1 Reptiles in restored agricultural landscapes: The value of
2 linear strips, patches and habitat condition
3
4 This is the submitted version of the manuscript. The final version is published in
5 Animal Conservation: http://dx.doi.org/10.1111/acv.12121
6
7 Authors: Sacha Jellinek a, Kirsten M. Parris b, Michael A. McCarthy c, Brendan A. Wintle d & Don
8 A. Driscoll e
9 a Centre of Excellence for Environmental Decisions, School of Botany, University of Melbourne,
10 Parkville, Melbourne, Vic 3072, Australia. Email: [email protected] (Corresponding
11 Author) Ph: +61 (0)410 917 123
12 b Centre of Excellence for Environmental Decisions, School of Botany, University of Melbourne,
13 Parkville, Melbourne, Vic 3072, Australia. Email: [email protected]
14 c Centre of Excellence for Environmental Decisions, School of Botany, University of Melbourne,
15 Parkville, Melbourne, Vic 3072, Australia. Email:
17 d Centre of Excellence for Environmental Decisions, School of Botany, University of Melbourne,
18 Parkville, Melbourne, Vic 3072, Australia. Email: [email protected]
19 e The Fenner School of Environment and Society, Hancock Building (43), Biology Place,
20 Australian National University, Canberra, ACT 0200, Australia. Email: [email protected]
21
22 Keywords: Agricultural landscapes, distance, ground layers, linear strip, patch, reptile,
23 revegetation, restoration, vegetation condition
24
1
1 Abstract 2 Habitat restoration, including revegetation of linear strips and enlargement of
3 remnant patches, may benefit native fauna in highly fragmented landscapes.
4 Such restoration has occurred around the world, even though the relative
5 importance of strips and patches of vegetation remains controversial. Using
6 reptile communities from south-eastern Australia, we assessed the conservation
7 value of revegetation in strips and alongside remnant patches compared with
8 remnant vegetation and cleared roadsides. We also examined the distance that
9 reptiles occurred from remnant patches into linear vegetation. We found that
10 reptile species richness and counts did not substantially differ between
11 revegetated, remnant and cleared habitats, or between linear strip and patch
12 treatments. This may indicate that species sensitive to land clearing have
13 already been lost from the landscape. These results imply that if specialist
14 species have already been lost, we may be unable to measure the effects of
15 agriculture on biodiversity. Furthermore, revegetation with the expectation that
16 fauna will recolonize may be unrealistic and translocations may be necessary.
17 Unexpectedly, we recorded higher species richness and counts of rare reptile
18 species in remnant linear strips as distance from remnant patches increased.
19 Ground layer attributes were important for increasing reptile species richness
20 and counts and in structuring reptile communities, explaining approximately
21 three times as much variation as remnant shape or vegetation type (remnant,
22 revegetated, cleared). Management agencies should protect and effectively
23 manage remnant linear strips if rarer reptiles are to be retained, paying
24 particular attention to ground layer attributes. The decision to include ground
2
1 layers in future revegetation activities will be more important than the shape of
2 restored areas.
3
4 Introduction
5 Habitat loss and fragmentation as a result of intensive agriculture is a major
6 threat to biodiversity conservation around the world (Harrison and Bruna, 1999,
7 Lindenmayer, 2009). Broad-scale habitat restoration through revegetation is
8 now necessary to maintain or increase biodiversity in these landscapes (Huxel
9 and Hastings, 1999, Ruiz-Jaen and Aide, 2005, Ryan, 2000). However, there is
10 debate about the effectiveness of revegetating for maintaining faunal species,
11 because revegetated areas generally do not contain the same faunal
12 communities as remnant habitats (Munro, Fischer, Barrett et al., 2011, Munro,
13 Fischer, Wood et al., 2009). There is also debate about whether it is better to
14 revegetate linear strips or to enlarge existing remnant patches (Bailey, 2007,
15 Simberloff and Cox, 1987, Tewksbury, Levey, Haddad et al., 2002).
16
17 There are benefits and drawbacks to enlarging patches of habitat and to
18 focusing on restoration of linear strips to link remnant areas together. Enlarged
19 patches would provide a larger habitat area and have greater interior habitat yet
20 would remain isolated from similar habitats, limiting species movement and
21 dispersal (Bennett, Crooks and Sanjayan, 2006, Hodgson, Thomas, Wintle et
22 al., 2009). Restored linear strips could allow the movement and dispersal of
23 native plants and animals, but have a greater edge-to-area ratio than patches
3
1 and may act as conduits of disease and invasive species, decreasing their
2 value for native species (Brown, Phillips, Webb et al., 2006, Simberloff, Farr,
3 Cox et al., 1992). Further, the distance that fauna can move along linear strips
4 is not well established (Haddad, Bowne, Cunningham et al., 2003).
5 Understanding the movement of a species is important because species with
6 limited dispersal abilities such as reptiles and amphibians would be less able to
7 move along linear strips or escape degraded habitats (Carthew, Horner and
8 Jones, 2009).
9
10 The response of faunal groups such as birds and mammals to habitat
11 fragmentation and habitat restoration are reasonably well studied, while groups
12 such as reptiles have received little attention in the ecological literature
13 (Bateman, Chung-MacCoubrey and Snell, 2008, Cunningham, Lindenmayer,
14 Crane et al., 2007, Kanowski, Reis, Catterall et al., 2006, Munro, Lindenmayer
15 and Fischer, 2007). While researchers report that some reptile communities
16 have persisted in fragmented landscapes (Cabrera-Guzman and Hugo
17 Reynoso, 2012, Kitchener, Chapman, Dell et al., 1980, Smith, Arnold, Sarre et
18 al., 1996), other studies report that some reptile species and communities have
19 declined (Dixo and Metzger, 2009, Driscoll, 2004, Mac Nally and Brown, 2001).
