Worldwide, efforts to restore habitat quality are rarely matched by efforts to evaluate the effects of those restoration attempts. Simply documenting usage of the newly created habitats by biota is not enough, because such areas may serve as sink populations. We need to monitor viability (growth, survival, reproduction) of individuals that colonize the newly created habitat, compared with conspecifics in non-restored areas.
In the Sydney region in south-eastern Australia, humans have degraded sandstone rock outcrops by removing natural rocks for landscaping urban gardens. We restored degraded rock outcrops by placing artificial rocks at sites where natural rocks had been removed. We measured growth rates and survival in velvet geckos Oedura lesueurii at control and restored sites over a 2-year period.
Gecko growth rates were unaffected by habitat restoration, but restoring sites with artificial rocks increased the overall numbers of lizards detected (both adults and juveniles). The apparent survival rates of adult male lizards (as estimated using mark) were not significantly affected by habitat restoration. However, apparent survival rates of juvenile geckos were higher at restored sites than at unrestored sites.
Synthesis and applications. Habitat restoration using artificial rocks has had measurable conservation benefits on these degraded rocky outcrops. Quantifying those benefits in terms of species' survival and growth rates enables management decisions about habitat restoration to be based upon evidence rather than wishful thinking or untested intuition.
The current scale of habitat degradation means that simply preserving existing areas of relatively unspoiled habitat is no longer enough. To conserve many systems that are currently under threat, we need to actively restore degraded habitats (Sinclair et al. 1995; Miller & Hobbs 2007; Kouki et al. 2011). Ideally, efforts at habitat restoration should be based on a detailed knowledge of habitat requirements for the biota of interest; and the effects of those manipulations on the biota need to be monitored carefully, over biologically meaningful time periods (Block et al. 2001; Lovett et al. 2007; Miller & Hobbs 2007). Attempts at habitat restoration often fail because of unsuspected factors such as responses of other (non-desired) species (Bice & Zampatti 2011), or low vagility (and hence, low colonization rates) of the target species (Dubey et al. 2012). Even if the ‘at risk’ species colonize the restored habitat, however, the net impact of our manipulations on those species may not be positive. The newly restored habitat may simply attract animals out of the few patches of less-degraded habitat and potentially expose them to new risks such as disease, parasites, predation or costs of increased social conflict (Nicol et al. 2004; Klein et al. 2007; Skwarska et al. 2009; Bice & Zampatti 2011). For example, some birds use artificial nesting boxes in preference to natural tree hollows, but may be at greater risk of predation as a result (Skwarska et al. 2009). Analogously, urban birds and other wildlife enthusiastically consume the supplemental food offered by householders, tourists and the like, but may thereby experience severe alterations in behaviour, foraging ability and nutritional balance (Orams 2002; Newsome & Rodger 2008).
The possibility of inadvertent collateral damage from efforts to conserve species places a premium on careful monitoring of the consequences of any conservation-oriented habitat manipulations (Block et al. 2001; Lovett et al. 2007; Miller & Hobbs 2007). Unfortunately, more effort typically is devoted to manipulating habitats than to monitoring the effects of the implemented changes (Block et al. 2001; Ruiz-Jaen & Aide 2005). In particular, we need to quantify not only the rate at which animals colonize the newly created habitat, but also the consequences of that colonization for individual viability. Ideally, the restored habitat should provide conditions that enable animals to achieve rates of survival, growth and reproduction at least as high as is obtainable in patches of the original (non-degraded) habitat (Block et al. 2001). Such data are difficult to obtain in the field, especially from vagile animals.
Our long-term research on rock-dwelling reptiles in south-eastern Australia provides an ideal opportunity to evaluate the impacts of habitat restoration. Sandstone outcrops in this region provide thermally distinctive microenvironments that support a unique reptile fauna, including the endangered broad-headed snake Hoplocephalus bungaroides Schlegel 1837 and its major prey, the velvet gecko Oedura lesueurii Dumeril & Bibron 1836 (Webb & Shine 1998; Pringle, Webb & Shine 2003). Illegal removal of exfoliated rocks, which are popular for garden landscaping, has denuded many areas of the rock crevices required by these animals (Schlesinger & Shine 1994a; Shine et al. 1998). We restored sites in a bush rock denuded area with artificial rocks, enabling us to examine the impacts of rock replenishment not only on gecko numbers, but also on the viability of individual geckos. The strong philopatry of these lizards (Webb, Pike & Shine 2008) facilitates direct comparisons between individuals that rely upon artificial versus natural rocks on the same or adjacent sites.
