A forgotten habitat? Granite inselbergs conserve reptile diversity in fragmented agricultural landscapes


*Correspondence author. E-mail: david.lindenmayer@anu.edu.au


  • 1Rocky outcrop ecosystems support unique biological communities, high levels of species endemism and are important in the conservation of biodiversity worldwide. Some rocky ecosystems occur in fragmented landscapes, and as such, play a key role in conserving reptile biodiversity in modified environments.
  • 2We present a case study of reptile diversity in granite landforms from south-eastern Australia, using a conceptual framework based on landscape ecology theory. We stratified inselbergs by landform and assessed the relevance of patch size, matrix, habitat complexity and hierarchy theories in explaining reptile responses. Regression modelling was used to relate species richness, abundance and diversity to theory and habitat variables.
  • 3We found all theories to be generally applicable in interpreting reptile responses in this system but certain habitat attributes needed to be measured carefully to accurately predict reptile responses. We found that reptile species richness and diversity were congruent with predictions of patch size (island biogeography theory) and habitat structure (complexity theory), although both concepts were confounded by landform. Matrix condition had a significant influence on reptile diversity with low predicted values in relictual landscapes.
  • 4At the outcrop patch-level, reptile diversity was negatively related to exotic grass cover, stem density, vegetation structure and grazing intensity, whereas native grass cover and total rock cover increased diversity.
  • 5Synthesis and applications. The conservation of rock-dwelling reptiles in fragmented agricultural landscapes worldwide can be guided by concepts based on landscape ecology and will involve strategic management of ‘inselberg landscapes’, by addressing issues relevant to both the outcrop and surrounding matrix. Thus, matrix (landscape-level) management should focus on maintaining maximum habitat heterogeneity, whereas outcrop (patch-level) management will require controlling grazing regimes, invasive weeds and woody regrowth, thereby maintaining solar infiltration levels necessary for reptile thermoregulation.


Rocky outcrops have a profound influence on the distribution and abundance of biodiversity worldwide (Anderson, Fralish & Baskin 1999; Porembski & Barthlott 2000). Such environments are well documented as being biological hotspots, and often support unique biotic communities and high levels of endemism (Poremski, Seine & Barthlott 1997). The importance of rocky environments is highlighted by the discovery of new endemic species. For example, eight endemic plant species were recently described from a dolomite ‘glade’ formation in Alabama (Allison & Stevens 2001). However, despite the apparent attention some rock outcrops have received, along with cliff face environments, these geological landforms rank among the most poorly surveyed ecosystems in the world (Larson, Matthes & Kelly 2000).

Granite inselbergs are one regularly occurring geological feature found in most vegetation types and climate zones throughout the world. They form isolated, insular habitats and are exposed at over 15% of all continental areas (Twidale & Romani 2005). Such a widespread distribution may explain why saxicolous species contribute substantially to the faunas of different continents (Howard & Hailey 1999). In the Serengeti grasslands, granite ‘kopjes’ support many endemic species, including reptiles (Trager & Mistry 2003), and in Africa and southern America, granite outcrops have been an important selective force in the evolution of mammal communities (Mares 1997). In Australia, granite inselbergs also have high conservation value. For example, in Western Australia granite inselbergs support a diverse range of specialized taxa, including approximately 2000 species of plants (Hopper, Brown & Marchant 1997), over 80 species of aquatic invertebrates (Bayly 1997), at least 30 species of mammals (Morris 2000), and at least 17 species of reptiles (Withers & Edward 1997; Wilson & Swan 2008). However, few studies have quantified the role granite outcrops play in conserving reptile diversity. This is surprising, considering inselbergs have been the subject of extensive botanical research in recent years (Hopper et al. 1997; Porembski & Barthlott 2000 and references therein).