20 Reptiles are strongly influenced by vegetation type and structure, and this may
21 have a stronger effect than remnant shape or size in highly modified landscapes
22 (Jellinek, Driscoll and Kirkpatrick, 2004, Michael, Cunningham and
23 Lindenmayer, 2010, Schutz and Driscoll, 2008).
24
4
1 In order to maintain reptiles that are sensitive to loss of native vegetation, it is
2 essential to determine how effective revegetation is for conserving or recovering
3 reptile communities in enlarged and connected habitats. In two agricultural
4 landscapes in south-eastern Australia we asked if reptile species richness and
5 counts, and counts of individual species was influenced by: (i) the shape of the
6 sampled area (linear strip or patch), the type of the sampled area (remnant,
7 revegetated or cleared) and the interactions of shape and type; (ii) increasing
8 distance from patches along remnant, revegetated and cleared linear strips; (iii)
9 environmental variables and how these account for variation in reptile
10 community composition compared with shape and site-type. This information
11 could assist management agencies and community groups to plan restoration
12 activities that will best maintain reptile species in modified agricultural
13 landscapes.
14
15 Materials and Methods
16 Study region
17 This study took place in two regions of Victoria, south-eastern Australia: Benalla
18 (Fig. 1 a) and Wimmera (Fig. 1 b). These two regions have been cleared for
19 intensive agriculture for over 150 years, and have less than 10% of their native
20 vegetation remaining (Duncan, Moxham and Read, 2007, Radford, Bennett and
21 Cheers, 2005). Cropping and livestock production are the dominant agricultural
22 industries in these regions (Jellinek, Parris and Driscoll, 2013a).
23
5
1 Within each of the two regions (Benalla and Wimmera), we surveyed two
2 treatment shapes: (1) linear strips and (2) enlarged patches of remnant native
3 vegetation (Fig. 2 a, b, c). Linear strips were defined as areas of habitat along
4 roadsides or fence lines that were approximately 20 - 40m wide and at least
5 500m long. Linear strips always originated from a remnant patch that was a
6 similar size and shape to remnant patches in the patch treatment. Patches were
7 approximately square or oval-shaped and at least 4 hectares in size. Within
8 linear strips there were four site-types, and within enlarged patches there were
9 two site-types (Fig. 2 a). The site-types in our linear strip treatment were (i) a
10 revegetated linear strip, which had been cleared but was revegetated using
11 native trees and shrubs 8 - 14 years ago; (ii) a cleared linear strip, which had
12 been cleared of trees and shrubs and contained a grassy ground layer; (iii) a
13 remnant linear strip, containing remnant native vegetation that had not been
14 cleared; and (iv) a patch of remnant native vegetation, from which the linear
15 strips originated (Fig. 2 b). In some cases it was not possible to have a single
16 remnant patch from which the three linear strips originated and situations arose
17 where two of the three linear strips originated from one remnant patch while the
18 third originated from a nearby remnant patch (<1 km away). In replicates where
19 only one remnant patch was used, the patch was surveyed using two lots of five
20 pitfall lines to ensure consistent trapping effort.
21
22 Within the patch treatments, there were two site-types: (v) a remnant patch,
23 containing remnant native vegetation that had not been cleared; and (vi) an
24 adjacent revegetated patch, containing native trees and shrubs 8 - 14 years old
6
1 that had been cleared prior to being revegetated (Fig. 2 c). Each treatment set
2 was replicated in five locations (Fig. 2 a) within each of the two regions. We use
3 the term "site" to refer to a site-type within a location and region.
4
5 Pitfall trapping
6 We established pitfall traps in the Wimmera and Benalla region in November
7 and December 2007 and 2008 and operated them from January until March
8 2008 and 2009 respectively. We established five pitfall lines with a total of ten
9 pitfall traps at each site-type (a - e, Fig. 2 a). Each treatment was surveyed
10 twice daily for 5 consecutive days (4 nights) every month for three months. We
11 identified captured reptiles, marked them with nail polish on the underside of
12 their hind leg, and released them near the site of capture.
13
14 Each pitfall line was comprised of two 20 litre buckets joined by a 16m long,
15 30cm high drift fence (made of 200 micrometer black polyurethane plastic).
16 Buckets were dug into the ground such that the rim of the bucket was level with
17 the top of the soil and the drift fence bisected the top of each bucket. To protect
18 captured animals we placed PVC pipe, soil and pine blocks into each bucket.
19
20 Visual habitat estimates were conducted over an area of approximately two
21 hectares within each site-type around the trapping area. All surveys were
22 undertaken by the same observer who was experienced in habitat assessment
23 techniques. Density measures were initially recorded using the following
24 system: 1 = 0 - 5%, 2 = 5 - 25%, 3 = 25 - 50%, 4 = 50 - 75%, 5 = 75 - 95% and 7
1 6 = 95 - 100% (O’Shea and Kirkpatrick, 2000), and then converted to the mid-
2 point of the percent cover. Variables recorded were: density of tallest stratum of
3 vegetation, density of second tallest stratum (mid-stratum density), ground
4 cover, rock cover, cover of fallen timber, litter cover, tussock grass cover, cover
5 of other grasses, herb cover, shrub cover and percentage of bare ground. We
6 also recorded all of the plant species within each site-type, from which we
7 calculated the proportion of all species present that were native.
8
9 Data analysis
10 Treatment relationships
11 We analyzed the response of reptile species richness and counts to treatment
12 type using linear mixed models with a Poisson distribution in a Bayesian
13 framework (Lunn, Thomas, Best et al., 2000, McCarthy, 2007, Spiegelhalter,
14 Best, Carlin et al., 2002). The graphed data was consistent with the expectation
15 that count data follow a Poisson distribution (Hilborn and Mangel, 1997). Our
16 response variables were reptile species richness and counts of individual
17 species of reptiles that were captured in more than half of the site-types, as we
18 had sufficient data to analyze these species (Appendix A, Appendix B). Region
19 and location were fitted as random effects and site-type (explained below) was
20 fitted as a fixed effect (Appendix C). Reptile species richness and individual
21 species abundance counts were calculated by pooling the total captures in the
22 five pitfall lines at each site-type. Reptile counts were the total number of
23 uniquely marked individuals during the entire period of study (2008 and 2009).
