Ecography 36: 001–011, 2013

doi: 10.1111/j.1600-0587.2013.00388.x
© 2013 The Authors. Ecography © 2013 Nordic Society Oikos
Subject Editor: Miguel Araújo. Accepted 31 May 2013

Are protected areas maintaining bird diversity?

Laura Rayner, David B. Lindenmayer, Jeffrey T. Wood, Philip Gibbons and Adrian D. Manning­
L. Rayner (laura.rayner@anu.edu.au), D. B. Lindenmayer, J. T. Wood, P. Gibbons and A. D. Manning, Fenner School of Environment and
Society, The Australian National Univ., Canberra, ACT 0200, Australia. DBL and PG also at: ARC Centre of Excellence for Environmental
Decisions, National Environmental Research Program, The Australian National Univ., Canberra, ACT 0200, Australia.­

Evaluating the effectiveness of protected areas for sustaining biodiversity is crucial to achieving conservation
outcomes. While studies of effectiveness have improved our understanding of protected-area design and management,
few investigations ( 5%) have quantified the ecological performance of reserves for conserving species. Here, we
present an empirical evaluation of protected-area effectiveness using long-term measures of a vulnerable assemblage
of species. We compare forest and woodland bird diversity in the Australian Capital Territory over 11 yr on protected
and unprotected areas located in temperate eucalypt woodland and matched by key habitat attributes. We examine
separately the response of birds to protected areas established prior to 1995 and after 1995 when fundamental changes
were made to regional conservation policy. Bird diversity was measured in richness, occurrence of vulnerable species,
individual species trajectories and functional trait groups. We found that protected areas were effective in maintaining
woody vegetation cover in the study region, but were less effective in the protection of the target bird species
assemblage. Protected areas were less species rich than unprotected areas, with significant declines in richness across
sites protected prior to 1995. Small, specialised and vulnerable species showed stronger associations with unprotected
areas than protected areas. Our findings indicate that recently established reserves (post-1995) are performing similarly
to unprotected woodland areas in terms of maintaining woodland bird diversity, and that both of these areas are more
effective in the conservation of woodland bird populations than reserves established prior to 1995. We demonstrate
that the conservation value of protected areas is strongly influenced by the physical characteristics, as well as the
landscape context, of a given reserve and can diminish with changes in surrounding land use over time. Both protected
areas and off-reserve conservation schemes have important roles to play in securing species populations.

Conserving biodiversity through protected areas has been at improving biodiversity management efforts, and rectifying
the core of global conservation strategies for more than a failures to achieve conservation goals.
century (Pimm et al. 2001). Today, over 160 000 protected To assess the effectiveness of protected areas, studies
areas covering between 10.8 and 12.7% of the Earth’s ter- predominantly focus on one of three subjects: design, man-
restrial surface comprise the global protected-area network agement processes, or ecological integrity (sensu Ervin
(Bertzky et al. 2012, WDPA 2012). The primary objective 2003). We reviewed the empirical literature on protected-
of a protected area is ‘to achieve the long-term conserva- area effectiveness (539 studies) and found that studies of
tion of nature with associated ecosystem services’, where design, management and ecological integrity accounted for
‘conservation’ refers to ‘the in-situ maintenance of 39, 44 and 17% of articles respectively (Rayner unpubl.).
ecosystems… and of viable populations of species in their Importantly, only a small subset of studies ( 5%) directly
natural surroundings’ (Dudley 2008, pp. 8–9). Protected quantified the effectiveness of protection for sustaining
areas draw heavily on limited conservation resources in striv- biodiversity. This finding supports calls from the scientific
ing to achieve this objective (Brooks et al. 2004). Yet, community for greater research focus on the ecological effec-
the world continues to experience unprecedented levels of tiveness of protected areas to provide direct measures of con-
biodiversity loss (WWF 2012) and ongoing destruction of servation outcomes and enhance adaptive decision making
natural habitat (FAO 2011), sometimes within areas desig- (Gaston et al. 2008, Jones et al. 2011).
nated for biodiversity protection (DeFries et al. 2005). Here, we define ‘ecological effectiveness’ as the ongoing
Consequently, the capacity for protected areas to protect maintenance or recovery of biodiversity within areas implic-
and maintain biological diversity is contested (Joppa itly or explicitly established for its protection. A comprehen-
et al. 2008), bringing their effectiveness as a global conser­ sive assessment of ecological effectiveness requires, at the
vation tool into question. It is therefore essential to mea- least, comparative and temporal data. That is, the status of
sure protected-area effectiveness as the first step towards biodiversity in the presence and absence of protection should

