Conserving dryland lizards by reducing predator-mediated apparent competition and direct competition with introduced rabbits


Grant Norbury, Landcare Research, PO Box 282, Alexandra, New Zealand (fax + 64 3448 9939; e-mail


  • 1Native skinks Oligosoma spp. in New Zealand’s dry grasslands have declined in range and abundance since the arrival of humans. I hypothesized that introduced rabbits Oryctolagus cuniculus exacerbate this decline by supporting introduced mammalian predators for which they are primary prey; by sudden declines in abundance that cause predators to switch to skinks; by grazing vegetation, thereby reducing skink refuges from predators; and by reducing skink food and shelter from climatic extremes. The first three effects cause enhanced skink predation and represent indirect or apparent competition. The fourth effect represents direct competition.
  • 2Interactions between two introduced predators (ferrets Mustela furo and cats Felis catus), rabbits and two secondary prey (McCann’s skink O. maccanni and common skink O. nigriplantare polychroma) were studied. By measuring skink consumption, and predator, rabbit and skink numbers, offtake of skinks by predators under varying rabbit and skink densities was calculated.
  • 3Predation by ferrets and cats was inversely density-dependent because predation accelerated at low skink densities. As skink densities rose, predation became an increasingly less important source of mortality. Therefore, predation could potentially exterminate skink populations if densities fell below some critical range.
  • 4Predation of skinks increased markedly after sudden declines in rabbit abundance, because predators remained abundant but switched to feeding on skinks. Although a temporary effect, repeated cycles of intense rabbit control and population recovery may have chronic detrimental effects on skink population viability.
  • 5Optimal rabbit management for maintaining viable skink populations is likely to require avoidance of large swings in rabbit abundance by maintaining populations at low, stable levels. Fewer rabbits mean fewer predators and greater refuge (less apparent competition), and improved food and shelter (less direct competition). If large swings in rabbit abundance cannot be avoided, the effects of prey-switching could be reduced by controlling predators when rabbit numbers decline.
  • 6Better understanding and management of primary–secondary prey systems, where the primary prey consume the habitat of secondary prey, will require recognition of the extra complexities that underlie these systems.


There has been a recent upsurge of interest in the interactions between primary and secondary prey of predators (Pech, Sinclair & Newsome 1995; Smith & Quin 1996; Sinclair et al. 1998; Courchamp, Langlais & Sugihara 1999b, 2000). Depending on their nutritional importance, primary prey can be the main driver of predator abundance and their consumption of secondary prey (the numerical and functional responses, respectively). In a management context, these interactions become particularly important if the secondary prey are of high conservation value, and the primary prey or predators are pests. This is the case for a number of indigenous secondary prey species in Australia and New Zealand, where introduced rabbits Oryctolagus cuniculus (L.) are the primary prey of two or three introduced mammalian predators. Rabbit populations fluctuate widely in parts of these countries, creating similar variation in predator population dynamics. Rabbits can reach plague densities and have devastating impacts on the vegetative habitat for some secondary prey species. These dynamic characteristics of rabbit populations add further complexity to primary–secondary prey interactions (see Pech & Hood 1998). These interactions were explored in a predator–prey system in the dry grasslands of New Zealand.

New Zealand’s ecological communities have changed profoundly since Polynesian rats Rattus exulans (Peale) arrived at least 1700 years ago (Holdaway 1999). These changes accelerated during Polynesian settlement around 700 years ago, and again during European settlement around 200 years ago. Habitat for wildlife has since been altered, and many plants and animals have been introduced from around the globe. These include mammalian predators such as cats Felis catus (L.), ferrets Mustela furo (L.), stoats M. erminea (L.), weasels M. nivalis (Erxleben) and rats R. rattus (L.) and R. norvegicus (Berkenhout), which have seriously affected many New Zealand native species (Clout 1999).

Reducing mammalian predator abundance, by trapping and poisoning, can provide significant protection for vulnerable wildlife (O’Donnell, Dilks & Elliott 1996) but sustainable, cost-effective, predator control remains elusive. Recovery programmes may benefit from indirect management, for example enhancement of habitat characteristics that provide refuge from predators; manipulation of the abundance of primary prey species (rabbits and rodents); or reintroduction of wildlife above hypothetical threshold densities, below which populations may fall to extinction (Sinclair et al. 1998). However, judicious use of these options requires a clearer understanding of how predators and their prey interact.

