The ecological effectiveness of protected areas: a case study for South African birds

Authors

  • M. Greve,

    1. Department of Botany and Zoology, Centre for Invasion Biology, Stellenbosch University, Matieland, South Africa
    Search for more papers by this author
    • *Current address: Ecoinformatics & Biodiversity Group, Department of Biological Sciences, Aarhus University, Aarhus-C, Denmark.

  • S. L. Chown,

    1. Department of Botany and Zoology, Centre for Invasion Biology, Stellenbosch University, Matieland, South Africa
    Search for more papers by this author
  • B. J. van Rensburg,

    1. Department of Zoology and Entomology, Centre for Invasion Biology, University of Pretoria, Pretoria, South Africa
    Search for more papers by this author
  • M. Dallimer,

    1. Biodiversity and Macroecology Group, Department of Animal and Plant Sciences, University of Sheffield, Sheffield, UK
    Search for more papers by this author
  • K. J. Gaston

    1. Biodiversity and Macroecology Group, Department of Animal and Plant Sciences, University of Sheffield, Sheffield, UK
    Search for more papers by this author

  • Editor: Res Altwegg

  • Associate Editor: Nick Isaac

Correspondence
Kevin J. Gaston, Biodiversity and Macroecology Group, Department of Animal and Plant Sciences, University of Sheffield, Sheffield S10 2TN, UK Tel: 0 114 222 0030; Fax: 0 114 222 0002
Email: k.j.gaston@sheffield.ac.uk

Abstract

While the importance of individual protected areas (PAs) to biological conservation is widely acknowledged, rather few empirical studies have explicitly attempted to assess their ecological effectiveness. Significantly, this includes consideration of how well they represent the biodiversity of taxonomic groups for which the designation of these areas was not a primary or intentional goal. Here, we provide one of the most detailed comparisons to date of the avian biodiversity found inside and outside PAs, focusing on three PAs distributed widely across South Africa. Typically, bird assemblages were richer, with a higher density, and a different structural and functional composition inside than outside the PAs. Importantly, insectivore richness was much higher inside than outside, and the converse was true of granivores. Overall, these findings suggest that PAs do indeed provide valuable repositories for native biodiversity, with species richness, density and species composition being substantially different beyond their bounds. With human land-use increasing in South Africa, and habitat transformation recognized as a major and growing threat to biodiversity, such differences are expected to become greater.

Introduction

Existing protected areas (PAs) have often been criticized for insufficiencies in their coverage, biases in their distribution, and inadequacies in their design and management (e.g. Ervin, 2003a,b; Goodman, 2003; Rodrigues et al., 2004; Bonham, Sacayon & Tzi, 2008). However, while in no sense a panacea, they remain central to the majority of local, regional, and global strategies for the conservation of biodiversity. Indeed, many tens of thousands of PAs have been designated, substantial additions continue to be made, and a large proportion of the overall conservation budget is expended in establishing, maintaining, and improving them (James, Gaston & Balmford, 1999; Chape et al., 2005; Lockwood, Worboys & Kothari, 2006).

Given this heavy reliance placed on PAs, surprisingly few studies have explicitly attempted to assess their ecological effectiveness, either in terms of the representation or the maintenance of key biodiversity features (Gaston et al., 2006, 2008a,b). In part, this may result from an inherent assumption that the existence of individual PAs commonly provides a net ecological and conservation good in instances where the only alternative is a situation entirely different from a conserved state. Certainly, in many cases this is so. For example, tendencies for the last remaining patches of natural habitat in a landscape, and for the last remaining local, regional or global populations of particular species, to be entirely or largely confined to PAs support such a contention (e.g. Cowling et al., 2004; Wei et al., 2004; Gaston et al., 2008a; Jackson & Gaston, 2008). However, these situations are far from universal (with some so-called ‘paper parks’ constituting extreme counter-examples). Indeed, it is often not clear to what extent particular PAs better or differently represent biodiversity features, and especially features of high conservation concern, than do surrounding areas that are not so protected (Gaston et al., 2006, 2008a).

Significantly, this paucity of understanding includes the effectiveness of PAs in representing the biodiversity of taxonomic groups for which their designation was not a primary or intentional goal. Such ‘bycatch’ is assumed to be an important benefit of PAs, but remarkably few studies have thus far been conducted to ascertain how successfully it is achieved (Caro, 2002; Sinclair, Mduma & Arcese, 2002; Betrus, Fleishman & Blair, 2005; Thiollay, 2006; Devictor et al., 2007).

