Introduction

Biological invaders pose a significant threat to native ecosystems and biotas on a global scale (Soulé 1990; Vitousek et al. 1996). This is particularly evident for island biotas that comprise a reduced suite of sympatric species, as island size and resources limit species diversity (Case and Bolger 1991). New Zealand’s terrestrial fauna evolved in the absence of predatory mammals (Worthy et al. 2006). The introduction of Pacific rats (kiore; Rattus exulans) to New Zealand approximately 1250–1300AD (Anderson 1991; Wilmshurst and Higham 2004) is widely believed to have initiated large-scale biogeographic changes to native fauna, exacerbated by subsequent mammalian introductions (Cassels 1984; Towns et al. 2006; Worthy and Holdaway 2002). Predation and competition by a range of introduced mammals continues to pose a significant threat to indigenous vertebrates in New Zealand (Hoare et al. 2007a; Innes et al. 2010; Towns and Daugherty 1994).

Eradication of introduced mammals from New Zealand’s offshore islands is an important conservation tool for the protection of native fauna (Baker et al. 2020; Towns and Broome 2003), and has provided secure refugia for populations of a range of species that are unable to coexist with mammals (Worthy 1987). Lizard populations released from predation pressure through eradication of rodents have shown dramatic positive responses, increasing up to 30-fold within ten years of eradication (Towns 1991, 2002; Newman 1994; Towns et al. 2003; Monks et al. 2014).

However, protection of native fauna on the main islands of New Zealand is much more complicated for three main reasons: (1) there is a much larger suite of mammalian species on the main islands (in particular, four rodent species, three mustelid species, possums (Trichosurus vulpecula), hedgehogs (Erinaceus europaeus) and feral cats (Felis catus); King and Forsyth 2021), (2) there is the added complexity of periodic masting tree and tussock species driving dynamics of many communities and leading to rodent irruptions, spikes in mustelid populations and occasional prey switching to native species (Dilks et al. 2003; Monks and O’Donnell 2017; O'Donnell and Phillipson 1996; Pryde et al. 2005b), and (3) the competitive and predatory effects among species of introduced mammals come in to play (Bridgman 2012; Bridgman et al. 2013).

A seminal review of the factors driving population declines in forest birds (Innes et al. 2010) found that predation by introduced mammals alone was sufficient to explain patterns of decline among forest birds. Further, stoats (Mustela erminea), rats (Rattus spp) and possums were implicated as the key predators responsible for forest bird declines (Innes et al. 2010). Conservation research in Aotearoa has traditionally been biased towards forest birds and, as such, the work drives national thinking about conservation. Consequently, the two major landscape scale initiatives directed at conservation of our terrestrial fauna, ‘Predator Free 2050’ and the national predator control programme (formerly Tiakina Ngā Manu/Battle for our Birds), focus on eradication or suppression (respectively) of possums, stoats and rats (Elliott and Kemp 2016; Russell et al. 2015).

Aotearoa is home to 124 endemic lizard species, 52 geckos in seven genera and 72 skink species in the genus Oligosoma (Hitchmough et al. 2021). Lizards are preyed upon by the full suite of predatory mammals present in New Zealand (reviewed by Hoare et al. 2007a, b). This suite of predators includes mice (Mus musculus) (Towns 1992; Newman 1994), hedgehogs (Jones et al. 2005), weasels (Mustela nivalis) (Miskelly 1997) and feral cats (Jones et al. 2005). Relatively scant data on lizards (in comparison with birds) suggests that focussing predator control efforts on a small set of meso-predators and neglecting other invasive mammalian predators (such as mice, weasels, hedgehogs and feral cats) is insufficient to protect vulnerable lizard populations (Hoare et al. 2007a, b, Tocher 2009, Nelson et al. 2016). In contrast, the only study to comprehensively control the full range of invasive mammals produced recovery of large-bodied grand and Otago skinks (Oligosoma grande and O. otagense, respectively) in eastern Otago (Reardon et al. 2012). Collectively, these studies point to meso-predator release, the process by which declines in higher order predators release lower order predators from predation and competition (Courchamp et al. 1999, 2000, 2003), potentially rendering current national conservation strategies ineffective for some native taxa. Indeed, White et al. (2006) recognise that “inappropriate management of invasive species may lead to adverse changes, such as potential extinctions or the expansion of other invasives” citing a range of evidence (Courchamp et al. 1999, 2000, 2003, Crooks and Soulé 1999, Mack and Lonsdale 2002). Further, Linklater and Steer (2018) point out that the effect of eliminating five introduced mammals species on the other 26 introduced mammalian predator and herbivore species will be complex, with negative outcomes likely. As such, an increased understanding of the functional relationships between native and introduced species is essential for effective management of biodiversity (Baker et al. 2020; White and King 2006) as aspired to by the national biodiversity strategy in Aotearoa New Zealand, Te Mana o te Taiao (New Zealand Government 2020).