Early View (EV): 1-EV

be compared through time. This is critical for determining 2001). Consequently, these woodlands are an internationally-
whether ecosystems are showing direct and sustained benefits recognised endangered ecoregion (Hoekstra et al. 2005). We
from protection. This approach is distinct from most avail- used a 20  40 km area in the north-east region of
able studies on protected-area design (e.g. location, cover- the Australian Capital Territory (ACT) in south-eastern
age), management (e.g. funding, enforcement) and integrity Australia (Fig. 1a) as an exemplar of this highly-modified,
(e.g. intactness, threats), which often evaluate effectiveness continent-scale ecosystem type. Prior to European settle-
using proximal measures that are assumed to be reliable sur- ment in 1824, this region was dominated by temperate euca-
rogates of biodiversity conservation, but are rarely tested lypt woodland (or temperate broad-leaf/mixed forest, sensu
(Brooks et al. 2004). Jenkins and Joppa 2009). The type of temperate woodland
Important studies of ecological effectiveness exist, but found in this area (box-gum grassy woodland) is listed
few have measured the impact of protection on species as critically endangered at the national level (ACT 2011).
populations; a core biodiversity indicator (Brooks et al. However, woodland remnants of the ACT are typically
2004, Gaston et al. 2006). For example, several studies larger, more intact, and more adequately protected than
examine the response of vegetative cover to protection over woodlands found anywhere else in Australia (ACT 2004).
time using retrospectively collected, remotely-sensed, time- These remnants make a significant contribution to the
series data (Mas 2005, Andam et al. 2008). However, global protection of this biome (Jenkins and Joppa 2009).
such methods cannot be used to track species populations, For this study, sites in protected areas were located
which rely on field-based monitoring records that are often within the Canberra Nature Park, a series of protected areas
difficult to obtain (Lindenmayer and Likens 2010). within a matrix of urban and rural land uses (Fig. 1b). These
Consequently, studies that utilise long-term monitoring protected areas contain the highest quality longitudinal
records to assess protected-area effectiveness are scarce records of birds and provide the best representation of
(Gaston et al. 2006). As a result, the capacity for protected protected temperate woodland for the region, including one
areas to ensure the persistence of species populations woodland sanctuary (www.mfgowoodlandexperiment.org.
is poorly understood, and this has been identified as a sig- au) and large areas of Eucalyptus melliodora/Eucalyptus
nificant knowledge gap in protected-area research (Gaston blakelyi Grassy Woodland, an endangered ecological com-
et al. 2008). munity in the ACT (ACT 2011).
In this paper, we studied ecological effectiveness using a
dataset for forest and woodland birds observed in temperate
eucalypt woodland within and without protected areas span- Bird surveys and site selection
ning 11 yr. We pose four primary questions: 1) do protected
areas support higher bird species richness than unprotected We sourced records of bird occurrence from a long-term
areas over time? 2) Do protected areas support more birds woodland bird monitoring project undertaken by the
of conservation concern than unprotected areas over time? Canberra Ornithologists Group (COG). The design and
3) Do protected areas support more species with stable or execution of this project was based on statistical principles
increasing trends than unprotected areas? 4) Do protected ensuring a sound data collection framework (see Bounds
areas benefit birds with particular functional traits? et al. 2010 for further information). From this database, we
Where possible, we assessed separately, the response of selected 92 study sites nested within 10 broad locations for
birds to protected areas that were established pre- and analysis (Fig. 1c). Sites were circular plots of 50 m radius. All
post-1995, following a change in legislation that re- sites were surveyed over an 11-yr period between 2000 and
prioritised reserve selection criteria in the study region. This 2010. Where possible, bird surveys were conducted season-
allowed us to account for developments in global reserve-se- ally, providing four surveys per year for each site. However,
lection standards. Further, to characterise protected-area some variation in survey effort occurred in the dataset. Bird
conditions and explore potential environmental predictors surveys were 10 min point counts recording all bird species
of species-site occurrence, we investigated the physical seen or heard within a 50 m radius of a permanent marker
characteristics (area and elevation), landscape context (rela- point (Bounds et al. 2010).
tive topographic position and urban proximity) and envi­ To compare bird responses, we located sites on both
ronmental processes (woody vegetation cover and plant protected and unprotected land (hereafter ‘reserved sites’
productivity) associated with our study sites. In doing so, and ‘unreserved sites’). All reserved sites were managed
our study demonstrates the strength of long-term ecological strictly for biodiversity protection under IUCN protection
monitoring in evaluations of protected-area performance. class I–IV. Unreserved sites were of mixed land tenures and
varied in the level of active biodiversity management, from
very little (e.g. grazed leasehold land) to substantial (e.g.
Material and methods meeting statutory obligations on defence land). Reserved
and unreserved sites were matched in three key characteris-
Study region and protected areas tics: 1) all sites occurred in either temperate grassy wood-
land (86 sites) or dry forest contiguous with temperate
Eucalypt woodland once covered over 1.5 million km2 of grassy woodland (six sites), 2) all sites were dominated
the Australian continent (NLWRA 2001), this is the equiva- by eucalypt tree species with Eucalyptus blakelyi, Eucalyptus
lent to the area of 19% of the lower 48 states of the USA. melliodora, Eucalyptus bridgesiana, Eucalyptus macroryncha
Since European settlement, almost 28% has been cleared, and Eucalyptus polyanthemos being the most common over-
and less than 8% of the remainder is protected (NLWRA storey trees, and 3) all sites were located within 25 km of the

2-EV
Figure 1. Map of study area. (a) Position of the Australian Capital Territory (ACT) within Australia. (b) Distribution of protected
areas within the ACT, showing Canberra Nature Parks (shaded grey) and all other protected areas (cross-hatched). (c) Location of sites
in pre-1995 reserves (squares), post-1995 reserves (triangles) and unreserved land (circles). Distribution of Canberra Nature Parks
(shaded grey) and urban development (grey lines) within the study region. Location of Canberra, the capital city, is denoted by a star.

city of Canberra and were likely to have been subject to some Within our sample of reserved sites, we examined sepa-
level of modification over time as a result of fire, grazing rately, the response of birds to reserves established prior to
and/or invasion from weeds and feral species. More remote 1995 and reserves established since 1995. The year 1995
woodland sites were not available due to limited habitat pro- coincides with a period of change in the motivation behind
tection and a paucity of longitudinal bird data. reserve establishment in the ACT, shifting reserve objectives

Reservation categories Table 1. Reservation status of sites. The number of sites represented
by the three reservation categories (pre-1995 reserved sites, post-
1995 reserved sites and unreserved sites) and their corresponding
The number of reserved and unreserved sites changed over years of gazettal. Note that the total number of sites is fixed (n  92).
time as 12 sites gained reservation status during the survey Twelve sites gained reservation status during the survey period
period (Table 1). This meant that, in any given year, data for (2000–2010). The total for unreserved sites is the number of sites
‘reserved sites’ included surveys from sites added recently to that were unprotected for the duration of the survey period.
the reserve system where, one could argue, the effects of Year of Pre-1995 Post-1995 Unreserved
legal protection may not have had time to manifest. Despite establishment reserved sites reserved sites sites
this caveat, we argue that recently reserved sites make an 1975 14 0 78
important contribution to this study. Theoretically, the 1987 10 0 68
addition of reserved sites in a study where the total number 1995 0 22 46
of sites is fixed, should favour the long-term conservation 2003 0 9 37
performance of reserves if representativeness is the guiding 2004 0 3 34
principle behind land acquisition. Total 24 34 34

