Introduction

The predator/prey relationship plays an integral role in community structure: the number of predators and the rate they consume prey determine ecosystem dynamics (Miller et al. 2006). In the process of habitat fragmentation, the dynamic interaction among co-exiting species may be disrupted (Cardinale et al. 2006). Historically, habitat edges have been viewed as problematic in this relationship. They have variously been described as a barrier to predators (Bider 1968), and funnels for predator movement (Dyer et al. 2001). However, in a human-modified environment, predators (e.g. foxes; Molsher et al. 2000; birds; Catterall et al. 1991) may successfully cross forage between a remnant’s perimeter and adjacent modified ecosystems. This may increase opportunities for these predators in edge areas (Berry 2001).

Edge habitats are typically inhabited by a suite of omnivorous birds with large body size compared to those in the core of the remnant. These attributes allow for a larger number of avian predators to inhabit edge areas (Major et al. 2001; Shirley and Smith 2005). However, there is a dearth of information on the impact of such species on reptile prey (Barrows et al. 2006).

In Australia, reptiles are key species in endemic ecosystems (Brown and Nelson 1993), however, in fragmented landscapes research on the impact on reptiles, particularly small species that are ready prey to birds, is limited (e.g., Anderson and Burgin 2002; Barrows and Allen 2007).

Anderson and Burgin (2002) observed that the widespread, sympatric and common species of small (snout − vent length ≤55 cm) skinks, Lampropholis delicata and Lampropholis guichenoti, were significantly more sparse in the edge areas, compared to the core of small woodland remnants in north western Sydney (Australia). Micro-climate (Baiada 2002) and habitat characteristics (Anderson and Burgin 2002) have been previously shown not to have a significant impact on the distribution of skinks between edge and core habitat of remnants. In this paper we investigate if predatory birds influence the abundance of small skinks in these open woodland remnants.

Method

Site description

The study was undertaken in remnant woodlands of the Cumberland Plain. Since European settlement these woodlands have been extensively cleared for agriculture. More recently, urban encroachment has accelerated the woodlands’ fragmentation (Benson and Howell 1990; James 1997). As a consequence, by 1990 less than 6% of the original vegetation remained in patchy and isolated remnants (Tozer 2003). The study was undertaken in such remnants on the Hawkesbury campus of the University of Western Sydney, near Richmond, Australia (150°75′ E, 33°62′ S; Fig. 1) commencing with summer of 2000 and finishing at the end of summer 2002.

Fig. 1
figure 1

Map of the study area showing edge (1–4) and core sites (5–8). Inset shows the study site in relation to Sydney, New South Wales and Australia

Full size image

The eight remnant woodlands sampled ranged in size between 20 and 176 ha and abutted areas of agriculture, power line easements and/or roadways (see Fig. 1). Initial fragmentation occurred almost 120 years ago and, although the intensity of agriculture has waned over time, current land use has been in place since that time, while fire has been excluded for at least 40 years (Burgin in press). Previous research in these open woodland remnants showed that habitat characteristics that are generally used to determine small reptile distribution did not vary between the edge and core of these remnants (Anderson and Burgin 2002). Baiada (2002) also found that microclimate conditions in these remnants did not differ significantly along a gradient between the edge and core of the remnants.

The climate of the area is temperate with cool, dry winters and hot, wet summers. During the study, average monthly rainfall was 103.2 mm, with unseasonably high rainfall during February 2002 (274.5 mm), while the lowest rainfall (14.3 mm) occurred in September 2001. The average ambient air temperature over the sampling period was 21.1°C. The highest monthly temperature was recorded in December 2001 (40°C) while May 2001 temperature dropped to 0°C. Between mid-December 2001 and early January 2002 bushfires threatened the area.

Sampling

Four perimeter and four core areas of remnants were randomly selected from among those available on the campus. Kremsater and Bunnell (1999) determined that species distribution is most influenced within 50 m of the edge of remnants. We sampled within the outer 10 m wooded perimeter of the remnant, and the core sampling was conducted at ≥150 m from the edge.