Quantitative data on the distribution and abundance of species are essential in mitigating the impacts of human activities on biodiversity (Lindenmayer & Burgman 2005; Krebs 2008). This is applicable to granite landforms, which occur in human-modified landscapes throughout the world and are quarried for building and landscape gardening industries or utilized for intensive recreational activities, such as rock-climbing (Twidale 2000). Hence, the lack of ecological knowledge on biota associated with granite landforms has global implications for biodiversity conservation, but particularly in Australia, which contains more reptile taxa than any other continent and where endemic species are still being described from rocky environments (Horner 2007; Wilson & Swan 2008). It is evident from the literature that reptiles respond poorly to habitat fragmentation (Brown & Bennett 1995; Brown 2001; Mac Nally & Brown 2001; Driscoll 2004; Cunningham et al. 2007). However, despite the growing literature on the effects of habitat fragmentation on reptiles, no study has applied landscape theory to explain patterns of reptile diversity in granite landforms. Just how important are granite outcrops in conserving reptile diversity in modified landscapes and what are the factors responsible for influencing species diversity? Here, the cumulative effects emerging from a naturally patchy ecosystem occurring within human-induced fragmented landscapes is of primary interest.

We used a conceptual framework based on landscape ecology theory to investigate reptile responses in fragmented agricultural landscapes. Specifically, we investigate: (i) patch size theory, where patch size influences species richness and diversity (MacArthur & Wilson 1967; Rosenzweig 1995), (ii) matrix theory, where the type and condition of the surrounding landscape influences species richness and diversity within patches (Ricketts 2001; Lindenmayer & Franklin 2002), (iii) complexity theory, where habitat structural complexity influences species richness and diversity (MacArthur & MacArthur 1961), and (iv) hierarchy theory, where both local and landscape factors influence species distributions and diversity (Allen & Starr 1988; Mackey & Lindenmayer 2001). To explore the relevance of these theories in explaining patterns of reptile diversity, we measured a range of explanatory variables relating to: (i) landscape-level factors (e.g. geographical location, isolation and surrounding matrix condition), (ii) patch-scale factors (e.g. geomorphology, vegetation type, patch size, grazing intensity and outcrop structural complexity), and (iii) plot-scale attributes (e.g. percentage cover of vegetation strata, floristic groups, surface rock and fallen timber, etc.).

Based on these theories, we hypothesized that large, structurally complex outcrops in high-quality landscapes will support the greatest reptile diversity. However, relatively small, structurally simple outcrops in highly modified landscapes also may have important conservation value for some species. Although there was little evidence to suggest any of the outcrops in the study area had been damaged by reptile collectors, we hypothesized that other human-induced disturbances such as livestock grazing will have negative effects on reptile diversity. We believe that our work will provide much-needed new information on the ecological role of granite outcrops and will help land managers in developing strategic methods for managing granite landforms for reptile conservation in fragmented agricultural landscapes.


Over much of Australia, granite inselbergs are analogous to African kopjes as they are insular and surrounded by grassland plains. However, inselberg insularity in Australia has been exacerbated by excessive modification to the matrix through the development of crops and open pastures (Lindenmayer, Crane & Michael 2005). Small outcrops (less than 5 ha in size), are often overgrazed, cleared of native vegetation and harbour pest animals, such as the European rabbit Oryctolargus cuniculus (Mawson 2000) and invasive plants (Pigott 2000). In contrast, larger outcrops are generally less affected by human-induced activities and as such, are in better condition.

study area

We studied an area within the South-western Slopes biogeographical region of New South Wales bordered by the towns of Tarcutta (0556007E, 6078043N) in the north, Albury (0494981E, 6008873N) in the south, Holbrook (0528452E, 6047058N) in the east and Walbundrie (0463118E, 6046968N) in the west. Intrusive granite formations are a dominant component of the regions’ geology (Twidale 2007). The predominant vegetation type on granite inselbergs in the region is temperate woodland (sensu Hobbs & Yates 2000); dominated by white box Eucalyptus albens, Blakely's red gum E. blakelyi, yellow box E. melliodora, red stringybark E. macrorhyncha and long-leaved box E. goniocalyx. Other outcrops with different soil profiles support communities dominated by currawang Acacia doratoxylon, tumbledown red gum Eucalyptus dealbata and white cypress pine Callitris glaucophylla.

study design

We surveyed reptiles in 44 granite inselbergs located within 40 grazing landscapes and four areas managed for nature conservation. We surveyed along a gradient of patch sizes, ranging from small (< 1 ha, N = 7), medium (1–10 ha, N = 32) and large (> 10 ha, N = 5). The mean and median outcrop patch sizes were 11 and 4 ha, respectively, and the percentage cover of vegetation ranged from zero to 45%.

