* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Download AND Gehyra variegata) IN REMNANT HABITAT
Overexploitation wikipedia , lookup
Latitudinal gradients in species diversity wikipedia , lookup
Restoration ecology wikipedia , lookup
Wildlife corridor wikipedia , lookup
Conservation biology wikipedia , lookup
Biogeography wikipedia , lookup
Island restoration wikipedia , lookup
Holocene extinction wikipedia , lookup
Assisted colonization wikipedia , lookup
Biodiversity action plan wikipedia , lookup
Theoretical ecology wikipedia , lookup
Mission blue butterfly habitat conservation wikipedia , lookup
Occupancy–abundance relationship wikipedia , lookup
Source–sink dynamics wikipedia , lookup
Habitat destruction wikipedia , lookup
Decline in amphibian populations wikipedia , lookup
Reconciliation ecology wikipedia , lookup
Extinction debt wikipedia , lookup
Molecular ecology wikipedia , lookup
Biological Dynamics of Forest Fragments Project wikipedia , lookup
Biological Conservation 71 (1995) 25-33 0 1994 Elsevier Science Limited Printed in Great Britain. All rights reserved 0006-3207/95/$9.50 ELSEVIER 0006-3207(94)00017-4 PERSISTENCE OF TWO SPECIES OF GECKO (Oedura reticulata AND Gehyra variegata) IN REMNANT HABITAT Stephen Molecular Evolution and Systematics, Sarre Research School of Biological Sciences, Australian National University, PO Box 475, Canberra, ACT, 2601, Australia Graeme T. Smith CSIRO, Division of Wildlife and Ecology, LMB 4, Midland, WA 6056, Australia & Jacqueline A. Meyers CSIRO, Division of Wildlife and Ecology, PO Box 84. Lyneham, (Received 7 January generalists, dispersal, 2602, Australia 1994; revised version received and accepted 28 February 1994) INTRODUCTION Abstract Identifving the characteristics that make a species vulnerable to extinction jtillowing habitat fragmentation is one of the most pressing problems in conservation biology. One suggestion is that habitat specialists are less able to cope with rapid changes to their habitat or to move through the modljied landscape than habitat generalists and so will be more vulnerable to extinction. We examine this hypothesis by comparing the distributior of two species of gecko, Gehyra variegata and Oedura reticulata, in patches of remnant woodland. The former species is a habitat generalist relative to the latter and shows a markedly higher level of persistence (97% remnant occupancy vs 72%). Logistic regression modelling of the presence or absence of 0. reticulata revealed a sig@cant correlation between the number of smoothbarked eucalypts (both species preferred by 0. reticulata) in the remnant and the presence of 0. reticulata. This suggests that the probability of extinction for a given population is related to the amount of suitable habitat in the remnant and is a function of processes operating at the population level rather than on a regional basis. PitfalI trapping in this study and evidence from another long-term study suggest that the movement of 0. reticulata between remnants is negligible. As a consequence, this species has been unable to form a metapopulation at equilibrium. In contrast, it is likely that G. variegata is maintaining its widespread distribution through a metapopulation structure. These results demonstrate the importance of the ability to form a metapopulation for a species to maintain persistence in recently and highly fragmented ecosystems. Keywords: persistence, habitat cialists, lizards, fragmentation, tion, extinction. ACT, Clearing natural vegetation for agriculture or other development has occurred at an unprecedented rate in recent centuries. The survival of species within a landscape of fragmented natural habitat will depend on their ability to exploit new habitats such as agricultural lands and to cope with the changes in their habitat. These changes include a reduction in the area of available habitat, changed spatial relationships of habitat, increased fluxes of solar radiation, wind and water and the disruption of ecological processes such as nutrient cycling, energy transfer, competition and predation (Saunders et al., 1991). One major consequence of clearing native vegetation is that continuous or near continuous populations of species are broken into smaller and more insular local populations. Small populations are likely to be more vulnerable to extinction through stochastic demographic, genetic and environmental processes (Shaffer, 1987) although what constitutes a minimum viable population size is contentious (Shaffer, 1981; Soul& 1987; Nunney & Campbell, 1993) and will vary between species and ecosystems. The relative importance of extinction pressures related to population size, as opposed to the ‘driven’ causes of extinction (Caughley, 1994), is difficult to assess. In many cases of humaninduced habitat fragmentation, the minimum viable population size may be irrelevant, because, regardless of the population size of the species, it is unable to cope with the array of changed conditions and becomes locally or even regionally extinct. The probability of the