British Ecological Society is collaborating with JSTOR to digitize, preserve and extend access to Journal of Applied Ecology.

The Effect of Habitat Fragmentation and Livestock Grazing on Animal Communities in Remnants of Gimlet Eucalyptus salubris Woodland in the Western Australian Wheatbelt. II. Lizards Author(s): G. T. Smith, G. W. Arnold, S. Sarre, M. Abensperg-Traun and D. E. Steven Reviewed work(s): Source: Ecology, Vol. 33, No. 6 (Dec., 1996), pp. Published by: British Ecological Society Stable URL: http://www.jstor.org/stable/2404771. Accessed: 09/09/2012 17:12 Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at. http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.. British Ecological Society is collaborating with JSTOR to digitize, preserve and extend access to Journal of Applied Ecology. http://www.jstor.org

Journal of Applied Ecology 1996, 33, The effect of habitat fragmentation and livestock grazing on animal communities in remnants of gimlet Eucalyptus salubris woodland in the Western Australian wheatbelt. II. Lizards G.T. SMITH*, G.W. ARNOLD*, S. SARREt, M. ABENSPERG-TRAUN* and D.E. STEVEN* *CSIRO Division of Wildlife and Ecology, LMB 4 PO Midland WA 6056, Australia, 2. Molecular Evolution and Systematics, Research School of Biological Sciences, Australian National University, PO Box 475, Canberra ACT 2601, Australia Summary 1. The study examined relationships between habitat and biogeographic variables and the presence of lizard groups and individualizard species in remnants of gimlet Eucalyptus salubris woodland in the Western Australian wheatbelt. 2. The lizard species found in various gimlet woodland remnants are sub-sets of those found prior to fragmentation. 3. Regression analysis showed that woodly litter, percentage shrub cover and number of trees were the only habitat variables to influence species richness of the lizard taxa. Area, connectivity and distance to the nearest native vegetatation were the only biogeographic variables to influence species richness of geckos, 'other' lizards and total lizards. 4. Three individual species showed no significant relationships with any variables, whereas three species had significant relationships to variables related to cover/shelter only. 5. Disturbance from sheep grazing and trampling had no influence on the species richness of the different lizard taxa, but may have influenced the persistence of individual species in some remnants. 6. Implications of our findings for management of remnant vegetation are discussed. Key-words: biogeography, eucalypt woodland, habitat, lizards, remnant, Western Australia. Ecology (1996) 33, Ecological Society Introduction The development of agricultural lands has resulted in the destruction of the natural ecosystems, with the remaining original biota being confined mostly to small isolated remnants. The rate of change from a predominantly natural to an anthropogenic landscape has varied from millennia to decades (Angelstam& Arnold 1993). One of the fastestransitions has been in the Western Australian wheatbelt, where most of the agricultural development has taken place in the last 90 years, with over 50% of the clearing taking place in the last 50 years (Smith 1987). In any given district, as little as 7% of the original vegetation remains, mostly in small remnants, with the majority having been affected by human activities to some degree (Wallace & Moore 1987). The fragmentation of the original ecosystems has raised a number of conservation questions in relation to the ability of species and communities to persist in the new ecosystems dominated by agriculture (Saunders, Hobbs & Margules 1991). In particular, the functional significance of biogeographic and habitat variables in the persistence of species in small remnants (Lord & Norton 1990; Cutler 1991; Fahrig & Merriam 1994; Nunney & Elam 1994; Sarre 1995); the value of such remnants in short to medium-term conservation; and their long-term value in reconstructing quasi-natural ecosystems (Hobbs & Saunders 1993), highlighthe need for intergrated research projects at a variety of spatial and temporal scales. The changing status of vertebrate species in the 1302

1303 G.T. Smith et al. Western Australian wheatbelt has been reviewed by Kitchener et al. (1980a,b, 1982); Burbidge & Mc- Kenzie (1989) and Hobbs et al. (1993b). In contrast to mammals and birds, lizards have not suffered any extinctions and in general, lizard communities have persisted well, even in remnants as small as 30 ha (Kitchener et al. 1980a). Comparisons with communities on offshore islands suggest that mainland remnants are not super-saturated (remnants that retain a larger number of species than will be present at equilibrium, because insufficient time has passed since fragmentation for all species that will become extinct to be lost) and that in relation to area, few species should be lost in the future (Kitchener et al. 1980a; Kitchener & How 1982). This situation may not apply to very small remnants (<10 ha) which are characteristic of much of the wheatbelt. In these remnants, disturbancesuch as fire and chronic disturbances from livestock grazing, soil compaction, nutrient input, erosion, etc. will result in continuing habitat degradation (Saunders et al. 1991). In such situations, habitat specialists (species that are only found in areas of 'natural' vegetation) will be disproportionately lost compared to generalists (species that can survive in degraded areas such as road verges, paddocks, etc.) as has been shown by Kitchener (1982), Kitchener & How (1982), and Humphreys & Kitchener (1982) and in a more detailed two species comparison by Sarre, Smith & Myers (1995) and Sarre (1995). Overall, generalists have less specific habitat requirements that can be met in a wider range of environments, as well as, or as a consequence, have greater ability to move between remnants, using intervening sub-optimal habitat or agricultural land. The impact of livestock grazing and trampling on lizards has received little attention in Australia (Friedel& James 1995). However, there is increasing evidence to suggest that grazing is an important factor in the decline of biodiversity (Friedel& James 1995). The few studies in North America (Bury& Busack 1974; Jones 1981) suggest that grazing also would be detrimental for most species, given the wide ranging disturbance it causes (Fleischner 1994). In this paper we investigate whether species richness of lizards is correlated with biogeographic variables relating to area and spatial isolation of the remnants, or alternatively, that species richness is correlated with the degree of habitat change caused by livestock grazing and trampling. Methods STUDY AREA The study was carried out in a 1680 km2 area between Kellerberrin and Trayning, 200 km east of Perth, Western Australia. The district has low relief ( < 100 in), with scattered rock outcrops and was originally covered with a complex mosaic of heath, shrublands and woodlands (details in Hobbs, Saunders & Arnold 1993). Now, the area is dominated by wheat and sheep farming and only seven percentage of the original vegetation remains, mostly in small (< 100 ha) remnants (Arnold& Weeldenburg 1991). Within this study area, 24 remnants (areas of native vegetation surrounded by agricultural land) of gimlet woodland were selected, to provide a range of degrees of isolation from other remnants; area of the remnants ranged from 0 5 ha to 25 ha. All remnants had been grazed by livestock to varying degrees. In addition to these remnants, two quasi-control remnants were selected. One site was a 174 ha remnant (remnant 6) containing 90 ha of gimlet woodland which had not been grazed, but had been logged in some areas about 50 years ago. The other was a 4 ha area of gimlet woodland within a 138 ha remnant (remnant 46) that had not been grazed or logged. The distribution of remnants on the study area is shown on Fig. 1 in Abensperg-Traun et al. (1996). SAMPLING Presence of lizards was determined by standard searching methods (litter raking, lifting rocks, logs, bark, etc., digging out burrows) and spotlight searches for nocturnal species. Searches in each remnant were carried out in October 1992, February/March 1993 and October 1993. During each sampling period, every remnant was searched for a minimum of one hour. Spotlight searches for geckos were carried out during the summer of 1991/92. Searches were centred on the habitat quadrat (see below) and covered an area of approximately one hectare. In the four remnants less than 1 hectare in area (see Table 1), the intensity of searching was not increased; that is, no areas were searched more than once. Some small species were caught in invertebrate pitfall traps (see Abensperg- Traun et al. 1996). In order to provide a baseline for the studies on the small remnants, a pitfall trap grid was established on the larger 'control' remnant. The assumption was that, in this remnant, the lizard fauna would closely approximate that of gimlet woodlands prior to clearing, because it 'was a relatively large, undisturbed area. The pitfall trap grid was 4 x 14 with 25-m intervals between traps. The traps were plastic pails (28 cm diameter, 39 cm deep) with a 7-m drift fence of flywire mesh 24 cm high (Friend et al. 1989). The grid was operated for 10 consecutive days in October 1991 and 1992 and for four consecutive days, five times between November and March in 1991/92 and 1992/93, giving a total of 3134 trap nights. The traps were visited each morning and all animals were recorded and then released. BIOGEOGRAPHIC AND HABITAT VARIABLES The biogeographic variables (at the landscape scale as opposed to local or habitat variables) measured were:

1304 Lizard response to fragmentation and grazing Table 1. Lizard species recorded in gimlet woodland remnants. 1 = numbers of each species caught in pitfall traps, x = presence recorded from observation. 2 = presence (x) from systematic searches. % = the percentages of specimens collected in woodland, from Chapman & Dell (1985) CC CC Ce~ es C)~~~~~~~~~~~~~~~L C)~~~~~~0C Cd. ~ ~ ~ ~ ~ ~ ~ ~ ~ 7 M C)~~~~~~~~~~~~~~~~~~~~~~~~~C. I-')~~~~~~~~~~C 61 22 21 31 15 1 11 23 7 1 7 x 2 3 9 x 7 x 17 174 62 x x x x x x x x x x x x 12 174 39 x x x x 4 7 46 x x x x x x x x x x x 1 138 587 x x x 3 1 159 x x x x x 5 25 168 x x 2 0-5 170 x x x x 4 0-5 171 x x x x x 5 2 175 x x 2 0-5 183 x x x 3 2 185 x x 2 3 216 x x 2 3 217 x x x x x x 6 3C5 220 x 1 3 223 x x x x x 5 615 563 x x x 3 213 92 x x x 3 4 303 x x 2 3 313 x x x 3 317 403 x x x x x x 6 821 442 x x x 3 5 2 443 x x 2 08 444 x x x x 4 1 448 x x x x 4 327 461 x x x x 6 8 469 x x x 3 2 No. of sites 2 2 2 25 4 23 1 5 10 8 1 2 2 17 1 2 1 2 1 % 68 30 47 62 45 100 86 0 37 54 43 34 13 24 64 32 13 50 0 area of remnant (ha); connectivity index (number of remnants that a remnant is connected to by linear strips of native vegetation); the area of native vegetation within a 5 km radius of the remnant; distance from the edge of the remnanto the nearest native vegetation. Details are given in Abensperg-Traun et al. (1996). Within each experimental remnant a representative quadrat, 50 x 50 m, was established within which microhabitat variables were measured in 25 sub-quadrats (0 25 M2) laid out in a 5 x 5 grid with 10-m intervals. Three quadrats were set up in the larger 'control' remnant and two in the small 'control' remnant. The variables used were: presence or absence of a cryptogamic (lichen) crust, leaf litter cover rating, rating of woody litter (branches and tree trunks >5 mm diameter), percentage cover of weeds, number of sheep pellets, length of tree logs. For the whole quadrat, the following variables were measured: number of trees, percentage cover of shrubs and mallee (multistemmed eucalypts). Details of these variables and the statistics used are given in Abensperg-Traun et al. (1996).