3-EV
from those of scenic value to conservation value (ACT 1994). environmental descriptor in Supplementary material
Consequently, this provides a unique opportunity to exam- Appendix 2, Table A2.
ine the effect of changed global standards in reserve selec-
tion. Thus, three reservation categories were created for
analysis: pre-1995 reserved sites (n  24), post-1995 reserved Data structure and statistical analyses
sites (n  34), and unreserved sites (n  34) (Table 1).
In total, we used 3768 bird surveys over 11 yr in our analysis.
Bird surveys were not available for one unreserved location,
Species of conservation concern Majura Training Area (8 sites), between 2001 and 2003. All
other sites were surveyed in all years. The spatio-temporal
We assigned conservation status to each bird species based structure of the dataset is presented in Supplementary mate-
on regionally-relevant current literature. Here, species of rial Appendix 3, Fig. A3.
conservation concern included species that were declared For our analyses, we examined all species detected in
as vulnerable at a regional level (ACTFFC 2011) or have  1% of surveys, excluding waterbirds (n  60). These
exhibited a long-term declining trend in the region over the species are listed in Supplementary material Appendix 1,
last decade (Bounds et al. 2010). We present the conserva- Table A1. We pooled bird surveys annually to give three
tion status of each species in Supplementary material measures per site: 1) richness, the total number of species
Appendix 1, Table A1. detected; 2) proportion, the number of surveys in which a
species was detected as a proportion of the total number of
surveys conducted in that year; and 3) presence, the species
Species traits detected in at least one of the surveys conducted for each
year. We derived all of these measures from presence/absence
Many authors encourage the use of functional trait analysis data. For all analyses, c2 test statistics quoted are derived
in ecological studies for a deeper understanding of commu- for Wald statistics.
nity responses to environmental conditions (Devictor and
Robert 2009). This is particularly relevant where changes Trends in richness and species occurrence
to community composition might be missed by simple We fitted hierarchical generalised linear models (HGLM,
diversity measures such as richness (Filippi-Codaccioni Lee et al. 2006) to determine whether bird responses dif-
et al. 2010). We therefore assigned each species to functional fered across reservation categories over time. To do so, we
trait groups based on life-history attributes. These groups calculated longitudinal trend patterns for: 1) species rich-
included habitat specialisation (woodland dependent, ness, 2) species of conservation concern, and 3) each spe-
non-woodland dependent), bird mobility (resident and sed- cies separately. The first two analyses assumed a Poisson
entary species versus migratory, part-migratory and disper- distribution and we used the richness measure as the
sive species), body weight, nest type (e.g. hollow, cup, response, with survey effort included as a fixed effect. For
dome), nest location (e.g. arboreal, understorey, ground), individual species, we fitted quasi-binomial models with
main food type (e.g. invertebrates, seed, nectar), foraging the proportion measure as the response, accounting for
substrate (e.g. aerial, arboreal, shrub) and whether the variability in survey effort directly. For both models, we
species feeds on the ground. A species could belong to mul- included location and site as random effects to account for
tiple functional trait groups. We provide details of trait the influence of spatial autocorrelation that could result
assignment for individual species and reference material from the clustering of sites within locations. For 11 of the
in Supplementary material Appendix 1, Table A1. rarer species, there were insufficient data to estimate sepa-
rate location and site components of variance. In these
cases, we estimated the pooled variance of location and site
Environmental descriptors
combined. We fitted the HGLMs in GenStat statistical
We examined four broad-scale environmental variables for software package (14th ed.).
their relationship with area protection and species functional
traits. These were: woody vegetation cover, potential pro­ Functional trait analysis
ductivity, plant productivity, and proximity to the urban We used RLQ analysis (Doledec et al. 1996) to relate envi-
boundary. We chose these variables for two primary reasons: ronmental conditions and species functional traits to pat-
1) their documented influence on woodland bird com­ terns in species-site occurrence (using the presence measure).
munities (Chace and Walsh 2006, Huth and Possingham RLQ analysis explains variation in species composition using
2011, Montague-Drake et al. 2011); and 2) their strength and scores derived from the ordinations of three separate matri-
increasing availability as data types to be used in protected- ces: (R) site by environmental descriptors, (L) site by species,
area design, management and performance assessments (Mas and (Q) species by functional traits. We used reservation sta-
2005, Radeloff et al. 2010). We also compared these vari- tus as an environmental descriptor in the R matrix, to high-
ables, as well as reserve area, elevation and landscape posi- light the environmental conditions and species traits that are
tion, across reservation categories. In doing so, we examined: most strongly associated with reserved and unreserved sites.
1) differences in physical reserve characteristics as an indica- A challenge with RLQ analysis is dealing with both spa-
tor of biases in reserve establishment, and 2) changes in eco- tial and temporal autocorrelation within the dataset.
logical processes as an indicator of disturbance. We provide Including location and year as descriptors in the ordina-
details on source and method of data collection for each tions would mask the contribution of our environmental

4-EV
conditions of interest because: 1) location remains
unchanged through time, and 2) surveys within years are
likely to be more similar due to abiotic factors. RLQ analy-
sis does not account for such random effects. To overcome
this issue, we performed two separate analyses: one using
data collected in 2000 and the second on data collected in
2010. This approach allowed us to maximise the covariance
between site and species scores using only the environmen-
tal descriptors directly relevant to our aims. Furthermore,
this approach allowed us to compare the relative influence
of explanatory variables across two points in time (by stan-
dardising the RLQ eigenvectors for each year), while mini-
mising the variance explained by location that would be
inflated by pooling all survey years together. As we used
only two years in this analysis (2000 and 2010), reservation
categories were constrained to ‘reserved’ versus ‘unreserved’
categories.
We assessed the statistical strength of the analysis using a
permutation test (1000 permutations) and by comparing
variance explained by the RLQ analysis to separate R, L Figure 2. Results of HGLM showing changes in mean species
and Q ordinations. We conducted RLQ analysis in the R richness over time on sites reserved pre-1995, sites reserved
statistical program (R Development Core Team), using the ­post-1995 and unreserved sites. Standard errors for the model
ade4 software package. predictions, which include the random effect of location, are
shown in grey.
Environmental differences among reservation categories
We used HGLMs to examine differences in environmental
descriptors across reservation categories. We compared static category (c 23 32, p  0.001) (Fig. 3). Like total species
environmental descriptors across reserves using analysis richness, there was a significant interaction between
of variance (ANOVA) with location as a blocking variable. reserve category and time (c 23 64, p  0.001), driven by a
We calculated all static environmental descriptors as means decline in the number of species of concern on pre-
at the reserve-level. Where environmental descriptors were 1995 reserved sites (Fig. 3). In contrast, the number of
derived from time-series data (temporal environmental species of conservation concern remained stable on
descriptors), we used HGLMs to compare variation across post-1995 reserved sites and unreserved sites during the sur-
reservation categories through time. These models included vey period (Fig. 3, Supplementary material Appendix 4,
location as a random effect. We calculated all temporal Table A4).
environmental descriptors at the site-level. We performed
both ANOVA and HGLMs in GenStat statistical software
package (14th ed.).

Results
Species richness

We found a significant difference in species richness between

reservation groups (c 23 596, p  0.001) with unreserved
sites supporting higher mean species richness for most
years (Fig. 2). There also was a strong interaction between
reservation category and time (c 23 76, p  0.001) with a
decline in richness on pre-1995 reserved sites and an increase
in richness on unreserved sites (Fig. 2). While these trends
were highly significant, differences in expected (mean)
richness among reservation categories were small for most
years (Supplementary material Appendix 4, Table A4).
There was no significant change in species richness on post-
1995 reserved sites during the survey period (Fig. 2).

Figure 3. Results of HGLM showing changes in mean richness of

Number of species of conservation concern bird species of conservation concern over time on sites reserved pre-
1995, sites reserved post-1995 and unreserved sites. Standard
The number of species of conservation concern was consis- errors for the model predictions, which include the random effect
tently lower on reserved sites, regardless of year or reservation of location, are shown in grey.