regional extinction of a species in a fragmented landscape will also be affected by the interaction between sub-populations of that species. If a set of populations interacts on a regional basis (i.e. habitat spemetapopula25 Stephen 26 Sarre, Graeme T. Smith, form a metapopulation: Levins, 1970; Gilpin, 1987; Hanski & Gilpin, 1991) then the extent and nature of the extinction rate will be affected (Opdam, 1990; Harrison, 1991; Verboom et al., 1991). Harrison (1991) has described four basic models of equilibrium and non-equilibrium metapopulation structures, but as yet few studies of metapopulation dynamics have been applied to recently fragmented ecosystems. Thus the importance of metapopulations to the conservation of species in fragmented ecosystems remains unclear. The interaction of subpopulations and the probability of their extinction is likely to vary between species. Consequently, the ability to predict the vulnerability of a given species to fragmentation is one of the most pressing problems facing conservation biologists. The breadth of habitat that a species can occupy may be one characteristic useful for predicting the vulnerability of a species to fragmentation, the general expectation being that habitat generalists will exhibit low vulnerability to habitat fragmentation (Opdam, 1990; Harris & Silva-Lopez, 1992; Hobbs, 1992). The basis for this expectation is that species with narrow niches are less able to exploit new habitats or disperse through the modified landscape (Harris & Silva-Lopez, 1992) than species with more general habitat requirements. In this paper we explore the relative abilities of two species of gecko (Oedura reticulata and Gehyra variegata) to persist in a fragmented environment. 0. reticulata is a habitat specialist relative to G. variegata (Kitchener & How, 1982; Kitchener et al., 1988). The hypothesis tested is that the habitat generalist G. varie gata will exhibit a higher degree of persistence in habitat remnants than the habitat specialist. A second, related prediction derived from island biogeographic theory (MacArthur & Wilson, 1967) is that the rate of extinction of both species will be related to the size of the habitat remnant, the distance to the nearest source population and the time since isolation. Specifically, the extinction rate is expected to increase with decreasing remnant size, increasing distance between remnants and increasing time since isolation. Jacqueline MATERIALS A. Meyers AND METHODS Life history characteristics of the two species of gecko The reticulated velvet gecko 0. reticulata (Bustard 1969) is restricted to the southwest of Western Australia although its distribution within the wheatbelt region is extensive (Chapman & Dell, 1985). It is also limited in its range of habitat, being arboreal and mainly confined to mature smooth-barked Eucalyptus species with dead wood (How & Kitchener, 1983). In contrast, the tree dtella G. variegata (Dumeril & Bibron 1836) is widely distributed over the southern half of Australia (Cogger, 1992) although its current systematic status is uncertain primarily because of chromosomal variation (King, 1979; Moritz, 1986; Cogger, 1992). Moritz (1986) identified G. variegata as the chromosomal form 2n = 40B, which is restricted to the southwest of Western Australia. Chromosomes from the Gehyra populations used in this study were not examined, but 2n = 40B is the only chromosome race found in the region (C. Moritz, pers. comm.) and so is probably the form of G. variegata studied here. Voucher specimens of the taxa studied will be lodged with the Western Australian Museum. G. variegata is a habitat generalist living in logs and fallen timber, tree trunks or shrubs (Bustard, 1968; Kitchener et al., 1988) and can also be found under rocks or in highly disturbed habitats (Kitchener et al., 1988; personal observations). A summary of the life history characteristics of both species obtained from a long-term population study (How & Kitchener, 1983; Kitchener et al., 1988) is provided in Table 1. Study area The study area is 1680 km* between Kellerberrin and Trayning in the Western Australian wheatbelt (Fig. 1). The district is gently undulating, with low relief (100 m) and occasional granite outcrops (Beard, 1980). Approximately 93% of the original vegetation has been cleared for agriculture since 1900, leaving over 450 vegetation remnants. These range in size from less than 1 ha to 1190 ha (Arnold & Weeldenburg, 1991). The dominant agricultural activity is wool and wheat production. Table 1. Summary comparison of basic life history parameters for the Konnongorring populations of 0. reticuluta and G. variegata (Modified from Kitchener et al., 1988) Species characteristics 0. retieulata G. variegata Habitat breadth” Adult sex ratio Sexual dimorphism Longevity (years) Annual population survival (x f SD) (%) Estimated survival to 24 months (%I) Adult females gravid each season (%) Clutch size Clutches/season ‘M adult size when sexually mature Age of sexual maturity (years) Mean SVL of reproductively