Mammalian predator–prey systems in New Zealand, particularly in grassland ecosystems, are characterized by well-defined functional and numerical relationships between introduced predators and their introduced primary prey (King 1983; Norbury & McGlinchy 1996). There are two well-known ‘predator–primary prey’ couplings: in southern beech (Nothofagus spp.) forests, the dominant predators are stoats, which eat mainly mice Mus musculus (L.) and birds (King & Moody 1982; Murphy & Dowding 1995); in dry grasslands, the dominant predators are cats and ferrets, which prey mainly on rabbits (Fitzgerald 1990; Lavers & Clapperton 1990). Some native prey are taken as by-catch, or as prey that is of secondary importance to the predators’ diet. The loss of native wildlife to predators may therefore be a function of the interaction between predators and their primary prey (Pech & Hood 1998; Courchamp, Langlais & Sugihara 1999b, 2000). Enhanced predation caused by unusually abundant predators that are supported by primary prey was referred to by Holt (1977, 1984) as predator-mediated apparent competition. It has been studied mainly from a theoretical perspective (Holt 1977, 1984; Holt & Lawton 1994; Abrams, Holt & Roth 1998) and has been proposed as an extinction process for some Australian rodents by Smith & Quin (1996), who call it hyperpredation.

New Zealand’s 28 species of skink have declined dramatically in range and number since humans arrived, and predators are believed to be partly responsible (Towns & Daugherty 1994; Patterson 2000). For example, the rare and endemic lizards, the grand skink Oligosoma grande (Gray) and the Otago skink O. otagense (McCann), have declined dramatically in range in the last 100 years (Whitaker & Loh 1995; Patterson 2000). The Otago skink now occupies less than 10% of its former range. These lizards are heavily preyed upon and their grassland habitat is often highly modified by grazing and agricultural development (Whitaker 1996). New Zealand’s Department of Conservation undertakes intensive research and management of the few remaining populations (Whitaker & Loh 1995). Even some populations of common grassland skinks, McCann’s skink O. maccanni (Hardy), common skink O. nigriplantare polychroma (Patterson & Daugherty) and cryptic skink O. inconspicuum (Patterson & Daugherty), although still widespread, appear to be seriously depleted (A. Whitaker & M. Tocher, personal communication). Because rabbit populations are controlled intensively across parts of New Zealand, rabbit control may have important knock-on effects on skinks and other native wildlife. For example, concerns about short-term prey-switching by predators after declines in rabbit populations when rabbit haemorrhagic disease was illegally introduced into New Zealand in 1997 led to a large-scale predator control programme by the Department of Conservation (Keedwell & Brown 2001). The rabbit–predator–skink system is particularly interesting because rabbits, together with other herbivorous pests and livestock, can deplete grassland habitat (Norbury & Norbury 1996). This habitat provides food, shelter (White 1991) and perhaps refuge from predators for many secondary prey. Decline in refuge is one reason postulated for the decline in New Zealand skinks (Patterson 2000).

I aimed to test the hypothesis that secondary prey are vulnerable to direct and indirect effects of primary prey. I examined the interactions between two predator species (cats and ferrets), their primary prey (rabbits), and two sympatric, and closely related, secondary prey species (McCann’s and common skinks). I combined numerical and functional responses of predators to rabbit abundance, determined the relationships between density and predation of skinks, examined the effects of rabbit control on skink predation, and recommend management strategies that offset the adverse direct and indirect effects of rabbits on skink population viability.

Materials and methods


Rabbits, ferrets and cats were counted at night on two pastoral sites (1000 ha of Earnscleugh station and 2500 ha of Bendigo station) in the Central Otago region of the South Island, New Zealand, every 3–4 months from March 1994 to February 2001. The sites were 20 km apart. Animals were also counted on a third site (6000 ha of Grays Hills station) in the Mackenzie Basin region of the South Island every 3–4 months from April 1994 to July 1996. Animals were counted by spotlight from the back of a slow-moving vehicle (10–15 km h−1) along 13–19-km transects. Counts were made by the same observer on two to three consecutive nights and began about 1 h after sunset. The terrain allowed counting within a strip approximately 100 m wide. Animal abundance was expressed as the number seen per kilometre. The rabbit populations were subject to two major perturbations during this study: the Bendigo and Grays Hills populations were poisoned with 1080 (sodium monofluoroacetate) applied to aerially sown carrot baits in September 1994, and all three populations were infected with rabbit haemorrhagic disease from September 1997 onwards.

In deriving the predator numerical response to rabbits, data collected from August to October were excluded because ferrets, in particular, are less trappable during that time (Clapperton 2001) and are more difficult to see at night. Data collected in the year immediately after the rabbit populations were perturbed by poisoning or by disease were also omitted because of the delayed decline in predator abundance following sudden declines in rabbit abundance (Norbury & McGlinchy 1996). Spotlight counts were considered too imprecise to allow comparisons between single estimates of predator and rabbit abundance, so for the remaining data average counts were taken during the remainder of each year.