Mice pose a serious threat to native biodiversity that is often overlooked, in part due to the difficulty and expense of suppressing mouse populations (Norbury et al. 2023). In the absence of other mammalian predators, introduced mice on Gough Island had a devastating impact on the island’s burrowing petrel species and the Critically Endangered Tristan albatross (Diomedea dabbenena) (Cuthbert and Hilton 2004; Cuthbert et al. 2016). Similarly, after the removal of all other introduced mammals from Mana Island, mice had a devastating impact on McGregor’s skinks (Oligosoma macgregori), which only recovered following the eradication of mice (Newman 1994; Miskelly 2023). Mice are also predators of some of New Zealand’s native birds, including rock wrens (Xenicus gilviventris) (Michelsen-Heath 1989; Weston et al. 2018) and have a major impact on lizards (Norbury et al. 2014, 2023; Nelson et al. 2016) and terrestrial invertebrates (Watts et al. 2022; Norbury et al. 2023) in mainland sanctuaries where all species except mice are excluded. Small body size enables mice to access small crevices that afford lizards and large terrestrial invertebrates some protection from larger-bodied predators (Lennon et al. 2021).

Mouse population increases are strongly associated with high levels of seedfall, especially in beech forest (e.g. Ruscoe et al. 2005), but also in podocarp forest (Ruscoe et al. 2004) and tussock (Wilson and Lee 2010). However, irruptions in mouse populations can be dampened by ship rats, Rattus rattus, primarily through intraguild predation, despite dietary overlap and the potential for competitive effects (Bridgman 2012; Bridgman et al. 2013). Therefore, suppression of rats through large-scale predator control strategies has the potential to trigger much larger mouse irruptions, especially in ecosystems dominated by mast-seeding species.

We examined the relationships between introduced mammalian predators that are either the subject of a control programme (rats) or not (mice) and a native lizard population (southern grass skinks, Oligosoma aff. polychroma Clade 5) over an 11-y period in a temperate beech forest ecosystem. Specifically, we asked: (1) is mouse abundance correlated with the relative abundance of rats, which are known predators of mice? (2) is skink abundance correlated with relative abundance of rats and/or mice? and (3) what is the overall population trend in southern grass skinks over the 11-y period? We also created a conceptual model to visualise the effects of controlling a selected suite of predators (excluding mice) on both mouse and lizard populations informed by our findings.

Methods

Site

The study was conducted in the Eglinton Valley in Fiordland (South Island, Aotearoa New Zealand). Skink monitoring occurred in c. 200 ha of modified grassland habitat atop the fluvioglacial outwash fans of the East Eglinton River (168° 01′ E, 45° 03′ S) c. 350 m a.s.l. (Lettink et al. 2011). The original grassland habitat was extensively modified through a long history of sheep grazing until 1998 and is now dominated by exotic pasture grasses supporting a relatively high-density skink population, possibly due to the increased vegetation biomass of valley-floor grasslands providing increased refuge, food and shelter for skinks (Norbury 2001; Lettink et al. 2011). The grassland is surrounded by temperate southern beech (Fuschospora spp. and Lophozonia menziesii) forest to the treeline at 1000–1200 m a.s.l.. Mast seeding events were defined as years in which beech seed density exceeded 500 seeds/m2 based on publicly available data available at https://docnewzealand.shinyapps.io/seedrain_shiny/. Predator control in the valley commenced with stoat trapping along a single transect running the length of the valley floor in 1997 (Dilks et al. 2003), but has progressively expanded to include bait station grids targeting rats timed in accordance with mast seeding of beech forest to protect bats and birds vulnerable to rat predation (Monks and O’Donnell 2017; Pryde et al. 2005b) and more recently aerial 1080 operations when rodent irruptions exceed levels for which ground-based control operations are effective (Edmonds et al. 2017). Mean annual rainfall averages 2300 mm per year at Knobs Flat, 7 km to the north of the study area, and mean maximum daily temperatures range from 3.3 °C in July to 14.7 °C in February (O’Donnell 2002).