1305 G. T. Smith et al. STATISTICAL ANALYSIS Initially, a correlation matrix was calculated to examine relationships between individual habitat and biogeographic variables. Relationships between fauna (dependent variables) and habitat and biogeographic (independent) variables were examined in two ways. First, a step-wise regression (Poisson distribution) procedure was used. This procedure selects the variables that maximize the reduction in residual variance. The regression was allowed to fit a maximum of four independent variables. The test for addition of a variable was a statistically significant reduction in variance with the addition. The relationship between the presence/absence of individual species and habitat and biogeographic variables was analysed using step-wise logistic regression. Secondly, remnants were placed arbitarily into one of three disturbance classes (DC 1, DC2, DC3) based on sheep faecal pellet density and weed cover, which can be used as measures of grazing intensity. DC1 = low weed and pellet scores (n = 5), DC2 = moderate weed and pellet scores (11 sites), DC3 = high pellet and weed scores (n = 10). Details are given in Abensperg-Traun et al. (1996). To control for the effects of remnant biogeography on disturbance classifications, analysis of covariance using the four biogeographic variables was used to test for differences in the lizard variables across the three disturbance (DC) classes. Poison distribution was used for the data. Genstat 5 2 was used for all calculations. Results Nineteen lizard species comprising six geckos (Gekkonidae), eight skinks (Scincidae), two dragons (Agamidae), one legless lizard (Pygopodidae) and two monitors (Varanidae) were recorded in the gimlet woodlands (Table 1). In the larger 'control' remnant (remnant 6), pitfall trapping captured 14 species, the most common were the geckos, Diplodactylus granariensis, D. pulcher, Gehyra variegate and Oedura reticulate and the skink Lerista muelleri (Table 1). The presence of three other species (Tiliqua occipitalis, Pygopus lepidopodus, Varanus tristis) was recorded from opportunistic observations. During the systematic searches, 12 of these species were recorded in this remnant. The species not recorded in the systematic searches were rare and were only recorded once, either from other observations or pitfall trapping (Table 1). In the smaller 'control' remnant (remnant 46) 11 species were recorded. With the exception of Heteronotia binoei, which was not found in the larger 'control' remnant, they were the more abundant species captured in the larger 'control' remnant (Table 1). Ten species were recorded in the 24 experimental remnants. The most common species were Gehyra variegata, Qedura reticulatand Tiliqua rugosa which were recorded in 23, 21 and 15 remnants, respectively. The other species were found in one to eight remnants. Individual remnants had from one to six species. Only four habitat and three biogeographic variables had significant correlation coefficients with the number of species of geckos, skinks, other lizards (dragons, legless lizards and monitors) and total lizard species and are given in Table 2. Geckos had significant positive correlations with woody litter and area and a negative correlation with distance to nearest native vegetation. Skinks had positive correlations with lichen, woody litter,% shrub cover, connectivity and area and a negative correlation with percentage weed cover. The larger lizards in the other lizard group had positive correlations with percentage shrub cover, area, connectivity and a negative correlation with distance to nearest native vegetation. The combined lizard groups had positive correlations with lichen, woody litter, % shrub cover, area and connectivity and negative correlations with percentage weed cover and distance to nearest vegetation. Stepwise linear regression analysis between habitat and biogeographic variables and species numbers in the various lizard groups showed a number of significant relationships (Table 3). Geckos had positive relationships with woody litter and the number of trees and a negative relationship with the distance to the nearest native vegetation. Skinks only had a positive relationship with the percentage of shrub cover as did other lizards which also had a positive relationship with area and connectivity. Total lizards had positive relationships with woody litter and area and a negative relationship with the distance to the nearest native bush (Table 3). The relationship between species number and area of remnant was further examined for the experimental remnants by linear regression using the equation S = C + Z (In A) where S is number of species, A is the area of remnant, and C and Z are constants. The resultant equation was S = 2 75 + 0 687 In A with an R2 value