5-EV
Longitudinal trends significant temporal trend was always negative, and 3) spe-
cies that showed contrasting temporal trends dependent on
Of the 60 species detected in  1% of surveys, 35 showed reservation category (Table 2).
significant temporal trends in response to our reservation 1) Increasers. Twelve species were identified as increasers
categories (Table 2, Supplementary material Appendix 4,
in this study (Table 2). Of these, four showed significant
Table A4). Temporal trends for a further six species were
increasing trends across all reservation categories. These
strong, but not significant (p  0.05–0.1), leaving 18 species
exhibiting no significant change in detection over time, species were Australian king-parrot Alisterus scapularis,
for any reservation category. An examination of the long- Australian raven Corvus coronoides, crested pigeon Ocyphaps
term trends of individual species and their responses to lophotes, and noisy miner Manorina melanocephala. Of
reservation categories revealed three types of patterns: the 12 increasers, only three were defined as woodland-
1) increasers, species for which a significant temporal dependent: Australian king-parrot, sacred kingfisher
trend was always positive, 2) decliners, species for which a Todiramphus sanctus and the white-eared honeyeater

Table 2. Results of HGLM showing individual species trends (Slope) including standard errors (SE) on pre- and post-1995 reserved sites, and
unreserved sites. Only species with strong or significant trends (p  0.1) are presented (n  42). Significance is indicated by the Wald statistic
(c2df ) and p-value as follows: *p  0.05, **p  0.01, ***p  0.001. Strong trends (p  0.05–0.10) are presented in italics. Non-significant
trends for individual reservation categories are presented in grey. Species are listed in order of detection frequency. Scientific names for
­species are provided in Supplementary material Appendix 1, Table A1.

Reservation  Pre-1995 Post-1995 Unreserved

year reserved trend reserved trend trend
Surveys
2
Common name present (%) c3 Slope SE Slope SE Slope SE
Crimson rosella 43.58 6.50 20.011 0.023 20.005 0.019 0.050 0.020
Striated pardalote 35.75 41.08*** 20.133 0.027 0.078 0.020 0.019 0.020
Weebill 34.90 61.84*** 20.129 0.025 0.075 0.022 0.102 0.021
Australian magpie 30.41 19.40*** 0.046 0.023 0.006 0.022 0.084 0.021
Grey fantail 24.04 41.31*** 20.198 0.033 20.008 0.023 20.049 0.021
Galah 23.67 50.21*** 0.086 0.026 0.008 0.029 0.145 0.023
Buff-rumped thornbill 18.90 21.96*** 20.161 0.037 0.020 0.023 0.038 0.027
Noisy miner 18.37 67.34*** 0.264 0.045 0.158 0.035 0.157 0.044
Common starling 15.61 7.67 20.123 0.050 20.053 0.044 0.014 0.027
Superb fairy-wren 15.10 43.82*** 20.228 0.035 20.072 0.059 20.025 0.024
Sulphur-crested cockatoo 14.78 19.54*** 0.108 0.031 0.010 0.027 0.079 0.030
Spotted pardalote 14.68 13.22** 20.080 0.034 0.057 0.023 0.035 0.028
Rufous whistler 13.16 14.11** 20.166 0.058 20.008 0.025 20.061 0.025
Red wattlebird 11.65 16.31*** 0.027 0.025 0.023 0.037 0.147 0.038
Pied currawong 11.44 18.41*** 0.038 0.025 20.015 0.049 0.154 0.039
Striated thornbill 11.12 18.86*** 20.165 0.064 20.051 0.027 20.089 0.030
Australian raven 10.56 29.90*** 0.035 0.033 0.041 0.038 0.152 0.029
White-throated treecreeper 10.06 26.26*** 20.251 0.053 20.051 0.028 20.020 0.034
Yellow-rumped thornbill 9.98 17.59*** 20.159 0.045 0.067 0.034 0.035 0.028
White-plumed honeyeater 8.17 10.36* 20.326 0.212 20.165 0.061 20.023 0.030
Mistletoebird 6.61 23.77*** 20.229 0.062 0.077 0.036 20.088 0.038
Magpie-lark 6.40 20.72*** 20.042 0.049 20.005 0.043 0.188 0.042
White-throated gerygone 6.24 17.53*** 20.423 0.117 20.001 0.034 20.076 0.036
White-winged chough 5.79 24.93*** 20.080 0.057 0.101 0.034 0.196 0.052
Red-rumped parrot 5.73 21.90*** 20.545 0.129 0.329 0.162 20.005 0.031
Crested pigeon 5.65 53.58*** 0.230 0.047 0.271 0.068 0.170 0.046
Common myna 5.39 68.24*** 20.319 0.045 0.162 0.119 0.201 0.049
Speckled warbler 4.64 26.20*** 20.107 0.047 0.180 0.058 0.157 0.047
Willie wagtail 3.90 7.30 20.408 0.156 0.036 0.055 20.002 0.038
White-eared honeyeater 3.77 32.25*** 20.049 0.057 0.126 0.056 0.263 0.051
Brown-headed honeyeater 3.69 8.79* 20.268 0.125 0.074 0.038 20.030 0.055
Grey shrike-thrush 3.40 8.66* 20.266 0.103 20.067 0.054 20.025 0.040
Tree martin 3.37 28.17*** 0.017 3.000 20.230 0.155 20.182 0.036
Silvereye 3.24 20.21*** 20.249 0.056 0.010 0.057 20.012 0.059
Brown thornbill 3.13 6.27 20.162 0.072 20.049 0.064 20.032 0.043
Australian king-parrot 2.81 10.29* 0.080 0.040 0.167 0.144 0.516 0.236
Golden whistler 2.71 9.34* 20.207 0.069 0.003 0.053 20.035 0.057
Leaden flycatcher 2.42 10.53* 20.262 0.089 0.075 0.056 20.023 0.069
Dusky woodswallow 1.80 7.44 20.283 0.118 20.151 0.117 20.003 0.051
Grey currawong 1.65 7.57 20.194 0.079 20.099 0.823 0.010 0.081
Sacred kingfisher 1.04 12.05** 0.049 0.163 0.163 0.077 0.302 0.110