8
1
2 We undertook three forms of the analysis using different subsets of site-types:
3 (1) revegetated linear strip, remnant linear strip, cleared linear strip and remnant
4 patch; (2) revegetated patch and remnant patch; and (3) revegetated linear strip
5 and revegetated patch. We also used GLM to determine whether there was an
6 effect of distance from remnant patches on reptile species in linear strips
7 (Appendix D). Reptile species richness and counts were again fitted as the
8 response variable and site-type and distance along the linear strip from the
9 patch (100, 200, 300, 400 and 500m) were fitted as fixed effects (Appendix A).
10 Reptile species richness and counts were calculated at each of the five pitfall
11 lines for this analysis. Extra Poisson variation was included in the models to
12 take into account the variation between capture rates in each pitfall line.
13
14 To account for imperfect detection in individual species counts, we utilized an
15 N-mixture model (Royle, 2004) to simultaneously model the influence of site-
16 type on counts and detectability of individuals (Appendix E). All inference
17 around individual species counts was conducted using the N-mixture
18 formulation to model the detection process, incorporating the Poisson
19 hierarchical model structure described above.
20
21 As we were using occupancy data to calculate and model richness, we opted to
22 analyze the single-visit detection probability at a site-type (within location) using
23 a zero-inflated binomial model (Kery and Schaub, 2011, Tyre, Tenhumberg,
24 Field et al., 2003) (Appendix F). This tested whether imperfect detection would
9
1 have impacted estimates of site-level species richness. We included six
2 vegetation variables (rock cover, mid-stratum density, litter cover, bare ground,
3 the proportion of native plants, and tussock cover) as covariates on reptile
4 species richness to enable unbiased estimation of detectability.
5
6 We used WinBUGS (Lunn et al., 2000) to estimate posterior distributions for the
7 Poisson regression model coefficients. The 2.5th, 50th and 97.5th percentiles of
8 the distribution were computed to provide 95% Bayesian credible intervals (CIs)
9 for model coefficients. We used WinBUGS to generate 150,000 samples from a
10 posterior distribution of each mixed effects model for each species, after
11 discarding the first ‘burn-in’ of 10,000 samples. All explanatory variables were
12 centered by subtracting the mean from each variable to minimize
13 autocorrelation between successive samples and improve efficiency of the
14 Monte Carlo Markov Chain sampling. For each model, we ran three Markov
15 chains with a suitable number of iterations so that convergence was reached for
16 all variables on the basis of the Gelman-Rubin statistic (i.e., r < 1.05).
17 Independent variables were considered to display ‘some evidence’ of a
18 difference if the proportion of CI overlap was no more than half the average
19 length of the two overlapping arms (p < 0.05) (Cumming and Finch, 2005), and
20 the two overlapping arms did not differ in length by more than a factor of two.
21 Independent variables were considered to display ‘quite strong evidence’ of a
22 difference if there was no overlap, or a gap between credible intervals (p < 0.01)
23 (Cumming, 2009, Cumming and Finch, 2005).
24
10
1 Relationships with habitat variables
2 Habitat variables were analyzed using a Pearson’s correlation analysis and
3 correlated variables (r > 0.4) were removed from subsequent analysis, leaving
4 the variable that had the greatest ecological influence based on biological
5 knowledge (Elith and Leathwick, 2009). This left six variables: mid-stratum
6 density, percentage of bare ground, rock cover, tussock grass cover, litter cover
7 and the proportion of native plants (Appendix G). We analyzed relationships
8 between reptile species richness and counts and the remaining habitat
9 variables measured at a site-type using a linear mixed model with a Poisson
10 distribution in WinBUGS (Appendix H). We also analyzed the relationships
11 between the two most abundant reptile species; Morethia boulengeri and
12 Menetia greyii and these habitat variables. To account for habitat differences
13 between regions and locations, region and location were fitted as random
14 effects. We fitted all combinations of main effects for habitat variables and
15 identified the best models using Deviance Information Criteria (DIC) values.
16 Models with DIC values within 2 points of the best model were considered to be
17 very similar to the best model, so were kept (Spiegelhalter et al., 2002). We
18 ensured that the variables in each of these models were not ‘pretending
19 variables’ (Anderson, 2008).
20
21 We assessed the importance of variables by calculating the multiplicative effect
22 (with 95% credible intervals) of each of the six uncorrelated habitat variables on
23 reptile species richness and counts and the count of individual reptile species
24 (Spiegelhalter et al., 2002). The multiplicative effect was calculated as the
11
1 exponent of the standardized coefficient in Poisson regression models. The
2 multiplicative effect size of 1 means ‘no clear evidence’ of a change in species
3 richness or count associated with a variable. Multiplicative effect sizes that
4 overlap 1 are likely to show ‘some evidence’ of a change in species richness
5 and counts. Effect sizes greater than 1 indicate a positive effect of the
6 explanatory variable on species richness or counts, while effect sizes less than
7 1 indicate a negative effect.
8
9 Influence of environmental variables on community composition
10 We used a redundancy analysis (RDA) to investigate how much of the variation
11 in reptile community composition could be explained by the six uncorrelated
12 environmental factors and/or the site-type and treatment shape. The RDA was
13 conducted in CANOCO version 4.0 (ter Braak and Šmilauer, 1998). The species
14 matrix was constructed using the total counts of each species transformed to
15 produce the Hellinger distance between sites (Rao, 1995). We divided the
16 treatment matrix into two separate variables: treatment shape (linear strips and
17 patches); and vegetation state (revegetated areas, remnant areas, and cleared
18 areas). See Jellinek et al. (2013a) for further details.
19
20 Results
21 We caught a total of 399 individuals from 14 different reptile species in the
22 Benalla region, and a total of 341 individuals from 17 different species in the
23 Wimmera (Appendix B). Five reptile species were only recorded in the Benalla
12
1 region and eight reptile species were only recorded in the Wimmera region.