of 0 210 and F = 5 85, P < 0 05. The mean and standard deviation of the number of species in each lizard group in the three disturbance classes show an overall decline in species number with increasing levels of disturbance (Table 4). However, after analysis of covariance to remove the effects of biogeographic variables, there were no significant differences in the means for any lizard groups between disturbance classes. Geckos and other lizards had no significant biogeographic covariates whereas for skinks and total lizards, connectivity was significant (Table 4). There were sufficient data on the presence or absence of six species to examine their relationship with the habitat and biogeographic variables using stepwise multiple logistic regression. G. variegata, which was absent from only one remnant, was not included in the analysis. Heteronotia binoei, Lerista macropisthopus and L. muelleri had no significant relationships with any habitat or biogeographic vani-

1306 Lizard response to fragmentation and grazing Table 2. Significance of correlation coefficients for significant relationships between number of species in the different groups of lizards and habitat and biogeographic variables. Other lizards = dragons, legless lizards and monitors; -= negative relationship. See Methods for list of variables Habitat Geckos Skinks Other lizards Total Lichen ** * Woody litter ** ** Weed cover (%) -* Shrub cover (%) *** *** Area ** ** *** Connectivity *** Distance to nearest vegetation -* *P < 0-05, **P < 0-01, ***P < 0.001. Table 3. Summary of stepwise linear regression analysis showing significant relationships between dependent lizard variables and independent habitat and biogeographic variables. Direction of effect is indicated by + ve positive or - ve negative. DNB = Distance to nearest native vegetation; Con = Connectivity Independent variables associated with Dependent lizard variables Soil surface Vegetation structure Remnant biogeography Geckos Woody litter** Trees (no.)* DNB - ve*** Skinks None Shrub cover (%)** None Other lizards None Shrub cover (%)*** Area + ve*** Con. + ve* Total lizards Woody litter*** None Area + ve*** DNB - ve* *P < 005; **P < 001, ***P < 0001. Table 4. Means (? SD), before (in parenthesis) and after adjustment for significant biogeographi covariates, deviances and list of significant biogeographicovariates for lizard variables after analysis of covariance (Poisson distribution) of remnants grouped into disturbance classes. Sample sizes are given in parentheses. No analysis on 'other' lizard due to no data for DC2 and DC3 Disturbance classes Deviance Significant Lizard assoc. with biogeographic variables 1(5) 2(11) 3(10) disturbance covariates Gecko spp. 210 + 102 228 + 053 198? 041 020 None (3 5? 1-3) (22? 1-0) (1 9? 06) Skink spp. 2 53 + 1 17 1 48 + 0 40 1 09 + 0-36 1 91 Connectivity (3 91 + 1.6) (0-8 + 12) (1.0 + 0 9) + ve*** Other lizard spp. (2-5 + 2-0) (0) (0) Total lizard spp. 4-37 + 1-39 3-94? 0-69 3 20? 0-62 0.99 Connectivity (86 + 37) (3.0 + 1.4) (29 + 12) + ve*** Area + ve* P < 005; **P < 001; ***P < 0001. able. Qedura reticulate and Menetia greyii both had significant positive relationshps with the amount of leaf litter (deviance 1468, P < 0 001 and 5 38, P < 0 05, respectively). Tiliqua rugosa had a significant positive relationship with the percentage shrub cover (Deviance 10 58, P < 0 01). Discussion Our study was essentially a snap-shot of the presence or absence of species some decades after fragmentation had been completed and after varying degrees of livestock grazing and trampling had impacted

1307 G. T. Smith et al. on the remnants. The gimlet woodlands we studied formed part of a widespread salmon gum E. salmonophloia/gimlet association. It was most common in the alluvial valleys, which were the first areas to be cleared for agriculture, hence our lack of recently isolated remnants (Beard& Sprenger 1984; McArthur 1993). The major habitat change was the reduction or elimination of the understorey shrub cover and woody litter, and the number of trees. The latter is in part, due to logging in the early 1940s and the lack of regeneration due to livestock and rabbit grazing. Livestock trampling causes considerable changes in the soil surface, reducing water infiltration, and root growth (Proffitt et al. 1993; Pettit, Froend & Ladd 1995). This effect was seen most clearly in the decline in the presence of a lichen crust and to a lesser extent in the decrease in the quantity of leaf litter and woody litter. These changes, together with the input of nutrients from livestock and the surrounding paddocks resulted in a profusion of weeds in the more degraded remnants (Hobbs & Atkins 1988; Hester & Hobbs 1992). In relation to the impact of livestock grazing and trampling, the least disturbed remnants were larger, had greater connectivity and were closer to other areas of native vegetation. However, we have no data on the history of grazing in these remnants and hence the present grazing intensity may not reflect past farming practices. The question of how representative the lizard fauna recorded in remnant gimlet woodland was of the original fauna, was addressed by comparing the data with those from Chapman & Dell (1985), who calculated the percentage of specimens in the Western Australian Museum collected from all types of vegetation associations throughouthe wheatbelt; results from different types of woodland associations were combined. Fifteen species were collected most commonly in woodland; of these species, only Diplodactlus mainii (50% of