• conservation biology;
  • fragmentation;
  • hierarchical experimental design;
  • landscape models;
  • variegated landscapes


  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  • 1
    Understanding the ecological effects of processes operating at multiple spatial scales on multiple species is a key challenge in ecology. It underpins both basic research and the increasing recognition of scale-dependence in conservation biology.
  • 2
    Organisms are affected by ecological processes operating at multiple spatial scales. Hence the study of multiple species and spatial scales is a key challenge in ecology.
  • 3
    A spatially nested experimental design was used to investigate the habitat relationships of reptiles in a grazing landscape in southern New South Wales, Australia. Regression modelling was used to relate the presence of species and species richness to habitat variables.
  • 4
    Different species were predicted by different habitat variables, and some species reflected habitat structure at one, but not all, spatial scales. The four-fingered skink Carlia tetradactyla was associated with box woodlands, areas with few rocks and many spiders, and areas with a northerly aspect and high tree cover. Boulenger's skink Morethia boulengeri was more likely to be found in areas with many ants and beetles, and also in areas with a northerly aspect and high tree cover. The striped skink Ctenotus robustus and olive legless lizard Delma inornata were significantly more likely to be detected in areas with a simple microhabitat structure. Species richness was highest in box woodlands (Eucalyptus albens and Eucalyptus melliodora) and at sites characterized by a high variability of habitat structure.
  • 5
    Different habitat variables varied over different spatial scales. For example, invertebrate abundance varied mostly over tens of metres, while grass/forb cover varied mostly over hundreds to thousands of metres.
  • 6
    Synthesis and applications. The multitude of species’ responses and spatial habitat variability highlighted the potential limitations of the fragmentation paradigm because it may oversimplify ecological complexity. On this basis, conservation in variegated grazing landscapes may need to consider changes to livestock management across entire landscapes, rather than concentrate solely on the preservation of particular patches.


  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Studying multiple species and spatial scales is a key challenge in applied ecology (Meentemeyer & Box 1987; MacNally et al. 2002). Because organisms respond to environmental conditions at a range of spatial scales (Forman 1964), multi-scaled studies can provide important insights for conservation management. For example, Lindenmayer (2000) demonstrated that Leadbeater's possum Gymnobelideus leadbeateri (McCoy) in south-eastern Australia responded to its environment at scales ranging from continental to the individual tree hollow chosen for nesting, and highlighted that no single study at any arbitrarily chosen scale could have delivered the same set of conservation recommendations as the multi-scaled approach employed. Similarly, it is important to consider the habitat requirements of multiple species simultaneously when addressing conservation issues. A recent study by MacNally et al. (2002) demonstrated that biodiversity surrogate schemes based on few or single taxa often were unable to encompass the needs of a wide range of organisms, because different species and faunal assemblages responded differently to a given set of environmental conditions.

We present a study on multiple species of reptiles at multiple spatial scales in a grazing landscape of south-eastern Australia. Compared with birds and mammals, reptiles are neglected in conservation biology (MacNally & Brown 2001). However, they are an interesting group because of their fundamentally different autoecology and their ectothermy as a key life-history trait (Pough 1980; Heatwole & Taylor 1987). There is little quantitative information on the habitat requirements of most reptiles in south-eastern Australia. Much information is based on anecdotal evidence or field guides (Jenkins & Bartell 1980; Bennett 1997; Cogger 2000), or is concerned with forest-dwelling reptiles (Lunney & Barker 1986; Lunney, Eby & O’Connell 1991; Brown & Nicholls 1993; Kutt 1993; Goldingay, Daly & Lemckert 1996) or species at mining sites (Halliger 1993; Letnic & Fox 1997). There are few habitat studies in grazed woodlands (but see Dorrough & Ash 1999; Brown 2001), which can support remnants of threatened woodland communities under pressure from grazing and land clearance (Prober & Thiele 1995; Gibbons & Boak 2002). Grazing can pose a significant threat to ground-dwelling reptiles in Australia (Sadlier & Pressey 1994; Hadden & Westbrooke 1996; Lemckert 1998) and other parts of the world (Busack & Bury 1974; Bock, Smith & Bock 1990).

Three specific questions were addressed in our study. (i) Which structural, thermal and food-related factors best describe the distribution of reptiles at landscape and microhabitat scales? (ii) At what spatial scale do these factors vary, and what are the consequences for the spatial distribution of reptiles? (iii) Which factors are related to reptile species richness?


  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

study site

Field work was conducted on three private properties near Jugiong in southern New South Wales, Australia (34°58′S, 148°29′E). The study site was characterized by undulating topography ranging between 300 and 700 m above sea level. Prior to European settlement, most of the area was covered by Eucalyptus woodlands [Blakely's red gum/yellow box Eucalyptus blakelyi (Maiden)/Eucalyptus melliodora (A. Cunn., ex. Schauer); white box Eucalyptus albens (Benth.); red stringybark Eucalyptus macrorhyncha (F. Muell., ex. Benth.); minor occurrences of apple box Eucalyptus bridgesiana (R. T. Bak) and red box Eucalyptus polyanthemos (Schauer) and the native fruit tree kurrajong Brachychiton populneus (Schott)]. Approximately 85% of the original tree cover had been cleared, and the study area was grazed by sheep Ovis aries (L.) and cattle Bos taurus (L.). The remaining 15% of remnant woodland cover was generally in small patches (mostly < 15 ha) or semi-isolated trees or clumps of trees in the pastures. Box woodlands are a threatened tree community in Australia (Yates & Hobbs 1997). On the basis of tree cover, the study area could have been classified as variegated (sensuMcIntyre & Hobbs 1999) because discrete boundaries between woodland patches and grazing pastures were uncommon. Understorey vegetation was rare due to grazing, and most grasses and forbs were introduced species.

experimental design

The study was based on a hierarchical experimental design to assess possible ecological effects at multiple spatial scales simultaneously (Meentemeyer & Box 1987). Three levels of experimental unit were specified within the hierarchy: (i) landscape units (n = 16), (ii) sites (n = 48) and (iii) plots (n = 144; Fig. 1). Landscape units were chosen on the basis of their aspect (northerly or southerly), topographic position (ridge or valley) and tree cover (high or low, i.e. above or below average in the study area). There were two replicates for each treatment combination (e.g. 2 × ‘north & ridge & high tree cover’, etc.). This stratification encompassed a wide range of conditions that may influence reptile occurrence, such as temperature and vegetation structure (Heatwole & Taylor 1987). A landscape unit was an equilateral triangle with a side length of 250 m. Depending on the topographic position of the landscape unit, one side of the triangle followed the topographic contour lines of the ridge or valley, respectively. The third corner point was located at the mid-slope (Fig. 1). The spatial extent of landscape units was chosen to match the spatial grain of the underlying topographic variability in the landscape, i.e. the typical distance from valley or ridge, respectively, to the mid-slope was approximately 250 m.