2 These reptiles were caught over 10,050 trap nights, with an average of 13
3 individuals caught per 100 trap nights. More than 90% of the reptile species
4 were caught during the first 7 to 8 days of trapping at each location, with only
5 one species caught after that time. The most abundant species recorded were
6 Morethia boulengeri, Menetia greyii, Carlia tetradactyla and Ctenotus robustus.
7
8 Out of the 22 reptile species we surveyed, the probability of detection for 20 of
9 those species was >0.5 (Appendix B). Of these 20 species, 8 had a very high
10 probability of detection (>0.9) (Tyre et al., 2003). Moreover, when we estimated
11 the expected species richness at a site and compared that value to the
12 observed species richness, we found low variability (on average 1 species)
13 between the observed and expected species richness. These two analyses
14 suggest that our detection effort was sufficient to obtain solid inference about
15 how reptile species richness varies with habitat and treatment variables.
16 Similarly, our analysis of imperfect detection in individual species counts
17 showed that none of the individual species we analysed (M. boulengeri, M.
18 greyii, C. tetradactyla, C. robustus, Pogona barbata, Christinus marmoratus and
19 Lerista bougainvillii) showed a substantially different trend to site-type to those
20 reported below.
21
22 Treatment relationships
23 Reptile species richness and counts showed no clear evidence of an effect of
24 treatment across the treatment shapes or site-types. There was no evidence 13
1 that region influenced reptile species richness or counts in the different
2 treatments. While the southern rainbow skink (C. tetradactyla) showed no clear
3 evidence of being more abundant in any of the treatment types, it was the only
4 species that showed a trend towards having a higher counts in remnant linear
5 strips compared to other treatments (Fig. 3).
6
7 There was a trend towards lower reptile species richness and counts with
8 increasing distance from remnant patches in revegetated and cleared linear
9 strips, but not in remnant linear strips. Reptile species richness and counts
10 showed a declining trend in cleared and revegetated linear strips and an
11 increasing trend in remnant linear strips as distance from remnant patches
12 increased.
13
14 After the most abundant species (M. boulengeri, M. greyii, C. tetradactyla and
15 C. robustus) were removed from the analysis, to establish how less abundant
16 species responded to distance effects, reptile species richness and counts
17 showed some evidence of being higher in remnant linear strips compared to
18 revegetated linear strips 300 - 500m away from remnant patches (Fig. 4 a, b).
19 Rarer reptile species richness and counts showed a trend towards increasing in
20 remnant linear strips and decreasing in cleared linear strips as distance from
21 remnant patches increased. There was some evidence of decreasing rare
22 reptile species richness and counts in revegetated linear strips as distance from
23 remnant patches increased (Fig. 4 a, b, Appendix I). The count of C.
24 tetradactyla showed a trend of increasing in remnant linear strips and
14
1 decreasing in revegetated and cleared linear strips as distance from remnant
2 patches increased.
3
4 Relationships with habitat variables
5 Reptile species richness at a site was most influenced by location and region
6 (Appendix J), while habitat variables showed no clear evidence of a relationship
7 with reptile species richness (Fig. 5 a). Rock cover, litter cover, bare ground,
8 location and region had the greatest influence on reptile counts (Appendix J).
9 Rock cover showed some evidence of association with increasing reptile counts
10 (mean effect size = 5.2) (i.e., sites with the greatest rock cover had 5.2 times
11 the count of reptiles as sites with the least rock cover), as did litter cover (mean
12 effect size = 1.4). The percentage of bare ground showed some evidence of a
13 relationship with decreasing reptile counts (mean effect size = 0.3) (Fig. 5 b).
14
15 Habitat variables such as rock cover, the proportion of native plants and the
16 percentage of bare ground, along with the effect of location and region, had the
17 greatest influence on Morethia boulengeri counts (Appendix K). Bare ground
18 showed some evidence of association with decreasing M. boulengeri counts
19 (mean effect size = 0.5), while other habitat variables showed no clear
20 relationship with the counts of this species (Fig. 5 c). Menetia greyii was most
21 influenced at a site by mid-stratum density, litter cover, tussock cover, the
22 proportion of native plants, location and region (Appendix K). Mid-stratum
23 density and litter cover both showed some evidence of an association with
24 increasing M. greyii counts (mean effect size = 1.9 and 2.1) (Fig. 5 d). 15
1
2 Influence of environmental variables on community composition
3 The environmental and treatment (shape, site-type) variables used in this
4 analysis explained 29% of the variation in reptile community composition in the
5 Benalla region (Table 1). Environmental variables accounted for most of this
6 variation (21.5%) and this did not differ in the pure environmental matrix. This
7 meant that treatment variables were not strongly correlated with environmental
8 variables. The most influential environmental variables in the Benalla region
9 were rock cover and herb cover (Table 2). The treatment variables explained
10 around 8% of the total variation in reptile community composition, largely due to
11 the interaction between treatment shape and vegetation state (Table 2).
12 Treatment shape also had a substantial influence in the treatment matrix but not
13 in the pure treatment matrix (Table 2).
16
1
2 In the Wimmera region the overall environmental and treatment matrix
3 accounted for 35% of the variation in reptile community composition. Again, this
4 was largely a result of the environmental matrix accounting for 27% of the total
5 variation (Table 1). The environmental variables most strongly associated with
6 reptile community composition were litter cover, herb cover and mid-stratum
7 density, although mid-stratum density was substantially more influential in the
8 pure environmental matrix. Rock cover was also marginally influential in the
9 environmental matrix (Table 2). The treatment variables in the Wimmera
10 accounted for around the same amount of variation as those in the Benalla area
11 (7.5%), but were most influenced by vegetation state (revegetated, remnant,
12 and cleared areas) and to a lesser degree treatment shape (Table 2).
13
14 Discussion
15 This study is one of few to examine the relative importance of revegetation in
16 linear strips and revegetation adjacent to remnant patches. Unexpectedly, we
17 found little evidence that revegetation affected overall reptile species richness
18 and counts, regardless of whether it was a revegetated patch or linear strip.