The richness and abundance of non-target species assemblages might be greater within PAs compared with their surroundings for a variety of reasons, including (i) non-randomness in where PAs were originally designated, resulting in initial conditions being more favourable compared with their surroundings; (ii) temporal improvement of conditions within PAs (e.g. from active management); (iii) temporal decline in conditions outside PAs (e.g. habitat loss and change) (Gaston et al., 2008a). These effects will often be extremely difficult to pick apart, but regardless, greater richness and abundance of non-target native taxa could reasonably be used as one broad indicator (among many possibilities; see table 1 in Gaston et al., 2006) of the ecological effectiveness of PAs. Note that we regard non-randomness in the spatial location of PAs, and differences in their land cover, that result in better representation of biodiversity features as important determinants of their effectiveness. We consider such features to be indicators of the success of conservation planning rather than confounding biases that simply need to be controlled for.

In this paper, we provide one of the most detailed comparisons to date of the biodiversity found inside and outside PAs for a taxonomic group for which the designation of these areas was not a primary or necessarily intentional goal. We focus on three PAs distributed widely across South Africa, and in two cases established principally for the conservation of large mammals (Matthews et al., 2001; Chown et al., 2003). For each, we use several different measures of their ecological effectiveness, contrasting the richness, abundance, and composition of avian assemblages in these and the surrounding areas.

Methods

Study areas

PAs and their adjacent lands were selected in three regions for this study: Jonkershoek Nature Reserve (9800 ha) in the Fynbos Biome, Karoo National Park (76 788 ha) in the semi-arid Nama-Karoo Biome (Mucina & Rutherford, 2006), and Tembe Elephant Park (30 000 ha), which includes both Savanna and Forest Biome elements (Fig. 1; Mucina & Rutherford, 2006). These three areas were selected non-randomly to obtain an understanding of the effectiveness of PAs in three very different South African regions.

Figure 1.

 Map showing the positions of the three sampling regions across South Africa. The vegetation structure of the different vegetation types in the three regions is shown in the photos. For Jonkershoek Nature Reserve and Tembe Elephant Park, the vegetation both inside (‘In’) and outside (‘Out’) protected areas is shown. In Karoo National Park, vegetation structure inside and outside protected areas was very similar.

Jonkershoek Nature Reserve (hereafter Jonkershoek, Fig. 1) was proclaimed in 1992, although it had been under the management of the Cape provincial department since the 1960s from when it was not utilized for agricultural activities. The area receives winter rainfall and the main vegetation type in the reserve is Mesic Mountain Fynbos (Moll et al., 1984; McDonald, 1985). Areas directly adjacent to the reserve have been transformed by afforestation (largely with Pinus radiata).

The vegetation of the semi-arid Nama-Karoo, including Karoo National Park (henceforth KaNP) (Fig. 1), is dominated by shrubs and grasses (Midgley & van der Heyden, 1999). Summer rainfall dominates, but is unpredictable (Cowling & Hilton-Taylor, 1999). Five vegetation types, influenced by topography, have been recognized in the KaNP and the surrounding areas: Stipagrostis ciliata Dwarf Shrubland, which occurs in lowlands; Karoo Slope Mosaic, which is found on mountain slopes; Aristida diffusa-Rhus burchelli Grassy Shrubland, which occupies the middle plateaus; and Montane Open Shrubland and Montane Dwarf Shrubland, both of which occur on the plateau of the escarp (Mucina & Rutherford, 2006; H. Bezuidenhout & S. D. Holness, unpubl. data). The KaNP was proclaimed in 1979, and further farms, used previously for stock farming, have been added to the reserve since the 1990s (South African National Parks, 2005). Areas adjacent to the park are typically used for extensive livestock agriculture, and it is thought that overstocking and the imposition of perennial grazing regimes in a highly seasonal or periodic landscape are having negative effects (Hoffman et al., 1999; Vernon, 1999).