6-EV
Lichenostomus leucotis (Supplementary material Appendix 1, Functional trait relationships
Table A1). All increasers were represented on unreserved
sites, while pre- and post-1995 reserved sites represented Both the 2000 and 2010 RLQ analyses performed best
nine and six species respectively. with a single common set of variables (permutation test
2) Decliners. Twelve species were identified as decliners in p  0.001; Table 3). These included two environmental
this study (Table 2). Of these, two showed significant descriptors: woody vegetation cover within 500 m of sites
decreasing trends across all reservation categories. These spe- and distance from the urban boundary, and three species
cies were the striated thornbill Acanthiza lineata and superb functional traits: habitat specialisation, body size and
fairy-wren Malurus cyaneus. Of the 12 decliners, 11 were whether the species feeds on the ground (Fig. 4). The first
defined as woodland-dependent species (Supplementary axes of the 2000 and 2010 RLQ analyses explained 80
material Appendix 1, Table A1). The only decliner that was and 87% of total variance in environmental conditions
not dependent on woodland was the white-plumed hone- and species functional traits across reservation, respectively.
yeater Lichenostomus pencillatus. Almost all decliners were Thus, results here are presented along a single axis of eigen-
represented on pre-1995 reserves (except the tree martin values standardised for separate years.
Hirundo nigricans) while post-1995 reserves and unreserved In both the 2000 and 2010 RLQ analyses, we observed
sites supported seven and six species, respectively. clear differences in the functional groups and environmental
3) Contrasting trends. Eleven species exhibited contrast- conditions associated with reserved versus unreserved sites
ing trends that were dependent on reservation category (Fig. 4). Reserved sites were more strongly associated with
(Table 2). The most common pattern of contrasting trends woody vegetation cover, and large, ground-feeding bird
was for a species to decline on pre-1995 reserved sites, species that were not strictly dependent on woodland habi-
while increasing on post-1995 reserved sites (n  4) or unre- tat. Unreserved sites were typically located further from the
served sites (n  1) or both (n  6) (Table 2). Such species urban boundary and more strongly associated with smaller,
included the striated pardalote Pardalotus striatus and non-ground-feeding birds that were woodland-dependent.
speckled warbler Chthonicola sagittata (Table 2). Only one While differing slightly in strength, the direction of these
species, the mistletoebird Dicaeum hirundinaceum, showed a associations remained consistent across the two time periods
decreasing trend on both pre-1995 reserved sites and unre- analysed.
served sites, while increasing on post-1995 reserved sites
(Table 2). Seven of the species listed here are dependent on
woodland (Supplementary material Appendix 1, Table A1). Environmental descriptors across reservation
categories
General trends across reservation category Static differences in environmental descriptors at the
Post-1995 reserved sites and unreserved sites showed similar reserve-level
patterns in the number of species with increasing (n  17 We found that pre-1995 reserves were significantly higher in
and 19, respectively) and decreasing (n  7 for both) trends elevation and relative topographic position than post-1995
(Table 2). Pre-1995 reserved sites showed very different reserves (Table 4). Compared to pre-1995 reserves, post-
results with nine species exhibiting an increasing trend 1995 reserves were generally larger with lower levels of
and 23 species exhibiting a decreasing trend (Table 2). Of woody vegetation cover based on the calculated reserve
these 23 decreasing species, 14 were unique to pre-1995 means. However, these results were not statistically signifi-
reserved sites, showing stable or increasing trends across cant (Table 4).
other reservation categories (Table 2). There were no declin-
ing species unique to post-1995 reserved sites or unreserved Static differences in environmental descriptors at the
sites. That is, neither of these reservation categories sup- site-level
ported species with decreasing trends that did not also Potential productivity was significantly higher on unreserved
show a decrease in another category. sites when compared to reserved sites (Table 4). This trend

Table 3. Results of RLQ analysis. Eingenvalues for the first two axes of: (a) individual ordinations of the R (environmental variables of each
site), L (species detection at sites) and Q (bird species traits) matrices, and (b) RLQ analysis, including covariance and correspondence with the
L matrix, and projected variance of the R and Q matrices. Percent variance explained by each analysis component is shown in parentheses.

2000 2010
Simulated p-value:  0.001 Simulated p-value:  0.001
Axis 1 (%) Axis 2 (%) Axis 1 (%) Axis 2 (%)
(a) Individual ordinations:
R (Hill–Smith PCA) 1.83 (60.87) 0.78 (25.85) 1.81 (60.24) 0.69 (23.09)
L (CA) 0.28 (8.93) 0.20 (6.39) 0.34 (11.75) 0.18 (6.09)
Q (Hill–Smith PCA) 1.88 (46.89) 1.02 (25.44) 1.82 (45.45) 1.03 (25.78)
(b) RLQ analysis:
RLQ axis eigenvalues 0.06 (79.58) 0.01 (20.38) 0.07 (86.60) 0.01 (13.02)
Covariance 0.24 0.12 0.26 0.10
Correlation: L 0.15 (28.41) 0.12 (25.77) 0.16 (27.79) 0.12 (27.41)
Projected variance: R 1.42 (77.59) 2.55 (98.14) 1.63 (89.96) 2.36 (94.27)
Projected variance: Q 1.77 (94.14) 2.72 (94.17) 1.55 (85.19) 2.57 (90.37)

7-EV
Figure 4. Standardised eigenvectors for axis 1 of RLQ analyses relating bird species occurrence on reserved and unreserved sites to environ-
mental variables and species traits for the years 2000 and 2010. Variables with positive standardised eigenvectors are more strongly associ-
ated with reserved sites. Variables with negative standardised eigenvectors are more strongly associated with unreserved sites. Increasing
difference from zero indicates increasing contribution to the variance explained by the analysis.

was reversed for measured plant productivity. Mean cover this trend was only significant and particularly strong for post-
of woody vegetation at the site-level was significantly higher 1995 reserved sites (Table 4).
on reserved sites than unreserved sites. There also was a
­significant difference in urban proximity across reservation
categories, with pre-1995 reserves situated nearest to the Discussion
urban boundary, followed by post-1995 reserves, and unre-
served sites situated the furthest from the urban boundary. Whether originally located on species-rich sites, or actively
managed to prevent habitat degradation, protected areas
Temporal change in environmental descriptors at the are expected to foster positive relationships with biodiver-
site-level sity (Jackson et al. 2009). In this study, we assessed the
While potential productivity (calculated from abiotic measures) ecological effectiveness of protected areas for conserving
increased across all reservation categories over time, measured bird diversity by examining the long-term response of
plant productivity decreased. Cover of woody vegetation species to the presence and absence of protection. We
decreased consistently over time and across all reservation posed four key questions to empirically test the long-term
­categories. The urban boundary encroached on all sites, but effect of protected areas on: 1) species richness, 2) species of

Table 4. Comparison of environmental conditions across reservation categories. Differences in the physical characteristics of reserves
(pre- and post-1995 reserved sites) are expressed using means with standard errors (SE). Differences and trends in environmental conditions
across reservation categories (pre- and post-1995 reserved sites and unreserved sites) are expressed using the estimate (Est.) and slope
respectively, including standard errors (SE). Significance is indicated using the Wald statistic (c2df ) and p-value as follows: *p  0.05, **p  0.01,
***p  0.001. Non-significant relationships are presented in grey. Details on source and method of data collection for each environmental
covariate are provided in Supplementary material Appendix 2, Table A2.