Skink monitoring

We established eight transect lines, each containing 25 Onduline artificial retreats spaced 10 m apart, in grassland habitat either side of the East Branch Eglinton River confluence with the main river based on research into optimal design of retreats (O’Donnell and Hoare 2012a). Start points for lines were randomly allocated and each line ran east to west across the grassland. Each Onduline retreat was 670 mm × 420 mm.

Monitoring occurred across the austral summer from early spring (late August or early September) to early autumn (March or April) from February 2009 to March 2020. We aimed to check retreats at ambient temperatures of 12–18 °C any time during the day except early evening (Hoare et al. 2009). Ambient temperature in the shade at 1.4 m above the ground was recorded at the start of each transect. A retreat check involved quickly lifting the sheet of Onduline, counting the number of skinks beneath the retreat and visually classifying them as neonates (young of the year; southern grass skinks are born in December in the Eglinton Valley) or older based on size. Mice seen beneath the covers were also counted.

Predator monitoring

Stoats, rats and mice were monitored by Department of Conservation staff quarterly (February, May, August and November) via footprint tracking tunnels in beech forest surrounding the grassland in which skinks were monitored using a standard protocol (Gillies and Williams 2013). Briefly, tracking tunnels were organised in transect lines of 10 tunnels at 50 m spacing and baited with peanut butter on day 0, checked for rodents on day 1 at which time every second tunnel (i.e. 5 tunnels per line at 100 m spacing) was rebaited with rabbit meat and then these tunnels were checked for mustelid tracks on day 4. Tracking was recorded as either 1 (tracked) or 0 (untracked) for both rats and mice on day 1 and for stoats on day 4 at activated tracking tunnels. We included predator tracking data from the 41 tracking tunnel lines within beech forest of the Eglinton Valley in analyses, but excluded lines from the alpine zone above the beech forest that have less relevance to predator pressure on skinks of the valley floor.

Statistical analyses

We analysed relative abundance of skinks and predators using generalised linear models in R Studio v.1.4 (R Core Team 2019). Model fit was compared using AIC; residuals and assumptions of the best model were checked using the DHARMa package (Hartig 2021).

We used the glmmTMB package (Brooks et al. 2017) to investigate the association between mouse and rat relative abundance in February, May, August and November from 2009 to 2020. Rodent relative abundance was calculated as the number of tunnels tracked by rats and mice per line of ten tunnels (i.e. a continuous variable from 0 to 10). We included mouse abundance as the response variable, rat abundance, year and season as fixed factors and line as a random effect. We fitted the models with Poisson (with and without zero inflation) and negative binomial (1 and 2) distributions. The final model included rat abundance, season and year without interactions and was fitted with Poisson with zero inflation distribution (Table 1).

Table 1 Model output table for AICc-based model selection procedures used to select model distribution for a model investigating the effects of rat abundance, season and year on mouse abundance in which line is also included as a random effect
Full size table

We investigated the association between skinks and rodents using the glm function from the lme4 package (Bates et al. 2015) fitted with a gamma distribution. We first looked at skink and rodent relative abundance in months when both taxa were monitored from 2009 to 2019 (range = 2–6 months per year). We omitted data from 2011 from analyses because of insufficient simultaneous monitoring of skinks and rodents (monitoring data overlapped for only one month in this year). In all models, skink abundance was included as the response variable. Initially, we considered abundance of both rats and mice as predictors and included year as an additional fixed factor. Based on initial results, we further explored the impact of mice on lizards at different points in time by modelling mouse abundance, year and their interactions as predictors.

Finally, we used data from November skink counts in all years to evaluate trends in skink counts over the 11-y period of the study. For this analysis we included skink counts (number of skinks detected per line of 25 artificial retreats each November, divided by the number of times the retreats were checked in that month) as the response variable, year as a continuous predictor variable and transect line as the random variable in a linear mixed effects model fitted in the package lme4. We used skink counts from November for this trend analysis, because this is the month before adult female skinks give birth in the Eglinton Valley (Lettink et al. 2011; JM pers. obs.), and because ambient temperatures are often within the 12–18 °C optimal range for monitoring (Hoare et al. 2009) during this month.