Figure 1. Summary of the hierarchical experimental design used to survey reptiles. The example represents a landscape unit with (i) aspect = northerly and (ii) topographic position = ridge. Canopy cover was also used to stratify sites, and could be either ‘low’ or ‘high’. Three treatments (aspect, topographic position, canopy cover) with two levels each resulted in 23= 8 treatment combinations at the landscape unit level. There were two replicates of each treatment combination. Hence, there were 16 landscape units, 16 × 3 = 48 sites, 48 × 3 = 144 plots, and 144 × 2 = 288 pitfall traps.

Download figure to PowerPoint

Three sites were located at every corner point of each landscape unit (16 landscape units × 3 = 48 sites). Thus, while there were two possible topographic positions at the landscape unit level (ridge vs. valley), three topographic positions were identifiable at the site level (ridge, valley and mid-slope), and each topographic position was represented 16 times at the site level (Fig. 1). Within each site, three plots measuring 10 × 10 m were located in a triangle, with their centres spaced approximately 25 m apart (48 sites × 3 = 144 plots). The distance between plots was chosen to match the spatial grain of small lizards in the study area (sensuWiens 1989; Kotliar & Wiens 1990); most movements of small lizards occurred within a diameter of 20 m (Turner, Jennrich & Weintraub 1969; James 1991). Each plot contained a plastic drift fence (10 m × 30 cm) and a pitfall trap (9·6-l plastic bucket) at either end of the fence (Fig. 1). This plot size was considered to be a meaningful area to measure microhabitat variables in relation to small reptiles.

survey protocol

The aim of the survey was to record the majority of reptile species inhabiting the plots. Two repeat surveys were conducted: one in late spring 2001 (November–December), and one in late summer 2002 (February–March). These seasons coincided with the warmest period of the year when reptiles are most active. Repeat surveys ensured that sampling was completed during a wide range of weather conditions (Read & Moseby 2001). In both surveys, a set protocol was followed. In a given week, a set of four randomly chosen landscape units was pitfall trapped for 5 days (four nights). Traps were checked daily, and captured animals were identified and released on site. In addition, each plot was actively searched for reptiles on one afternoon during a given survey. This involved looking under rocks and logs, and searching through leaf litter and grass. Throughout both surveys, each landscape unit, site and plot were visited 13 times (10 days of pitfall trapping, 2 days of active searching and 1 day of habitat assessment). Incidental observations of reptiles during these visits were recorded. Data from all surveys and techniques were combined to obtain presence data and species richness data for each plot, site and landscape unit.

covariates used in the analysis

The percentage of rocks, logs, canopy, grasses/forbs and litter, as well as the number of sticks, were assessed visually at every plot. The number of individuals belonging to the most numerous invertebrate groups captured in pitfall traps was recorded every morning. For each plot, the maximum number of spiders, beetles and ants captured in a morning was recorded. The dominant grass and forb species in each plot were recorded, and a vegetation condition score ranging from 1 (completely weed infested) to 6 (pristine native grass/forb layer) was derived for each plot. Scores obtained ranged from 1 to 4. The dominant tree species was recorded in every plot and at every site, and four categories were created for analysis: (1) ‘box’, white box, yellow box or apple box; (2) ‘gum’, Blakely's red gum; (3) ‘stringybark’, red stringybark; (4) ‘no trees’, no trees present in the plot or site.

For each plot, a topographic position score based on its location within a landscape unit was recorded. This could be valley (score = 3), mid-slope (2) or ridge (1). Plots within the same site were assigned the same topographic position score (Fig. 1). Within each landscape unit, a temperature data logger (accuracy 1 °C) was placed in a shaded location at one of the ridge or valley sites, respectively. Throughout the February–March trapping session, temperatures were recorded every 30 min. From this information, the daily maximum temperatures were obtained for each landscape unit, and the mean daily maximum temperature over the survey period was calculated. This gave an indication of overall differences in the thermal regimes of different landscape units. For subsequent analyses, all plots within a given landscape unit were assumed to have the same mean maximum temperature. We were aware that temperature was likely to vary between plots within a landscape unit, but instrumentation was limited by cost. Finally, the elevation and the cosine of the aspect were calculated for every plot because these variables are related to the temperature regime; northerly plots had a high cosine (maximum +1) while southerly plots had a low cosine (minimum −1).

In some cases, insufficient data on lizard presence were obtained at the plot level (n = 144) for statistical analysis. In these cases, data from the three plots within a site were pooled, and analyses were conducted at the site level (n = 48). To obtain meaningful measures of the above continuous covariates at the site level, the mean of a given covariate at the three plots located within the same site was calculated.

data analysis

Principal component analysis (PCA) was used for continuous plot- and site-level covariates to reduce the number of covariates, and obtain a smaller number of independent variables (Everitt & Dunn 1991; Manly 1994). Prior to PCA, highly skewed variables were log-transformed.

Generalized linear mixed modelling (Schall 1991; Breslow & Clayton 1993) was used to relate the presence of a given species or species richness at a given plot or site to a set of explanatory variables. Mixed models differentiate between random effects and fixed effects. Random effects are used to account for nesting in the experimental units. Here, random effects were specified to account for the physical structure of experimental units, for example it was reasonable to expect that plots within a given site may be more similar to one another than plots in different sites. Similarly, sites within a given landscape unit may be more similar to one another than sites in different landscape units. This meant that a nested specification of random effects was appropriate (landscape unit/site/plot).

Fixed effects are variables of direct interest to the study. For each response variable measured at the plot or site level, three separate models using a different set of fixed effects were constructed. (i) The landscape model used only the stratifying variables measured at the landscape unit level (aspect, topographic position, canopy cover) as explanatory variables. (ii) The microhabitat model used the first four principal components derived from the plot- or site-level habitat variables as explanatory variables (fitted as linear terms and as quadratics). (iii) The vegetation association model used the dominant vegetation association at a plot or site (box, gum, stringybark, or no substantial tree cover) as the only explanatory variable. Two-way interactions were allowed for the landscape and microhabitat models. No overall model incorporating microhabitat, landscape and floristic variables was developed, because of computational difficulties when interactions between micro- and landscape scale were modelled (i.e. unstable models). Hence, the models relating to different spatial scales in this study need to be interpreted separately, although we are acutely aware that interactions between phenomena at the microhabitat and landscape scale may occur.

The presence/absence of species was modelled for common and moderately common species using logistic regression (Collett 1991), and species richness was modelled using a Poisson distribution. Analysis of deviance tables were examined to decide which variables to retain in the model, and the significance level was chosen to be P < 0·05. Where appropriate, residual plots and the distribution of residuals were examined to check model assumptions.