active female (mm) 1.276 -1: 1 female > male -14 82.5 f 6.8 33350 c. 100 2 5161 -1: 1 female = male -14 86.7 +_3.4 39-50 100 1 2 90.0 -2.8 49.9 k 3.2 ‘Estimated from the frequency 94.6 -4.8 64.4 + 4.1 of use of perch sites. SVL = snout vent length Persistence of lizards in remnant habitat 21 Since 0. reticulata is most commonly found on smooth-barked eucalypts such as E. salubris and E. salmonophloia, the survey was restricted to woodland remnants containing these two species. All remnants surveyed were unfenced and located within paddocks used for sheep and wheat production. The former distribution of vegetation associations containing these species (as well as those containing E. wandoo, which is has been mapped also favoured by 0. reticulata) (McArthur, in press; M. G. Brooker, pers. comm.) and is presented in Fig. 2. This figure suggests that the likely distribution of 0. reticulata within the study area before European settlement was widespread but discontinuous. The current known distribution of 0. reticulata in the study area (Fig. 2) supports the notion that the species was widespread. G. variegata also shows a wide current distribution (Fig. 2) and given its broad habitat preferences would have been distributed over most of the region prior to clearing. Fig. 1. The location of the study area (dark shading) near Kellerberrin in Western Australia. The lightly shaded area indicates the extent of the Western Australian wheatbelt. Sampling The presence or absence of both species of gecko was determined by surveying 32 vegetation remnants (31 for G. variegata because its absence could not be confirmed in one remnant) in October-November 1989, January 1991 and February 1992. These remnants represent the total number of remnants containing homogeneous stands of E. salubrislE. salmonophloiu in the study area that have not been cut (i.e. harvested for timber) or burnt since isolation and that were isolated before 1937. Although some variation in smooth-barked eucalypt size structure was observed between remnants, an @ G. variegata present 0 0. reticulata present Fig. 2. The distribution of Oedura reticulata and Gehyra variegata within the study area. The dark shaded areas represent the habitat remnants, and the light shaded areas the distribution of habitat suitable for 0. reticulata before clearing (from McArthur, in press; M. G. Brooker, pers. comm.) 28 Stephen Sarre, Graeme T. Smith, intensive survey of habitat usage by 0. reticulata in three of the remnants (168, 170, 175, Sarre, in prep), found no preference by 0. reticulata for any of the tree characteristics (height, height of bole, diameter, degree of senescence, crown depth, number of trees with touching crowns, crown density and foliage cover) measured. Consequently, no assessment of smoothbarked eucalypt variation was made in the surveys described here. Each remnant was surveyed by two observers for a minimum of 2 h (or until both species were observed) from dusk onwards using head torches. More intensive surveys in several of the remnants (Sarre, unpublished data) showed that even at very low population densities both species are usually observed within 30 min of commencing the survey. Intensive hand searching was carried out during the day if one or both of the species were not seen during the night surveys. If the species were still not found it was deemed absent from the remnant. Environmental variables and data analysis Six variables thought to be important to the persistence of 0. reticulata (but not G. variegata, which was present in all but one of the remnants) were measured. These were owner, area (ha), total number of trees, total number of smooth-barked eucalypts (E. salubris + E. salmonophloia), distance from the nearest woodland (m) and years since isolation). The variable ‘owner’ was included to account for possible variation in management practices between owners that may affect the persistence of 0. reticulata. In the majority of cases, the remnants had either been managed by the same family or by the same pair of owners since clearing. The area of each remnant and the distance between remnants were calculated from aerial photographs. The total Fig. 3. The vegetation absent. The shaded Jacqueline A. Meyers number of trees and the total number of smoothbarked eucalypts were obtained by total counts within each remnant while the time since clearing was obtained from extensive surveys of all current and past land holders (G. W. Arnold & J. R. Weeldenburg, pers. comm.). If an exact clearing date was not available, then the midpoint of the estimated limits of the time of clearing was used. A further three variables, the proportion of the total number of trees that were smoothbarked eucalypts, the density of smooth-barked eucalypts (tree/ha) and the density of total trees, were also calculated to determine if habitat density was important to the persistence of 0. reticulata. Because the response variable has a binomial distribution, logistic regression analysis was used to determine the relationship between the presence