Pre-1995 Post-1995
Environmental measure Significance reserved sites reserved sites Unreserved sites
Static data (reserve category) F1,5 Mean SE Mean SE – –
Reserve area (ha) 0.93 378.30 114.40 546.70 132.10 – –
Reserve elevation (m) 7.06* 705.04 12.08 656.01 13.95 – –
Reserve landscape position 26.14** 21.19 0.16 0.06 0.18 – –
Site woody cover 2.75 0.38 0.10 0.12 0.12 – –
Pooled temporal data (reserve category) c23 Est. SE Est. SE Est. SE
Potential productivity 2226*** 1.772 0.054 1.730 0.058 1.849 0.060
Plant productivity 2796*** 1.530 0.039 1.494 0.040 1.481 0.045
Woody cover (within 500 m) 107*** 3.525 0.407 2.852 0.413 2.386 0.483
Distance to urban boundary 867*** 6.679 0.328 7.022 0.321 7.759 0.393
Temporal data (reserve category  year) c23 Slope SE Slope SE Slope SE
Potential productivity 23*** 0.010 0.005 0.015 0.004 0.008 0.004
Plant productivity 244*** 20.019 0.002 20.020 0.002 20.012 0.002
Woody cover (within 500 m) 176*** 20.059 0.008 20.091 0.009 20.060 0.014
Distance to urban boundary 418*** 20.009 0.012 20.087 0.004 20.007 0.004

8-EV
conservation concern, 3) species persistence and 4) species related to the abundance of hyper-aggressor Manorina
functional groups. melanocephala; a species linked to declines in avian diver-
We found that, compared to unprotected areas, protected sity and abundance (MacDonald and Kirkpatrick 2003).
areas collectively supported: 1) lower species richness, In our study, Manorina melanocephala exhibited highest
2) fewer species of conservation concern, 3) more species detection and strongest increasing trends within reserves
with declining trends, and 4) larger-bodied, generalist spe- (Supplementary material Appendix 4, Table A4) and
cies. By these measures, we can conclude that unprotected could be driving localised species declines. This example
areas sampled in this study are more effective in maintaining demonstrates that while the maintenance of vegetation
woodland bird diversity than protected areas within our characteristics within protected areas may provide broader
study region. This is a striking and counter-intuitive result landscape functions for biota (e.g. connectivity), additional
which we explore within the physical and ecological context on-site management actions (e.g. population control of
of protected areas below. Specifically, we draw attention to interspecific competitors) may be required to secure vulner-
three key findings. able species populations.
1) Reserve placement influenced ecological effectiveness 3) Urban encroachment threatens ecological effective-
over time. Our results indicated that the long-term response ness. In this study, smaller woodland-dependent species were
of birds to protected areas was strongly influenced by the associated with unprotected sites situated furthest from
period of protected-area establishment. Long-established urban boundaries (Fig. 4, Table 4). While studies have shown
protected areas may be limited in their capacity to meet positive relationships between human population density
conservation objectives due to reserve design that was and avian richness (due to a mutually positive response to
not ecologically-driven. For example, creating reserves to primary productivity, Luck 2010), such relationships are
protect scenic values (Margules and Pressey 2000) and estab- highly scale-dependent, with more localised effects tending
lishing reserves on less productive land (Joppa and Pfaff to be negative (Pautasso 2007). Indeed, the direct effects
2009). Indeed, we found that reserves established before of urban encroachment on protected areas and their associ-
1995 in this study were significantly higher in elevation, ated biodiversity are often negative (Radeloff et al. 2010).
higher in topographic position, had lower potential pro­ The mechanisms which underpin the negative relationship
ductivity and, on average, were smaller in size than post- between small, woodland-dependent species and urban
1995 reserves (although this last finding was not statistically proximity identified in this study are unclear, but may be
significant). associated with habitat or species composition changes
In 1994, an amendment to the Nature Conservation Act within reserves close to the urban fringe (Ikin et al. 2012).
1980 in the ACT (ACT 1994) introduced a statutory This is of particular concern as urban encroachment is
requirement to conserve endangered ecological communi- advancing rapidly toward the best-performing reserves in
ties. Since that time, the characteristics of newer reserves our study region (Table 4). Here, buffering protected
have changed (Table 4) and our analyses indicated that areas from the impacts of urban development will become
their ecological effectiveness has improved. This was exem- increasingly important as urban areas expand and opportu-
plified by increasing species richness, stabilisation of species nities for establishing future reserve sites contract (Ewers and
populations of conservation concern and fewer individual Rodrigues 2008).
species declines, particularly of less common specialist spe- Together, our findings demonstrate that protected areas
cies, on post-1995 reserves when compared to pre-1995 are dynamic systems, exhibiting their own temporal and
reserves. Hence, this study demonstrated that overcoming spatial response to environmental gradients. We have shown
traditional biases in reserve placement and re-prioritising that the conservation performance of protected areas can
designation objectives can enhance the ecological effective- diminish over time with changes in landscape context. It is
ness of protected areas. also likely that protected area effectiveness will be influenced
2) Reserves protect habitat and ecological processes. by increasing environmental pressures associated with
Despite an overall decline in woody vegetation cover across climate change (Hole et al. 2011, Araujo et al. 2011, Bagchi
study sites (Table 4), our reserved sites supported signifi- et al. 2013). For example, survey data for this study were
cantly higher vegetative cover and productivity than collected during a period of severe drought in Australia
unreserved sites (Fig. 3, Table 4). Other studies have found (2001–2009) and one could suggest that reserve perfor-
protected areas to be effective in representing and main­ mance may improve during years of higher rainfall. How­
taining vegetation cover (Andam et al. 2008) and plant pro- ever, projected climate changes include increased frequency
ductivity (Tang et al. 2011). Such outcomes demonstrate and severity of drought for our study region, indicating
that, where enforcement is adequate, protected areas can that the results presented here may provide a good indication
play a critical role in preserving habitat and maintaining of future trends. Further research is needed to quantify the
ecological processes through legislative controls of destruc- effect of weather on bird distributions in this region, and to
tive processes, such as land clearing, and can therefore determine whether protected areas are likely to facilitate
be associated with enhanced protected-area performance (Thomas et al. 2012) or inhibit (Araujo et al. 2011) species’
(Stoll-Kleemann and Job 2008). range expansions.
However, this may not strictly be the case for all wood- Given that protected area performance may fluctuate
land birds. For example, some argue that vegetation thick- through space and time, studies of ecological effectiveness that
ening favours the more specialised woodland bird species incorporate comparative and temporal data are better equipped
(Kutt and Martin 2010), but Montague-Drake et al. (2011) to track changes in protected systems relative to un-protected
found that overstorey cover and productivity were positively systems, identify the processes threatening ­protected-area