Finally, the components of variance were determined for habitat variables measured at the plot level. This was done to ascertain at which spatial scale most of the variation occurred for a given variable, for example did rock cover vary more strongly between plots (over tens of metres) or between sites or landscape units (over hundreds or thousands of metres)? This analysis was conducted separately from the regression models to highlight the potential complexity in habitat heterogeneity. To determine the components of variance for a given habitat variable, the habitat variable of interest was fitted as a function of spatially nested random effects (landscape unit/site/plot) using a linear mixed model (Laird & Ware 1982). Ninety-five per cent confidence intervals for the random effects associated with each spatial scale were constructed. These were converted into percentages and graphed to visualize what proportion of variance in a certain habitat variable was explained at each of the three spatial scales.


  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Eighteen species of reptiles were observed (Table 1).

Table 1.  Overview of the reptiles observed in the study area. The habitat description is a subjective summary of field observations during this study. Scientific names follow Cogger (2000)
SpeciesHabitat descriptionStatus in study area
  • 1

    , Status uncertain but infrequently observed in this study;

  • 2

    2, observed incidentally, but too large to be pitfall trapped and hence not analysed statistically;

  • 3

    , statistical model reported in Table 4.

Common long-necked tortoise Chelodina longicollis (Shaw)Rivers, creeks, ponds and farm damsUncommon1,2
Marbled gecko Christinus marmoratus (Gray)Woodlands with leaf litter and fallen timber or rocksRestricted distribution, but high densities locally
Stone gecko Diplodactylus vittatus (Gray)Rocky woodlands with fallen timberUncommon1
Olive legless lizard Delma inornata (Kluge)Open grassy habitat; shelters under rocks or logsModerately common3
Four-fingered skink Carlia tetradactyla (O’Shaughnessy)Woodlands and grassy areas; shelters under rocks and logsVery common3
Striped skink Ctenotus robustus (Storr)Open grassy habitat; shelters under rocks or logsCommon3
Copper-tailed skink Ctenotus taeniolatus (White, ex. Shaw)Woodland habitats with fallen timber and rocky areasUncommon
Cunningham's skink Egernia cunninghami (Gray)Boulder outcrops in woodlands; also observed near buildings and building debrisRestricted distribution, but high densities locally2
Tree skink Egernia striolata (Peters)Rock outcrops and areas with large logs, both in woodlands and grassy areasUncommon1
Southern water skink Eulamprus heatwolei (Wells and Wellington)Rocky areas within a few metres of flowing waterRestricted distribution, but high densities locally
Three-toed skink Hemiergis decresiensis (Cuvier)Cool areas with relatively dense vegetationUncommon1
Boulenger's skink Morethia boulengeri (Ogilby)Woodlands and grassy areasModerately common3
Jacky lizard Amphibolurus muricatus (White, ex. Shaw)Woodlands and grassy areasUncommon1
Bearded dragon Pogona barbata (Cuvier)Woodlands and grassy areas; observed along roadsides and basking on fence postsModerately common2
Lace monitor Varanus varius (White, ex. Shaw)WoodlandsModerately common2
Red-bellied black snake Pseudechis porphyriacus (Shaw)Woodlands and grassy areas; banks of creeksCommon2
Eastern brown snake Pseudonaja textilis (Duméril, Bibron and Duméril)Woodlands and grassy areasModerately common2
Dwyer's snake Suta spectabilis dwyeri (Krefft)Woodlands and grassy areasUncommon1,2

principal components analyses

In both plot-level and site-level analyses, the first four principal components accounted for approximately 70% of variation in the habitat variables and were used in later analyses. The loadings of habitat variables and the resulting interpretation were identical for the plot and site levels (Table 2).

Table 2.  Results of the principal component analyses at the plot and site levels (* = log-transformed prior to analysis). The analysis was conducted to reduce the number of related covariates, and obtain a smaller number of independent variables
Principal componentRelated habitat variablesPlot-level analysisSite-level analysis
Loading (plot-level analysis)Variance explained (%)Loading (site-level analysis)Variance explained (%)
(1) Vegetation structurePercentage grass and forb cover0·38531·1 0·36734·2
Grassy vs. complex with treesPercentage canopy cover*−0·375 −0·380 
Percentage litter cover*−0·388 −0·380 
Number of sticks*−0·358 −0·338 
(2) Rockiness vs. spidersPercentage of rock cover0·43216·2 0·41917·8
Number of spiders−0·416 −0·406 
(3) Landscape position and temperatureElevation0·46611·1 0·46411·6
Warm northerly valleys vs. cool southerly ridgesCosine of the aspect−0·467 −0·474 
Average daily maximum temperature−0·362 −0·369 
Topographic position score−0·365 −0·347 
(4) Invertebrate abundanceNumber of beetles0·575 8·6 0·552 7·9
Abundance of beetles and antsNumber of ants*0·543  0·542 

The first principal component described a gradient from areas with a large amount of tree-related features (sticks, litter, canopy cover) to grassy areas. The second principal component described an increase in rocks and a reduction in spiders. The third component reflected the temperature regime, in which low values were associated with warm areas in northerly valleys, while high values were related to cool areas on southerly ridges. High values of the fourth principal component were related to increased invertebrate abundance (particularly ants and beetles).

logistic regression models

At the plot level (n = 144), logistic regression models were obtained for the two most common species, the four-fingered skink and the striped skink. The landscape model indicated that four-fingered skinks were more likely to be observed at plots located within landscape units characterized by a large amount of canopy cover (Table 3). The microhabitat model showed that this species was most likely to occur in plots with a low score of the second principal component (i.e. plots with few rocks and many spiders; Table 3 and Fig. 2). In addition, the four-fingered skink was more likely to occur in plots of the gum or box vegetation association (Table 3).

Table 3.  Plot-level logistic regression models. Only the four-fingered skink and striped skink were encountered sufficiently frequently to allow an analysis at this level (n = 144). For these species, the landscape model and microhabitat model as well as a model in relation to vegetation associations are shown. PC, principal component
ResponseModel termCoefficientSEP-value
Four-fingered skink
Landscape modelIntercept−1·88240·53710·0007
High canopy cover1·74410·75960·0376
Vegetation association modelIntercept−1·00660·27650·0004
Box or gum dominated1·74180·46770·0003
Microhabitat modelIntercept−1·08700·45250·0182
Score (PC2)−0·58340·19820·0041
Striped skink
Landscape modelNo significant relationships detected
Vegetation association modelNo significant relationships detected
Microhabitat modelIntercept−1·75040·2238< 0·0001
Score (PC1)0·66640·0918< 0·0001
Score (PC4)−0·66430·18140·0004
Score (PC4)^2−0·77470·1035< 0·0001

Figure 2. Graphical representation of the plot-level logistic regression models summarized in Table 3. Only the four-fingered skink and striped skink were encountered sufficiently frequently to allow an analysis at this level (n = 144). The graphs show the probability of encountering a species in a given plot. A flat line indicates no response to a certain habitat gradient or principal component, while a curve indicates a statistically significant response to a given habitat gradient.