specimens collected in woodlands), Delma fraseri (39%) and Ctenophorus reticulatus (43%) were not recorded in our remnants. Pitfall trapping in other vegetation associations (including other woodland associations) in the study area (G.T. Smith, unpublished data) suggest that woodland is not the preferred habitat for these three species in our study area. The other seven species recorded in our study had percentage capture rates in wheatbelt woodlands ranging from 0 to 43% (Chapman& Dell 1985), but the number of individuals were low. Comparison of the 19 species found in this study with the species recorded in woodland by Chapman & Dell (1985), as well as those species found in other undisturbed woodlands in the study area (G.T. Smith, unpublished data), sugges that the assemblage of species we found in gimlet woodlands, probably closely resembled that found in extensive areas of gimlet woodland prior to fragmentation. If this is correct, then even our large 'control' remnant (17 species) may have lost species (or the populations were so small that we did not find them) and the smaller remnants have lost even more species, assuming that all species were in the area of the remnant prior to fragmentation. The 10 species found in the experimental sites consisted of two specialists (0. reticulate, C. ocellatus) and eight generalists. Five specialists and four generalists were not found in the experimental sites. These findingsupport those of Kitchener (1982), Kitchener & How (1982) and Humphreys & Kitchener (1982) on the greater ability of generalists to persist in the smaller degraded remnants. In part, loss of larger species, whether generalists or specialists, is a function of available habitat and hence population size. The relationship of body mass to size of home range found by Turner, Jennrich & Weintaub (1969) suggests that most of our experimental remnants could support, at best, only a few larger lizards. The presence of two large generalists (T. rugosa, V. gouldii) in some of the remnants is more a reflection of the remnants being within the foraging ranges of these species, than of the remnants having the resources to support them permanently. Both species commonly use road-side verges and have no difficulty in crossing agricultural land. The most common species, G. variegate and 0. reticulate, were found in all but one and three remnants, respectively. The widespread occurrence of G. variegate, a generalist, was expected as it is commonly found in a wide variety of habitats, both disturbed and undisturbed. Despite small populations (<20) in some remnants, Sarre et al. (1995) suggested that it is able to maintain its widespread distribution through a metapopulation structure. On the other hand, 0. reticulate, a specialist, has persisted well because it is arboreal and thus to some extent buffered, so far, from the changes that have taken place in the remnants (Kitchener, How & Dell 1988). Sarre et al. (1995) found that persistence of this species was related to the number of trees; in our smaller but overlapping sample of remnants, the number of trees in a remnant was not a significant factor. However, there is evidence to sugges that the populations in the remnants are not part of a metapopulation and their long-term persistence in doubt (Sarre 1995; Sarre et al. 1995). Longevity in both species is in excess of 16 years and it may well be that the presence of the species in some remnants may representhe final, nonbreeding members of the population (How & Dell 1994). The other common species, T. rugosa is widespread throughouthe wheatbelt in all types of undisturbed and disturbed habitat. In most remnants, only one or two animals were recorded, and they cannot be considered a 'population'. In most cases their presence in the smaller remnants was probably ephemeral and thus the question of their persistence in small remnants is of little immediate conservation concern. Among the vertebrates, lizards are the group that has coped best with the fragmentation of habitat (Kitchener et al. 1980a; Burbidge & McKenzie 1989), primarily because of their low energy and space

1308 Lizard response to fragmentation and grazing requirements compared to mammals and birds (Nagy cover is related to the importance of structural com- 1987; Turner et al. 1969). These characteristics allow plexity for many lizard species (Pianka 1966; Jones larger populations to persist in smaller reserves, which 1981). possibly have reduced resources. Initial studies of liz- Initial inspection of the data in relation to the disards on 23 nature reserves (34-5119 ha) in the wheat- turbance classes suggested a definite relationship belt by Kitchener et al. (1980a) found significant cor- between the intensity of the effects caused by grazing relations between lizard richness and the area of and trampling and the number of species present. reserves as well as the area of woodland within the However, analysis of covariance with biogeographic reserves. Only 'other' lizards and total lizards showed variables held constant, showed no significant effect a significant positive relationship between species rich- of disturbance, but with area and connectivity as signess and area, a relationship that is driven by the nificant covariates. Analysis of covariance with dislarger spatial requirement of the larger lizards in the turbance held constant confirmed the importance of 'other' lizard group. The species-area relationship for area and connectivity. The confounding influences of the experimental remnants was not significantly