At 4 weekly intervals, skinks were sampled over three periods (autumn, spring 2001, and summer 2002). Sampling was undertaken in the mid-morning when small skinks actively forage, generally over 4 consecutive days with some adjustment for inclement weather. Each day two pairs of sites (i.e. edge and core sites arbitrarily paired for efficiency of sampling) were sampled. The sequence of sampling was re-randomised each month. The point of transect survey was used instead of pitfall trapping (e.g. Singh et al. 2002), in part, because of the similarity in distribution of trees and understorey throughout these long-established remnants and limited weed intrusion allowed for equivalent sight lines in both habitat types. Previous experience in these remnants (Anderson and Burgin 2002) had also shown that traps were time consuming to set up and yielded relatively fewer animals than the point of transect survey approach.

Two transects of 20 m length and at least 3 m apart, with points at every 5 m, were randomly chosen within each site on each day of sampling. These distances were based on unpublished observations of home range of these small skinks. To record skink numbers, the researcher crouched for 5 min at each point along the transect to observe the area immediately ahead and to either side. All skinks observed were identified and numbers recorded. Since sampling was restricted by line of vision and accuracy of identification was compromised beyond approximately 1.8 m, animals beyond 1.6 m were not included in the census to ensure accuracy of identification.

All small skinks observed were recorded and used to establish the initial distribution and diversity patterns of skinks within the woodlands. However, because the numbers of two species were too small to be amenable to statistical treatment, they were excluded from further analyses.

Diurnal bird sampling is traditionally conducted in the early morning as they leave their roosting sites (Pizzey and Knight 1997), however, we sampled mid-morning when skinks were active and therefore potentially available for predation. Birds were also sampled on transect lines, although transects were extended to 50 m with points 10 m apart. Species were identified visually and verified using Pizzey and Knight (1997), and by call recognition. The species numbers, and the number within each species, were recorded. Birds were subsequently categorised as predators, based on information gleaned from the literature (Barker and Vestjens 1984a, b; Higgins 1999; Higgins and Davies 1996; Higgins et al. 2001), and observations made during the study. All potential predators were recorded, although one species (European starling Sturnus vulgaris) was not included in the analyses because it was only encountered in low numbers in edge habitats and not in core habitat.

To investigate predation, decoys were used to mimic juvenile, sub-adult and adult small skinks. These were constructed of plastic or latex, and cast in a basking position. They were overlayed with 3-M Blu-tac™ and painted either ‘metallic grey’ or ‘metallic brown’ with folk art paint. Finally, each ‘animal’ was also threaded with ‘15 lb’ fishing line to tether it to the substrate.

Fifteen decoys (five of each size) were tethered at random points along a single transect (1 transect/week). To mimic basking skinks, the decoys were placed on logs, fallen branches and tree stumps, and they were left in position for 7 days. After collection, decoys were examined for unambiguous beak mark impressions (to calculate attack rates), and the size of the decoy (juvenile, sub-adult, adult) were recorded. The procedure was repeated on new transects in subsequent weeks. These data were collected for 4 weeks each season.

The sampling periods were modified in some seasons. Due to extreme weather conditions sampling in summer 2002 commenced in December 2001, but was interrupted by bush fires and heavy rains which precluded fieldwork until the end of February 2002. This delayed the commencement of the autumn 2002 sampling, and it was abandoned after 2 weeks, due to extreme weather conditions. Overall, decoys were in place for 126 days during the period December 2000 and March 2002.

In spring 2001, and summer and autumn 2002 decoys were further examined for the position of the ‘bite mark’ on the body (head, torso, tail). Tail attacks were of particular interest because a major anti-predator behaviour displayed by the skinks is tail autotomy. Calculations were made to determine ‘escape rate’ differences between the edge and core of remnants.

If multiple marks were observed only one attack was counted (usually the most conspicuous), unless one of the marks included the tail. This was because an attack on the tail usually denoted a ‘miss’ by the predator since these skinks readily drop their tails when they are in a position to escape by this means (pers. obs.).

Data analysis

Reptile and bird distribution were analysed using multivariate analysis within PRIMER, Version 5 (2001) (Clarke and Ainsworth 1993; Clarke and Warwick 2001). Multi-dimensional scaling stress tests were used to test homogeneity of variances. When not homogeneous (tested using Cochran C), a log10 transformation was used. Similarity was tested using one-way ANOSIM.

Interactions between the skinks and birds were analysed using the BIOENV function, also within PRIMER. This function allowed for the generation of multiple correlations (using Spearman’s weighted rank correlation) with various combinations of predatory species to identify a best fit explanation for the distribution pattern of skinks.