Download figure to PowerPoint

Striped skinks were not related to any of the landscape variables or specific vegetation associations. However, this species was significantly related to the first and fourth principal component of plot-level microhabitat variables, striped skinks being more likely in areas where vegetation structural complexity was low, and in areas with an intermediate abundance of beetles and ants (Table 3 and Fig. 2).

At the site level (n = 48) sufficient data were available to develop logistic regression models for four species: the four-fingered skink, striped skink, Boulenger's skink and olive legless lizard. The landscape model for the four-fingered skink indicated it was significantly more likely to be observed at sites within landscape units with a northerly aspect and a high amount of canopy cover (Table 4). The four-fingered skink was also significantly more likely at sites dominated by box trees and significantly less likely in areas with no trees (Table 4). However, this species was related to none of the microhabitat gradients. In contrast, the striped skink was not related to any landscape variables or vegetation associations, but appeared to respond significantly to the first principal component at the microhabitat scale. The coefficients indicated that it was more likely at sites with a relatively low structural complexity in vegetation (Table 4 and Fig. 3). Boulenger's skink was significantly more likely to occur at sites within landscape units characterized by a northerly aspect and a large amount of tree cover. It also appeared to respond to vegetation associations at the site level, and was significantly less likely in areas dominated by stringybark eucalypts. In addition, Boulenger's skink varied along one of the microhabitat gradients, and it was more likely at sites with a high score of the fourth principal component, i.e. areas with a high invertebrate abundance (Table 4 and Fig. 3). The olive legless lizard did not respond to any of the landscape variables used to stratify landscape units. However, it varied with vegetation associations and was more likely at sites with no tree cover. The species was also more likely at sites with a relatively high score of the first principal component, i.e. in areas with relatively low structural complexity (Table 4 and Fig. 3).

Table 4.  Site-level logistic regression models for the four-fingered skink, striped skink, Boulenger's skink and olive legless lizard (n = 48). For these species, the landscape model and microhabitat model as well as a model in relation to vegetation associations are shown. PC, principal component
ResponseModel termCoefficientSEP-value
Four-fingered skink
Landscape modelIntercept0·00000·53541·000
Northerly aspect−0·33650·75720·6647
High canopy cover0·69310·75720·3780
Northerly aspect × high canopy cover8·84611·0709< 0·0001
Vegetation association modelIntercept1·38630·53780·0151
Gum dominated7·81640·9119< 0·0001
No substantial tree cover−1·62740·68020·0232
Microhabitat modelNo significant relationships detected
Striped skink
Landscape modelNo significant relationships detected
Vegetation association modelNo significant relationships detected
Microhabitat modelIntercept−0·01810·52960·9729
Score (PC1)1·20820·1947< 0·0001
Score (PC1)^2−0·47660·0762< 0·0001
Boulenger's skink
Landscape modelIntercept−1·09860·61300·0826
High canopy cover−9·10420·8670< 0·0001
Northerly aspect−0·51080·86700·5666
Northerly aspect × high canopy cover10·71371·2260< 0·0001
Vegetation associations modelIntercept−1·03610·33530·0042
Stringybark dominated−8·16660·9485< 0·0001
Microhabitat modelIntercept−1·46980·39550·0008
Score (PC4)0·86370·37480·0281
Olive legless lizard
Landscape modelNo significant relationships detected
Vegetation associations modelIntercept−10·20280·5417< 0·0001
Gum dominated or no substantial tree cover9·02420·6436< 0·0001
Microhabitat modelIntercept−1·96300·3745< 0·0001
Score (PC1)0·39830·17110·0266

Figure 3. Graphical representation of the site-level logistic regression models summarized in Table 4 (n = 48). The graphs show the probability of encountering a species in a given plot. A flat line indicates no response to a certain habitat gradient or principal component, while a curve indicates a statistically significant response to a given habitat gradient. The striped skink and olive legless lizard responded to the first principal component, no species responded to the second and third component, and only Boulenger's skink responded to the fourth principal component.

Download figure to PowerPoint

species richness

Reptile species richness at a given site was not significantly related to any of the landscape variables used to stratify landscape units, but it was higher at sites dominated by box trees than in other areas (Table 5). No significant relationship was detected with any of the first four principal components at the site level. However, species richness reflected microhabitat variability within a site, being greatest in areas with a large standard deviation of the first principal component as measured at the plot level. Thus, sites with the greatest variability in vegetation structure supported more species (Table 5).

Table 5.  Reptile species richness models at site level (n = 48). Landscape and microhabitat models as well as a model in relation to vegetation associations are shown. * = binary variable created to distinguish between the 24 sites with the highest standard deviation of Scoreplot (PC1) and the 24 sites with the lowest standard deviation of Scoreplot (PC1). The microhabitat model indicated that species richness was significantly higher in the 24 sites with above median microhabitat variability
ResponseModel termCoefficientSEP-value
Reptile species richness
Landscape modelNo significant relationships were detected   
Vegetation association modelIntercept0·46360·10740·0002
Box dominated0·59730·19760·0050
Microhabitat modelIntercept0·34830·14640·0237
High standard deviation of the three scores of PC1 measured at each plot within a given site*0·46260·18700·0190

components of variance of habitat variables

The components of variance at the plot, site and landscape unit levels were determined for microhabitat variables (Fig. 4). Most of the variability in invertebrate abundance occurred at the plot level, i.e. over tens of metres. Other variables varied to a similar extent at all three spatial scales (e.g. rock cover), or varied most strongly between landscape units (e.g. grass cover). The spatial variability of tree-related variables was more complex. Most tree-related variables (e.g. leaf litter, canopy cover and log cover) varied most strongly between landscape units, but could also vary substantially between plots, i.e. over tens of metres (Fig. 4). A graphical analysis of the temperatures recorded by the data loggers indicated that northerly landscape units experienced higher daily maximum temperatures than southerly ones, and valley landscape units experienced higher daily maximum temperatures than ridge landscape units (results not shown).


Figure 4. Ninety-five per cent confidence intervals of the components of variance for several continuous variables measured at the plot level. ‘Max. spiders’, ‘max. ants’ and ‘max. beetles’ refer to the maximum number of invertebrates captured in 1 day over 2 weeks in a given plot. Note that some variables varied most strongly at a fine spatial scale (e.g. invertebrate abundance, the bottom row of the graph) while others varied most strongly at larger spatial scales (e.g. grass cover at the landscape unit scale).