Materials and methods
The velvet gecko inhabits rocky outcrops in Australia from south-eastern New South Wales to south-eastern Queensland (Cogger 2000). This small [up to 65 mm snout–vent length (SVL)] nocturnal lizard shelters by day in crevices beneath small rocks and under exfoliated bedrock (Schlesinger & Shine 1994a; Webb & Shine 1998; Croak et al. 2008, 2010). At dusk, the lizards emerge from shelters to forage in leaf litter (Cogger 2000). Adult male velvet geckos are territorial and defend their retreat sites from other males and juveniles (Downes & Shine 1998). Velvet geckos form the main prey base for the highly endangered broad-headed snake (Webb & Shine 1998).
We conducted a 2-year mark recapture project (2009–2010) on velvet geckos at 12 sites in the Dharawal Conservation area (see Field Methods below, Fig. 1). The sites were rock outcrops (on average 50 × 50 m in surface area) separated by mixed eucalyptus forest dominated by scribbly gum Eucalyptus haemastoma, Sydney peppermint E. pipperita and red bloodwood E. gummifera.
We deployed 50 artificial rocks made from fibre-reinforced cement (each measuring 512 mm long × 352 mm wide × 46 mm thick) that provide crevices similar (both structurally and thermally) to those under natural rocks (Croak et al. 2008, 2010) on flat ground and in open areas at each of six treatment sites throughout the area (see Croak et al. 2008, 2010 for details on construction and deployment criteria). Each habitat-restored (treatment) site was paired with a nearby unrestored (control) site that contained a similar number of natural rocks as was present on the control site. Although we matched sites as closely as possible, control sites had a slightly higher average number of natural rocks than did treatment sites (Fig. 2a). Treatment and control site pairs were, on average, 500 m apart (range 500 m to 13 km); we never recorded geckos moving from one site to another. We monitored faunal usage by sampling all rocks (both natural and artificial) at all (i.e. both treatment and control) sites on a monthly basis from the beginning of 2009 until the end of 2010. We sampled the sites by turning all rocks by hand and capturing any animals found beneath. We gave any velvet geckos encountered a unique toe-clip sequence for individual recognition. We also recorded the sex, SVL (mm) and tail length (mm) of the lizards. We then carefully returned rocks to their original positions and allowed the lizards to retreat beneath them.
We classified animals as adults based upon the minimum size for reproductive adult males (SVL ≥ 50 mm) and females (SVL ≥ 56 mm) with males and females differentiated by the presence of enlarged male gonads and cloacal spurs as reported by Webb, Pike & Shine (2008). We used an analysis of variance (anova) with site type as a factor and numbers of male, female and juvenile geckos as the dependent variables, to compare numbers of individual geckos captured at control and treatment sites. We used analysis of covariance (ancova) with rock number as the covariate, site type as a factor and numbers of male, female and juvenile geckos as dependent variables to assess gecko numbers at the different site types relative to the total number of rocks available.
We used the Cormack–Jolly–Seber method (Cormack 1989) to estimate recapture (P) and survival (φ) probabilities from the mark–recapture data at control versus treatment (restoration) sites using the software package Program mark v 6.1 (White & Burnham 1999). We included site (restoration vs. control) as a factor (group) in the input file. Because survival probabilities of O. lesueurii can vary ontogenetically (Webb 2006), we included gecko age (adult vs. juvenile) and sex as factors in the input file. We classified animals as described above. Prior to data analysis in mark, we formulated a series of a priori candidate models to determine whether survival and recapture probabilities varied over time, among treatment groups (control vs. restoration), between age classes or as a function of interactions between these factors. To determine whether the most general model in our candidate model set [φ (t) p (t)] provided an adequate fit to the data (i.e. met the mark–recapture assumptions), we used the bootstrap GOF procedure in mark (Cooch & White 2011). Based on 500 bootstrap replicates, there was no significant deviation from the mark–recapture assumptions for our general model (P = 0·06).