of 0. reticulata in vegetation remnants and the variables. The continuous variables were fitted as both linear and quadratic functions to test for possible curvature in the response. Each independent variable was added to the null model and its level of significance determined using the change in deviance of the fitted model from that of the null model (Nicholls, 1991a,b). The variable that accounted for the greatest change in deviance was included in the model and then all other variables were added in turn to determine if they significantly improved the model. This process was continued until no additional variables significantly improved the model. The statistical package GLIM (McCullagh & Nelder, 1989) was used for all modelling. Movement between adjacent populations To obtain a measure of the extent of movement into and out of the remnants, 53 pitfall traps (45 X 30 cm) were installed at a spacing of 7 m, 15 m beyond the remnants surveyed for the presence/absence of Oedura reticulata. 0, 0. retiedata present; 0, 0. reticulata area is the distribution of habitat suitable for 0. reticulata before clearing (from McArthur, in press; M. G. Brooker, pers. comm.). Persistence of lizards in remnant habitat edge of remnant 170 (Fig. 3). A fibreglass drift fence (30 cm in height) was erected between the pitfalls so that the remnant was surrounded. Two in every three pitfalls were fenced to ensure that only those animals entering the remnant from the surrounding wheat fields could fall into them. The remaining pitfalls were fenced so that only the animals leaving the remnant could be captured. Both species have been caught in identical pitfalls used in other studies (G. Smith, personal observations). RESULTS Distribution of 0. reticulata The persistence of the two remnants varied markedly. from 28% of the remnants variegata was absent from 2). Consequently, a logistic and G. variegata species within the woodland 0. reticulata was absent (Table 2; Fig. 3), whereas G. only a single remnant (Table regression model could only 29 be developed for 0. reticulata. Five variables - remnant size, total number of trees, total number of E. salubris, total number of E. salmonophloia and total number of smooth-barked eucalypts were each significantly correlated with the presence of 0. reticufata, with the latter being the most significant (2 = 8.03, p < 0.005, d.f. = 1; Table 3). Interestingly, neither time since isolation nor distance to the nearest woodland were significant explanatory variables for the presence of 0. reticulata. The size of each remnant and the distance to the nearest woodland varied over several orders of magnitude (0.1-8.0 ha for the former and O-l 100 m for the latter; Table 2), providing a good range with which to examine the effects of these variables on persistence. In contrast, the difference between the ages of the remnants was small (54-78 years; mean = 69, SD = 8.0), which may preclude the possibility of obtaining a significant relationship between age and the presence of 0. reticulata if the relationship is weak. The 32 remnants sampled were on properties owned Table 2. The data collected for each vegetation remnant surveyed Remnant” 46a 168 170 171 175 183 185 216 217 220 222 223 269 277 303 313 319 442 443 444 446 449 461 469 S84 S85 S128 S136 s143 s147 S167 s173 Ownerh Dixon Beresford Anderson Anderson Gorfin Barnes Riley Ryan Laird Reedy Reedy Reedy Dixon Leake Moore Leake MacNeil Reed Ryan Ryan Neck Rolston Hocking Neck Barnes Barnes Anderson Enright Beresford Beresford Ryan Nelson Size (Ha) 1.0 0.6 0.5 1.4 0.6 4.0 3.0 4.0 3.5 2.9 2.8 6.5 2.3 4.0 3.0 3.7 1.8 5.4 0.8 1.0 5.6 3.7 8.0 2.2 0.4 0.8 0.1 0.1 0.5 0.3 0.4 0.1 Distance to nearest woodland (ml Year(s) of isolation 100 250 200 1920-10 1920 1912 1912 1930 1936 19:5 1909-20 1925 1920 1910-20 9 1O-20 910-20 92040 1930 9lCk-20 910-20 91620 909-25 909-25 910-20 93040 925-30 910-20 1936 1936 192040 1910-20 1926 1920 1909-25 1910-20 150 1000 350 400 600 0 300 250 300 0 50 1100 900 180 4.50 800 800 800 0 50 550 500 700 350 300 600 400 600 300 Years since isolation No. E. salub.’ No. E. salmd Total Trees Presence absence’ Oedura 60 70 78 78 60 54 75 75 65 70 75 75 75 60 60 75 75 75 73 73 75 55 63 75 54 54 60 75 65 70 73 75 177 138 89 140 76 443 304 570 951 279 190 270 291 314 206 307 120 350 68 49 200 520 497 191 6 60 14 8 15 17 13 22 12 0 28 0 10 0 0 0 13 4 7 45 9 159 4 0 88 21 0 0 124 1.50 50 5 8 0 0 0 0 0 3 0 190 138 119 160 108 449 320 570 970 297 219 315 300 473 270 313 208 405 69 52 334 800 606 196 33 60 14 8 40 21 24 22 1 1 1 1 1 1 1 1 1 0 0 1 1 1 0 1 1 1 0 1 1 1 1 1 1 1 0 0 0 1 0 0 Gehyra 1 1 1 1 1 1 1 1 1 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 “The number designated for each remnant by CSIRO (M. Brooker, pers. comm.), where prefixed by S there is no CSIRO number. ‘Current owner. ‘Eucalyptus salubris. dE. salmonophloia. ‘1, present; 0, absent. 