Download figure to PowerPoint


  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Our study revealed that different reptile species reflected different ecological variables, which in turn varied to different extents over different spatial scales.

habitat relationships

The four-fingered skink was the most common species. It was associated with box (and gum) woodlands, landscape units with a northerly aspect, and microhabitats characterized by abundant spiders and few rocks. These findings are largely consistent with published information. Two empirical studies (Rauhala 1993; Letnic & Fox 1997) and several field guides (Jenkins & Bartell 1980; Bennett 1997; Cogger 2000) suggest the four-fingered skink inhabits open woodlands (Rauhala 1995; Bennett 1997). Our finding that the species was more likely to occupy northerly aspects may be related to its thermal biology. For example, Bennett (1997) noted that in the Australian Capital Territory the species was restricted to ‘warm locations’.

The apparent preference of four-fingered skinks for areas with few rocks and abundant spiders is more difficult to explain. Most skinks in south-eastern Australia are arthropod generalists (Brown 1991), and perhaps the abundance of spiders was correlated with the abundance of other invertebrates. However, the four-fingered skink did not exhibit a significant response to the abundance of beetles and ants (Tables 2, 3 and 4), so the apparent response to spiders and rocks is probably more complex.

The striped skink occurred in areas with relatively low microhabitat structural complexity but did not respond to variables measured at larger spatial scales or vegetation associations (Figs 2 and 3, and Tables 3 and 4). In the coastal dunes of New South Wales, Twigg & Fox (1991) and Taylor & Fox (2001) also found that the striped skink was associated with structurally simple habitats. Other work suggests this species can also be associated with denser woodlands (Rauhala 1993; MacNally & Brown 2001). The preference of the striped skink for open microhabitats in our study area may be related to its relatively high preferred body temperature (Greer 1989), because low structural complexity of vegetation is associated with less shade and higher environmental temperatures (Sartorius, Vitt & Colli 1999).

Unlike the striped skink, Boulenger's skink appeared to respond to landscape variables as well as microhabitat variables. It was more likely to be found on northerly aspects with high vegetation cover, and tended to be absent in areas dominated by stringybark eucalypts. These findings were obtained in separate models (Table 3) but are probably related because southerly ridges with high vegetation cover were often dominated by stringybark eucalypts, while northerly ridges tended to be dominated by box woodland (J. Fischer, personal observation). This pattern is similar to the four-fingered skink, and is probably related to the relatively high preferred body temperature of Boulenger's skink and its extensive distribution throughout warmer parts of Australia (Greer 1989; Cogger 2000). Boulenger's skink did not respond to tree-related microhabitat variables, although it is considered to forage in leaf litter and may shelter under logs (Jenkins & Bartell 1980; Bennett 1997). This indicates that Boulenger's skink can tolerate substantial levels of disturbance (Brown & Nicholls 1993; Brown 2001; MacNally & Brown 2001).

No quantitative habitat studies are available on the olive legless lizard. Our results are consistent with the qualitative assessment of Jenkins & Bartell (1980) and Bennett (1997), and suggest that this species prefers structurally simple microhabitats. Field observations indicate the species can withstand moderate levels of disturbance from grazing (i.e. introduction of some weeds, fertilizers and soil compaction) as long as a sufficient number of half-buried rocks and logs are available as shelter sites. This assessment is similar to the closely related striped legless lizard Delma impar (Fischer) in the Australian Capital Territory (Dorrough & Ash 1999).

Our finding that different species required different habitats translated directly into the observed pattern in which species richness increased with structural habitat variability (Table 5). A similar finding has been made for birds (MacArthur & MacArthur 1961; Gilmore 1985) and reptiles in Australian chenopod grasslands (Read 1995), and may be explained by habitat partitioning (Schoener 1974; Austin 1999). Species richness was also positively related to box woodland associations. This is an important finding, as white box and yellow box woodlands are threatened tree communities due to clearing and grazing (Prober & Thiele 1995; Glanznig & Kennedy 2000).

a spatially nested survey design

The main advantages and disadvantages of conventional, single-scale surveys vs. spatially nested surveys can be summarized as in Table 6. The main advantage of conventional single-scaled studies is their simplicity, and the main disadvantage of spatially nested designs is their complexity. However, there are situations when a spatially nested design offers insights not obtainable from a single-scaled study. Multi-scaled work allows additional insights on any given species, and adds flexibility in dealing with several species simultaneously. An additional practical advantage of a hierarchical, spatially nested survey design is the ability to pool response data to higher levels. For example, we obtained insufficient data to analyse the presence of the olive legless lizard at the plot level (n = 144), but by pooling data from three plots to the site level we were able to construct a regression model (Table 4). Despite these advantages, our case study of a hierarchical survey design should not be seen as an ideal example, and several improvements are possible. Future projects should include an appraisal of how many levels to include in the hierarchy, and how many different explanatory variables at different spatial scales can realistically be modelled before computational difficulties will arise (e.g. in this study interactions between spatial scales were not modelled).

Table 6.  Summary of advantages and disadvantages of a conventional single-scale experimental design vs. a hierarchical, spatially nested experimental design
Conventional single-scale experimental designHierarchical, spatially nested experimental design
Easy to interpret and communicateMultiple scales in one study for a given species
Simple exploratory graphs can be used to summarize findings  (e.g. scatterplots)Particularly suited for surveying multiple species with different spatial grain
Computationally simpleMaximizes the number of experimental units by spreading survey effort
Can cater for uncertainty in obtaining sufficient data by allowing the pooling of response data to higher levels
Can explicitly highlight spatial relationships of continuous habitat variables
Can only answer if a species responds to phenomena at the spatial  scale deemed important a prioriThe amount of detail that can be investigated on a single species or at a single spatial scale is limited
Contains no spatial information on habitat variablesComputationally complex
Cannot target multiple species simultaneously if they have  substantially different spatial grainsThere are limits to how many scales can realistically be incorporated in a meaningful way in a single survey
Less flexible if insufficient data are obtained at the level chosen  as the experimental unit a prioriExploratory scatterplots and similar graphs are limited because they do not account for spatial nesting
 Higher complexity in design, interpretation and communication

implications for conceptual landscape models

Useful landscape models need to be readily interpretable but flexible enough to reflect ecological complexity. In terrestrial ecosystems, modified landscapes are often considered to be fragmented, and habitat fragmentation is a major research area in conservation biology (Saunders, Hobbs & Margules 1991). While many important insights have been gained from studies on habitat fragmentation, the fragmentation paradigm can sometimes be problematic. For example, some landscapes appear to be fragmented from an anthropocentric perspective, but may not be fragmented for all the biota inhabiting these environments (Ingham & Samways 1996; Laurance et al. 2002). Similarly, different species utilize different habitats, and a fragmented habitat for one species may be intact or continuous habitat for another (Gascon et al. 1999; Lindenmayer et al. 2002).