Where three or more birds were recorded at a site, a Chi-square analysis was used to investigate differences in predatory bird distribution between core and edge of remnants. The same analysis was used on predation based on number of attacks and decoy size. Spearman’s rho, rank correlation coefficient was used to examine the relationship between predator–prey abundance and number of attacks.

Results

Skink distribution

Overall 486 skinks, representing four species Lampropholis guichenoti, Lampropholis delicata, Cryptoblepharous virgatus and Saproscincus mustelina were encountered: 352 in the core and 134 on the edge of sites (Fig. 2). Within habitat type (edge, core), skink numbers were more similar among replicates than habitat types (global R = 1, P = 0.029), although core sites were more similar (group average 97%) than edge areas (88.8%). Most individuals (95%) were one of the two Lampropholis species, and both L. guichenoti (χ2 1, 0.0001 = 96.43) and L. delicata (χ2 1, 0.0001 = 68.36) varied significantly between the habitat types.

Fig. 2
figure 2

Total number of skinks of each species that were encountered in edge and core habitat of woodland remnants in peri-urban north-western Sydney in 2001–2002

Full size image

Bird distribution

A total of 445 predatory/omnivorous birds, encompassing seven species were countered: 311 in edge and 134 in core habitats (Fig. 3). There was a significant difference in bird numbers between habitat types (global R = 0.427, P = 0.029): substantially more predatory birds were present in edge than in core areas. In contrast to the skinks, the edge diversity and abundance were more similar (84.2%) than core sites (72.7%). Four of these six species that preyed on skinks (grey butcherbird, noisy miner; magpie and raven; see Table 1), and were common to both habitat types, were in significantly greater numbers in edge compared to core habitat. While individually species showed a strong relationship between bird and skink distribution, Spearman’s rank correlation coefficient indicated that the strongest correlation was a combination of five species (ρ = 0.894; Table 2).

Fig. 3
figure 3

Total number of birds in each species identified as predators of skinks encountered in edge and core habitats of woodland remnants in peri-urban north-Western Sydney, collected in 2001–2002

Full size image
Table 1 Overall number (n) of each avian species recorded in remnant woodland habitat in 2001–2002 that were identified as skink predators, the mean number and standard deviation (SD) among replicates in edge and core habitat types, and statistical comparison of bird numbers between habitat types (** = 0.01 significance level, *** = 0.001 significance level; NS = not significant)
Full size table
Table 2 Spearman’s rank correlation coefficient on up to six identified avian predators on small skinks, compared with skink census data
Full size table

Predation rate

Overall, 704 decoys were unambiguously attacked by birds: 411 on the edge, and 293 in the core of remnants. Attack rate was significantly different between edge and core habitats (χ2 2, 0.001 = 23.31). A similar number of ‘adult’ (259), ‘sub-adult’ (231) and ‘juvenile’ (214) decoys were attacked. There was no significant difference in attack due to size of decoy, or due to zone of attack.

Attack correlation

Analysis indicated that the number of attacks was negatively correlated with prey abundance (r = −0.455, P < 0.257), and there was a strong and significant negative correlation between predator and prey numbers (r = −0.810, P < 0.015), and the number of attacks was strongly and significantly positively correlated (r = 0.790, P < 0.02) with the number of predators.

Discussion

The two Lampropholis species that dominate the woodlands we studied are common, widespread, and generally the first species to re-establish, even after major disturbance such as logging (Goldingay et al. 1996; Lunney et al. 1991; Webb 1995), fire (Lunney et al. 1991; Taylor and Fox 2001), mine rehabilitation (Bragg et al. 2005; Taylor and Fox 2001), and grazing (Abensperg-Traun et al. 1996; Fischer et al. 2004). They also commonly inhabit urban gardens (Burgin 1993). However, in the small remnant woodlands of the Cumberland Plain where we undertook this study, their numbers in edge areas were approximately one third of overall numbers. Based on the previous study by Anderson and Burgin (2002), this is the established pattern of skink distribution in these woodlands.

In contrast to the skink distribution, the abundance and diversity of avian predators were greater in edge areas, compared to core habitat. Chace and Walsh (2006) suggested that the abundance of avian predators in edge habitats was due to their ability to cross-forage, and thus take advantage of adjacent urban areas. Barrows et al. (2006) demonstrated that this occurred. They noted that cross foraging between remnants and the adjacent human-modified landscape resulted in increased avian predator numbers, compared to those present pre-fragmentation. In their study this resulted in increased predation on the flat-tail horned lizard (Phrynosoma mcalli).