The functional responses of predators to rabbits were derived by measuring changes in predator diet along the transects on which rabbits were counted. The daily consumption rate of skinks by predators was estimated by counting the number of skinks in scats of predators that had been held in cage traps for up to 13–24 h. Predators were trapped every 4–6 weeks for 25 months on each site from March 1994 to May 1996. Sixty to 70 cage traps (360 × 330 × 630 mm) were set 300–400 m apart at permanent trap sites along the spotlight transects for four consecutive nights. Traps were baited with about 30 g of skinned rabbit meat, and checked daily. All animals captured were ear tagged and released. Predator scats were collected from the traps only on the first night an individual was caught during a given trapping session. Scats collected from the same individual on subsequent nights were more likely to contain the remains of bait, and so were discarded. Because predators were held in cages for up to 13–24 h before being released, the total amount of material defecated during that period was regarded here as a single scat. A scat therefore reflected differences in absolute food intake assuming the average confinement period was constant. Fewer cats than ferrets were caught and they tended to defecate in traps less often than ferrets. Therefore, the number of cat scats collected from traps was supplemented by collecting fresh cat scats found at random throughout each site.

Scats were soaked in water over night, washed into a 250-µm sieve, and sorted macroscopically into rabbit, skink, common gecko Hoplodactylus maculatus (Gray), bird, invertebrate, hedgehog Erinaceus europaeus (Barrett-Hamilton) and mouse. Nine-hundred and ninety-nine ferret scats were collected, of which 794 (79%) contained identifiable prey. Of the 159 cat scats collected, 155 (97%) contained identifiable prey. It was possible to count accurately the number of individuals of skinks only. The minimum number of skinks per scat collection was determined by counting the number of left and right front and hind feet, and dividing it by four. It was impossible to differentiate reliably between skink species.

Because rabbit carcasses persisted as a potential food source for several weeks after the poisonings in September 1994, the beginning of November 1994 was chosen as the end of the pre-poison period and the start of the post-poison period. The pre-poison period was therefore 9 months. The post-poison period was divided into two parts: the 8-month period immediately after the poisons, from the beginning of November 1994 to the end of June 1995; and the 11-month period following this from July 1995 to May 1996. This allowed us to monitor diet changes through two summer–winter periods following the poisonings.

Asymptotic and exponential decay functions were fitted to the numerical and functional responses using non-linear regression models. A least-squares loss function was used to calculate maximum likelihood estimates of the regression coefficients. Models were estimated using exact derivatives derived from the Gauss–Newton method (systat 6.0 for Windows: Statistics 1996).


Daily offtake of skinks by predators for a given rabbit density was calculated by multiplying the numerical and functional responses. Offtake was calculated for the 10-month period outside the skink breeding season (i.e. March to December), when skink numbers were most stable. Offtake per hectare was expressed by assuming that the number of predators counted at night reflected their absolute abundance within a 100-m wide strip on the count transect (see the Discussion for justification). This strip width was based on the fact that an observer’s ability to detect rabbits in this habitat declines dramatically beyond 40–60 m either side of spotlight transects (Fletcher et al. 1995). Offtake was partitioned by predator species by assuming that 70% of the total predator counts were ferrets and 30% were cats. This was based on the relative proportion of marked ferrets and cats known to be alive on the trapping transects. It was considered invalid to partition offtake among predator species according to the raw spotlight counts because, not only was it impossible to always identify a predator as a ferret or as a cat, but cats appeared to be easier to identify than ferrets because of their more brilliant eyeshine. The degree of partitioning between predator species, and the assumed strip width, did not alter the qualitative conclusions of the study. An example of how offtake by ferrets was calculated (e.g. at 60 rabbits spotlight km−1) follows. The numerical response (Fig. 1) shows that 60 rabbits km−1 supports 0·403 predators km−1 on average, or 0·0403 predators ha−1 assuming a strip width of 100 m. Assuming that 70% of the predators are ferrets means there are 0·0282 ferrets ha−1. The functional response (Fig. 2b) shows that an individual ferret consumes 0·104 skinks day−1 on average when there are 60 rabbits km−1, which is 31·2 skinks consumed ferret−1 over 10 months. Therefore, total offtake of skinks by ferrets at 60 rabbits km−1 = (0·0282 ferrets ha−1) × (31·2 skinks ferret−1 10 months−1) = 0·88 skinks eaten ha−1 10 months−1.

Figure 1.

Numerical response of predators (cats and ferrets combined) to rabbits, based on spotlight counts taken over 7 years at Earnscleugh and Bendigo pastoral stations, and over 28 months at Grays Hills station. Each point is a 9-monthly average from each study site. The line is the equation y = 0·54(x + 1·7)/(22·67 + x) fitted to the data (r2 = 0·97).

Figure 2.