Some workers have emphasized that patches in a fragmented landscape may be heterogeneous in themselves, and hence may differ in habitat quality (Forman 1995). However, in practice a conceptual framework of patches in a matrix lends itself to a binomial classification of land into habitat and non-habitat. In some situations, a binomial classification may be simplistic and could potentially lead to suboptimal conservation outcomes. This is because a binary classification of land may lead to inappropriate emphasis on the remaining ‘habitat’ fragments, while the matrix (i.e. the background patch type; Forman 1995) is neglected (Fischer & Lindenmayer 2002a; Haila 2002). This may not always be desirable because in some settings the matrix can be an important landscape element (Lindenmayer & Franklin 2002).

There are few alternative ways of conceptualizing modified landscapes other than the fragmentation model. The only widely known alternative to the fragmentation model is the variegation model by McIntyre & Barrett (1992), which recognizes gradients in habitat suitability. While this model can adequately describe habitat use by some animals in some landscapes (Barrett, Ford & Recher 1994; Fischer & Lindenmayer 2002b,c), the model is limited in that it does not distinguish between species-specific differences in habitat use and spatial grain (Lindenmayer, McIntyre & Fischer 2003).

These limitations do not preclude the usefulness of existing landscape models in many cases. Rather, which landscape model is appropriate will depend on the landscape under investigation and the organisms of interest. However, as demonstrated by this case study, in some situations a more flexible landscape model may be helpful in describing ecological patterns. Such a model would need to recognize species-specific differences in habitat and spatial grain, and allow different organisms to respond to ecological phenomena at a range of scales. For example, some species may respond to certain ecological conditions at one, but not all, spatial scales; in our study, the striped skink was more likely to be detected in structurally simple microhabitats measured at a spatial scale of square metres, but was not detected more frequently in sparsely vegetated landscape units that covered several hectares. While it is beyond the scope of this paper to develop a more flexible landscape model, conceptualizing landscapes as a set of overlapping species-specific habitat contour maps may be a useful starting point (Wiens et al. 1993; Wiens 1995). This approach has the advantage of being straightforward to understand because of widespread familiarity with topographic maps, whilst being able to summarize large amounts of ecological information across multiple species and spatial scales.

conservation implications

Our findings suggest that a range of different conditions were used by different reptiles. Therefore, simplification of habitat diversity, for example through more intensive grazing, the removal of fallen timber in paddocks or the logging of woodland remnants, may adversely affect reptile diversity. Conversely, the maintenance of habitat heterogeneity may help maintain a high diversity by providing a large number of different niches. The role of landscape heterogeneity as an adjunct to nature conservation has been noted elsewhere, for example in European farming landscapes (di Giulio, Edwards & Meister 2001; Benton, Vickery & Wilson 2003) and Indonesian forestry landscapes (Hamer et al. 2003).

Many reptiles encountered in this study depend on tree-related habitat features, such as woodlands, logs and leaf litter (Cogger 2000). Grazing pressure can substantially reduce tree regeneration in Australian pastoral landscapes (Saunders et al. 2003) and thus may pose a significant long-term threat to tree-related habitat features. In a recent survey near our study area, Spooner, Lunt & Robinson (2002) showed there was no natural tree regeneration in 87% of woodland remnants. Hence, although some reptiles may persist in moderately grazed areas in the short term (Osborne, Kukolic & Jones 1995; Dorrough & Ash 1999; Brown 2001), in the long term grazing may pose a significant threat to reptile habitat (Ehman & Cogger 1985; Sadlier & Pressey 1994; Hadden & Westbrooke 1996).

More generally, the potential limitations of existing landscape models have implications for conservation approaches in grazing landscapes. Traditionally, the conservation of biodiversity was thought to be best achieved through land allocation, for example the establishment of nature reserves (Margules & Pressey 2000; Rodrigues & Gaston 2001). This approach can be powerful in some environments but may have limitations in relatively heterogeneous production landscapes outside protected areas. This is because different species perceive different areas as habitat, and perceive the same landscape at different spatial scales. In the grazing landscapes of south-eastern Australia, this may mean that fencing off large woodland remnants or planting new trees in well-defined areas may not have the desired positive consequences for biodiversity conservation in the long term if other areas are neglected. Instead, more integrated approaches to landscape management may be needed in these landscapes. For example, it may be useful to temporarily exclude different parts of a landscape from stock to allow natural tree regeneration to occur, or reduce stocking rates throughout entire landscapes (Jansen & Robertson 2001).


Our results highlight the importance of maintaining landscape heterogeneity to provide habitat for a range of reptile species. Despite its limitations, our study may serve as a useful starting point to consider how multiple species and spatial scales may be incorporated in a single study. Additional multi-scaled and multi-species investigations may help to advance current conceptual models in landscape ecology and conservation biology.


  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

We are most grateful to K. Skinner, N. and S. Keatinge and V. MacWhinney for allowing us to survey their properties. Financial support for this study was obtained from the Australian Society of Herpetologists. We also thank a number of students who assisted with different aspects of field work, particularly J. Kondo and J. Burmester. Finally, comments by I. Fazey, H. B. Shaffer and two anonymous referees on earlier versions of the manuscript are gratefully appreciated.


