the greater yellowstone ecosystem. Conservation Biology
16, 895–908.
Noss, R.F., Strittholt, J.R., Vance-Borland, K., Carroll, C., Frost, P.,
1999. A conservation plan for the klamath-siskiyou ecoregion.
Natural Areas Journal 19, 392–411.
Olson, D.M., Dinerstein, E., Powell, G.V.N., Wikramanayake, E.D.,
2002. Conservation biology for the biodiversity crisis. Conservation
Biology 16, 1–3.
Parks, S.A., Harcourt, A.H., 2002. Reserve size, local human density,
and mammalian extinctions in US protected areas. Conservation
Biology 16, 800–808.
Pienkowski, M.W., Bignal, E.M., Galbraith, C.A., McCracken, D.I.,
Stillman, R.A, Curis, D.J., 1996. A simplified classification of land-
type zones to assist the integration of biodiversity objectives in
land-use policies. Biological Conservation 75, 11–25.
Prendergast, J.R., Quinn, R.M., Lawton, J.H., Eversham, B.C.,
Gibbon, D.W., 1993. Rare species, the coincidence of diversity
hotsposts and conservation strategies. Nature 365, 335–337.
Pressey, R.L., Humphries, C.J., Margules, C.R., Vane-Wright, R.I.,
Williams, P.H, 1993. Beyond opportunism: key principles for
systematic rerserve selection. Trends in Ecology and Evolution 8,
124–128.
Pressey, R.L., Nicholls, A.O., 1989. Efficiency in conservation evalu-
ation: scoring versus iterative approaches. Biological Conservation
50, 199–218.
Pressey, R.L., Possingham, H.P., Day, J.R., 1997. Effectiveness of
alternative heuristic algorithms for identifying minimum require-
ments for conservation reserves. Biological Conservation 80, 207–
219.
Purvis, A., Hector, A., 2000. Getting the measure of biodiversity.
Nature 405, 212–219.
Reid, W.V., 1998. Biodiversity hotspots. Trends in Ecology and
Evolution 13, 275–280.
Ricketts, T.H., Dinerstein, E., Olson, D.M., Loucks, C., 1999. Who’s
where in North America? Bioscience 49, 369–381.
Rodriguez, L.O., Young, K.R., 2000. Biological diversity of
peru: determining priority areas for conservation. AMBIO 29,
329–337.
Saetersdal, M., Birks, H.J.B., 1993. Assessing the representativeness of
nature reseves using multivariate analysis: vascular plants and
breeding birds in deciduous forests, western Norway. Biological
Conservation 65, 121–132.
S.A.S. User‘s Guide: Statistics Ver.5. 1985. SAS Institute Inc., Cary,
NC, USA.
Shafer, C.L., 2001. Inter-reserve distance. Biological Conservation 100,
215–227.
Soberon, J., Rodriguez, P., Vazquez-Dominguez, 2000. Implications of
the hierarchical structure of biodiversity for the development of
ecological indicators of sustainable use. Ambio 29, 136–142.
Stoms, D.M., Borchert, M.I., Moritz, M.A., Davis, F.W., Church,
R.L., 1998. Systematic process for selecting representative research
natural areas. Natural Areas Journal 18, 338–349.
Strittholt, J.R., Boerner, R., 1995. Applying biodiversity gap analysis
in a regional nature reserve design for the Edge of Appalachia,
Ohio (USA). Conservation Biology 9, 1492–1505.
V. Kati et al. / Biological Conservation 120 (2004) 471–480 479

Vane-Wright, R.I., Humphries, C.I., Williams, P.H., 1991. What to
protect? Systematics and the agony of choice. Biological Conser-
vation 55, 235–254.
Wessels, K.J., Freitag, S., van Jaarsveld, A.S., 1999. The use of land
facets as biodiversity surrogates during reserve selection at local
scale. Biological Conservation 89, 21–38.
Williams, P.H., Gibbons, D., Margules, C., Rebelo, A., Humph-
ries, C., Pressey, R.L., 1996. A comparison of richness
hotspots, rarity hotspots, and complementary areas for con-
serving diversity of british birds. Conservation Biology 10,
155–174.
Wilson, E.O., 1992. The Diversity of Life. Norton, New York.
480 V. Kati et al. / Biological Conservation 120 (2004) 471–480