Functional responses of predators to rabbits, based on the proportion of scat collections containing rabbit (open circles, cats; filled circles, ferrets) (a), and on the mean number of common skinks consumed per day by ferrets (b) and cats (c). Note the different scales of the y-axes in (b) and (c). The solid line in (b) is the equation y = 1·14 × e(−0·11x) + 0·10 (r2 = 0·94). Because the form of this equation could not be applied to cats, the upper dashed line in (c) is taken as the high predation scenario, and the lower dashed line is the low predation scenario based on the minimum value observed. Each data point is based on 52–147 ferret scats, and on 6–30 cat scats, collected from three sites during the 9-month period before rabbits were poisoned, during the 8-month period immediately after the poisoning, and during the 11-month period thereafter.

Sudden declines in rabbit abundance induce greater consumption of secondary prey species (prey-switching) before predator numbers eventually decline within 12 months (Norbury & McGlinchy 1996). This temporarily high offtake was calculated by multiplying the predator density at the original level of rabbit abundance by the per capita consumption of skinks at very low rabbit abundance (which was chosen to be 1 rabbit km−1). Reduction to 1 rabbit km−1 is considered a very successful rabbit control operation, so I have presented a worst-case scenario of rabbit control effects on skink consumption. As in the previous example, the numerical response (Fig. 1) was used for the original rabbit abundance, but this time it was multiplied by the functional response for the reduced level of rabbit abundance [i.e. a prey-switch to 1·12 skinks day−1 (Fig. 2b) or 336·2 skinks 10 months−1]. As an example, total offtake of skinks by ferrets after 60 rabbits km−1 have been reduced to 1 rabbit km−1 = (0·0282 ferrets ha−1) × (336·2 skinks ferret−1 10 months−1) = 9·48 skinks eaten ha−1 10 months−1. This higher offtake eventually declines as predator numbers decline.


Total offtake of skinks by predators has little bearing on the viability of skink populations unless offtake can be expressed as a proportion of skink density. For example, very high offtake may not be a problem for skink populations if it occurs when densities are sufficiently high to absorb predation. There were insufficient measures of skink abundance in the field to compare with estimated skink offtake, so the indices of rabbit abundance were converted to skink abundance according to assumed inverse relationships between rabbit and skink abundance: the relationships are inverse partly because rabbit grazing removes vegetative habitat for skinks. Patterson (1985) measured very high skink densities in a dry grassland where tall tussocks were intact because of light grazing by livestock, but found fewer skinks where tussocks were replaced by short, exotic, pastures. The same effect was apparent where burning reduced tussocks (Patterson 1984). Whether this relationship held on the present study sites was tested by measuring vegetation cover and skink abundance at 10 randomly chosen locations on each of the three sites. At each location in late March 1996, skinks were captured over 14 continuous days in 10 pitfall traps (plastic cups 11 cm wide and 10 cm deep) spaced 10 m apart in a straight line. The traps were partly filled with the preservative ethylene glycol. Vegetation cover in the vicinity of each transect was measured using the wheel point apparatus of Tidmarsh & Havenga (1955). This method involves rolling a wheel, with long protruding spikes or points, along the ground and recording what each point strikes. Gibson & Bosch (1996) calculated that 200 points account adequately for within-site variation in this vegetation type. Percentage cover of vegetation was calculated as the number of strikes on vegetation at ground level, divided by the total number of strikes, which also included leaf litter, rock and bare ground.

Observations by some New Zealand herpetologists (e.g. M. Tocher, A. Whitaker & L. McFarlane, personal communication) suggest it is not unreasonable to assume an inverse relationship between rabbit and grassland skink abundance, at least for O. n. polychroma, but it may not be strictly correct for O. maccanni because their densities may peak at intermediate levels of grazing rather than at very low grazing levels as observed for O. n. polychroma (M. Tocher & A. Whitaker, personal communication). However, this does not alter the main conclusions of the study. The shape of the inverse relationship between rabbits and skinks is unclear, and is likely to vary according to vegetation type and other sources of cover for skinks. Therefore linear, sigmoid, convex up and convex down relationships were tested between rabbit and skink abundance. Maximum skink densities of 500 ha−1 on average were based on the very high densities measured by Patterson (1985) in tall tussock grassland, and by myself in modified, but highly vegetated, short tussock grassland (unpublished data). To estimate predation rates of skinks, total skink offtake at a given rabbit density was expressed as a proportion of inferred skink density.



The number of predators counted showed a significantly asymptotic numerical response to the number of rabbits (Fig. 1), best described by the equation:

predators = (a × (rabbits + 1·7))/(b + rabbits) (eqn 1)

where a = 0·54 [95% confidence interval (CI), 0·40–0·68] and b = 22·67 (95% CI, 5·68–39·67). r2 = 0·97.