9-EV
­ erformance, and assess the success (or otherwise) of manage-
p data are needed to complement broader environmental data
ment interventions addressing those processes. and that, where biodiversity data are lacking, assumptions
Our results illustrate how studies that lack comparative of species persistence within protected areas should be made
data fail to examine the direct effect of protection on biodi- with great caution until such data are gathered.
versity and, therefore, are limited in their assessments of
­species representativeness. Similarly, studies that lack tempo- ­­­­­­
­­Acknowledgements – This research received funding support from
ral data fail to examine the sustained effect of protection on Conservation Planning and Research, Environment and Sustainable
biodiversity and therefore, will be limited in their assessments Development Directorate ACT Government and the Fenner School
of species persistence. Importantly, these two measures, repre- of Environment and Society. We thank the Canberra Ornithologists
sentativeness and persistence, are widely recognised as pri- Group (COG) for providing bird records, J. Kesteven and the
mary indicators of protected-area performance (Margules National Carbon Accounting Service (NCAS) for calculating fpi
and Pressey 2000) and should be addressed wherever possible data, and B. Mackey and colleagues for calcu­lating fPAR data. We
also thank K. Ikin, C. Sato for improve­ments to the manuscript; and
in assessments of ecological effectiveness. M. Evans and J. Stein for valuable contributions.

Conservation implications References

Our study joins a substantial body of work acknowledging [ACT] Australian Capital Territory Government 1994.
the benefits and importance of protected areas, while Australian Capital Territory Legislative Assembly: Nature
also identifying significant shortfalls in their conservation Conservation (Amendment) Bill 1994, Explanatory
performance (Vellak et al. 2009, Fandohan et al. 2011). We Memorandum. –  www.legislation.act.gov.au/es/db_14972/
believe our study has taken a first step in addressing these 19940616-16965/pdf/db_14972.pdf  accessed 7 May 2013.
shortfalls by quantifying the outcomes of past conservation [ACT] Australian Capital Territory Government 2004. Woodlands
efforts and establishing a contemporary ecological context to for Wildlife: ACT Lowland Woodland Conservation Strategy.
Action Plan No. 27. –  www.environment.act.gov.au/data/
inform future systematic planning (sensu Hockings et al. assets/pdf_file/0010/234487/actionplan27t1.pdf  accessed
2004). The similar, and sometimes superior, performance of 21 February 2013.
unprotected areas in comparison to protected areas was [ACT] Australian Capital Territory Government 2011. ACT State
unexpected, but not unique to this study region (e.g. of the Environment Report 2011. –  www.envcomm.act.gov.
China, Liu et al. 2001; Canada, Kharouba and Kerr 2010; au/actsoe2011/index.html  accessed 21 February 2013.
Europe, Araujo et al. 2011). Such findings suggest that the [ACTFFC] Australian Capital Territory Flora and Fauna Commit-
protection of modified habitats situated more favourably in tee 2011. Thirteenth Annual Report: 2010–2011. –  www.
the landscape (in terms of surrounding land-use threats and environment.act.gov.au/cpr/conservation_and_ecological_
communities/act_flora_and_fauna_committee  accessed 21
changing climate space) could maximise conservation bene- February 2013.
fits to a region, particularly where active restoration is Andam, K. S. et al. 2008. Measuring the effectiveness of protected
feasible. Perhaps more importantly, it highlights that options area networks in reducing deforestation. – Proc. Natl Acad.
for conservation are not limited to protected areas, and that Sci. USA 105: 16089–16094.
there is great potential for biodiversity gains through invest- Araujo, M. B. et al. 2011. Climate change threatens European
ments in off-reserve conservation schemes (Lindenmayer conservation areas. – Ecol. Lett. 14: 484–492.
and Franklin 2002, Cox and Underwood 2011). Bagchi, R. et al. 2013. Evaluating the effectiveness of conservation
The protected-area portfolio in the ACT has 99% of all site networks under climate change: accounting for uncer-
tainty. – Global Change Biol. 19: 1236–1248.
ecosystems represented to the minimum global conserva- Bertzky, B. et al. 2012. Protected Planet Report 2012: tracking
tion target of 10% (Taylor et al. 2011). But does this mean progress towards global targets for protected areas. – IUCN,
that it is ecologically effective for conserving biodiversity? Gland, Switzerland and UNEP-WCMC, Cambridge, UK.
Our study showed that the protection of ecosystems has Bounds, J. et al. 2010. A statistical analysis of trends in occupancy
been effective in the maintenance of some ecological pat- rates of woodland birds in the ACT, December 1998
terns and processes (i.e. vegetative cover and productivity), to December 2008: the ten-year data analysis. – Canberra
but has not been effective in ensuring the persistence of a Bird Notes 35: 158–192.
vulnerable assemblage of species (i.e. woodland birds). Brooks, T. M. et al. 2004. Protected areas and species. – Conserv.
Biol. 18: 616–618.
Thus, a key lesson from this study is that the use of perfor- Chace, J. F. and Walsh, J. J. 2006. Urban effects on native avifauna:
mance measures that are not necessarily proxies for ecologi- a review. – Landscape Urban Plann. 74: 46–69.
cal effectiveness (e.g. habitat classes, Brooks et al. 2004; or Cox, R. L. and Underwood, E. C. 2011. The importance of con-
areal coverage, Gaston et al. 2008) can severely limit our serving biodiversity outside of protected areas in Mediterranean
capacity to understand and describe the actual performance ecosystems. – PLoS One 6: e14508.
of protected areas for biodiversity, particularly at the DeFries, R. et al. 2005. Increasing isolation of protected areas
resolution needed for the on-ground adaptive management in tropical forests over the past twenty years. – Ecol. Appl.
of species populations. This is not to say that direct biodi- 15: 19–26.
Devictor, V. and Robert, A. 2009. Measuring community responses
versity data should replace generalised environmental vari- to large-scale disturbance in conservation biogeography.
ables in evaluations of protected area effectiveness. Indeed, – Divers. Distrib. 15: 122–130.
monitoring a potentially unrepresentative assemblage of Doledec, S. et al. 1996. Matching species traits to environmental
species will not, in itself, provide a comprehensive assess- variables: a new three-table ordination method. – Environ.
ment of performance. Rather, we argue that species-specific Ecol. Stat. 3: 143–66.