30 Stephen Sarre, Graeme T. Smith, Jacqueline A. Meyers Table 3. Summary of the fitted statistical model to estimate the presence of Oedura reticulata in remnant patches of vegetation The variables are Owner, current owner; Size, area of remnant (ha), Distance, distance from the nearest woodland (m); Age, years since isolation; E. salubris, the number of E. salubris per remnant; E. salmonophloia, the number of E. salmonophloia per remnant; Total trees, the total number of trees per remnant; Smooth-barked eucalypts, total number of E. salubris + E. salmonophloia. The model assumes a binomial distribution for 0. reticulata. The levels of significance (ns, p > 0.05, *0.05 Zp < 0.01; **O.Ol >p > 0.005; ***p <0.005) refer to the change in deviance associated with the term added at each step. Model Null model Owner Size Distance Age E. salubris E. salmonophloia Smooth-barked Total trees Smooth-barked Smooth-barked Total trees/size Current model Smooth-barked +owner +size +distance +age +total trees +total trees/size +smooth-barked +smooth-barked Final model Smooth-barked eucalypts eucalyptsitotal eucalyptsisize eucalypts eucalyptsitotal eucalypts/size eucalypts Deviance d.f. 38.02 15.28 33.22 36.80 37.83 31.19 Change in deviance d.f P 22.75 4.80 1.22 0.19 6.83 7.13 8.03 7.20 1.50 1.12 0.87 18 1 ns * 1 30.89 29.99 30.80 36.52 36.90 37.15 31 13 30 30 30 30 30 30 30 30 30 30 1 1 1 1 1 ns ns ** ** *** ** 1 1 1 ns ns ns 29.99 8.29 29.41 29.86 29.68 27.83 29.58 29.95 29.70 30 29 29 29 29 29 29 29 29 21.69 0.58 0.13 0.31 2.16 0.41 0.04 0.29 18 1 1 ns ns ns ns ns ns ns ns 29.99 30 8.03 by 19 different farmers with no more than three remnants per owner. This kind of ownership pattern minimises the possibility that any patterns observed in the regression analysis were merely artefacts of the land management practices of individual owners. The lack of significance of the variable ‘owner’ in the regression model supports this view. The smooth-barked eucalypt species (E. salubris and E. salmonophloia) preferred by 0. reticulata were the most numerous tree species (> 64%) in all remnants except S84 (42%) and S147 (37%) which contained several mallee form and roughbarked eucalypts. After the most significant variable ‘total number of smooth-barked eucalypts’ was fitted to the data, no other variable contributed significantly when added to the model. Thus the best explanatory model took the form p = ,-04492+0.008275 T/(1+,-04492+0008275T) (where p = probability that 0. reticulata is present, T = total number of smooth-barked eucalypts in the remnant) which estimates the probability that 0. reticulata will occur in a given remnant (Fig. 4). Figure 4 suggests that remnants with a total of 411 smooth-barked eucalypts will have a 95% probability of containing 0. reticulata after the mean time interval of 69 years, but it also indicates that the probability of the presence of the species drops sharply below this number of trees. This implies that there is a threshold amount of habitat, 1 1 1 1 1 beyond which the population can be reasonably certain of being present. Below this threshold, the probability of persistence drops to where a population in a remnant containing 55 smooth-barked eucalypts has only a 50% probability of persisting for 69 years. Movement between adjacent populations The pitfall trapping over the summer of 1990/91 around remnant 170 demonstrated that, even at a time of peak activity, neither species exhibited a high propensity for dispersal beyond the boundaries of the remnant. In a total of 3,445 pitfall nights between 5 November 1990 and 20 January 1991, a single G. variegata was captured on the inside of the drift fence (i.e. leaving the remnant). Neither species was captured in the pitfalls outside the drift fence although remnant 170 was only 200 m from the nearest woodland (a roadside verge known to contain 0. reticulata) and 700 m from remnant 171 (known to contain both species). Although this cannot be construed as evidence for zero movement between the remnants, it certainly suggests that rates of dispersal are low for both species. This evidence is supported by the findings of Kitchener et al. (1988) who, over a 3-year period of capturemark-recapture sampling in a habitat remnant, captured only one unmarked adult 0. reticulata and 10 unmarked adult G. variegata that were possible immigrants. Persistence of lizards in remnant habitat 31 0.8 j- / 0.2 .. 0.0 0 200 . I I I I 400 regression line ~~~~~~~~~ 95% confidence 600 Number of smooth-barked limits I I 800 1000 eucalypts Fig. 4. Logistic regression of the probability that Oedura reticuluta is present in a vegetation remnant and the total number of trees in the remnant. 