Tembe Elephant Park (hereafter Tembe) was proclaimed in 1983, before which it was a sparsely populated communal land. The park lies in a humid summer rainfall area in the Maputaland-Pondoland-Albany biodiversity hotspot (Matthews et al., 2001). The area in and around Tembe consists of a matrix of Mixed Woodland and Sand Forest patches (Fig. 1; Matthews et al., 2001). Areas adjacent to the park are being disturbed by frequent burning, subsistence agriculture, livestock grazing, and the utilization of selected plant species for traditional medicines or wood carvings. Much of this disturbance is relatively recent and restricted in spatial scope. Inside the park, elephants are influencing Sand Forest vegetation owing to a disruption of their preferred selection of Mixed Woodland for feeding (Matthews et al., 2001; Botes, McGeoch & van Rensburg, 2006).

Sampling design

Equal numbers of sample points were selected inside and outside each PA (16 for Jonkershoek, 20 for KaNP, and 20 for Tembe). Sites within PAs had natural vegetation. Sites outside PAs were situated close (<10 km) to their borders, and had been variously affected by anthropogenic land-use changes. The sampling design was essentially a stratified one, but this varied between Jonkershoek and the other two reserves (see supporting information for full details). At Jonkershoek, the areas selected outside the reserve have been afforested with pines. Thus, the vegetation type is now very different to Mountain Fynbos, but the elevation of the sites was chosen such that in the absence of pines, the vegetation would have been Mountain Fynbos. Both here and elsewhere in the Cape Floristic Region (CFR), it has been shown that pines can quickly come to dominate Mountain Fynbos if invasions are left unchecked, so reaching situations similar to those associated with afforestation (van Wilgen, 2009). Moreover, rehabilitation projects in Jonkershoek have shown that Mountain Fynbos re-establishes if pines are removed (Holmes et al., 2000). In consequence, the sites within and outside the PA are comparable in the sense of the vegetation type that would have occupied them in the absence of human disturbance. At KaNP and at Tembe, sites within and outside the PA were selected such that their vegetation types and elevations were matched based on field assessments, vegetation maps (Matthews et al., 2001; H. Bezuidenhout & S. D. Holness, unpubl. data), and spatial datasets (Mucina & Rutherford, 2006). For example, sites were matched to include Sand Forest and Mixed Woodland at Tembe Elephant Park, and Lowland, Slopes, and the Middle Plateau at KaNP.

At Jonkershoek, only one stratum was sampled: Mountain Fynbos (Table 1). Sites sampled outside the reserve were afforested. In August/September 2005, every point was sampled on five mornings, and in March/April on four mornings. For the KaNP, points inside the reserve were only selected in areas that had been part of the reserve for more than 10 years. They were positioned in each of three vegetation classifications: ‘Lowlands’, ‘Slopes and Middle Plateaus’, and ‘Plateau’ vegetation (Table 1). Sites sampled outside the reserve were under livestock or intensive game farming (the latter being heavily overgrazed and stocking several species exotic to the region). In October 2005 and in February/March 2006, each site was sampled on one morning and one afternoon. At Tembe, sample points were selected in two vegetation types: Mixed Woodland and Sand Forest (cf. van Rensburg et al., 2000). Bird counts in Tembe were made in November/December 2005 and April/May 2006. Sites outside Tembe were, or had recently been under subsistence cultivation. During each sampling period, each point was visited on three mornings and two afternoons.

Table 1.   Summary of number and characteristics of transects sampled in and around the three protected areas
 JonkershoekKaroo National ParkTembe Elephant Park
  1. The number of transects sampled is indicated for inside and outside protected areas. For example ‘2 × 16’ indicates that there were 16 transects inside, and 16 transects outside the protected area.

Sampling methodPoint transectsLine transectsPoint transects
Total no. of transects2 × 162 × 202 × 20
Names (& no.) of transects of each of the strata(a) Mountain Fynbos (2 × 16)(a) Lowlands (2 × 10)
(b) Slopes and Middle Plateau (2 × 7)
(c) Plateau (2 × 3)
(a) Mixed Woodland (2 × 10)
Sand Forest (2 × 10)
ClimateMediterranean winter rainfallSemi-arid summer rainfallSubtropical summer rainfall
Nature of land-usage in transects outside protected areasPinus radiata plantationsCattle, sheep, and intensive game farmingSubsistence agriculture and tree clearance