19 There was little evidence of an effect of site-type (remnant, revegetated or
20 cleared) on overall reptile species richness or counts, and only a minor effect of
21 treatment shape and site-type on reptile community composition. However, by
22 examining distance from patches we were able to provide an insight into the
23 use of linear strips by reptiles. Also unexpectedly, we found that species
17
1 richness and counts of rare reptiles was likely to be higher in remnant linear
2 strips compared to cleared and revegetated linear strips, and the difference
3 increased as distance from remnant patches increased. The common species
4 Carlia tetradactyla also showed this pattern. Such trends have not been
5 reported previously in the literature.
6
7 Effectiveness of revegetation and the influence of habitat variables
8 Overall reptile species richness and counts were not influenced by revegetated,
9 remnant or cleared habitats, or by linear strips and patches of habitat. This is
10 further supported by the minimal influence of treatments on reptile community
11 composition. This is an unexpected result, but probably occurred because many
12 of the reptiles we studied were generalist species, and able to persist in a
13 variety of different habitats (Brown, Bennett and Potts, 2008, Driscoll, 2004,
14 Jellinek et al., 2004). Historic records show the presence of other reptile species
15 that we failed to record (Rawlinson, 1966), suggesting that many of the more
16 specialist species have already been lost (Brown et al., 2008, Driscoll, 2004).
17 Even if revegetation is undertaken, reptile species may not recover because
18 they may be regionally extinct. Ongoing extensive sampling would be needed to
19 confirm these localized extinctions and increase detectability in species with low
20 detection probabilities.
21
22 If reptile species have already been lost from the communities we studied, and
23 restoration efforts are not attracting rarer reptile species into restored
24 agricultural landscapes, then novel management actions may be required. Such 18
1 actions may include translocating reptile species that were present prior to
2 clearing into remnant and restored habitat, providing that appropriate
3 microhabitats are available (Christie, Craig, Stokes et al., 2011). Similarly, long-
4 term biodiversity funds and restoration activities that assist landholders in
5 restoring ground layers would be desirable to enhance habitat for translocated
6 animals (Jellinek, Parris, Driscoll et al., 2013b).
7
8 Reptile species richness and counts, individual reptile species and reptile
9 community composition responded strongly to environmental variables and in
10 particular to ground layer attributes. Rock cover had a strong positive effect on
11 reptile species counts and community composition, as rocky areas contained a
12 disproportionate number of the key microhabitat elements that reptiles require
13 such as refuges (Croak, Pike, Webb et al., 2010, Michael et al., 2010). Ground
14 cover variables and mid-stratum density were also important in structuring
15 reptile community composition (Bateman et al., 2008, Lindenmayer, Wood,
16 MacGregor et al., 2008, Valentine and Schwarzkopf, 2009). The cover of litter,
17 tussocks, native plants and herbs, as well as a dense mid-stratum that would
18 contribute litter to the ground layer, probably help to reduce bare ground and
19 increase important microhabitats for reptiles (Leynaud and Bucher, 2005, Smith
20 et al., 1996). Although Driscoll (2004) speculated that high shrub density would
21 increase shading and be detrimental to reptiles, our results suggest that mid-
22 stratum density was not negatively correlated with reptile species richness or
23 counts.
24
19
1 A lack of some ground layers in relatively young revegetated areas (15 years)
2 may have limited the use of replanted habitats by some reptile species (Croak
3 et al., 2010, Cunningham et al., 2007, Marquez-Ferrando, Pleguezuelos,
4 Santos et al., 2009). The enhancement of habitat variables by replacing rocks
5 and other ground layers may benefit these reptile species by providing refuges
6 (Marquez-Ferrando et al., 2009, Michael, Lunt and Robinson, 2004). Many
7 ground layer structures can take a long time to develop (Vesk, Nolan, Thomson
8 et al., 2008). Similarly, weed control and the replanting of native grasses can
9 increase reptile species that have declined as a result of weed invasion
10 (Valentine and Schwarzkopf, 2009). However, to revegetate natural habitats
11 more effectively we need to adaptively manage these areas so they provide the
12 specific habitat requirements of rarer reptile species without reducing species
13 that require open habitats.
14
15 Influence of distance along linear strips
16 While a number of studies have shown that linear strips can increase movement
17 and dispersal for some invertebrate and mammal species (Haddad, 2000, Hill,
18 1995, Lindenmayer, Cunningham, Donnelly et al., 1994), few studies report on
19 the distance species are located along linear strips (Haddad et al., 2003). What
20 could explain the strong decline of rare reptiles in revegetated linear strips,
21 weaker decline in cleared linear strips and some evidence of an increase in
22 remnant linear strips?
23
20
1 Reptile declines with increasing distance from the patch is consistent with a
2 “peninsula effect” (Simpson, 1964). The decline of species at the peninsula tip
3 may be related to a decline in habitat quality (Means and Simberloff, 1987),
4 which could prevent successful reproduction at larger distances from the patch
5 or discourage dispersal. Alternatively, the peninsula effect may arise directly
6 from limited dispersal. If habitat quality is at fault by restricting the availability of
7 habitat for rare reptiles, restoration to improve use of linear remnants by
8 including suitable habitat variables may be possible. If dispersal limitation is the
9 problem, then wider linear strips of vegetation will be needed to establish
10 functional connectivity.
11
12 A competition/colonization trade-off, where species that are the weakest
13 competitors are the best dispersers (Driscoll, 2008, Huxel and Hastings, 1999,
14 Tilman, May, Lehman et al., 1994) may help to explain the observation of
15 increasing reptile species richness and counts with increasing distance along
16 linear remnants from the patch. In this scenario the strongest competitors (or
17 predators) could dominate patches (Driscoll, Kirkpatrick, McQuillan et al., 2010,
18 Tilman et al., 1994). These dominant species may exclude other species, but
19 have a limited tendency to move into linear strips, resulting in an increase in
20 rare species as distance away from patches increased (Driscoll, 2008). There
21 were no reptiles confined to patches, but other faunal groups could drive this
22 pattern via competition or predation.