Although ferrets and cats appear to be numerically driven by rabbits, some individuals can survive on other prey species when rabbits are absent (Fitzgerald & Veitch 1985). The constant 1·7 was therefore added so that when rabbit abundance was zero, predator abundance equalled the minimum value observed. If no constant was added, the equation predicted zero predators where there were no rabbits. A statistically significant estimate of this constant could not be obtained. The data point in the far right of Fig. 1 had little influence on the regression coefficients. If omitted, a = 0·55 (although the 95% CI increased to 0·36–0·74), b = 23·29 (95% CI, 1·75–44·83) and r2 = 0·96.


The proportion of ferret and cat scats containing rabbit declined when rabbit abundance was below about 10 spotlight km−1 (Fig. 2a). For ferrets, this was mirrored by a reciprocal increase in the mean daily consumption of skinks (Fig. 2b). For example, 19 skinks were found in one ferret scat after rabbit abundance was reduced by poisoning. The functional response of ferrets is best described by the equation:

no. skinks per scat = a × e(−b×rabbits) + 0·10 (eqn 2)

where a = 1·14(95% CI, 0·72–1·56) and b = 0·11(95% CI, 0·02–0·20). r2 = 0·94.

The constant 0·10 was added so that when rabbit abundance was very high, skink consumption equalled the minimum value observed. If no constant was added, the equation predicted zero skink consumption at high rabbit densities, which was not what was observed. A statistically significant estimate of this constant could not be obtained.

The overall daily per capita consumption of skinks by cats was greater and more variable than by ferrets (Fig. 2c). The relationship with rabbit abundance was similar to that for ferrets in that cats appeared to consume more skinks at low rabbit density. Indeed, 35 skinks were found in one cat scat after rabbit abundance was reduced by poisoning. However, a statistically significant description of these data at the 5% significance level was marginal at best (regardless of the value of the apparent outlier point at high rabbit density), and so it was concluded that skink consumption by cats did not show a clear response to rabbit density. Therefore a low predation or best-case scenario for cats was derived by holding skink consumption constant at the minimum rate observed (i.e. 0·32), and a high predation or worst-case scenario by setting skink consumption to the value obtained from a straight line joining the maximum values observed at high and low rabbit densities (Fig. 2c).


Skink offtake (per hectare) by ferrets over 10 months peaked when there were 6 rabbits spotlight km−1 and was maintained at about half this rate from about 30 rabbits km−1 onwards, and actually increased slightly (the solid line in Fig. 3a). Note, offtake did not decline to zero as rabbit densities increased. The dotted line depicts the maximum temporary increase in predation caused by prey-switching during the period of the delayed numerical response of predators to sudden declines in rabbit abundance (e.g. by poisoning or by an acute disease epidemic).

Figure 3.

Estimated offtake of skinks per hectare by ferrets (a) and cats (b) vs. rabbit abundance. Note the different scales of the y-axes. Offtake was derived from the product of the numerical and functional responses, and is expressed over the 10-month period outside the skink breeding season. The solid lines are ‘normal’ offtake (for cats, top solid line is high-predation scenario, bottom line is low-predation scenario) and the dotted lines are the maximum temporarily high offtake caused by prey-switching during the delayed numerical response of predators to sudden declines in rabbit abundance from their original levels.

Skink offtake by cats was generally greater than that of ferrets (Fig. 3b). In the low predation scenario, skink offtake gradually increased to an asymptotic level similar to the peak level observed for ferrets. The high predation scenario showed a more dramatic increase in offtake, that levelled off (but then slowly declined) at more than 10 times the offtake rates predicted in the low predation scenario. Again, note that offtake declined little as rabbit densities increased.


Skinks were never caught at vegetation cover less than 55%, but they occurred increasingly at greater cover (Fig. 4). Vegetation cover on these sites was largely a function of the intensity of grazing by livestock and rabbits. This supports the assumed inverse relationships between rabbit and skink abundance, partly because rabbits remove vegetative habitat of skinks. For clarity, the scatter of skink densities was omitted from the analysis where rabbit abundance was low and vegetation cover was high. Including the scatter reduced the predation rates only slightly and did not alter the conclusions. Regardless of which rabbit–skink relationship was used, predation rates were inversely correlated with skink density (= inversely density-dependent or depensatory; the solid lines in Fig. 5). Predation by ferrets was low over most skink densities and increased sharply only when densities dropped below about 20 ha−1. Predation increased markedly after sudden declines in rabbit abundance (the dotted lines in Fig. 5), particularly at low skink densities. Predation by cats in the low predation scenario was similar to that of ferrets, but was markedly higher in the high predation scenario. Predation by cats will vary, probably unpredictably, between these low and high scenarios. Again, cat predation increased after sudden declines in rabbit abundance.

Figure 4.

Numbers of skinks caught at 30 locations in Earnscleugh, Bendigo and Grays Hills stations over 14 days of trapping, vs. the cover of vegetation at each location.