  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  • Austin, M.P. (1999) The potential contribution of vegetation ecology to biodiversity research. Ecography, 22, 465484.
  • Barrett, G.W., Ford, H.A. & Recher, H.F. (1994) Conservation of woodland birds in a fragmented rural landscape. Pacific Conservation Biology, 1, 245256.
  • Bennett, R. (1997) Reptiles and Frogs of the Australian Capital Territory. National Parks Association of the ACT, Canberra, Australia.
  • Benton, T.G., Vickery, J.A. & Wilson, J.D. (2003) Farmland biodiversity: is habitat heterogeneity the key? Trends in Ecology and Evolution, 18, 182188.
  • Bock, C.E., Smith, H.M. & Bock, J.H. (1990) The effect of livestock grazing upon abundance of the lizard, Sceloporus scalaris, in southeastern Arizona. Journal of Herpetology, 24, 445446.
  • Breslow, N.E. & Clayton, D.G. (1993) Approximate inference in generalized linear mixed models. Journal of the American Statistical Association, 88, 925.
  • Brown, G.W. (1991) Ecological feeding analysis of south-eastern Australian scincids (Reptilia: Lacertilia). Australian Journal of Zoology, 39, 929.
  • Brown, G.W. (2001) The influence of habitat disturbance on reptiles in a box–ironbark eucalypt forest of south-eastern Australia. Biodiversity and Conservation, 10, 161176.
  • Brown, G.W. & Nicholls, A.O. (1993) Comparative census techniques and modelling of habitat utilization by reptiles in northern Victoria. Herpetology in Australia: A Diverse Discipline (eds D.Lunney & D.Ayers), pp. 283290. Royal Zoological Society of New South Wales, Sydney, Australia.
  • Busack, S.D. & Bury, R.B. (1974) Some effects of off-road vehicles and sheep grazing on lizard populations in the Mojave Desert. Biological Conservation, 6, 179183.
  • Cogger, H. (2000) Reptiles and Amphibians of Australia. Reed, Melbourne, Australia.
  • Collett, D. (1991) Modelling Binary Data. Chapman & Hall, London, UK.
  • Dorrough, J. & Ash, J.E. (1999) Using past and present habitat to predict the current distribution and abundance of a rare cryptic lizard, Delma impar (Pygopodidae). Australian Journal of Ecology, 24, 614624.
  • Ehmann, H. & Cogger, H. (1985) Australia's endangered herpetofauna: a review of criteria and policies. Biology of Australasian Frogs and Reptiles (eds G.Grigg, R.Shine & H.Ehmann), pp. 435447. Royal Zoological Society of New South Wales, Sydney, Australia.
  • Everitt, B.S. & Dunn, G. (1991) Applied Multivariate Data Analysis. Arnold, London, UK.
  • Fischer, J. & Lindenmayer, D.B. (2002a) Small patches can be valuable for biodiversity conservation: two case studies on birds in southeastern Australia. Biological Conservation, 106, 129136.
  • Fischer, J. & Lindenmayer, D.B. (2002b) The conservation value of paddock trees for birds in a variegated landscape in southern New South Wales. I. Species composition and site occupancy patterns. Biodiversity and Conservation, 11, 807832.
  • Fischer, J. & Lindenmayer, D.B. (2002c) The conservation value of paddock trees for birds in a variegated landscape in southern New South Wales. II. Paddock trees as stepping stones. Biodiversity and Conservation, 11, 833849.
  • Forman, R.T. (1964) Growth under controlled conditions to explain the hierarchical distribution of a moss, Tetrapis pellucida. Ecological Monographs, 34, 125.
  • Forman, R.T.T. (1995) Some general principles of landscape and regional ecology. Landscape Ecology, 10, 133142.
  • Gascon, C., Lovejoy, T.E., Bierregaard, R.O.J., Malcolm, J.R., Stouffer, P.C., Vasconcelos, H.L., Laurance, W.F., Zimmerman, B., Tocher, M. & Borges, S. (1999) Matrix habitat and species richness in tropical forest remnants. Biological Conservation, 91, 223229.
  • Gibbons, P. & Boak, M. (2002) The value of paddock trees for regional conservation in an agricultural landscape. Ecological Management and Restoration, 3, 205210.
  • Gilmore, A.M. (1985) The influence of vegetation structure on the density of insectivorous birds. Birds of Eucalypt Forests and Woodlands: Ecology, Conservation and Management (eds A.Keast, H.F.Recher, H.Ford & D.Saunders), pp. 2131. Royal Australasian Ornithologists Union and Surrey Beatty and Sons, Chipping Norton, Australia.
  • Di Giulio, M., Edwards, P.J. & Meister, E. (2001) Enhancing insect diversity in agricultural grasslands: the roles of management and landscape structure. Journal of Applied Ecology, 38, 310319.
  • Glanznig, A. & Kennedy, M. (2000) From words to action: addressing biodiversity loss, land degradation and native vegetation clearance in the 1990s: the Australian experience. _Law/vegman.html[visited April 2001]. Nature Conservation Council of New South Wales, Sydney, Australia.
  • Goldingay, R., Daly, G. & Lemckert, F. (1996) Assessing the impacts of logging on reptiles and frogs in the montane forests of southern New South Wales. Wildlife Research, 23, 495510.
  • Greer, A.E. (1989) The Biology and Evolution of Australian Lizards. Surrey Beatty and Sons, Chipping Norton, Australia.
  • Hadden, S.A. & Westbrooke, M.E. (1996) Habitat relationships of the herpetofauna of remnant Buloke woodlands of the Wimmera Plains, Victoria. Wildlife Research, 23, 363372.
  • Haila, Y. (2002) A conceptual genealogy of fragmentation research: from island biogeography to landscape ecology. Ecological Applications, 12, 321334.
  • Halliger, M. (1993) Reptiles on regenerating mine sites in western New South Wales. Herpetology in Australia: A Diverse Discipline (eds D.Lunney & D.Ayers), pp. 327332. Transactions of the Royal Zoological Society of New South Wales, Sydney, Australia.
  • Hamer, K.C., Hill, J.K., Benedick, S., Mustaffa, N., Sherratt, T.N., Maryati, M. & Chey, V.K. (2003) Ecology of butterflies in natural and selectively logged forests of northern Borneo: the importance of habitat heterogeneity. Journal of Applied Ecology, 40, 150162.
  • Heatwole, H.F. & Taylor, J. (1987) Ecology of Reptiles. Surrey Beatty and Sons, Chipping Norton, Australia.
  • Ingham, D.S. & Samways, M.J. (1996) Application of fragmentation and variegation models to epigaeic invertebrates in South Africa. Conservation Biology, 10, 13531358.
  • James, C.D. (1991) Population dynamics, demography, and life history of sympatric scincid lizards (Ctenotus) in Central Australia. Herpetologica, 47, 194210.
  • Jansen, A. & Robertson, A.I. (2001) Relationships between livestock management and the ecological condition of riparian habitats along an Australian floodplain river. Journal of Applied Ecology, 38, 6375.
  • Jenkins, R. & Bartell, R. (1980) Reptiles of the Australian High Country. Inkata Press, Melbourne, Australia.
  • Kotliar, N.B. & Wiens, J.A. (1990) Multiple scales of patchiness and patch structure: a hierarchical framework for the study of heterogeneity. Oikos, 59, 253260.
  • Kutt, A. (1993) Initial observations on the effect of thinning eucalypt regrowth on heliothermic skinks in lowland forest, East Gippsland, Victoria. Herpetology in Australia: A Diverse Discipline (eds D.Lunney & D.Ayers), pp. 187196. Transactions of the Royal Zoological Society of New South Wales, Sydney, Australia.