disturbance and biogeographic factors are difficult to different to that found by Kitchener & How (1982) separate because of the history and type of clearance, in a comparable study of small salmon gum/gimlet similar problems have been encountered in other taxa remnants, even though only 22% of the variance was (Norton, Hobbs & Atkins 1995). While this may be explained by area. This was in part due to the varia- true at the group level, it is almost certainly not true bility in the number of species (one to four) present in for individual species. The absence of Diplodactylus the smaller (< 1 ha) remnants, compared to the range pulcher and D. granariensis (both use other species' of species (one to six) in the larger remnants. Our (spiders, etc.) burrows for shelter), from the experinitial analyses included time since the remnant imental sites which had all been grazed, strongly sugreached its present status (only two broad time periods gests that the habitat changes caused by livestock could be used); because there were no significant grazing and trampling (directly or indirectly), is most relationships, it was dropped from subsequent analyses. likely the factor responsible for their absence. This This result mirrors that of Kitchener et al. (1980a), conclusion is supported by the decline in the numbers but the explanation must differ. Rather than habitat of mygalomorph spiders with increasing disturbance providing the explanation (ours is uniform) we suggest and the total absence of some species as well as some that the initial changes in the assemblage were rela- species of Lycosid spiders from the experimental remtively fast and that subsequent changes were and will nants (Abensperg-Traun et al. 1996). The dominance continue to be slow. In fact, some changes may have of generalists over specialists also suggest that distaken place during the initial period of clearing leading turbance from livestock grazing has had a detrimental to the present spatial configuration of remnants. In effect on the lizard fauna. Without detailed inforsome areas it may have taken 70-80 years from the mation on the stocking history of these remnants, it onset of clearing to reach the present state (Arnold& is impossible to preclude the possibility that more Weeldenburg 1991). The distance to the nearest area intense grazing in the past was responsible for the of native vegetation has significant negative relation- absence of some species today. ships with both geckos and total lizards, suggesting Whereas groups of lizards apparently were influthat for the former their dispersal abilities in the pre- enced by a small number of habitat variables, indisent landscape are limited and hence the probability vidual species showed few significant positive relationof recolonizing a remnant is low. Studies on Gehyra ships. The amount of cover provided by shrubs is variagata and Oedura reticulate (Sarre et al. 1995; important for T. rugosa as is the leaf litter cover for Sarre 1995) support this conclusion, which may apply M. greyii (G.T. Smith, unpublishe data), on the other also to some of the other smaller lizards. Connectivity hand, leaf litter is of no apparent importance to the was significant for 'other' lizards. This is related to arboreal 0. reticulate (Kitchener et al. 1988) but may the positive relationship with area, as the connecting reflecthe importance of old healthy trees. The paucity linear strips of vegetation are in effect increasing the of significant habitat variables is perhaps a reflection area of available habitat as well as providing corridors of the fact that, with exception of 0. reticulate, all for movement. Habitat variables showed a limited the species analysed were generalists, which have few number of significant relationships with the lizard specific habitat requirements. taxa. The significant positive relationship between The lizard taxa in this study had few significant geckos and the number of trees is not surprisingiven relationships with the biogeographic and habitat varithe dominance of the arboreal 0. reticulate and semi ables used in this study. The reasons for this are twoarboreal G. variagata in our sites and supports the fold. First, the historical pattern of clearing precluded finding of Sarre et al. (1995) that the number of trees the possibility of obtaining a set of remnants whose in a remnant is the best predictor of the probability of biogeographicharacteristics were independent of disoccurrence for 0. reticulata. The positive relationship turbance-related variables. In general, the older rembetween gecko and total lizards and woody litter and nants were smaller, more isolated and have suffered between skinks and other lizards and percentage shrub greater disturbance. Secondly, the lizard fauna is