Figure 5.

Relationship between predation rate of ferrets (a)–(d) and cats (e)–(h) vs. estimated skink density. Density estimates depend on the relationship between rabbit (R) and skink (S) abundance, depicted hypothetically in the inset graphs. The dotted lines represent maximum predation rates caused by prey-switching during the delayed numerical response of predators to maximum rabbit control, which eventually decline to the solid lines more than 12 months later. The solid lines represent a steady state of rabbit, vegetation and skink abundance (for cats, top solid line is high-predation scenario, bottom line is low-predation scenario).


I propose that introduced rabbits have partly contributed to the decline in skink populations in New Zealand’s dry grasslands in four ways: enhanced predator abundance; prey-switching and delayed numerical responses of predators; depletion of refuge; and depletion of food and shelter. The first three processes are forms of indirect or predator-mediated apparent competition.


Skink predation is inversely density-dependent because predators are more abundant when rabbits are more abundant, and thus skinks are less abundant. Also, predators continue to consume skinks, albeit at a reduced rate, even when skinks are scarce. This results in increasing offtake of skinks as densities decline. Inverse density-dependent predation is a characteristic of many secondary prey species (Pech, Sinclair & Newsome 1995; Sinclair et al. 1998). This presents a potentially dangerous situation for the viability of secondary prey populations because it can lead to local extinctions if prey densities fall too low. Although still widespread, some grassland skink populations are seriously depleted. New Zealand lizards are characterized by low reproductive output, which leads to concerns about their ability to persist when threatened by predation (Cree 1994). McCann’s and common skinks, for example, produce only a single litter of 3·0–3·8 young female−1 year−1 in my study area (Cree 1994). Unfortunately, we do not know how recruitment and population growth in these species vary with density, especially in the absence of predators. Superimposing this information on the density-dependent predation rates in Fig. 5 would indicate skink densities that are likely to be stable or unstable, and a range of densities that might be vulnerable to extinction.

Courchamp, Clutton-Brock & Grenfell (1999) invoked the Allee effect to explain inverse density-dependent predation. They suggest that anti-predator behaviour of prey, such as anti-predator signalling, becomes less effective at low population density. To my knowledge, the skink species I studied have no such anti-predator behaviour.

Predation rates of skinks by ferrets drop to low levels beyond about 20 skinks ha−1, and become an increasingly less important source of skink mortality. After declines in rabbit abundance, predation by ferrets and cats drops beyond about 100–200 skinks ha−1. This implies that skink populations are more likely to tolerate predation when they are sufficiently abundant. Banks (1999) found the same for high-density bush rat Rattus fuscipes populations that were preyed on by foxes Vulpes vulpes in Australia. Indeed, some local skink populations are extremely abundant (G. Norbury, personal observations) and appear to be secure. However, many other populations are declining.


Rabbit control may be one reason some skink populations are declining. Prey-switching during the period when predators are still abundant but have switched to a secondary prey species following rabbit control elevates predation rates markedly, although temporarily, particularly at lower skink densities. High rabbit numbers have been subjected to large-scale poisoning by farmers and pest control agencies every 3–5 years in New Zealand. This causes repeated pulses of predation of native wildlife. Repeated cycles of rabbit control and population recovery may have chronic detrimental effects on skink population viability.

I have presented a worst-case scenario of rabbit control effects by assuming very large reductions in rabbit abundance down to 1 rabbit km−1, and therefore maximum intake rates of skinks by predators. Depending on the quality of the rabbit control operation, or on the intensity of disease outbreaks such as rabbit haemorrhagic disease, very large population reductions sometimes occur, but not always. The time taken for predation rates to return to normal levels (the solid lines in Fig. 5) will be longer than the 12 months required for the predator population to decline, because of the longer delay in vegetation recovery and subsequent recovery of skink populations. These delays will depend on seasonal conditions and stocking regimes of domestic herbivores. Also, invasive species of exotic shrubs [e.g. sweet briar Rosa rubiginosa (L.) and broom Cytisus scoparius (L.)] can increase following reductions in rabbit grazing (G. Sullivan, personal communication) and might limit the recovery of skink habitat.


As reported by Patterson (1984, 1985), I found fewer skinks in areas where vegetation was depleted by grazing. This may represent direct competition between rabbits and skinks for vegetation that provides food and shelter. However, it is likely to be confounded by apparent competition because fewer skinks in depleted vegetation may also be a function of abundant rabbits supporting abundant predators, and loss of refugia probably exposes skinks to more predation. The importance of vegetation as refuge for skinks, or as a source of food and shelter, is best determined experimentally by measuring skink population dynamics at different vegetation densities, in the presence and absence of predators. On Round Island, Mauritius, where there are no lizard predators, some lizard species showed positive responses to competitive release following the eradication of rabbits and the subsequent recovery of vegetation (North, Bullock & Dullo 1994).