23
21
1 Conclusion
2 In response to our three main questions we found that: (i) Overall reptile
3 species richness and counts showed very little response to the treatments we
4 studied. This may suggest that more specialized reptiles have been lost from
5 these agricultural landscapes, leaving a robust subset of generalist reptile
6 fauna. Larger scale habitat restoration programs along with reintroductions of
7 rarer reptiles may be necessary to re-establish diverse reptile communities. (ii)
8 Linear strips are complex landscape elements whereby increasing distance
9 away from remnant patches influences rarer reptile and individual species
10 differently between linear remnants, replantings and cleared roadsides. Little is
11 known about how reptiles use linear strips, however it is possible that species
12 interactions may play an important role in structuring reptile communities in
13 linear habitat. (iii) Environmental variables are important in structuring reptile
14 communities, so restoring ground layer elements during restoration activities will
15 be important to provide adequate habitat for rarer reptile species.
16
22
1 Acknowledgements 2 3 SJ undertook this project thanks to the assistance of the Norman Wettenhall
4 Foundation, the University of Melbourne Albert Shimmins Postgraduate
5 Scholarship, Greening Australia Victoria, the Wimmera Catchment Management
6 Authority and VicRoads. This article was supported by the University of
7 Melbourne Albert Shimmins postgraduate writing-up award (SJ). DD was
8 supported by the Australian Government’s national Environmental Research
9 Program and the Australian Research Council Centre of Excellence for
10 Environmental Decisions. We thank the landholders who allowed us onto their
11 properties and volunteers who helped with this research and Professor Geoff
12 Cumming for assistance with the interpretation of credible intervals.
13
23
1 References 2 Anderson, D. R. (2008). Model based inference in the life sciences: A primer on 3 evidence. New York: Springer. 4 Bailey, S. (2007). Increasing connectivity in fragmented landscapes: An 5 investigation of evidence for biodiversity gain in woodlands. For Ecol 6 Manage, 238: 7-23. 7 Bateman, H. L., Chung-Maccoubrey, A. , Snell, H. L. (2008). Impact of non- 8 native plant removal on lizards in riparian habitats in the southwestern 9 united states. Restor Ecol, 16: 180-190. 10 Bennett, A. F., Crooks, K. R. , Sanjayan, M. A. (2006). The future of connectivity 11 conservation. In: Connectivity conservation: 676-694. Crooks, K. R. , 12 Sanjayan, M. A. (Eds.). New York: Cambridge University Press. 13 Brown, G. P., et al. (2006). Toad on the road: Use of roads as dispersal 14 corridors by cane toads (bufo marinus) at an invasion front in tropical 15 australia. Biol Conserv, 133: 88-94. 16 Brown, G. W., Bennett, A. F. , Potts, J. M. (2008). Regional faunal decline - 17 reptile occurrence in fragmented rural landscapes of south-eastern 18 australia. Wildl Res, 35: 8-18. 19 Cabrera-Guzman, E. , Hugo Reynoso, V. (2012). Amphibian and reptile 20 communities of rainforest fragments: Minimum patch size to support high 21 richness and abundance. Biodivers Conserv, 21: 3243-3265. 22 Carthew, S. M., Horner, B. , Jones, K. M. W. (2009). Do utility corridors affect 23 movements of small terrestrial fauna? Wildl Res, 36: 488-495. 24 Christie, K., et al. (2011). Movement patterns by egernia napoleonis following 25 reintroduction into restored jarrah forest. Wildl Res, 38: 475-481. 26 Croak, B. M., et al. (2010). Using artificial rocks to restore nonrenewable shelter 27 sites in human-degraded systems: Colonization by fauna. Restor Ecol, 28 18: 428-438. 29 Cumming, G. (2009). Inference by eye: Reading the overlap of independent 30 confidence intervals. Statistics in Medicine, 28: 205-220. 31 Cumming, G. , Finch, S. (2005). Inference by eye - confidence intervals and 32 how to read pictures of data. American Psychologist, 60: 170-180. 33 Cunningham, R. B., et al. (2007). Reptile and arboreal marsupial response to 34 replanted vegetation in agricultural landscapes. Ecol Appl, 17: 609-619. 35 Dixo, M. , Metzger, J. P. (2009). Are corridors, fragment size and forest 36 structure important for the conservation of leaf-litter lizards in a 37 fragmented landscape? Oryx, 43: 435-442. 38 Driscoll, D. A. (2004). Extinction and outbreaks accompany fragmentation of a 39 reptile community. Ecol Appl, 14: 220-240. 40 Driscoll, D. A. (2008). The frequency of metapopulations, metacommunities and 41 nestedness in a fragmented landscape. Oikos, 117: 297-309. 42 Driscoll, D. A., et al. (2010). Classic metapopulations are rare among common 43 beetle species from a naturally fragmented landscape. J Anim Ecol, 79: 44 294-303. 45 Duncan, D., Moxham, C. , Read, C. (2007). Effect of stock removal on 46 woodlands in the murray mallee and wimmera bioregions of victoria.). 47 Melbourne, Australia: Department of Sustainability and Environment.