Because of differences in vegetation structure and avian species richness sampling technique varied among regions, this approach precluded direct statistical comparisons being made between sites, but substantially improved the accuracy and efficiency of data collection. Line transects (KaNP) or point transects (Tembe and Jonkershoek) were conducted at each sample point (Bibby et al., 2000). In the KaNP, transects were 1 km long and detections of individual birds were recorded up to a distance of 50 m perpendicular to, and on either side of, the transect line. For point transects, the observer stood at one point for a specified time, and recorded the birds heard or seen. Birds flying through transects were not counted (cf. van Rensburg et al., 2000). The observer(s) (M. G., and, in Tembe – M. G. and Bongani Tembe) spent 2 min at each point transect before the count commenced to allow birds to become accustomed to them (Bibby et al., 2000). Each point count lasted for 7 min at Jonkershoek and 10 min at the structurally more complex and more species-rich Tembe. Count durations differed to maximize the number of species observed in the given time, while minimizing the chances of counting the same bird twice (Bibby et al., 2000).

Analyses

Sampling adequacy

Sample-based rarefaction curves were used to assess sampling adequacy (Gotelli & Colwell, 2001). Rarefaction curves were calculated using the Mau Tau moment-based interpolation method (Colwell, Mao & Chang, 2004). Sampling is considered to be adequate if the rarefaction curve approaches an asymptote (Longino, Coddington & Colwell, 2002). Analyses were conducted with EstimateS (Colwell, 2004).

Species richness and density

Species richness was calculated in EstimateS (Colwell, 2004) using Jacknife2 estimators (Magurran, 2004). Unlike several other species richness estimators, Jacknife2 does not require sample points to be compositionally similar, data to be normally distributed, or independence of species (Chao, 2004; Magurran, 2004), and the index provides conservative but accurate richness estimates (Magurran, 2004).

Jacknife2 estimates were calculated separately inside and outside each of the PAs. Two different estimates of richness are presented. The first was obtained without re-sampling and is more accurate (Colwell, 2005), although the generated data are dependent on the real data (Walther & Moore, 2005), and no variance estimates can be produced (Colwell, 2005). A second estimate was obtained using 500 randomizations and sampling with replacement. This method is more appropriate for the comparison of datasets (Colwell, 2005) and was used to compare richness inside and outside the PAs.

Bird density was calculated by dividing the mean number of birds recorded per transect by the transect area. Densities were not adjusted for detectability (Buckland et al., 2001) because detectability functions were, with one exception, monotone (supporting information).

Differences in species richness and density among sample points within and outside PAs, and between vegetation types (except in Jonkershoek, where only one vegetation type was sampled) and years (except for the KaNP dataset, where years were pooled due to small sample sizes) were assessed using general linear models in jmp v. 8.0.1. Interactions between variables were included where these were significant, and the response variable transformed where necessary to ensure that the assumptions of homogeneity of variance and normality of residuals were met (Quinn & Keough, 2002).

Assemblage composition

To investigate differences in bird species composition inside and outside the PAs, three approaches were taken. First, the proportions of species found inside a PA but not outside it, and vice versa, were calculated. Second, birds were assigned to one of six feeding guilds (frugivore, granivore, insectivore, mixed, nectarivore, predator) based on diet information provided by Hockey, Dean & Ryan (2005). For each feeding guild, the number of species was compared among the land allocations and vegetation types of each region. For both analyses, generalized linear models with a Poisson distribution and a log-link function were run in sas (PROC GENMOD). The deviance of the model was scaled to compensate for overdispersion. Percentage deviance explained was calculated by dividing the difference between the deviance of the null model (no predictors) and that of the model with predictors, by the deviance of the null model (Dobson, 2002). Analyses were not conducted for guilds that were absent or rare (Jonkershoek: predators; KaNP: predators and nectarivores; Tembe: nectarivores).

Finally, differences in bird species composition inside and outside PAs were compared using multivariate analyses implemented in primer v.5 software (Clarke & Gorley, 2001). A Bray–Curtis similarity index was used to calculate similarities in composition among assemblages (Magurran, 2004) and data were fourth-root transformed to down-weight common species (Clarke & Warwick, 1994). Non-parametric analyses of similarity (anosim; Clarke, 1993) were conducted to determine how treatments differed in species composition. Two-way crossed anosims were used to measure the contributions of inside versus outside a PA, sampling period and/or vegetation type on the composition of bird assemblages. For Jonkershoek, where only one vegetation type was sampled, protected/non-protected and sampling period were used as factors. For KaNP, data from both years were pooled, and protected/non-protected and vegetation type comprised the two factors. Global R values were used to determine the degree of similarity among treatments. The closer R is to 1, the more dissimilar species compositions of assemblages are. The composition of assemblages in protected and non-protected lands, different vegetation types, and sampling periods was plotted using non-metric multi-dimensional scaling ordinations (Clarke & Warwick, 1994). Six random restarts were used each with a different number of randomizations (10, 20, 30, 40, 50, 100) to ensure that the lowest stress value (i.e. the global optimum) was obtained (Clarke, 1993). The stress values presented in the results were obtained for all restarts.