Reducing predator abundance

In theory, reducing predator abundance can shift a population above an extinction density threshold and into positive population growth (Sinclair et al. 1998). However, predation rates of skinks by ferrets are already low across most skink densities and increase steeply only when skink densities are very low. Ferret control may therefore reduce predation of grassland skinks only marginally. However, this is not necessarily the case for cats, nor for periods following sudden reductions in rabbit abundance for both predator species, when predation rates are much higher. Here, there is much scope for reducing predator impacts, especially if predator control was immediately after rabbit control. The Department of Conservation initiated a predator-control programme concurreint with rabbit haemorrhagic disease reduced rabbit populations in the Mackenzie Basin (Keedwell & Brown 2001). This significantly reduced predation of nests of banded dotterels Charadrius bicinctus (Jardine and Selby) (Norbury 2000), another secondary prey species. A generic model, developed by Courchamp, Langlais & Sugihara (1999b), concluded that concurrent control of both rabbits and predators is the best strategy for maximizing the recovery of native species.

Maintaining low, stable rabbit populations

Rabbit populations can be described in three states: high, low or rapidly fluctuating density resulting from rabbit control. My results indicate that the worst state for skinks is likely to be fluctuating rabbit density because it elevates predation and exposes a greater range of skink densities to extinction. Large fluctuations in rabbit abundance should therefore be avoided. In the longer term, high rabbit abundance is worse for skinks because abundant rabbits support abundant predators while also depleting skink habitat. Courchamp, Langlais & Sugihara’s (2000) model of a generically similar system reached the same conclusion. The optimum rabbit management regime for maintaining viable populations of grassland skinks is therefore to reduce rabbit numbers to low, stable levels.

Rabbit haemorrhagic disease arrived in New Zealand in 1997 and has so far suppressed rabbit numbers in some areas but not in others (Parkes et al. 2002). During the initial epidemics, significant increases in predation were observed for some native birds [banded dotterels (Norbury, Heyward & Parkes 2002) and pukeko Porphyrio porphyrio (L.) (Haselmayer & Jamieson 2001)] which, like skinks, occur sympatrically with rabbits and are secondary prey. It is too early to know how the disease will affect rabbit populations in the long term, but it clearly has important implications for the viability of skink populations and other secondary prey. Rabbit haemorrhagic disease will assist some skink species if it suppresses rabbit populations in the long term [as predicted by Pech & Hood’s (1998) model for secondary prey in Australia], but it will be detrimental if it mimics poisoning events by creating short-term cycles of rabbit population decline and recovery.

One of the unquantified risks to secondary prey of fewer rabbits, and therefore fewer cats and ferrets, is the possible increase or mesopredator release of rats (Courchamp, Langlais & Sugihara 1999a). Rats are less reliant on rabbits for food, and are serious predators of skinks and many other indigenous prey (Innes 1990; Towns & Daugherty 1994). An increase in rats might offset the benefits of fewer rabbits. However, we have no reliable evidence of these effects and, until we do, we should monitor the response of indigenous prey and other predator species where rabbit populations are suppressed.

Improving habitat quality for secondary prey

In theory, the greatest gains for secondary prey can be made by changing the shape of the predation curve from one of increasing predation with declining prey, to one of decreasing predation with declining prey (Pech, Sinclair & Newsome 1995). This might be achieved by increasing refugia for prey, which for grassland species like skinks means increasing the vegetation biomass, changing species composition of vegetation, or increasing other forms of cover, like rocks. Improving habitat quality may also increase recruitment rates that are mediated through improved food supply and shelter. The role of prey refuge in predator–prey interactions has been studied mainly from a theoretical perspective, although there is some empirical evidence (McNair 1986; Sih 1987; Hawkins, Thomas & Hochberg 1993). Few experiments have tested the conditions under which prey refugia may be able to prevent extinctions. Improving prey refugia is a management imperative in some cases. To give a local example, the increased exposure of skinks to predators, particularly to cats, as they disperse through grass swards that have been depleted by grazing is thought to be a significant threat to the viability of New Zealand’s grand and Otago skink populations (Whitaker 1996). The utility of managing herbivory to mitigate predation is worthy of further investigation.

Reintroductions of secondary prey

Another way to avoid extinctions is to reintroduce prey to boost numbers above putative extinction densities and allow unassisted increases to higher stable states. In the case of grassland skinks, further work is required to quantify the range of these critical densities that needs to be exceeded. The problem with this approach is that, unless primary prey and prey refugia are managed with a view to minimizing predation risks, reintroduction programmes may be more prone to failure than success. Another risk with reintroductions is the possibility of introduced animals carrying pathogens that could have serious effects on resident endangered populations.