24
1 Elith, J. , Leathwick, J. R. (2009). Species distribution models: Ecological 2 explanation and prediction across space and time. In: Annual review of 3 ecology evolution and systematics: 677-697.). Palo Alto: Annual 4 Reviews. 5 Haddad, N. (2000). Corridor length and patch colonization by a butterfly, junonia 6 coenia. Conserv Biol, 14: 738-745. 7 Haddad, N. M., et al. (2003). Corridor use by diverse taxa. Ecology, 84: 609- 8 615. 9 Harrison, S. , Bruna, E. (1999). Habitat fragmentation and large-scale 10 conservation: What do we know for sure? Ecography, 22: 225-232. 11 Hilborn, R. , Mangel, M. (1997). The ecological detective: Confronting models 12 with data. New Jersey: Princeton University Press. 13 Hill, C. J. (1995). Linear strips of rain forest vegetation as potential dispersal 14 corridors for rain forest insects. Conserv Biol, 9: 1559-1566. 15 Hodgson, J. A., et al. (2009). Climate change, connectivity and conservation 16 decision making: Back to basics. J Appl Ecol, 46: 964-969. 17 Huxel, G. R. , Hastings, A. (1999). Habitat loss, fragmentation, and restoration. 18 Restor Ecol, 7: 309-315. 19 Jellinek, S., Driscoll, D. A. , Kirkpatrick, J. B. (2004). Environmental and 20 vegetation variables have a greater influence than habitat fragmentation 21 in structuring lizard communities in remnant urban bushland. Austral 22 Ecol, 29: 294-304. 23 Jellinek, S., Parris, K. P. , Driscoll, D. A. (2013a). Are only the strong surviving? 24 Little influence of restoration on beetles (coleoptera) in an agricultural 25 landscape. Biol Conserv, 162: 17-23. 26 Jellinek, S., et al. (2013b). Are incentive programs working? Landowner 27 attitudes to ecological restoration of agricultural landscapes. J Environ 28 Manage, 127: 69-76. 29 Kanowski, J. J., et al. (2006). Factors affecting the use of reforested sites by 30 reptiles in cleared rainforest landscapes in tropical and subtropical 31 australia. Restor Ecol, 14: 67-76. 32 Kery, M. , Schaub, M. (2011). Bayesian population analysis using winbugs : A 33 hierarchical perspective. Elsevier Science. 34 Kitchener, D. J., et al. (1980). Lizard assemblages and reserve size and 35 structure in the western australian wheatbelt - some implications for 36 conservation. Biological Conservation, 17: 25-62. 37 Leynaud, G. C. , Bucher, E. H. (2005). Restoration of degraded chaco 38 woodlands: Effects on reptile assemblages. For Ecol Manage, 213: 384- 39 390. 40 Lindenmayer, D. B. (2009). Forest wildlife management and conservation. In: 41 Year in ecology and conservation biology 2009: 284-310.). Oxford: 42 Blackwell Publishing. 43 Lindenmayer, D. B., et al. (1994). Factors influencing the occurrence of 44 mammals in retained linear strips (wildlife corridors) and contiguous 45 stands of montane ask forest in the central highlands of victoria, south- 46 eastern australia. Forest Ecology and Management, 67: 113-133. 47 Lindenmayer, D. B., et al. (2008). How predictable are reptile responses to 48 wildfire? Oikos, 117: 1086-1097. 25
1 Lunn, D. J., et al. (2000). Winbugs: A bayesian modelling framework: Concepts, 2 structure, and extensibility. Statistics and Computing, 10: 325-337. 3 Mac Nally, R. , Brown, G. W. (2001). Reptiles and habitat fragmentation in the 4 box-ironbark forests of central victoria, australia: Predictions, 5 compositional change and faunal nestedness. Oecologia, 128: 116–125. 6 Marquez-Ferrando, R., et al. (2009). Recovering the reptile community after the 7 mine-tailing accident of aznalcollar (southwestern spain). Restor Ecol, 8 17: 660-667. 9 Mccarthy, M. A. (2007). Bayesian methods for ecology. New York: Cambridge 10 University Press. 11 Means, D. B. , Simberloff, D. (1987). The peninsula effect: Habitat-correlated 12 species decline in florida herpetofauna. J Biogeogr, 14: 551-568. 13 Michael, D. R., Cunningham, R. B. , Lindenmayer, D. B. (2010). Microhabitat 14 relationships among five lizard species associated with granite outcrops 15 in fragmented agricultural landscapes of south-eastern australia. Austral 16 Ecol, 35: 214-225. 17 Michael, D. R., Lunt, I. D. , Robinson, W. A. (2004). Enhancing fauna habitat in 18 grazed native grasslands and woodlands: Use of artificially placed log 19 refuges by fauna. Wildl Res, 31: 65-71. 20 Munro, N. T., et al. (2011). Bird's response to revegetation of different structure 21 and floristics-are "restoration plantings" restoring bird communities? 22 Restor Ecol, 19: 223-235. 23 Munro, N. T., et al. (2009). Revegetation in agricultural areas: The development 24 of structural complexity and floristic diversity. Ecol Appl, 19: 1197-1210. 25 Munro, N. T., Lindenmayer, D. B. , Fischer, J. (2007). Faunal response to 26 revegetation in agricultural areas of australia: A review. Ecol Manag 27 Restor, 8: 199-207. 28 O’shea, E. M. , Kirkpatrick, J. B. (2000). The impact of suburbanisation on 29 remnant coastal vegetation in hobart, tasmania. Applied Vegetation 30 Science, 3: 243-252. 31 Radford, J. Q., Bennett, A. F. , Cheers, G. J. (2005). Landscape-level 32 thresholds of habitat cover for woodland-dependent birds. Biol Conserv, 33 124: 317-337. 34 Rao, C. R. (1995). A review of canonical coordinates and an alternative to 35 correspondence analysis using hellinger distance. Quaderns 36 d'Estadı´stica i Investigacio´ Operativa, 19: 23-63. 37 Rawlinson, P. A. (1966). Reptile of the victorian mallee. Proceedings of the 38 Royal Society of VIctoria, 71: 605-619. 39 Royle, J. A. (2004). N-mixture models for estimating population size from 40 spatially replicated counts. Biometrics, 60: 108-115. 41 Ruiz-Jaen, R. C. , Aide, M. T. (2005). Restoration success: How is it being 42 measured? Restor Ecol, 13: 569-577. 43 Ryan, P. A. (2000). The use of revegetated areas by vertebrate fauna in 44 australia: A review. In: Temperate eucalypt woodlands in australia: 45 Biology, conservation, management and restoration: 318-335`. Hobbs, R. 46 J. , Yates, C. J. (Eds.). Chipping Norton: Surrey Beatty & Sons.