We classified inselbergs based on geomorphology (Campbell 1997; Twidale & Romani 2005) and recognized these four landform types (Fig. 1): (i) bornhardts – domical hills arising abruptly from the surrounding landscape, characterized by domes and expanses of exposed bedrock; (ii) nubbins – conical hills rising gently from the surrounding landscape, characterized by densely fractured jumbled rock piles; (iii) castle koppies – moderately fractured intrusions in gently undulating landscapes, characterized by orthogonally jointed pillars and domes; (iv) Corestone boulders (tors) – spatially dispersed rock masses in flat or gently undulating landscapes, characterized by fractured boulders, pillars and few surface rocks. Nubbins, castle koppies and corestone boulders are derived landforms representing varying stages in the erosion process of the original bornhardt (Twidale & Romani 2005). To sample all landforms in the study area, multiple inselbergs were surveyed on individual properties.

Figure 1.

Granite inselberg landforms surveyed in the South-western Slopes of NSW; (a) bornhardt, (b) nubbin, (c) castle koppie, and (d) corestone boulders.

survey protocol

We actively searched for reptiles between October 2006 and February 2007 on clear days above 25 °C and between 1000 and 1500 hours using an area-constrained (up to 4-ha grid centred over the outcrop) and time-constrained (2 h) protocol. For outcrops smaller than 4 ha, the interfacing landscape was not surveyed. We inspected all available habitats within each site including: beneath surface rocks, fallen timber, behind bark and within rock crevices. We did not damage rock exfoliations. Rather, where possible, crevices were inspected by reflecting sunlight off a mirror or with the aid of a torch. We also recorded reptiles found in the open either basking or foraging. Species were identified visually due to the difficulty of installing pitfall traps, noosing or hand-capturing animals. We believe the sampling method was adequate for detecting the majority of species in the focal area, although a few fossorial (Typhlopidae) species and nocturnal (Gekkonidae) individuals may have gone undetected.

measurement of covariates

To assess habitat suitability for reptiles, we recorded a range of categorical and continuous explanatory variables at multiple spatial scales. Landscape-scale attributes included distance to other outcrops and landscape context (i.e. whether outcrops were part of a range or formed isolated ‘satellites’). We classified outcrops in four landscape contrasts (sensu McIntyre & Hobbs 1999): (i) isolated in a ‘relictual’ matrix, that is, landscapes containing less than 2% of the original overstorey vegetation and supporting crops or annual pasture; (ii) isolated in a ‘fragmented’ matrix, that is, landscapes containing less than 10% of the original overstorey vegetation and supporting a mixture of native and exotic pasture; (ii) connected in a ‘variegated’ matrix, that is, landscapes containing abundant scattered paddock trees, roadside vegetation corridors or tree plantings, and (ii) connected in ‘continuous’ native vegetation. These were assessed visually from a GIS layer radiating 200 m and 500 m from the outcrop/matrix interface.

At the (outcrop) patch scale, we measured a range of variables (Table 1). Structural complexity was assessed visually by estimating rock structural classes, based on volume, to the nearest 5%. The structural classes were: (i) low-lying expanses of embedded rock, termed ‘pavements’; (ii) objects less than 0·5 m3, (corresponding with the kinds of objects that could feasibly be rolled over), termed ‘rocks’; (iii) objects ranging between 0·5 m3 to 2 m3, termed ‘boulders’; (iv) ‘pillars’ ranged between 2 m3 to 5 m3; (v) ‘blocks’ ranged between 5  to 10 m3; and (vi) ‘domes’ exceeded 10 m3. We tallied structural classes for each outcrop and calculated a complexity score ranging from one – simple outcrop (dominated by one or two structural classes) to five – complex outcrop (outcrops containing combinations of five or six structural classes). Mid-range values corresponded to outcrops with varying combinations of between two and five structural classes.