1309 G.T. Smith et al. dominated by small species with low spatial requirements and general habitat requirements; factors that increase the probability of these species surviving in small degraded habitats. The resultsuggesthat while biogeographic variables may influence the persistence of some taxa (species) and that disturbance from grazing and trampling has a detrimental effect on species persistence, the factors are confounded within our limited sample of remnants. Data from studies of lizards in the study area suggest that the present system of nature reserves, although it may be inadequate for many taxa (Hobbs et al. 1993b) has been adequate for the regionalizard fauna up to now. However, remnants will continue to be affected by external factors that cause degradation of remnants and which in time, will impinge on the ability of lizards to persist regardless of stochastic processes. Despite the fact that these remnants have a depauperate lizard fauna dominated by generalists; the present study shows clearly that even the smallest and most degraded remnants have some conservation value and that area may not be the most impoertant factor influencing persistence in a remnant. In the light of the present community awareness of the need to restore and repair damage to the ecosystem caused by agricultural development, small remnantsuch as the ones in this study have value, in that they can be used as nodes or nuclei in revegetation plans to redress both agricultural and conservation problems. Acknowledgements We thank the community in our study area for their help in providing free access to their properties and Earthwatch volunteers for help in setting up the pitfall traps. David Fox and Bert De Boer provided valuable statistical advice and Sonya Crute, Donelle Wallis, and Debbie Lister, typed the manuscript. Mike Brooker and John Dell made valuable comments on an earlier draft. The study was partially funded by the Australian National Parks and Wildlife Service (Save the Bush Programme, Project No. R036). References Abensperg-Traun, M., Smith, G.T., Arnold, G.W. & Steven, D.E. (1996) The effect of habitat fragmentation and livestock grazing on animal communities in remnants of gimlet Eucalyptusalubris woodland. I. Arthropods. Journal of Applied 1281-1301. Angelstam, P. & Arnold, G.W. (1993) Contrasting roles of remnants in old and newly impacted landscapes: lessons for ecosystem reconstruction. Nature Conservation 3. Reconstruction of Fragmented Ecosystems. Local and Regional Perspectives (eds D. A. Saunders, R. J. Hobbs & P. R. Ehrlich), pp. 109-125. Surrey Beatty & Sons, Chipping Norton. Arnold, G.W. & Weeldenburg, J.R. (1991) The distributions and characteristics of remnant native vegetation in parts of the Kellerberrin, Tammin, Trayning and Wyalkatchem shires of Western Australia. CSIRO Division of Wildlife and Ecology, Technical Memorandum No. 33, Gungahlin, ACT, Australia. Beard, J.S. & Sprenger, B.S. (1984) Geographical data from the vegetation survey of Western Australia. Vegetation survey of Western Australia. Occassional Paper No. 2. Vegmap Publications, Perth. Burbidge, A.A. & McKenzie, N.L. (1989) Patterns in the modern fauna decline of Western Australia's vertebrate fauna: causes and conservation implications. Biological Conservation, 50, 143-198. Bury, R.B. & Busack, S.D. (1974) Some effects of off-road vehicles and sheep grazing on lizard populations in the Mohave desert. Biological Conservation, 6, 179-183. Chapman, A. & Dell, J. (1985) Biology and zoogeography of the amphibians and reptiles of the Western Australian wheatbelt. Records of the Western Australian Museum, 12, 1-46. Cutler, A. (1991) Nested faunas and extinction in fragmented habitiats. Conservation Biology, 5, 496-505. Fahrig, L. & Merriam, G. (1994) Conservation of fragmented populations. Conservation Biology, 8, 50-59. Fleischner, T.L. (1994) Ecological costs of livestock grazing in Western North America. Conservation Biology, 8, 629-644. Friedel, M.H. & James, C.D. (1995) How does grazing of native pastures affec their biodiversity? Conserving Biodiversity. Threats and Solutions (eds R. A Bradstock, T. D. Auld, D. A. Keith, R. T. Kingsford, D. Lunney & D. P. Sivertsen), pp. 249-59. Surrey Beatty & Sons, Chipping Norton. Friend, G.R., Smith, G.T., Mitchell, D.S. & Dickman, C.R. (1989) Influence of pitfall and drift fence design on capture rates of small vertebrates in semi-arid habitats of Western Australia. Australian Wildlife Research, 16, 1-10. Hester, A.J. & Hobbs, R.J. (1992) Influence of fire and soil nutrients on native and non-native annuals at remnant vegetation edges in the Western Australian Wheatbelt. Journal of Vegetation Science, 3, 101-108. Hobbs, R.J. & Atkins, L. (1988) The effect of disturbance and nutrient addition on native and introduced annuals in the Western Australian wheatbelt. Australian Journal of Ecology, 13, 171-179. Hobbs, R.J. & Saunders, D.A. (1993) (eds) Reintegrating Fragmented Landscapes. Towards Sustainable Production and Nature Conservation. Springer-Verlag, New York. Hobbs, R.J., Saunders, D.A. & Arnold, G.W. (1993a) Integrated landscape ecology: a Western Australian perspective. Biological Conservation, 64, 231-258. Hobbs, R.J., Saunders, D.A., Lobry de Bruyn, L.A. & Main, A.R. (1993b) Changes in the biota. Reintegrating Fragmented Landscapes. Towards Sustainable Production and Nature Conservation (eds R. J. Hobbs & D. A. Saunders), pp. 65-106. Springer-Verlag, New York. How, R.A. & Dell, J. (1994) The zoogeographic significance of urban bushland remnants to reptiles in the Perth region, Western Australia. Pacific Conservation Biology, 1, 132-40. Humphreys, W.F. & Kitchener, D.J. (1982) The effect of habitat utilization on species-area curves: implications for reserve area. Journal of Biogeography, 9, 391-396. Jones, K.B. (1981) Effects of grazing on lizard abundance and diversity in Western Arizona. The Southwestern Naturalist, 26, 107-115. Kitchener, D.J. (1982) Predictors of vertebrate species richness in nature reserves in the Western Australian wheatbelt. Australian Wildlife Research, 9, 1-7. Kitchener, D.J., Chapman, A., Dell, J., Muir, B.G. & Palmer, M. (1980a) Lizard assemblages and reserve size and structure in the Western Australian wheatbelt - some implications for conservation. Biological Conservation, 17, 25-62.