Results

Although sample-based rarefaction curves started flattening off for all three study regions, typically they did not reach an asymptote. Raw species richness values, which were greater inside than outside Jonkershoek and KaNP and similar inside and outside Tembe (Table 2), should therefore be treated with caution. Jacknife2 species richness estimates without resampling were larger inside than outside the PAs in all three cases (Table 2). Similar results were found for the resampled data (Table 2), although the generalized linear models of species richness indicated a more complex outcome. For Jonkershoek and the Karoo, species richness was typically lower outside than inside the PA (Table 3). In the Karoo, slopes also displayed higher richness than the plateaus. In Tembe, no differences in richness were found between areas inside and outside the reserve, different vegetation types or sampling years.

Table 2.   Total abundance, recorded species richness and estimated species richness (Jacknife2, obtained without re-sampling and with re-sampling with replacement) inside (I) and outside (O) protected areas
 Jonkershoek Nature ReserveKaroo National ParkTembe Elephant Park
IOIOIO
Abundance937497122070720321827
No. of Species recorded332557499595
Jacknife2 (no re-sampling)46329162136119
Jacknife2 (with re-sampling)35.2 ± 5.126.8 ± 4.561.8 ± 9.352.9 ± 7.083.0 ± 9.377.6 ± 8.8
Table 3.   Results from general linear models comparing avian species richness (Jacknife 2) between land allocations (inside and outside protected areas), vegetation types and sampling periods in Jonkershoek Nature Reserve, the Karoo National Park, and Tembe Elephant Park
 d.f.FEffectR2
  • Richness was log-transformed for the Karoo dataset to ensure homogeneity of variance and normality in the residuals. For the Karoo dataset, samples were pooled across the 2 years due to small sample sizes.

  • *

    P<0.05.

  • ***

    P<0.001.

  • I, inside protected areas; O, outside protected areas; Slope, middle plateau and slopes; >, significantly larger; <, significantly smaller; ∼, no significant difference between parameters.

(a) Jonkershoek Nature Reserve
Model63577.54*** 0.36
Land allocation119.28***I>O 
Year114.59***2005>2006 
(b) Karoo National Park
Model399.83*** 0.45
Land allocation15.73*I>O 
Vegetation type211.89***Lowland∼Slope Lowland∼Plateau Slope>Plateau 
(c) Tembe Elephant Park
Model   0.05
Land allocation791.29  
Vegetation type11.10  
Year11.26  

In both Jonkershoek and KaNP, bird densities were higher inside than outside the PA, while no significant difference was found for Tembe (Table 4). In Jonkershoek, abundances also differed between years, although this depended on whether sites were inside or outside the PA, while in the Karoo the slopes had higher abundances than either of the other two vegetation types.

Table 4.   Results from general linear models comparing avian density between land allocations (inside and outside protected areas), vegetation types, and sampling periods in Jonkershoek Nature Reserve, the Karoo National Park, and Tembe Elephant Park
 d.f.FEffectR2
  1. Density was square root-transformed for the Jonkershoek dataset and log-transformed for the Karoo and Tembe datasets to ensure homogeneity of variance and normality in the residuals. For the Karoo dataset, samples were pooled across the 2 years due to small sample sizes.
    **P<0.01 . ***P<0.001.
    a One mixed woodland site inside the reserve was removed from the Tembe dataset for the density calculations because during two visits high numbers (>100 and >500) of migratory European swallows were found perching in the transect.
    I, inside protected areas; O, outside protected areas; >, significantly larger; <, significantly smaller; ∼, no significant difference between parameters.