Table 1.  Outcrop patch-scale variables used in linear regression models to explain overall reptile abundance, species richness and diversity on granite inselbergs in the South-western Slopes of New South Wales. *See text for detailed descriptions
Explanatory variableVariable typeDescription
Abiotic variables
 Elevation (metres above sea level)ContinuousThe highest point on the outcrop obtained from a Global Positioning System
 AspectCategoricalMeasured using a compass
 SlopeContinuousMeasured using a clinometer
 Landform typeCategoricalCorestone, nubbin, koppie, bornhardt
 Outcrop complexity*CategoricalScore ranging from 1 to 5
 Outcrop patch size (ha)ContinuousCalculated from a GIS aerial layer
 Rock class cover*ContinuousVisually estimated to the nearest 5%
Biotic variables
 Vegetation typeCategoricalEucalypt woodland, Acacia doratoxylon, Callitris sp. woodland
 Vegetation structureCategoricalOld growth, regrowth, including seedling regrowth and multi-stemmed regrowth, cleared
 Vegetation remnant patch size (ha) ContinuousCalculated from a GIS aerial layer.
 Grazing intensityCategoricalHeavy – set stocking; light – seasonal or rotational grazing; none – no livestock grazing

At the plot-scale, we randomly established a 100 × 100 m (1 ha) grid and visually estimated such continuous variables as: (i) percentage cover of overstorey and understorey vegetation layers, (ii) percentage ground cover (e.g. exotic/native forbs, exotic/native grass, leaf litter, fine woody debris, bare earth, lichen and rock cover), (iii) large logs/fallen trees (density and estimate of volume), and (iv) the number of trees, tree stems, shrubs, dead trees and tree stumps.

statistical analysis

We explored relationships between explanatory variables and overall abundance, species richness and diversity using Generalized Linear Models (GLMs) (McCullagh & Nelder 1989). For count data, we used Poisson regression and linear logistic regression for presence or absence data (rarer species). We accounted for extra variation due to clustering in count data by estimating a dispersion parameter and using variance ratio statistics for inference. We assessed dependence between observations within farms by fitting area (farm) as a random effect in a linear mixed model framework. The resulting correlations were low (0·26 for species abundance, 0·3 for species richness and 0·28 for species diversity), and therefore, we dealt with random effects only in generalized linear regression models. We rescaled explanatory variables by transforming the original data by taking natural logarithms where necessary. We constructed models first by fitting explanatory variables separately, ignoring all other variables, and secondly, by jointly considering all likely candidate variables. In the case of multiple regressions, we selected ‘best’ models using Akaike and Schwartz Information Criteria (AIC and SIC, respectively) for all possible regressions.

To measure species diversity, we used two indices; the Shannon–Wiener diversity (Shannon index) and the Simpson diversity index. When indices were congruent, we reported the most significant value.

In Shannon's diversity index, inline image log Pi, where Pi is the proportion of species ‘i’. For Simpson's index, inline image, where the parameters were the same as described for the Shannon–Wiener index.


species richness and abundance

We detected a total of 699 individuals from 44 inselbergs, representing 12 species from five families (Table 2). Two outcrops contained no species. The mean number of reptiles per site was 2·89 species (range = 0–9, mean density = 0·72 species ha−1) and overall mean abundance was 15·89 individuals per site (range = 0–35, mean density = 3·97 individuals ha−1). Four species accounted for over 93% of all observations (Table 2).

Table 2.  Species composition, site ubiquity (number of sites present) and species (relative) abundance of reptiles recorded from 44 granite inselberg in the South-western Slopes of New South Wales. (Figures in parentheses indicate percentage of total sites and total abundance; nomenclature follows Wilson & Swan 2008)
FamilySpeciesSite ubiquity (%)Total no. of individuals (%)
GekkonidaeMarbled gecko Christinus marmoratus7 (15·9)11 (1·57)
Eastern stone gecko Diplodactylus vittatus5 (11·4)5 (0·72)
PygopodidaeOlive legless lizard Delma inornata1 (2·3)1 (0·14)
ScincidaeWall skink Cryptoblepharus carnabyi15 (34·1)48 (6·87)
Southern rainbow skink Carlia tetradactyla10 (22·7)15 (2·15)
Large striped skink Ctenotus robustus26 (59·1)80 (11·44)
Cunningham's skink Egernia cunninghami2 (4·5)2 (0·28)
Crevice skink Egernia striolata36 (81·8)480 (68·67)
South-eastern slider Lerista bougainvillii1 (2·3)1 (0·14)
Boulenger's skink Morethia boulengeri15 (34·1)45 (6·44)
PythonidaeCarpet python Morelia spilota metcalfei2 (4·5)2 (0·28)
ElapidaeEastern brown snake Pseudonaja textilis7 (15·9)9 (1·29)
Total abundance44699