1310 Lizard response to fragmentation and grazing Kitchener, D.J., Chapman, A., Muir, B.G. & Palmer, M. (1980b) Conservation value for mammals of reserves in the Western Australian wheatbelt. Biological Conservation, 18, 179-207. Kitchener, D.J., Dell, J., Muir, B.G. & Palmer, M. (1982) Birds in Western Australian wheatbelt reserves - implications for conservation. Biological Conservation, 22, 127-163. Kitchener, D.J. & How, R.A. (1982) Lizard species in small mainland habitat isolates and islands off south-western Western Australia. Australian Wildlife Research, 9, 357-363. Kitchener, D.J., How, R.A. & Dell, J. (1988) Biology of Qedura reticulate and Gehyra variegate (Gekkonidae) in an isolated woodland of Western Australia. Journal of Herpetology, 22, 401-412. Lord, J.M. & Norton, D.A. (1990) Scale and spatial concept of fragmentation. Conservation Biology, 4, 197-202. McArthur, W.M. (1993) History of landscape development. Reintegrating Fragmented Landscapes: Towards Sustainable Production and Nature Conservation (eds R. J. Hobbs & D. A. Saunders), pp. 10-22. Springer-Verlag, New York. Nagy, K.A. (1987) Field metabolic rates and food requirement scaling in mammals and birds. Ecological Monographs, 57, 111-128. Norton, D.A., Hobbs, R.J. & Atkins, L. (1995) Fragmentation, disturbance, and plant distribution: mistletoes in woodland remnants in the Western Australian wheatbelt. Conservation Biology, 9, 426-438. Nunney, L. & Elam, D.R. (1994) Estimating the effective population size of conserved populations. Conservation Biology, 8, 175-184. Pianka, E.R. (1966) Convexity, desert lizards and spatial heterogeneity. Ecology, 47, 1055-1059. Pettit, N.E., Froend, R.H. & Ladd, P.G. (1995) Grazing in remnant woodland vegetation: changes in species composition and life form groups. Journal of Vegetation Science, 6, 121-130. Proffitt, A.P.B., Bendotti, S., Howell, M.R. & Eastham, J. (1993) The effect of sheep trampling and grazing on soil physical properties and pasture growth for red-brown earth. Australian Journal ofagricultural Research, 44, 317-331. Sarre, S. (1995) Size and structure of populations of Oedura reticulata (Gekkonidae) in woodland remnants: regional distribution of a currently common species. Australian Journal of Ecology, 20, 288-298. Sarre, S., Smith, G.T. & Meyers, J.A. (1995) Persistence and metapopulation structure of two species of gecko (Qedura reticulate and Gehyra variagata) in remnant habitat. Biological Conservation, 71, 25-33. Saunders, D.A., Hobbs, R.J. & Margules, C.R. (1991) Biological consequences of ecosystem fragmentation: a review. Conservation Biology, 5, 18-34. Smith, G.T. (1987) The changing environment for birds in the south-west of Western Australia: some management implications. Nature Conservation: The Role of Remnants of Native Vegetation (eds D. A. Saunders, G. W. Arnold, A. A. Burbidge & A. J. M. Hopkins), pp. 269-277 Surrey Beatty & Sons, Chipping Norton. Turner, F.B., Jennrich, R.I. & Weintaub, J.O. (1969) Home ranges and body size of lizards. Ecology, 50, 1076-1081. Wallace, K.J. & Moore, S.A. (1987) Management of remnant bushland for nature conservation in agricultural areas of southwestern Australia - operational and planning perspectives. Nature Conservation. 1. The Role of Remnants of Native Vegetation (eds D. A. Saunders, G. W. Arnold, A. A. Burbidge & A. J. M. Hopkins), pp. 259-268. Surrey Beatty & Sons, Chipping Norton. Received 25 November 1994; revision received 6 December 1995