Estimated predation rates of skinks assume that the number of predators seen along the count transects reflects their absolute abundance within a 100-m-wide strip. Support for this assumption is that the derived predator densities are similar to those estimated from the same sites using mark–recapture procedures (G. Norbury, unpublished data). Mark–recapture estimates were obtained only during the first third of the study, so there were insufficient estimates to construct the numerical response. The predation rates I calculated for both predator species assume that ferrets and cats comprise 70% and 30%, respectively, of all the predators counted at night. Again, this is based on their relative abundance using mark–recapture procedures. Neither assumption affects the shape of the predation curves nor the qualitative conclusions of the study.

Boutin (1995) criticized functional responses derived from the number of prey items in scats because no account is taken of the variable rate at which predators deposit scats. Rates of deposition can vary with changes in food intake, diet composition, and so on. Because a scat in this study was a collection of all the scat material deposited by a predator for up to 13–24 h, it is a reasonably good measure of absolute intake. However, it assumes the mean period the trapped animals were confined was constant. I have also assumed that a scat collection reflected the food consumed over the previous 24 h. I cannot validate this assumption without further experimentation, but it is unlikely to alter the qualitative conclusions.

While ferrets and cats are the main predators of skinks in New Zealand’s dry grasslands, they are not the only predators. For example, hedgehogs, stoats, rats and Australasian harriers Circus approximans (Peale) also prey on skinks (Pierce & Maloney 1989; Patterson 2000). Thus, absolute skink offtake is greater than the levels indicated here.

Estimated predation rates also assume some kind of inverse relationship between skink and rabbit abundance. I have argued that this is not an unreasonable assumption based on published accounts, anecdotal observations by herpetologists, and on some empirical data presented in Fig. 4. Predation rates are insensitive to the shape of the inverse relationship or to the scatter of points in Fig. 4.

One of the problems with predicting population viability based on predation rates is that it assumes random predation among sexes and ages. This is unlikely always to be the case, and will have different outcomes for population viability, especially if breeding females are more vulnerable to predators. Clearly, the utility of this study for conserving dryland skinks would be improved by experimental validation. In the meantime, there are few robust data that couple mortality rates with skink densities. A preliminary estimate of mortality of the same skink species studied here (0·17 annum−1; M. Tocher, unpublished data) at a density of about 158 ha−1 at the nearby Macraes Flat is within the range of total predation rates predicted at the same skink density (Fig. 5).


This study supports the proposition that the loss of some native wildlife to predators is a function of the interactions between predators and their primary prey, and the extent to which primary prey directly affect the habitat of secondary prey. Programmes aimed at conserving secondary prey currently focus on reducing predator density. These programmes may therefore benefit from management of each predator’s primary prey and improvement in the quality of habitat refugia for secondary prey. Ideally, these alternative management options should be tested experimentally in conjunction with population modelling. Sinclair et al. (1998) outline a protocol for experimental programmes aimed at protecting indigenous fauna from predation.

Sinclair et al. (1998) applied density-dependent theory to field data gathered from seven Australian native mammal species and found evidence of threshold prey densities. Although the concept of threshold densities appeals to managers, it remains largely untested and is founded on the density-dependent paradigm of population regulation, whose utility for understanding population dynamics has been questioned (Krebs 1995). A focus on density-dependent processes does not provide a comprehensive analysis of how predators and their prey interact in New Zealand grassland systems. Given the importance of climate, livestock grazing and other density-independent processes, predation and recruitment rates are unlikely to be constant for a given population density. This may drive populations to extinction for no apparent reason other than some stochastic external effect. Climate is especially important for skinks because they are cold-blooded. The utility of this study for protecting native skinks is limited to average outcomes from a wide array of stochastic and often unpredictable processes. This means that the predictions will not always be correct, but on average they should be. There may be greater utility in developing more dynamically interactive models (Pech & Hood 1998; Choquenot & Parkes 2001) that encompass stochastic environmental fluctuations that affect predation processes. Nonetheless, a wider view of predation, and how to mitigate both its direct and indirect effects, may be required. This is particularly important in New Zealand because predation is a key process that threatens not only critically endangered species, but also species that are still reasonably common but are declining.


This work was funded by MAF and the Foundation for Research, Science and Technology (Contract CO9X0009). Special thanks to Richard Heyward for analysing predator scats and assisting with spotlight counts. Thanks also to Roger Gibson for assisting with vegetation sampling. Dale Norbury, Richard Heyward, Phil Cowan, Ben Reddiex, Mandy Tocher, Chris Jones, Vilis Nams, Graham Nugent, Christine Bezar, Oliver Sutherland, Franck Courchamp and an anonymous referee made helpful comments on the manuscript. The work was approved by the Animal Ethics Committee of Landcare Research, and the skinks were collected under a permit issued by the Department of Conservation.