Results

A total of 19,000 retreat checks were completed between February 2009 and March 2020. That is, each of the 200 individual retreats was checked 95 times during this 11-year period. During the monitoring period we made a total of 9671 skink sightings beneath the retreats. Although not originally intended as a monitoring tool for mice, we also opportunistically recorded 787 live (n = 773) or freshly dead (n = 14) mice beneath the same retreats.

The 41 tracking tunnel lines were checked a total of 1009 times between February 2009 and November 2019, resulting in 10,090 tunnel checks for rodents and 5045 for stoats. 485 tunnels were tracked by rats, 2312 by mice and 130 by stoats in this 11-year period. Stoat tracking was too infrequent for meaningful comparison with rodent or lizard abundance and not included in statistical analyses.

Rat tracking rates were generally < 10% (i.e. fewer than 1 in 10 tunnels were tracked), and irruptions are suppressed so that average tracking rates did not exceed 35% (Fig. 1). In contrast, mice irrupted in response to beech seeding and tracking rates reached 75% before populations crashed to < 10% tracking rates (Fig. 1). Mouse abundance was not significantly correlated with rat abundance (0.05 ± 0.03, p = 0.08), but the positive z score and a p-value of < 0.1 indicate a weak positive relationship between the abundance of both species (Table 2; Fig. 1). Mouse abundance was highest in spring and varied significantly among years (Table 2; Fig. 1).

Fig. 1
figure 1

Relative abundance of mice and rats (number of rodents detected per tracking tunnel per line, averaged by month) from 2009 to 2020 in the Eglinton Valley. Mouse abundance is shown as a grey line and rat abundance as a black line. The arrows represent mast years defined as years in which silver, mountain and red beech trees produced more than 500 seeds/m2

Full size image
Table 2 Results of generalised linear mixed models evaluating the effect of rat abundance on mouse abundance during a 10-year study (2009/10–2019/20) in the Eglinton Valley, Fiordland, New Zealand. Significant (p < 0.05) effects are bolded
Full size table

Skink abundance was negatively correlated with mouse abundance (− 1.31 ± 0.32, p < 0.001), but not rat abundance (0.19 ± 1.61, p = 0.91), during the austral spring and summer period of 2009–2019 (Table 3; Fig. 2). Skink abundance was significantly negatively correlated with mouse abundance in four years out of ten and not significantly correlated in the other six years (Table 4).

Table 3 Results of generalised linear mixed models evaluating the effect of rat and mouse abundance on skink abundance in summer/spring during a 10-year study (2009–2019) in the Eglinton Valley
Full size table
Fig. 2
figure 2

Relative abundance of mice (grey line; number of tunnels on which mice were detected per tracking tunnel per line, averaged by month) and relative abundance of skinks (black line; number of skinks per artificial retreat, averaged by month) from 2009 to 2020 in the Eglinton Valley. The arrows represent mast years defined as years in which silver, mountain and red beech trees produced more than 500 seeds/m2

Full size image
Table 4 Results of generalised linear mixed models evaluating the effect of mouse abundance on skink abundance in summer/spring during a 10-year study (2009–2019) in the Eglinton Valley
Full size table

We used the 2094 skink sightings from November checks (just prior to the birthing season) to evaluate population trend during the 11-y period from 2009 to 2019. Ambient temperatures during skink monitoring in this month fell between 9.6 and 18.0 °C. Skink counts per transect line (of 25 artificial retreats) ranged from 0 to 34.5 in November between 2009 and 2019 (mean = 11.9 ± 0.8 SE). This corresponds to a range of 0–1.38 skinks per retreat (Fig. 2). We detected a significant decline in skink counts during the 11-y period (t79 = − 5.12, p < 0.001), which corresponds to a change from an average of one skink being detected per every retreat checked at the start of the monitoring period to one skink being detected for every four retreats checked 11 years later (Fig. 2).

Discussion

Evidence of the impacts of mice on lizards and invertebrates is mounting. Our long-term correlative study of lizard and rodent abundance complements a recent study investigating detailed density-impact functions of mice and indigenous lizards and invertebrates in New Zealand, which found strong evidence that geckos, skinks and wētā are all strongly impacted when mouse abundance exceeds 5% (Norbury et al. 2023). Skink abundance declined over time and was negatively correlated with mouse abundance, but not correlated with rat abundance, suggesting that current predator control in place to protect vulnerable forest birds and bats is insufficient to protect ground-dwelling lizards because mouse populations are not being controlled. We created a visualisation of this dynamic and the consequence for protecting native lizards as a conservation goal (Fig. 3). However, we acknowledge that the full picture is undoubtedly more complicated, with complex food web responses resulting from suppression of a suite of invasive mammalian predators (Binny et al. 2021).