26
1 Schutz, A. J. , Driscoll, D. A. (2008). Common reptiles unaffected by 2 connectivity or condition in a fragmented farming landscape. Austral 3 Ecol, 33: 641-652. 4 Simberloff, D. , Cox, J. (1987). Consequences and costs of conservation 5 corridors. Conserv Biol, 1: 63-71. 6 Simberloff, D., et al. (1992). Movement corridors: Conservation bargains or poor 7 investments? Conserv Biol, 6: 493-504. 8 Simpson, G. G. (1964). Species density of north american recent mammals. 9 Syst Zool, 13: 57-73. 10 Smith, G. T., et al. (1996). The effect of habitat fragmentation and livestock 11 grazing on animal communities in remnants of gimlet eucalyptus salubris 12 woodland in the western australian wheatbelt. Ii.Lizards. J Appl Ecol, 33: 13 1302-1310. 14 Spiegelhalter, D. J., et al. (2002). Bayesian measures of model complexity and 15 fit. Journal Of The Royal Statistical Society Series B, 64: 583-639. 16 Ter Braak, C. J. F. , Šmilauer, P. (1998). Canoco reference manual and user’s 17 guide to canoco for windows: Software for canonical community 18 ordination (ver. 4). Microcomputer Power. 19 Tewksbury, J. J., et al. (2002). Corridors affect plants, animals, and their 20 interactions in fragmented landscapes. Proc Natl Acad Sci U S A, 99: 21 12923-12926. 22 Tilman, D., et al. (1994). Habitat destruction and the extinction debt. Nature, 23 371: 65-66. 24 Tyre, A. J., et al. (2003). Improving precision and reducing bias in biological 25 surveys: Estimating false-negative error rates. Ecol Appl, 13: 1790-1801. 26 Valentine, L. E. , Schwarzkopf, L. (2009). Effects of weed-management burning 27 on reptile assemblages in australian tropical savannas. Conserv Biol, 23: 28 103-113. 29 Vesk, P. A., et al. (2008). Time lags in provision of habitat resources through 30 revegetation. Biol Conserv, 141: 174-186. 31
27
1 Table 1. Percentage influence of environmental and treatment variables in
2 explaining the variance of reptile community composition in the Benalla and
3 Wimmera regions with redundancy analysis (RDA) and partial RDA showing the
4 percentage of total variance explained in the first axis and all ordination axes.
5 Env = environmental matrix; Treat = treatment matrix; Env-Treat =
6 environmental matrix with treatment matrix removed; Treat-Env = treatment
7 matrix with environmental matrix removed.
8
Benalla Wimmera
Matrix Variance explained Variance explained Variance explained Variance explained
(%) First Axis (%) All Axes (%) First Axis (%) All Axes
Env 10 21.5 16.3 27.3
Treat 4 8 4.2 7.5
Env - Treat 11.5 21.4 16.6 27
Treat - Env 5.3 7.8 5.7 7.2
Total 29.4 34.8
9
10
28
1 Table 2. The contribution of the environmental variables (Env) and the
2 treatment matrix (Treat) to the variation in the reptile community composition
3 matrix in the Benalla and Wimmera regions, explained by RDA and partial RDA.
4 Env-Treat = environmental matrix with treatment matrix removed (pure
5 environmental matrix); Treat-Env = treatment matrix with environmental matrix
6 removed (pure treatment matrix).
7
Benalla Wimmera Benalla Wimmera Variables Matrix Variation % Variation % Matrix Variation % Variation %
Rock Env 26.7 11.8 Env-Treat 24.1 11.4
Herb Env 16.7 14.7 Env-Treat 20.7 14.3
Mid-stratum Env 10.0 14.7 Env-Treat 6.9 20
Tussock Env 6.7 5.9 Env-Treat 10.3 5.7
Litter Env 6.7 23.5 Env-Treat 3.4 17.1
Natives Env 6.7 8.8 Env-Treat 6.9 8.6
Treat interaction Treat 13.8 5.7 Treat-Env 13.3 2.9
Shape Treat 10.3 5.7 Treat-Env 6.7 5.9
Vegetation state Treat 3.4 11.4 Treat-Env 6.7 11.6
8 9
29
1 Fig. 1 Maps of the (a) Benalla and (b) Wimmera regions in Victoria, south-
2 eastern Australia. Shaded areas correspond to reserves; black circles represent
3 linear strip treatments; black triangles represent patch treatments surveyed as
4 part of this study.
5
6 Fig. 2 The sampling design used in this study. There were (a) two regions
7 (Wimmera and Benalla), two treatment shapes (linear strips and patches) and
8 five locations for each treatment shape within each region. Within each of the
9 (b) linear strip locations there were four site-types: (i) revegetated linear strip,
10 (ii) cleared linear strip, (iii) remnant linear strip, and (iv) remnant patch. Within
11 each of the (c) patch treatments there were two site-types: (v) remnant patch
12 and (vi) revegetated patch. Each site-type had five pitfall lines (a - e) and each
13 line (represented by circles) contained two pitfall traps.
14
15 Fig. 3 Counts of Carlia tetradactyla in linear strips treatments. Counts were the
16 total number of uniquely marked individuals during the study period.
17 CLS = cleared linear strip, RemLS = Remnant linear strip, RemP = Remnant
18 patch, RevLS = Revegetated linear strip. Bars represent 95% credible intervals.
19
30
1 Fig. 4 Rare reptile species richness and counts in linear strip treatments as
2 distance from the remnant patches increased: (a) overall count and (b) species
3 richness. Reptile counts were the total number of uniquely marked individuals
4 during study period. White columns = remnant linear strips, light grey columns =
5 revegetated linear strips, dark grey columns = cleared linear strips. Bars
6 represent 95% credible intervals.
7
8 Fig. 5 The multiplicative effect of: rock cover, mid-stratum density, litter cover,
9 bare ground, the proportion of native plants, and tussock cover on (a) reptile
10 species richness, (b) reptile counts (the total number of uniquely marked
11 individuals during study period), (c) Morethia boulengeri counts and (d) Menetia
12 greyii counts. Numbers below the habitat variable label represent the best-
13 supported model with that variable included (Appendix H, Appendix I). Bars
14 represent 95% credible intervals.
15
31
1
2 Fig. 1 a
3
4
5 Fig. 1 b
6
32
1
2 Fig. 2 a
3
33
1
2 Fig. 2 b
3
4 Fig. 2 c
5
34
1
2 Fig. 3
3
4 Fig. 4 a
35
1
2 Fig. 4 b
3
4 Fig. 5 a
36
1
2 Fig. 5 b
3
4 Fig. 5 c
5
6 Fig. 5 d 37