0: remnants which contained (probability of presence = 1) or did not contain (probability of presence = 0) 0. reticulata. DISCUSSION The survey of woodland remnants in the Kellerberrin region demonstrates quite clearly that, in terms of persistence, the habitat generalist G. variegata has coped considerably better with the changes wrought by habitat clearance than the habitat specialist 0. reticulata. Although G. variegata was almost certainly more widespread than 0. reticulata before clearing, the remnants that did not contain 0. reticulata were all formerly part of larger regions of woodland that are likely to have contained both species. In most cases, remnants that did not contain 0. reticulata were in the same paddock or within a few hundred metres of woodland that did contain the species. Current absences are thus likely to be due to extinction rather than absence at the time of remnant formation. The significant correlation of total number of smooth-barked eucalypts with the presence of 0. reticulata suggests that habitat quantity is important in determining the likelihood that a population of 0. reticulata will have persisted to the present time. The high proportion of absences of 0. reticulata in the small remnants (low tree numbers) is consistent with island biogeographic theory (MacArthur & Wilson, 1967), which predicts that populations on small ‘islands’ will have a greater probability of extinction than populations on larger ‘islands’. The underlying basis for this prediction is that small ‘islands’, by implication, will contain small populations which are more vulnerable to stochastic genetic, demographic and environmental extinction processes (MacArthur & Wilson, 1967; Shaffer, 1981; Goodman, 1987). Although the presence/absence data for 0. reticulata are consistent with minimum viable population size theory, it is important to make the distinction between the effects of habitat per se and the population level phenomena that may be causing the ultimate demise of the species in local populations. There are many possible causes of local extinction besides those related to stochastic processes but which may also be related to remnant area. These include insufficient resources, increased predation and greater influence of external factors such as edge effects, where smaller remnants have a proportionately larger edge than large remnants. Our data do not discriminate between these possible causes, although it is interesting that the index of habitat size (total number of smooth-barked eucalypts) has a higher statistical significance than the correlated variable of remnant size. This implies that the limit on population size provided by habitat availability may be more critical to persistence than actual remnant area. In contrast to 0. reticulata, there appears to be no lower size limit to habitat remnants in which G. variegata can persist. Two explanations are possible for this result. First, G. variegata may have some innate 32 Stephen Sarre, Graeme T. Smith, Jacqueline characteristic that allows it to persist in small habitat remnants. This means that the changes to the ecosystem caused by fragmentation have not been detrimental to G. variegata (e.g. food and shelter have been sufficiently abundant and changes in predation have been insufficient to cause extinction), but it would also mean that this species has been able to maintain populations of sufficient size in even very small remnants to avoid extinction through stochastic processes. Demographic extinction models developed by Goodman (1987), Kinnaird and O’Brien (1991) and others suggest that, when environmental variation is included in extinction models, quite large population sizes are required for even moderately short persistence times. Evidence gathered from several of the remnants included in this study (Sarre, unpublished data) suggests that population sizes of G. variegata within the remnants are often very small (sometimes fewer than 20 individuals) and well below the thousands required by the models to persist for 23 generations. It is therefore unlikely that these populations are persisting for long in complete isolation. The second explanation for the higher persistence levels of G. variegata is that it is able to move between the remnants and consequently the district population has a spatial structure that approximates a metapopulation (a group of interacting local populations: Gilpin, 1987; Harrison, 1991) rather than a series of insular populations. Such a structure would promote high levels of remnant occupancy through the recolonisation of remnants following extinction events or the ‘topping up’ (Brown & Kodrik-Brown, 1977) of extant populations from adjacent or nearby populations. In either case, the high occupancy rate exhibited by G. variegata suggests an equilibrium rather than a non-equilibrium metapopulation structure. Several characteristics of G. variegata suggest that some movement between remnants is likely and that it could form a metapopulation in the Kellerberrin region. For example, it is a highly mobile species and exhibits rapid rates of recolonisation (Moritz, 1987) and high levels of gene flow over several kilometres (Moritz, 1992). It is also unlikely that cleared land will represent a complete barrier to movement given that they have been found in highly disturbed areas (Kitchener et al., 1980, 1988; Kitchener, 1982; S. Sarre, personal observations). However, the evidence from pitfalling (this study) and from the intensive markrecapture study of Kitchener et al. (1988) suggests that rates of