To explore the influence of rock type (artificial vs. natural) on gecko survival and recapture probability independent of site type, we assigned each individual gecko to one of three groups: geckos that only used artificial rocks, geckos that only used natural rocks and those that used both. Because survival varies with age and sex, we analysed each age group separately. Thus, we used the Cormack–Jolly–Seber method (Cormack 1989) to estimate recapture (P) and survival (φ) probabilities from the mark–recapture data collected from individual juvenile geckos using ‘only natural’, ‘only artificial’ or ‘both rock types’ (thus three factors = groups). We estimated recapture (P) and survival (φ) probabilities for individual male velvet geckos that used only natural versus only artificial rocks (we omitted the ‘both rock types’ category for male geckos because most of these lizards used only a single rock type). We omitted female geckos from our analysis, due to low sample sizes (N = 11 individuals using only natural rocks). Prior to data analysis in mark, we formulated a series of a priori candidate models to determine whether survival and recapture probabilities varied over time, among treatment groups, or interactions between these factors. To determine whether the most general model in our candidate model sets [φ (t) p (t)] met the mark–recapture assumptions, we used the bootstrap GOF procedure in mark (Cooch & White 2011). Based on 500 bootstrap replicates, we found no significant deviation from the mark–recapture assumptions for our general models (P = 0·07 for juveniles, P = 0·09 for males).
For all analyses, we used the Akaike's information criterion (AIC) to select the most parsimonious model (Burnham and Anderson 1998). We adjusted the AIC values for over-dispersion using the variance inflation factor c, which we calculated from the GOF statistics (Cooch & White 2011). The AIC value adjusted for over-dispersion and finite sample size is termed the corrected quasi-likelihood AIC (denoted as QAICc). We ranked models by comparing ∆QAICc (the difference between the model QAICc and the lowest QAICc from the set of models), and we used normalized AIC weights to evaluate the relative strength of evidence of models (Burnham & Anderson 1998).
We used ancova to compare growth rates of velvet geckos captured under artificial vs. natural rocks. This analysis included rock type as a factor, mean SVL (average of initial and final length) as a covariate (to allow for ontogenetic changes in growth rate) and growth rate (mm day−1) as the dependent variable.
Our control sites had an average of 26 natural rocks, and our treatment sites had an average of 18 natural rocks and 50 artificial rocks (Fig. 2a). On average, we captured 27 individual geckos under natural rocks at each control site, seven individual geckos under natural rocks at each treatment site and 60 individuals under artificial rocks at each treatment site (Fig. 2b). Over the 2-year period, we captured 563 individual velvet geckos: 312 juveniles, 80 females and 171 males. Treatment sites had more male (F1,9 = 5·25, P = 0·04, Adjusted R2 = 0·30) and female geckos (F1,9 = 9·94, P = 0·01, Adjusted R2 = 0·47) than did control sites. Although numbers of juveniles increased to a similar degree as in adults, high variances meant that the increase in juvenile numbers in habitat-restored sites was not statistically significant (F1,9 = 1·93, P = 0·20, Adjusted R2 = 0·08; Fig. 3a).
When we controlled for the number of rocks at each site, there were no significant differences between control and habitat-restored sites in the numbers of male, female or juvenile geckos (males, slope: F1,7 = 0·47, P = 0·52, effect of site type: F1,8 = 1·71, P = 0·23, effect of rock number: F1,8 = 0·32, P = 0·59 females: slope: F1,7 = 0·44, P = 0·53, effect of site type: F1,8 = 2·04, P = 0·19, effect of rock number: F1,8 = 0·06, P = 0·70 Juveniles: slope: F1,7 = 0·02, P = 0·90, effect of site type: F1,8 = 1·93, P = 0·2, effect of rock number: F1,8 = 0·93, P = 0·36).
Survival at restored vs. unrestored sites
The model φ (g), P (t) was highest ranked in the candidate model set and had most support (∆QAICc = 0·00, QAICc weight = 0·9981; Table 1). In this model, survival was group dependent, and the probability of recapture was time dependent. From this model, survival rates for juveniles were higher at restored sites (φ = 0·90, SE = 0·01) than at unrestored sites (φ = 0·80, SE = 0·03; Fig. 3b). Survival rates of males were similar at restored (φ = 0·93, SE = 0·01) and control sites (φ = 0·92, SE = 0·02; Fig. 3b). Survival rates of adult females showed a non-significant trend for increased survival at treatment sites (restored sites φ = 0·92, SE = 0·01; control sites φ = 0·78, SE = 0·09; Fig. 3b).
Table 1. Candidate models used to determine whether survival (φ) and recapture probability (P) of geckos was influenced by treatment (restoration versus control), sex or interactions between age and treatment
The letters g, c, r and t refer to group, control site, restoration site and time respectively. The term ‘.’ denotes constant survival. The most parsimonious model was φ(g), P(t), where survival was group dependent and the probability of recapture was time dependent. Models are ordered according to the adjusted Akaike's information criterion (QAICc), with model parsimony increasing with decreasing QAICc weighting. Models with ∆QAICc < 2·0 have the greatest statistical support.