patch size effects

We found overall mean abundance was not related to either outcrop area (P = 0·13) or remnant vegetation area (P = 0·84). However, area of both the outcrop and remnant vegetation was positively related to species richness and diversity (P < 0·001, Fig. 2). Rock outcrop and remnant vegetation patch area differed in relation to landform. Bornhardts had significantly greater mean outcrop area (57 ± 25·23 ha, range = 2–200 ha) and remnant vegetation (176·52 ± 18·23, range = 1·4–498 ha) than derived landforms (P < 0·001). Patch areas did not differ significantly between the three derived landforms, although corestone boulders contained the least amount of outcrop (mean = 3·16 ± 6·12 ha, range = 0·2–15 ha) and remnant vegetation (mean = 8·82 ± 4·33 ha, range = 0–87·7 ha).

Figure 2.

Significant relationships between outcrop and remnant vegetation patch size (area in ha) and species richness and diversity. (Dsw, Shannon-Wiener diversity index, and diamonds represent actual observations).

matrix effects

We identified a significant relationship between the landscape context within a 200-m radius of an inselberg and reptile species richness (P < 0·05, Fig. 3) and diversity (P < 0·001, Fig. 3), but not overall mean abundance (P = 0·76). Likewise, landscape context within 500 m of an inselberg had a significant effect on mean overall abundance and diversity (P < 0·05, Fig. 3). The general trend in predicted values decreased in response to declining matrix condition within 200 m of an inselberg, with lower values observed from landforms situated in a ‘relictual’ matrix, compared with landforms situated within ‘variegated’ or continuous landscapes. We identified a similar pattern at a spatial scale of 500 m, although landforms in connected landscapes had lower overall mean abundance than other matrix types (Fig. 3).

Figure 3.

Significant relationships between landscape context within a 200-m and 500-m radius of the inselberg / matrix interface and overall abundance, species richness and diversity. (Dsw, Shannon-Wiener diversity index; Ds, Simpson diversity index, vertical bars represent two standard errors).

habitat complexity effects

We compared habitat structure among inselbergs based on outcrop complexity and vegetation structure. Outcrop structural complexity had a significant effect on overall abundance (P < 0·001, Fig. 4) and diversity (P < 0·05, Fig. 4), but was weakly related to species richness (P = 0·051). We found overall abundance increased with structural complexity (Fig. 4). Castle koppies were the most structurally complex landforms, followed by bornhardts. On average, corestone boulders were the least complex landform, although they spanned the entire complexity spectrum. Landform type had no significant effect on species richness (P = 0·46) or diversity (P = 0·12), although overall abundance varied significantly among inselbergs (P = 0·014) with more individuals observed from koppies and corestone boulders (Fig. 4).

Figure 4.

Significant relationships between outcrop structural complexity and granite inselberg landform and predicted mean species richness, overall abundance and diversity. (Ds, Simpson diversity index and vertical bars represent two standard errors).

Vegetation structure had no significant effect on species richness (P = 0·26), overall abundance (P = 0·53) or diversity (P = 0·38). All three vegetation structural classes were similar, although mean species richness was highest in dense regrowth sites (3·32 ± 0·59 species) and lowest in cleared sites (2·42 ± 0·44 species). In contrast, overall abundance was highest in cleared sites (17·18 ± 2·47 individuals) and lowest in regrowth sites (13·33 ± 3·26 individuals).