movement between remnants if they do occur will be low. Consequently, any metapopulation structure is likely to be similar to the type ‘C’ structure described by Harrison (1991) where extinction and dispersal rates are in approximate equilibrium. In this case both rates are likely to be low. In conclusion, the contrast between the two species is stark. The habitat generalist (G. variegata) has a high level of remnant occupancy and exhibits the characteristics of a metapopulation at equilibrium. It is not possible to say definitively whether this equilibrium has been maintained by high rates of dispersal or low rates A. Meyers of extinction, but it is likely that dispersal rates are low, suggesting that extinction rates are also low. In contrast, 0. reticulata exhibits comparatively low rates of remnant occupancy, suggesting a high rate of extinction relative to dispersal. If this species forms a metapopulation, then over the 54-78 years since fragmentation, extinction rates have exceeded recolonisation rates. It is likely (given the lack of statistical significance of the variable ‘distance to the nearest woodland’, anecdotal evidence from Kitchener et al. (1988) and this study and estimates of maternal gene flow (S. Sarre, unpublished data)) that dispersal rates of 0. reticulata approach zero. The implications of this are that the proportion of remnants containing 0. reticulata will continue to decline until only populations with a very low probability of extinction will remain. An important consequence of such a pattern of extinctions is that even if dispersal is occurring, it will decline with a decline in occupancy because there are fewer source populations and greater distances between the remaining populations. Consequently, the contribution made by dispersal to reducing the extinction rate is likely to decline with time. Recent debate surrounding metapopulation structure (Hastings & Wolin, 1989; Bengtsson, 1991; Ebenhard, 199 1; Hanski & Gilpin, 1991; Hansson, 199 1; Harrison, 1991; Kindvall & Ahlen, 1992) has tended to concentrate on metapopulations that are in equilibrium. This is understandable because equilibrium systems are particularly amenable to study questions relating population size and environmental variation to rates of extinction and colonisation. Non-equilibrium systems have received scant attention and yet in both a practical and theoretical sense an understanding of them is clearly critical to the regional conservation strategies adopted for many species. This study shows the contrast in persistence between two species, both of which were probably widespread and numerous before European settlement but show contrasting patterns in their ability to cope with the subsequent rapid and major environmental changes. The ability to form an equilibrium metapopulation may be critical to the persistence of such species in habitat remnants. ACKNOWLEDGEMENTS We thank B. Brooker, G. Hanna, A. Heideman, A. Sarre, G. Sarre, I. Sarre, D. Stevens and J. Weeldenburg for their assistance in the field and G. Arnold, M. Brooker, R. How, D. Saunders and D. Shaw for their logistic support and access to unpublished data. We also thank A. Georges, C. Margules, A. Nicholls, A. Sarre and D. Shaw for their comments on drafts of this manuscript, A. Nicholls for statistical advice and Tara Goodsell for producing the maps. S. Sarre thanks J. Short and C. Hannan for their very kind accommodations during field work. This project was supported by a CSIRO Institute of Natural Resources and Environment Postgraduate Project Award and an Australian Museum Post-graduate Research Award. S. Sarre is the recipient of a Commonwealth Post-graduate Research Award. Persistence of lizards REFERENCES Arnold, G. W. & Weeldenburg, J. R. (1991). The distributions and characteristics of remnant native vegetation in parts of the Kellerberrin, Tammin, Tray&g and Wyalkatchem Shires of Western Australia. CSIRO, Lyneham. Beard, J. S. (1980). The vegetation of the Kellerberrin area Perth. Western Australia. Vegmap Publications, Bengtsson, J. (1991). Interspecific competition increases local extinction rate in a metapopulation system. Nature, Lond., 340,713-15. Brown, J. H. & Kodric-Brown, A. (1977). Turnover rates in insular biogeography: effect of immigration on extinction. Ecology, 58, 44559. Bustard, H. R. (1968). The ecology of the Australian gecko, Gehyra variegata in northern New South Wales. J. Zoo/., Lond., 154, 113-38. Bustard, H. R. (1969). The population biology of the gekkonid lizard Gehyra variegata (Dumeril & Bibron) in exploited forests in northern New South Wales. J. Anim. Ecol., 38, 3555 1. Caughley, G. (1994). Directions in conservation biology. J. Anim. Ecol., 63, 2 1544. Chapman, A. & Dell, J. (1985). Biology and zoogeography of the amphibians and reptiles of the Western Australian wheatbelt. Rec. West. Aust. Mus., 12(l), 146. Cogger, H. G. (1992). Reptiles and amphibians of Australia. Reed International Books, Chatswood. Ebenhard, T. (1991). Colonization in metapopulations: a review of theory and