(a) Jonkershoek Nature Reserve
Model6316.78*** 0.46
Land allocation125.77***I>O 
Year110.71***2005>2006 
Land allocation × year113.85***I 2005>O 2005
I 2006>O 2006
I 2005∼I 2006
O 2005>O 2006
 
(b) Karoo National Park
Model397.54*** 0.39
Land allocation111.82**I>O 
Vegetation type25.39**Lowland < Slope
Lowland∼Plateau
Slope>Plateau
 
(c) Tembe Elephant Parka
Model781.24 0.05
Land allocation12.22  
Year10.97  
Vegetation type10.51  

Assemblage composition differed substantially between areas inside and outside PAs. The first measure, based on species not shared inside and outside, indicated that in Jonkershoek 47% of the 38 species were not shared (34% found only inside, 13% only outside the PAs), in KaNP 32% of the 63 species were not shared (22% found only inside the PAs, 10% only outside), and in Tembe 42% of the 120 species found were not shared (21% in each case). Second, the species richness of feeding guilds differed significantly inside and outside the PAs (Fig. 2). In KaNP, whether sites were inside or outside the PA had the smallest influence on feeding guilds – more species of mixed feeders were observed in the PA than on surrounding farms. However, in Jonkershoek and Tembe, the number of insectivores declined outside the PAs, while granivores increased. The number of nectarivores in Jonkershoek also declined outside the PA, while mixed feeders increased outside Tembe's borders. Third, differences in the assemblages among areas inside and outside the PAs were clearly reflected in the ordinations and the significant R values from the analyses of similarity (Fig. 3). For Jonkershoek, both whether sites were inside or outside the PA and sampling period were significant, although the former had the larger effect. In KaNP, whether sites were inside or outside the PA had a relatively small though significant influence on composition by comparison with vegetation type. In Tembe, bird assemblages inside and outside the PA differed significantly. Moreover, although bird assemblages varied between the two vegetation types, this was much less pronounced outside the PA, providing further evidence for a distinct change in the avifauna.

Figure 2.

 Mean number of species per transect in different feeding guilds inside and outside (a) Jonkershoek Nature Reserve, (b) Karoo National Park, and (c) Tembe Elephant Park. Bars represent 95% confidence intervals, and significant differences between transects inside and outside protected areas, as calculated from generalized linear models, are indicated (*P<0.05, **P<0.01, ***P<0.001).

Figure 3.

 Non-metric ordination plots of bird assemblages in (a) Jonkershoek Nature Reserve, (b) Karoo National Park, and (c) Tembe Elephant Park based on land allocation, sampling period and vegetation type. (I, inside protected area; O, outside protected area; **P<0.01, ***P<0.001).

Discussion

The current PA system faces a variety of significant challenges (Rodrigues et al., 2004; Pressey et al., 2007; Gaston et al., 2008a). Alongside changes in the environmental pressures that are being exerted, these include that PAs are expected to deliver conservation objectives, which often differ markedly from the grounds on which they were originally designated and the ends to which they may have been managed for long periods. Perhaps most obviously, PAs are expected to convey benefits of representation and persistence to a broad sample of biodiversity, although a small subset may have comprised the historical focus. Given the constraints on designating new areas, and on altering or abandoning current ones, much reliance has per force to be placed on these existing areas sampling other components of biodiversity in a form of ‘bycatch’.

Many PAs in South Africa, especially the larger and older ones, were established principally for the conservation of large mammals (some carrying the names of the species they were intended to conserve, such as Mountain Zebra National Park), and many of the species are entirely or largely absent from the wider landscape (Chown, 2010). Later, the focus shifted to the conservation of vegetation types or biomes poorly represented in the network of PAs and under considerable pressure from landscape transformation (Siegfried, 1989). Here, we have demonstrated that in three very different biomes the composition of avian assemblages also differs markedly between PAs and the surrounding matrix. Although regional landscape context will have an influence on the nature and extent of the differences (e.g. Wethered & Lawes, 2005), bird assemblages were either richer, with a higher density, and/or a different structural and functional composition inside than outside PAs. Albeit the number of areas examined is small, the differences between avian assemblages inside and outside were most marked where land-use changes outside the PAs were most extreme (Jonkershoek). Nevertheless, where the vegetation structure was similar, bird abundances remained markedly lower (KaNP). Although the differences inside and outside the PAs are modest, they are nonetheless indicative of a significant relationship between formal conservation designation and assemblage structure. In principle, this could have arisen because of non-randomness in where PAs were originally designated, temporal improvement in the favourability of conditions within PAs (e.g. from active management), and/or temporal decline in the favourability of conditions outside PAs (e.g. habitat loss and change). Nonetheless, particularly given the extent of land-use change that is occurring outside of PAs (Chown et al., 2003), it seems very likely that protection has a significant influence on richness across the region. Indeed such a pattern has been observed at a substantially coarser resolution across South Africa as a whole, where, for a given level of available environmental energy, avian species richness is typically lower in regions that have less area under formal protection (Evans et al., 2006).