φ(g), P(g × t)
Survival of juvenile velvet geckos under artificial and natural rocks
The model φ (t), P (g) was highest ranked in the candidate model set and had the most support (∆QAICc = 0·00, QAICc weight = 0·9893; Table 2). In this model, survival was time dependent, and the probability of recapture was group dependent; that is, juveniles using artificial rocks were more likely to be recaptured (P = 0·49, SE = 0·02; Fig. 4) than were individuals using natural rocks (P = 0·26, SE = 0·04; Fig. 4) or both rock types (P = 0·22, SE = 0·06; Fig. 4).
Table 2. Candidate models used to determine whether survival (φ) and recapture probability (P) of juvenile velvet geckos was influenced by their choice of shelter site (only artificial rocks, only natural rocks or a combination of both rock types)
The letters g and t refer to group and time respectively. The most parsimonious model was φ(t), P(g), where survival was time dependent and the probability of recapture was group (artificial, natural or both) dependent. Models are ordered according to the adjusted Akaike's information criterion (QAICc), with model parsimony increasing with decreasing QAICc weighting. Models with ∆QAICc < 2·0 have the greatest statistical support.
φ(t), P(g × t)
φ(g), P(g × t)
Survival of male velvet geckos under artificial and natural rocks
The model φ(.), p (g × t) was highest ranked in the candidate model set and had the most support (∆QAICc = 0·00, QAICc weight = 0·720; Table 3). In this model, survival was constant, with male geckos showing similar rates of survival under both artificial and natural rock types (φ = 0·89, SE = 0·01). Probability of recapture was group × time dependent, indicating that recapture probability differed between groups over time.
Table 3. Candidate models used to determine whether survival (φ) and recapture probability (P) of male velvet geckos was influenced by their choice of shelter site (only artificial rocks or only natural rocks)
The letters g and t refer to group and time, respectively. The term ‘.’ denotes constant survival. The most parsimonious model was φ(.), P(g × t), where survival was constant between animals using natural and artificial rocks, and the probability of recapture was dependent on group (natural or artificial) and time. Models are ordered according to the adjusted Akaike's information criterion (QAICc), with model parsimony increasing with decreasing QAICc weighting. Models with ∆QAICc < 2·0 have the greatest statistical support.
φ(.), P(g × t)
φ(g), P(g × t)
φ(t), P(g × t)
Of 563 individuals captured during this study, we caught 281 individuals more than once. Growth rates of these recaptured geckos were not significantly different (slopes: F1,277 = 0·58, P = 0·45, main effect: F1,278 = 0·84, P = 0·36; Fig. 5) between artificial and natural rocks.
Our results show that velvet geckos used artificial rocks placed on degraded rock outcrops. The addition of artificial rocks to degraded sites enhanced gecko abundance, and on average, we captured more than twice as many geckos at habitat-restored sites than at control sites (means of 67 vs. 27 lizards, respectively; Fig. 3a). More importantly, individuals that used artificial rocks did not incur any survival or growth costs from using such sites. Apparent survival rates of juveniles were higher at restored sites than at control sites, but rates of apparent survival of adults were not significantly affected by habitat restoration. Growth rates of free-living geckos were not affected by habitat restoration, suggesting that competition for resources did not increase in the restored sites.
The effects of habitat restoration on velvet geckos will depend upon the social system of this species. For example, resident adult males actively repel other adult males (and juveniles) from retreat sites (Schlesinger & Shine 1994b). This territorial exclusion may limit the numbers of geckos at a site, even if unoccupied shelter sites are available (as in many other species in which adult males vigorously exclude potential rivals: for example Le Boeuf 1974; Shine et al. 1981; Emlen & Nijhout 1999). Nonetheless, habitat restoration increased the number of adult male geckos on a site (Fig. 3a). The increased numbers and stable survival that we observed for male geckos (Fig. 3a vs. Fig. 3b) suggest that adding additional cover items allowed more of the local male lizards to remain in our study area for longer periods (and hence, become more accessible to capture). As previous studies have shown, adult geckos spend much of the year in deep crevices in the bedrock where they are inaccessible to investigators (Pike, Webb & Shine 2010). Thus, the primary impact on male geckos of adding artificial cover items was to modify their seasonal patterns of habitat use rather than to increase their survival rates (which were already very high, even in the absence of habitat restoration; Fig. 3b).