The species that we observed to prey on skinks (e.g., magpies, butcherbirds, kookaburras) use manmade structures (e.g., power lines, fences) in their daily routine, at least as effectively as they use natural habitat. Sewell and Catterall (1998) suggested that manmade structures provided these bird predators with an advantage. We observed birds perched on powerlines and fences, scanning the interface between the remnant edge and the peri-urban matrix surrounding the remnants and, when movement was identified, the bird swooped to collect their prey, usually a small skink.

The adjacent modified habitat, particularly urban backyards, provides supplementary food for these birds. For example, noisy miners were common in the remnants. Although typically considered honey eaters, they supplement their diet with arthropods and small reptiles (Grey et al. 1998). Parsons et al. (2006) observed that noisy minors were more abundant in gardens where meat was available than in those where only seed or nectar was present. They concluded that the minors were more carnivorous than previously assumed. We also observed them prey on skinks. Other species that are common visitors to urban gardens include the starling, magpie, kookaburra and butcherbird (Catterall et al. 1991, 2001; White et al. 2005) and none that foraged at the edge of woodland remnants were obligate skink predators. All were dietary generalists (see Barker and Vestjens 1984a, b). This cross-foraging tactic enables predators to persist despite low prey numbers (Andrén and Angelstam 1988; Norrdahl and Korpimaki 2000). Predation pressure would therefore not be released due to predators dispersing from the site because of depleted prey numbers (Norrdahl and Korpimaki 2000).

Previous research (e.g., Bagchi and Sankar 2003) has shown that most predators take prey in proportion to their abundance. This was not our observation. The use of decoys to mimic small skinks, showed that the pattern of bird attack was consistent with greater predation pressure in edge areas compared to the core of remnants, despite the density of skinks being much lower in the edge.

The avian predators showed no preference for zone of attack (head, body, tail). Caudal autotomy is a defence mechanism against predation in many species (Downes and Shine 2001; Lin et al. 2006), and many avian predators direct attacks preferentially towards the prey’s head (Langkilde and Shine 2004; Shepard 2007). There was no evidence that this occurred: decoy skinks were equally likely to sustain a strike on the head, body or tail.

There was also no preference shown for animal size, although it had previously been determined (Anderson and Burgin 2002) that there were substantially greater numbers of sub-adults in the edge compared to core habitats of these remnants. In the absence of evidence for selection by the avian predators, it is assumed that the sub-adults disperse outwards from the core and thus the edge acts as a ‘sink’ similar to that described by Delibes et al. (2001). In this case, skinks may be attracted to edges because of the relatively low density of skinks present and, as a result, are exposed to relatively greater predator pressure than in the core remnant from which they migrated.

Conclusions

We have demonstrated that in the remnants studied, at least in edge areas, the resident small skinks were under substantial predation pressure from birds commonly found to cross-forage between bushland remnants and the urban/agricultural matrix that surrounds them. These avian predators are widespread and abundant in the peri-urban/urban matrix (e.g., see Catterall 2004), and small skinks are also common throughout most human modified landscapes of Australia (see Cogger, 2000 for examples of distributions).

It is generally assumed that the two Lampropholis species studied are widespread and abundant, even in urban areas (Burgin 1993; Cogger 2000) and, therefore, would not attract the concern of more vulnerable species. However while L. guichenoti is apparently genetically similar across its range (Burgin 1989), there is evidence that L. delicata, as described, encompasses a genetically diverse group (Burgin 1989; Donnellan 1985; Hutchinson et al. 1990) with the species boundaries only delineated for a few taxa (see Mather 1990; Mather and Hughes 1992). There are also many other small skinks that have a restricted distribution in areas of increasing human population, particularly along the coastal fringe of Queensland and New South Wales (see Cogger 2000). It can therefore be assumed that predation pressure is occurring wherever there are small skinks on the edge of habitat remnants and avian predators that can cross-forage at the interface of remnants and its surrounding matrix, particularly when such areas are in close proximity to urban areas. Under such circumstances, the remnant edges act as a sink for these small skinks, and presumably other species that these generalist avian predators consume.