multi-spatial scale effects

Using general linear models, we found that species richness, abundance and diversity responded significantly to geographical context (P < 0·05). Isolated inselbergs supported lower reptile diversity than inselbergs connected to other rocky areas along a mountain range. In addition, a number of vegetation-related variables were related to species richness, abundance and diversity, whereas the climate and topographic variables were unrelated (e.g. elevation, slope and aspect). Vegetation community had a positive effect on species richness (P = 0·005, Fig. 5). Eucalypt woodland accounted for over 80% of all sites and contained the maximum number of 12 species, although predicted mean species richness was higher in C. glaucophylla and A. doratoxylonE. dealbata-dominated communities (Fig. 5). Native grass cover did not significantly influence species richness (P = 0·78) but had a positive affect on overall abundance (P < 0·05). In contrast, exotic grass cover decreased species richness (P < 0·05, Fig. 5), overall abundance (P < 0·05) and diversity (P < 0·05). No other variables, except percentage rock cover, related to overall abundance. Total abundance, species richness and diversity significantly (P < 0·05) increased with increasing rock cover, although only overall abundance related significantly to the rock structural class ‘pillars’ (P < 0·001). Grazing intensity had a significant negative effect on species richness (P < 0·001, Fig. 6) and diversity (P < 0·001) but not on overall abundance (P = 0·42). Higher species diversity was observed in inselbergs which were ungrazed or lightly (seasonally) grazed by domestic livestock compared to outcrops experiencing sustained (set stocking) grazing.

Figure 5.

Significant relationships between vegetation community, grazing intensity and exotic grass cover and predicted mean species richness and overall abundance. (Vertical bars represent two standard errors).

Figure 6.

Most parsimonious composite habitat model for overall reptile abundance. Predicted values are based on the fixed values (provided in parenthesis) of composite variates (vertical bars represent two standard errors).

composite models

After considering significant variables separately, we constructed the most parsimonious models from these sets of habitat measures. Many multi-scale variables that were significant when fitted individually in linear models became non-significant in the composite models. The most parsimonious mixed model explaining 56% of the variation detected in overall abundance included positive relationships with increasing outcrop structural complexity and negative relationships with increasing tree stem density and percentage cover of exotic grass (Fig. 6). Sixty-six per cent of the variation detected in species richness was explained by the absence of livestock grazing, increasing outcrop patch size and old-growth remnant vegetation (Fig. 7).

Figure 7.

Most parsimonious composite habitat model for species richness. Predicted values are based on the fixed values (provided in parenthesis) of composite variates (vertical bars represent two standard errors).


We used landscape ecology theory, traditionally applied to fragmented remnant vegetation, in our study of insular granite inselbergs. We examined hypotheses relating to patch size, matrix, habitat complexity and hierarchy theories and assessed their relevance in explaining reptile responses in a naturally patchy ecosystem. We found these theories to be broadly relevant in interpreting reptile responses in granite inselbergs.

species richness and abundance

Our species list (Table 2), represents approximately 50% of all documented species in the study area (Michael 2004; Cunningham et al. 2007). Many of these species however, are unlikely to use granite outcrops due to affiliations with other habitats, including grasslands or wetlands. From previous research, we know that inselbergs provide habitat for threatened species, such as Aprasia parapulchella (Michael 2004), and declining species, including Morelia spilota metcalfei (Michael & Lindenmayer 2008). Now, we have quantified the role that small rocky outcrops play in conserving a range of terrestrial and saxicolous species in agricultural landscapes (Table 2). Ninety-five per cent of the outcrops surveyed contained reptile species, suggesting novel management approaches may be needed to maintain maximum reptile diversity on granite inselbergs in modified landscapes.