  • Laird, N.M. & Ware, J.H. (1982) Random-effects models for longitudinal data. Biometrics, 38, 963974.
  • Laurance, W.F., Lovejoy, T.E., Vasconcelos, H.L., Bruna, E.M., Didham, R.K., Stouffer, P.C., Gascon, C., Bierregaard, R.O., Laurance, S.G. & Sampaio, E. (2002) Ecosystem decay of Amazonian forest fragments: a 22-year investigation. Conservation Biology, 16, 605618.
  • Lemckert, F. (1998) A survey for threatened herpetofauna of the south-west slopes of New South Wales. Australian Zoologist, 30, 492500.
  • Letnic, M.I. & Fox, B.J. (1997) The impact of industrial fluoride fallout on faunal succession following sand mining of dry sclerophyll forest at Tomago, NSW. I. Lizard recolonisation. Biological Conservation, 80, 6381.
  • Lindenmayer, D.B. (2000) Factors at multiple scales affecting distribution patterns and their implications for animal conservation: Leadbeater's possum as a case study. Biodiversity and Conservation, 9, 1535.
  • Lindenmayer, D.B. & Franklin, J. (2002) Conserving Forest Biodiversity. Island Press, Covelo, USA.
  • Lindenmayer, D.B., Cunningham, R.B., Donnelly, C.F., Nix, H.A. & Lindenmayer, B.D. (2002) Effects of forest fragmentation on bird assemblages in a novel landscape context. Ecological Monographs, 72, 118.
  • Lindenmayer, D.B., McIntyre, S. & Fischer, J. (2003) Birds in eucalypt and pine forests: landscape alteration and its implications for research models of faunal habitat use. Biological Conservation, 110, 4553.
  • Lunney, D. & Barker, J. (1986) Survey of reptiles and amphibians of the coastal forests near Bega, NSW. Australian Zoologist, 22, 17.
  • Lunney, D., Eby, P. & O'Connell, M. (1991) Effects of logging, fire and drought on three species of lizards in Mumbulla State Forest on the south coast of New South Wales. Australian Journal of Ecology, 16, 3346.
  • MacArthur, R.H. & MacArthur, J.W. (1961) On bird species diversity. Ecology, 42, 594598.
  • McIntyre, S. & Barrett, G.W. (1992) Habitat variegation, an alternative to fragmentation. Conservation Biology, 6, 146147.
  • McIntyre, S. & Hobbs, R. (1999) A framework for conceptualizing human effects on landscapes and its relevance to management and research models. Conservation Biology, 13, 12821292.
  • MacNally, R. & Brown, G.W. (2001) Reptiles and habitat fragmentation in the box–ironbark forests of central Victoria, Australia: predictions, compositional change and faunal nestedness. Oecologia, 128, 116125.
  • MacNally, R., Bennett, A.F., Brown, G.W., Lumsden, L.F., Yen, A., Hinkley, S., Lillywhite, P. & Ward, D.A. (2002) How well do ecosystem-based planning units represent different components of biodiversity? Ecological Applications, 12, 900912.
  • Manly, B.F. (1994) Multivariate Statistical Methods: A Primer, 2nd edn. Chapman & Hall, London, UK.
  • Margules, C.R. & Pressey, R.L. (2000) Systematic conservation planning. Nature, 405, 243253.
  • Meentemeyer, V. & Box, E.O. (1987) Scale effects in landscape studies. Landscape Heterogeneity and Disturbance (ed. M.G.Turner), pp. 1534. Springer-Verlag, New York, NY.
  • Osborne, W., Kukolic, K. & Jones, S. (1995) Management of threatened vertebrates in native grasslands: a case of clutching at straws? Management of Relict Lowland Grasslands: Proceedings of a Workshop and Public Seminar (eds S.Sharp & R.Rehwinkel), pp. 8995. ACT Parks and Conservation Service, Canberra, Australia.
  • Pough, F.H. (1980) The advantages of ectothermy for tetrapods. American Naturalist, 115, 92112.
  • Prober, S.M. & Thiele, K.R. (1995) Conservation of the Grassy White Box Woodlands: relative contributions of size and disturbance to floristic composition and diversity of remnants. Australian Journal of Botany, 43, 349366.
  • Rauhala, M. (1993) The Reptile, Amphibian and Mammal Fauna of the Stony Creek Nature Reserve. Technical Report 6. Australian Capital Territory ACT Parks and Conservation Service, Canberra, Australia.
  • Rauhala, M.A. (1995) The Reptile, Amphibian and Mammal Fauna of the Gigerline Nature Reserve. Technical Report 11. Australian Capital Territory ACT Parks and Conservation Service, Canberra, Australia.
  • Read, J.L. (1995) Subhabitat variability: a key to the high reptile diversity in chenopod shrublands. Australian Journal of Ecology, 20, 494501.
  • Read, J.L. & Moseby, K.E. (2001) Factors affecting pitfall capture rates of small ground vertebrates in arid South Australia. I. The influence of weather and moon phase on capture rates of reptiles. Wildlife Research, 28, 5360.
  • Rodrigues, A.S.L. & Gaston, K.J. (2001) How large do reserve networks need to be? Ecology Letters, 4, 602609.
  • Sadlier, R.A. & Pressey, R.L. (1994) Reptiles and amphibians of particular conservation concern in the western division of New South Wales: a preliminary review. Biological Conservation, 69, 4154.
  • Sartorius, S.S., Vitt, L.J. & Colli, G.R. (1999) Use of naturally and anthropogenically disturbed habitats in Amazonian rainforest by the teiid lizard Ameiva ameiva. Biological Conservation, 90, 91101.
  • Saunders, D.A., Hobbs, R.J. & Margules, C.R. (1991) Biological consequences of ecosystem fragmentation: a review. Conservation Biology, 5, 1832.
  • Saunders, D.A., Smith, G.T., Ingram, J.A. & Forrester, R.I. (2003) Changes in a remnant of salmon gum Eucalyptus salmonophloia and York gum E. loxophleba woodland, 1978–97. Implications for woodland conservation in the wheat–sheep regions of Australia. Biological Conservation, 110, 245256.
  • Schall, R. (1991) Estimation in generalized linear models with random effects. Biometrika, 78, 719727.
  • Schoener, T.W. (1974) Resource partitioning in ecological communities. Science, 185, 2739.
  • Spooner, P., Lunt, I. & Robinson, W. (2002) Is fencing enough? The short-term effects of stock exclusion in remnant grassy woodlands in southern NSW. Ecological Management and Restoration, 3, 117126.
  • Taylor, J.E. & Fox, B.J. (2001) Assessing the disturbance impact on vegetation and lizard communities of fluoride pollution interacting with fire and mining in eastern Australia. Austral Ecology, 26, 321337.
  • Turner, F.B., Jennrich, R.I. & Weintraub, J.D. (1969) Home ranges and body size of lizards. Ecology, 50, 10761081.
  • Twigg, L.E. & Fox, B.J. (1991) Recolonization of regenerating open forest by terrestrial lizards following sand mining. Australian Journal of Ecology, 16, 137148.
  • Wiens, J.A. (1989) Spatial scaling in ecology. Functional Ecology, 3, 385397.
  • Wiens, J.A. (1995) Landscape mosaics and ecological theory. Mosaic Landscapes and Ecological Processes (eds L.Hansson, L.Fahrig & G.Merriam), pp. 126. Chapman & Hall, London, UK.
  • Wiens, J.A., Stenseth, N.C., Vanhorne, B. & Ims, R.A. (1993) Ecological mechanisms and landscape ecology. Oikos, 66, 369380.
  • Yates, C.J. & Hobbs, R.J. (1997) Temperate eucalypt woodlands: a review of their status, processes threatening their persistence and techniques for restoration. Australian Journal of Botany, 45, 949973.