In contrast, habitat restoration increased the numbers of juvenile geckos by increasing their apparent survival (Figs 3a, b). After dispersing from communal nest sites, juveniles settle under rocks, and use one or two rocks as shelter sites for much of the year (Webb 2006). Notably, apparent survival was higher at restored sites, but was unaffected by the type of rock that a gecko sheltered under (Table 2). Hence, adding rocks translated into increasing shelter site availability for this age class. The same may well have been true of adult female geckos in our study. Their numbers increased, and their survival rates tended to be higher in habitat-restored sites (the trend was strong, although the effect fell short of statistical significance because of low sample size: see Fig. 3a). Because juveniles and adult female geckos are less territorial than adult males (Schlesinger & Shine 1994b), their numbers are likely to be limited by resource (food, shelter) availability as well as by social interactions. Low rates of juvenile survival in unrestored habitats mean that there is considerable opportunity for habitat restoration to elevate survival rates for this age class.
In summary, our data show that habitat restoration increases gecko abundance, at least partly because juvenile survival depends upon shelter site availability. Lizards in these higher-density populations in rock-restored sites exhibit growth rates as high as their conspecifics living under natural rocks in control sites. Thus, replacing stolen ‘bush-rocks’ with artificial rocks is an effective conservation tool for this system. The population viability of velvet geckos is not a major conservation issue, because these lizards are abundant over a broad area (albeit, as a series of genetically distinctive subpopulations: Dubey et al. 2012). However, these lizards are the main prey of juvenile broad-headed snakes, an endangered species endemic to the Sydney region (Webb & Shine 1998). Broad-headed snakes have become endangered due to habitat degradation (especially, rock theft and forest overgrowth: Shine et al. 1998; Pringle, Webb & Shine 2003; Pike, Webb & Shine 2011) and have disappeared from much of their former range (Shine & Fitzgerald 1989). Maintaining abundant stocks of velvet geckos are clearly crucial to conserving broad-headed snakes.
Ecosystems worldwide are affected by anthropogenic degradation through activities such as deforestation, urbanization, farming and mining (Cooke & Johnson 2002; Lal 2008; McKinney 2008; Bradshaw 2012). The protection of remaining ‘intact’ systems coupled with habitat restoration is needed to conserve threatened flora and fauna and evolutionary processes at work in those systems (Sinclair et al. 1995; Knowlton & Graham 2011; Perez et al. 2012). Thus, habitat restoration has been attempted in a variety of systems (Webb & Shine 2000; Croak et al. 2010; Roni, Hanson & Beechie 2011; Hugel 2012; Jones & Kress 2012). However, we have surprisingly little information on the impacts of restoration activities on biota, and in many cases, we do not even have counts of abundance through a sufficiently long time period that we can assess overall impacts, nor data from adjacent control (undisturbed) areas to use for comparison (Block et al. 2001; Lovett et al. 2007; Miller & Hobbs 2007). Even when such data exist, a simple increase in numbers of some target species in rehabilitated habitat is not necessarily indicative of success. In the worst case, our provision of ‘attractive’ habitat within an otherwise degraded mosaic may simply lure animals out of the surrounding matrix into areas where their high densities render them liable to threatening processes such as disease and parasite transmission, social stress or predation (Sugden & Beyersbergen 1986; Reitsma 1992; Gustavo et al. 2006). To ensure that restored habitats are not simply ‘sinks’ for surrounding areas, we need to monitor vital processes – such as rates of growth and survival – in the restored habitat patches as well as in ‘natural’ (unrestored) areas. Our studies on velvet geckos are encouraging as they confirm that we can provide a real and measurable benefit to wild populations by reversing the habitat-degradation processes that threaten so many populations of this species. However, more work is needed to assess the responses of this system as a whole to our restoration attempts.
We thank numerous volunteers, especially Matt Greenlees for assistance with fieldwork, Melanie Elphick for assistance with figures and proof reading and Mathew Crowther for assistance with ARC GIS. The Australian Research Council, Forests NSW, National Parks and Wildlife, The Hawkesbury/Nepean Catchment Authority, Zoos Victoria and the Australian Reptile Park provided funding. Meagan Hinds provided encouragement. Anita Zubovic facilitated site access. The University of Sydney Animal Ethics Committee approved all procedures; approval number L04/12-2008/3/4927. All research was conducted under NPWS Scientific Licence SL100472.