outcrop and vegetation patch size effects

Patch size is a central part of the island biogeography theory (MacArthur & Wilson 1967), and has been applied in landscape ecology to interpret patterns of biodiversity in human-induced fragmented landscapes (Haila 2002). Implicit in the fragmentation model are a number of assumptions including: (i) habitat patches are perceived to be comparable to oceanic islands, (ii) the area outside the patch is considered inhospitable for most organisms, and (iii) natural pre-fragmentation conditions were uniform (Haila 2002; Lindenmayer & Fisher 2006). Many studies have found this model to be ambiguous, often producing multi-directional responses between particular species and limited in its ability to accurately forecast reptile diversity (Hadden & Westbrooke 1996; Smith et al. 1996; Mac Nally & Brown 2001; Jellinek et al. 2004). This is because such assumptions are not readily transferable to human-modified landscapes. For example, fragmented remnant vegetation may lack resources that are important for reptiles such as fallen timber (Mac Nally et al. 2001), or contain an abundance of others, particularly rocky outcrops on hilltop remnants. Such spatial disparity in resources may produce negative patch size responses evident in other studies. This, however, was unlikely in our study, as all patches contained abundant rock cover. Therefore, island biogeography theory may have some application in the interpretation of diversity patterns in granite and other geological landforms. We found reptile diversity was related to both outcrop and remnant native vegetation area (Fig. 2), a relationship comparable with oceanic islands, albeit at a much finer spatial scale. Therefore, by protecting the largest outcrops, maximum reptile diversity can be maintained. However, this relationship should be interpreted with caution as we found patch size to be confounded by landform (and associated structural complexity), thus making their individual effects difficult to disentangle.

matrix effects

Matrix theory recognizes that the area surrounding ‘patches’ of vegetation (or in this case insular rock outcrops), is important to some species and can influence patch occupancy (Ricketts 2001; Haila 2002; Lindenmayer & Franklin 2002). For example, the presence of paddock trees and tree plantings contribute significantly to landscape function, connectivity and the habitat value of remnant vegetation for some species of reptiles (Cunningham et al. 2007). We found reptile diversity to be greater in continuous or variegated landscapes than in relictual landscapes (Fig. 3). Scattered paddock trees, roadside vegetation corridors and linear tree plantings all play an important role in ameliorating the effects of land clearing by reducing isolation of remnant vegetation (Driscoll 2004) and increasing landscape connectivity (Fisher et al. 2005). This finding has implications on how vegetation types are managed surrounding inselbergs worldwide and implies that although granite outcrops are insular systems, they often contain a high proportion of matrix-derived species (Watson 2002). Therefore, the management of granite outcrops should be considered in the context of ‘inselberg landscapes’, that is, by addressing issues pertinent to both the outcrop and the surrounding matrix.

habitat complexity effects

MacArthur & MacArthur (1961) predicted that bird species diversity would increase with increasing foliage height diversity and hence increasing habitat complexity, although many studies have since found the relationship to be multi-directional (see August 1983). We found reptile diversity was positively related to outcrop structural complexity but not to vegetation structural classes (Fig. 4). However, structural complexity was highly correlated with landform. The relationship between outcrop structural complexity and landform is complex, being dependent on attributes developed during plutonic formation (e.g. mineral composition and fracture density), level of subsurface weathering, and erosion following scarp retreat (Twidale & Romani 2005). Hence, landform significantly influenced overall abundance (Fig. 4). Cornerstone boulders and koppies contained more crevices than other landforms, providing abundant resources for crevice-dwelling species. Therefore, reptile diversity will be enhanced by maintaining maximum structural complexity in the landscape via managing an array of complex landforms in a given area.

spatial-scale effects

Hierarchy theory predicts that factors at a range of spatial scales will influence the distribution and diversity of most organisms (Allen & Starr 1988). We found that few landscape-level variables (except geographical context) had a significant effect on any level of biological organization. Instead, species diversity was strongly associated with patch and plot-level attributes, including vegetation community, grazing intensity and grass cover (Fig. 5).

We found species richness to be greater on inselbergs dominated by C. glaucophylla or A. doratoxylon and lowest in eucalypt woodland (Fig. 5). Eucalypt woodland occurs on productive loam soils, and hence, was one of the first vegetation types cleared for agriculture in the region (Benson 1999). In contrast, C. glaucophylla and A. doratoxylon communities grow on sandy or course-grained granitic soils (Stelling 1998), and hence, were spared from extensive clearing. Historical impoverishment may therefore account for differences in species diversity. However, some C. glaucophylla stands have been logged in the past and have subsequently regenerated, forming dense stands. This has created additional problems involving reptile thermoregulation.