observations. Biol. J. Linn. Sot., 42, 105-21. Gilpin, E. (1987). Spatial structure and population vulnerability. In Viable populations for conservation, ed. M. E. Soule. Cambridge University Press, Cambridge, pp. 12540. Goodman, D. (1987). The demography of chance extinction. In Viable populations for conservation ed. M. E. Soule. Cambridge University Press, Cambridge, pp. 1 l-34. Hanski, I. & Gilpin, M. (1991). Metapopulation dynamics: brief history and conceptual domain. Biol. J. Linn. Ser., 42, 3-16. Hansson, L. (1991). Dispersal and connectivity in metapopulations. Biol. J. Linn. Sot., 42, 899103. Harris, L. D. & Silva-Lopez, G. (1992). Forest fragmentation and the conservation of biological diversity. In Conservation biology. The theory and practice of nature preservation and management, ed. P. L. Fiedler & S. K. Jain. Chapman & Hall, New York, pp. 197-237. Harrison, S. (1991). Local extinction in a metapopulation context: an empirical evaluation. Biol. J. Linn. Sot., 42, 73-88. Hastings, A. & Wolin, C. L. (1989) Within-patch dynamics in a metapopulation. Ecology, 70, 126 l-6. Hobbs, R. J. (1992). The role of corridors in conservation: solution or bandwagon? TREE, 7, 389-92. How, R. A. & Kitchener, D. J. (1983). The biology of the gecko Oedura reticulata Bustard, in a small habitat isolate in the Western Australian wheatbelt. Aust. Wildl. Res., 10, 543-56. Kindvall, 0. & Ahlen, I. (1992). Geometrical factors and metapopulation dynamics of the bush cricket, Metrioptera bicolor Phihppi (Orthoptera: Tettigoniidae). Conserv. Biol., 6, 520-9. King, M. (1979). Karyotypic evolution in Gehyra (Gekkonidae: Reptilia), I. The Gehyra variegata ~ punctata complex. Aust. J. Zool., 27, 373-93. Kinnaird, M. F. & O’Brien, T. G. (1991). Viable populations for an endangered forest primate, the Tana River crested in remnant habitat 33 mangabey Cercocebus galeritus galeritus. Conserv. Biol., 5, 203-13. Kitchener, D. J. (1982). Predictors of vertebrate species richness in nature reserves in the Western Australian wheatbelt. Aust. Wildl. Res., 9, 1-7. Kitchener, D. J., Chapman, A., Dell, J. & Muir, B. G. (1980). Lizard assemblage and reserve size and structure in the Western Australian wheatbelt ~ some implications for conservation. Biol. Conserv., 17, 25-62. Kitchener, D. J. & How, R. A. (1982). Lizard species in small mainland habitat isolates and islands off south-western Western Australia. Aust. Wildl. Res., 9, 357-63. Kitchener, D. J., How, R. A. & Dell, J. (1988). Biology of Oedura reticulata and Gehyra variegata (Gekkonidae) in an isolated woodland of Western Australia. J. Herpetol., 22, 401-12. Levins, R. (1970). Extinction. In Some mathematical problems in biology, ed. M. Gerstenhaber. American Mathematical Society, Providence, RI, pp. 777107. MacArthur, R. H. & Wilson, E. 0. (1967). The theory of island biogeography. Princeton University Press, Princeton, New Jersey. McArthur, W. M. (in press). Soil landscape map of the Kellerberrin region. Land resource Map, No. 12. Department of Agriculture, Perth. McCullagh, P. & Nelder, J. A. (1989). Generalised linear models. Chapman & Hall, London. Moritz, C. M. (1986). The population biology of Gehyra (Gekkonidae): chromosome change and speciation. Syst. Zool., 35, 4667. Moritz, C. (1987). The population biology of Gehyra (Gekkonidae: Reptilia), II. Individual movements and colonization by Gehyra variegata and G. nana. Aust. J. Zool., 35, 587-96. Moritz, C. M. (1992). The population biology of Gehyra (Gekkonidae), III. Patterns of microgeographic variation. J. Evolut. Biol., 5, 661-76. Nicholls, A. 0. (1991a). Examples of the use of Generalised Linear Models in analysis of survey data for conservation evaluation. In Nature conservation: cost-effective biological surveys and data analysis, ed. by C. R. Margules & M. P. Austin. CSIRO, Canberra, pp. 5463. Nicholls, A. 0. (19916). An introduction to statistical modelling using GLIM. In Nature conservation: cost-effective biological surveys and data analysis, ed. C. R. Margules & M. P. Austin, CSIRO, Canberra, pp. 191-201. Nunney, L. & Campbell, K. A. (1993). Assessing minimum viable population size: Demography meets population genetics. TREE, 8, 2349. Opdam, P. (1990). Dispersal in fragmented populations: the key to survival. In Species dispersal in agricultural habitats, ed. R. G. H. Bunce & D. C. Howard. Belhaven Press, London, pp. 3-17. Saunders, D. A., Hobbs, R. J. & Margules, C. R. (1991). Biological consequences of ecosystem fragmentation: a review. Conserv. Biol., S(l), 1-15. Shaffer, M. (1981). Minimum population sizes for species conservation. Bioscience, 31, 13 14. Shaffer, M. (1987). Minimum viable populations: coping with uncertainty. In Viable populations for conservation, ed. M. E. Soule. Cambridge University Press, Cambridge, pp. 69-86. Soul& M. E. (ed.) (1987). Viable populations for conservation. Cambridge University Press, Cambridge. Verboom, J., Lankester, K. & Metz, J. A. J. (1991). Linking local and regional dynamics in stochastic metapopulation models. Biol. J. Linn. Sot., 42, 39-55.