10-EV
Dudley, N. 2008. Guidelines for applying protected area manage- Lindenmayer, D. B. and Likens, G. E. 2010. The science and
ment categories. – IUCN, Gland, Switzerland. application of ecological monitoring. – Biol. Conserv. 143:
Ervin, J. 2003. Protected area assessments in perspective. 1317–1328.
– Bioscience 53: 819–822. Liu, J. G. et al. 2001. Ecological degradation in protected areas:
Ewers, R. M. and Rodrigues, A. S. L. 2008. Estimates of reserve the case of Wolong Nature Reserve for giant pandas. – Science
effectiveness are confounded by leakage. – Trends Ecol. 292: 98–101.
Evol. 23: 113–116. Luck, G. W. 2010. Why is species richness often higher in more
Fandohan, B. et al. 2011. Effectiveness of a protected areas network densely populated regions? – Anim. Conserv. 13: 442–443.
in the conservation of Tamarindus indica (Leguminosea- MacDonald, M. A. and Kirkpatrick, J. B. 2003. Explaining bird
Caesalpinioideae) in Benin. – Afr. J. Ecol. 49: 40–50. species composition and richness in eucalypt-dominated rem-
[FAO] Food and Agriculture Organization of the United Nations nants in subhumid Tasmania. – J. Biogeogr. 30: 1415–1426.
2011. State of the world’s forests 2011. –  www.fao.org/docrep/ Margules, C. R. and Pressey, R. L. 2000. Systematic conservation
013/i2000e/i2000e00.htm  accessed 21 February 2013. planning. – Nature 405: 243–253.
Filippi-Codaccioni, O. et al. 2010. Toward more concern for Mas, J. F. 2005. Assessing protected area effectiveness using
specialisation and less for species diversity in conserving surrounding (buffer) areas environmentally similar to the target
farmland biodiversity. – Biol. Conserv. 143: 1493–1500. area. – Environ. Monit. Assess. 105: 69–80.
Gaston, K. J. et al. 2006. The ecological effectiveness of protected Montague-Drake, R. M. et al. 2011. A reverse keystone
areas: The United Kingdom. – Biol. Conserv. 132: 76–87. species affects the landscape distribution of woodland
Gaston, K. J. et al. 2008. The ecological performance of protected avifauna: a case study using the noisy miner (Manorina
areas. – Annu. Rev. Ecol. Evol. Syst. 39: 93–113. melanocephala) and other Australian birds. – Landscape Ecol.
Hockings, M. et al. 2004. Management effectiveness: assessing 26: 1383–1394.
management of protected areas? – J. Environ. Pol. Plann. [NLWRA] National Land and Water Resources Audit 2001.
6: 157–174. Australian Native Vegetation Assessment 2001. – Canberra.
Hoekstra, J. M. et al. 2005. Confronting a biome crisis: global Pautasso, M. 2007. Scale dependence of the correlation between
­disparities of habitat loss and protection. – Ecol. Lett. 8: 23–29. human population presence and vertebrate and plant species
Hole, D. G. et al. 2011. Toward a management framework for richness. – Ecol. Lett. 10: 16–24.
networks of protected areas in the face of climate change. – Pimm, S. L. et al. 2001. Environment – can we defy nature’s
Conserv. Biol. 25: 305–315. end? – Science 293: 2207–2208.
Huth, N. and Possingham, H. P. 2011. Basic ecological theory Radeloff, V. C. et al. 2010. Housing growth in and near United
can inform habitat restoration for woodland birds. – J. Appl. States protected areas limits their conservation value. – Proc.
Ecol. 48: 293–300. Natl Acad. Sci. USA 107: 940–945.
Ikin, K. et al. 2012. The influence of native versus exotic streetscape Stoll-Kleemann, S. and Job, H. 2008. The relevance of effective
vegetation on the spatial distribution of birds in suburbs protected areas for biodiversity conservation: an introduction.
and reserves. – Divers. Distrib. 19: 294–306. – Gaia-Ecol. Perspect. Sci. Soc. 17: 86–89.
Jackson, S. F. et al. 2009. Statutory protected areas and avian species Tang, Z. Y. et al. 2011. Effectiveness of protected areas in maintain-
richness in Britain. – Biodivers. Conserv. 18: 2143–2151. ing plant production. – PLoS One 6: e19116.
Jenkins, C. N. and Joppa, L. 2009. Expansion of the global terrestrial Taylor, M. F. J. et al. 2011. Building Nature’s Safety Net 2011:
protected area system. – Biol. Conserv. 142: 2166–2174. the state of protected areas for Australia’s ecosystems and
Jones, J. P. G. et al. 2011. The why, what, and how of global wildlife. –  http://awsassets.wwf.org.au/downloads/bi037_
biodiversity indicators beyond the 2010 target. – Conserv. building_natures_safety_net_2011_6jul11.pdf  accessed 21
Biol. 25: 450–457. February 2013.
Joppa, L. N. and Pfaff, A. 2009. High and far: biases in the location Thomas, C. D. et al. 2012. Protected areas facilitate species’
of protected areas. – PLoS One 4: e8273. range expansions. – Proc. Natl Acad. Sci. USA 109:
Joppa, L. N. et al. 2008. On the protection of “protected areas”. 14063–14068.
– Proc. Natl Acad. Sci. USA 105: 6673–6678. Vellak, A. et al. 2009. Past and present effectiveness of protected
Kharouba, H. M. and Kerr, J. T. 2010. Just passing through: areas for conservation of naturally and anthropogenically
global change and the conservation of biodiversity in protected rare plant species. – Conserv. Biol. 23: 750–757.
areas. – Biol. Conserv. 143: 1094–1101. [WDPA] World Database on Protected Areas 2012. –  www.
Kutt, A. S. and Martin, T. G. 2010. Bird foraging height predicts unep-wcmc.org/world-database-on-protected-areas-wdpa_76.
bird species response to woody vegetation change. – Biodivers. html  accessed 7 May 2013.
Conserv. 19: 2247–2262. [WWF] World Wide Fund for Nature 2012. Living Planet Report
Lee, Y. et al. 2006. Generalized linear models with random effects: 2012: biodiversity, biocapacity and better choices. – Gland,
unified analysis via H-likelihood. – Chapman and Hall/CRC. Switzerland,  http://awsassets.panda.org/downloads/1_lpr_
Lindenmayer, D. B. and Franklin, J. 2002. Conserving forest 2012_online_full_size_single_pages_final_120516.pdf 
biodiversity. – Island Press. accessed 21 February 2013.

Supplementary material (Appendix ECOG-00388 at

 www.oikosoffice.lu.se/appendix ). Appendix 1–4.

11-EV