We found greater reptile diversity on outcrops that were not grazed or were seasonally grazed, than on outcrops that were set stocked (Fig. 5). The non-significant relationship with overall abundance was confounded by high densities of E. striolata on corestone boulders. Grazing pressure can have a detrimental effect on reptiles (Busack & Bury 1974; Bock, Smith & Bock 1990) by reducing structural complexity and altering plant species composition (Hadden & Westbrook 1996; Brown 2001). We therefore propose that grazing regimes on granite outcrops will need to be strategically controlled to promote native vegetation regeneration and suppress invasive, agrestral plants.

We found that rock and native grass cover significantly influenced species richness, overall abundance and diversity, and exotic grass cover decreased them (Fig. 5). Exotic plants can often be a surrogate for other forms of disturbance such as grazing intensity and fertilizer use. Jellinek, Driscoll & Kirkpatrick (2004), Hadden & Westbrook (1996) and Smith et al. (1996) all found lizard species richness to be negatively associated with exotic plant species. The invasibility of granite outcrops by introduced plant species depends on a number of factors, including landform, outcrop size and exotic species composition in the surrounding landscape (Porembski 2000). In this study, low-lying, spatially dispersed outcrops, such as corestone boulders supported more agrestral species than large, structurally complex outcrops such as bornhardts and koppies (D. R. Michael et al., unpublished data), suggesting weed control will be dependant on and guided by inselberg landform.

composite models

Both stem density and vegetation structure (dense regrowth), were found to be significantly negatively related to species richness and overall abundance (Figs 6 and 7). We suggest this result was likely to be associated with increased shading effects resulting from canopy closure, a process which can alter thermal conditions, and hence, affect thermoregulatory behaviour (Pringle, Webb & Shine 2003; Webb, Shine & Pringle 2005). A study in Kansas (USA) found grazing removal, and the subsequent succession in vegetation cover and habitat suitability, caused the decline and local extinction of over 70% of the reptile fauna over a 50-year period (Fitch 2006). In addition, Pringle et al. (2003) found that nocturnal reptiles showed strong negative responses to vegetation structure, being absent from shaded rock outcrops. Duncan et al. (2008) found prescribed burning to be a useful tool in reducing woody intrusion on a dolomite outcrop ecosystem in North America. Therefore, ecological thinning or reintroducing fire regimes may be necessary on densely vegetated inselbergs to maintain solar radiation levels required for reptile thermoregulation (Huey 1991; Webb et al. 2005).

management implications

Future management of granite outcrops and other rock types in modified landscapes will need to focus on entire ‘inselberg landscapes’ (i.e. by incorporating both the rock outcrop and surrounding matrix). Landscape-level management should focus on protecting large, structurally complex outcrops and enhancing the condition of the matrix by maintaining habitat heterogeneity. Patch-level management will require the strategic control of domestic livestock grazing, invasive exotic plants and dense regrowth, thereby maintaining solar infiltration gradients necessary for reptile thermoregulation. In many cases, fencing outcrops to exclude grazers, reintroducing fire regimes or conducting ecological thinning experiments could help create conditions conducive to reptile conservation.


We have quantified the role granite inselbergs play in conserving reptile diversity in fragmented, agricultural landscapes. We have also illustrated that conceptual landscape ecology theories, traditionally applied to fragmented remnant vegetation, can contribute to a broader understanding of reptile diversity patterns. These concepts, which include patch size, matrix condition and habitat structural complexity, are useful in studying the biota associated with otherwise often overlooked rock-dominated environments such as granite inselbergs. We believe that additional investigations are required to disentangle the effects of patch size, habitat complexity and landform. Furthermore, additional experiments are needed to investigate the influence of canopy-shading on the ecology of saxicolous reptiles, and to further draw attention to the importance of such ‘forgotten’ habitats in the future persistence of reptiles in modified landscapes worldwide.


We gratefully acknowledge the numerous landholders in the region who allowed us access to the amazing granite landforms on their properties. We also thank a number of volunteers for assisting with aspects of fieldwork, namely, Hugh MacGregor, Greg Slade and Nigel Jones. D.M. would like to thank his family for their generous support over the many years and Simon Thirgood, Reed Noss and an anonymous referee who made insightful comments on an earlier version.