Consistent with the findings of some other studies (e.g. Herremans & Herremans-Tonnoeyr, 2000; Sinclair et al., 2002; Thiollay, 2006), for at least two of the PAs studied here (Jonkershoek and KaNP) the interaction with this status extends also to the densities of birds. This suggests that a greater abundance or diversity of resources is available within the PAs, or that one or more pressures on populations (e.g. predation) are reduced. In the case of birds in this particular region, the former seems the more likely. For Tembe, the extent of human impact, although increasing, remains spatially restricted (Matthews et al., 2001).

Alongside the simple differences in species richness and abundance inside and outside the PAs, species composition, and especially the occurrence of particular functional groups of species, change with protection status. Typically, insectivore richness was much higher inside than outside PAs and the converse was true of the granivores. The loss of insectivores has been widely recorded in areas where substantial landscape disturbances have taken place. For example, in the UK declines in insectivorous farmland birds have been attributed to deterioration in insect populations owing to agricultural practices (Benton et al., 2002; Newton, 2004). Elsewhere, similar losses of insectivorous species have been recorded (e.g. Fjeldså, 1999; Raman & Sukumar, 2002; Little et al., 2005; Waltert et al., 2005; Newmark, 2006), suggesting that a common outcome of heavy human use of the landscape is a decline in insect richness and/or populations (e.g. Benedick et al., 2006; Dennis et al., 2008; Braschler et al., 2010); but note that losses of insectivores can also be a consequence of other drivers (Şekercioglu et al., 2002). For two of the three PAs investigated here, other studies have demonstrated substantial declines in insect diversity. Farmland practices in the CFR have substantially lower arthropod diversity (Witt & Samways, 2004); while outside the Tembe Elephant Park, landscape disturbance has led to a fall in the diversity and abundance of dung beetles (Botes et al., 2006). In the Nama Karoo, the outcome of land management is more dependent on the particular type of extensive livestocking regime. Nonetheless, intensively utilized and transformed areas typically have reduced insect richness and abundance (Gebeyehu & Samways, 2003).

Vegetation type also affects species composition – particularly in the composition of birds in Tembe, as has been found previously (van Rensburg et al., 2000). In the Karoo, the slopes and middle plateaus display higher richness and abundance than the surrounding landscapes. This could be due to the fact that these areas display a higher diversity of habitats and may function as a transition zone between the plateau and the lowlands, possibly allowing them to support more birds (van Rensburg, Levin & Kark, 2009).

In sum, this study suggests that PAs can provide valuable repositories for non-target native biodiversity. Species richness, density and composition all differ inside and outside PAs. In South Africa, and elsewhere in Africa, human land-use is increasing, and habitat transformation is recognized as a major and growing threat to biodiversity (Driver et al., 2005; Newmark, 2008). Formally PAs are likely therefore to become an increasingly important component of regional and national conservation strategies. Given the evidence presented here of substantial differences in diversity among areas utilized by humans and those protected from such use, provisions for the human extraction of resources from within PAs (e.g. the South African National Environmental Management: Protected Areas Act) should be reconsidered (Terborgh, 1999). Moreover, how the benefits of these areas can be retained given the impacts of climate change on species abundances and distributions also requires further consideration.

Acknowledgements

For access to research sites, we thank CapeNature and Mountains to Oceans (in particular Mark Wilmot), South African National Parks and all the staff at the Karoo National Park (esp. Gerhard Pretorius), the Karoo farmers (Wynand and Rikkie Vivier, Wynand and Ruth van Rensburg, Richard Wilmot, Les Wright, and Clinton and Werner Köster), Ezemvelo KZN Wildlife and staff at Tembe (esp. Wayne Matthews) and the residents and chiefs of Bekabantu and Tshanini communities. Bongani Tembe and Suzaan Kritzinger-Klopper are thanked for assistance in the field. Sanet Hugo assisted with analyses. We are grateful to an associate editor and two anonymous reviewers for comments on the paper. K.J.G. is grateful to the Royal Society and the Leverhulme Trust for support. M.G. was partly funded by the Wilhelm Frank Trust, Stellenbosch University.

Ancillary