Fig. 3
figure 3

a A simplified conceptual model of a beech forest ecosystem in which controlling all invasive mammalian predators of lizards is needed to achieve the conservation goal of protecting lizards. Mast seeding is included as a key driver of rodent (rat and mouse) irruptions, with flow-on effects on higher order predators (stoats and weasels). b A graphical representation of the effect of current landscape scale management strategies to control selected predators (black ‘not allowed’ symbols) on mice, Mus musculus (represented by change in text box and font size), which are not included in control efforts, and the consequent effect on the conservation goal of protecting lizards (represented by decreased oval size)

Full size image

In the study area, targeted control of predators is ramped up following beech mast seeding with the aim of averting predicted population irruptions of rats and stoats (Dilks et al. 2020; Elliott and Kemp 2016; O'Donnell et al. 2017). This involves deployment of the toxin 1080 across the whole forest (c. 30,000 ha). Rats are highly susceptible to 1080 poisoning and usually population irruptions are averted by timely deployment of the toxin before irruptions occur (O’Donnell et al. 2017). Consequently, the lack of correlation between rat indices and both skinks and mice in this study is likely due primarily to rat numbers being suppressed before irruptions occur under the present predator control regime. However, the extent to which skinks and rats overlap in their habitats occupied is unknown. Previous radio tracking studies of rats in the study area did not mention rats using grasslands, although neither study quantified habitat use patterns (Pryde et al. 2005a; Smith et al. 2009). In contrast, although some mice are killed in these operations, their populations are not usually susceptible to 1080 (Fisher and Airey 2009), and irruptions ensue (Figs. 1 and 2). Further, mice are known to use grassland habitats (Harper 2010; Wilson and Lee 2010) and we regularly detected them in the Eglinton grassland during skink monitoring. Increases in mouse numbers are undoubtedly a product of both increased food availability following periodic beech mast seeding (Wilson and Lee 2010) and release from predation by rats (Bridgman et al. 2013), which are actively managed via predator control in mast years.

The decline in skink detections following irruptions in mouse numbers in the study area is not direct evidence for predation by mice. However, the frequent detection of chewed skink remains on and under artificial retreats during these periods (JM, pers. obs.) implies the negative correlation between skink numbers and mice reflects a functional response to increased predation, an interpretation supported by recent research investigating density impact functions for mice and lizards (Norbury et al. 2023). The cyclic nature of mouse irruptions and negative impacts on skinks results in recovery of southern grass skinks to some degree from periodic crashes (Fig. 2). However, the negative population trend we detected over the 11-y period of this study suggests that the skink population may have been unable to fully recover to previous levels before another mouse irruption seriously impacted the population. This ‘see-sawing’ of native populations within an overall negative trend of decline is akin to that witnessed in both mohua (Mohoua ochrecephala) and long-tailed bat (Chalinolobus tuberculatus) populations in beech forest ecosystems in relation to the abundance of rats, to which they are particularly vulnerable (Elliott 1996; Pryde et al. 2005b). Native bird populations that have recovered in response to the predator control are unlikely to be contributing to the overall trajectory of decline because native lizards coevolved with avian predators and display appropriate antipredator behaviours to them (Hoare et al. 2007b) and the vast majority of native birds are forest dwelling rather than occupying open habitat favoured by the skinks (Dilks et al. 2003). However, episodic predation by native and introduced predators other than mice may also contribute to the skinks’ apparent inability to fully recover from impacts of mouse irruptions (e.g. O’Donnell and Hoare 2009; Reardon et al. 2012). For example, black-fronted terns, Chlidonias albostriatus, have been observed preying on skinks in this grassland when a colony is nesting in the adjacent riverbed during the chick-rearing phase (O’Donnell and Hoare 2009). Further, although rats were generally suppressed to < 0.1 rats per tracking tunnel, rat tracking rates briefly rose to 0.35 rats per tracking tunnel during some periods of the study (Fig. 2) suggesting that even predators targeted by the predator control operation may be contributing to the lizard population decline we observed.

Although southern grass skink populations can reach high densities in the Eglinton Valley grassland (Lettink et al. 2011), the negative population trend over 11 years adds to mounting evidence that native skink populations on mainland Aotearoa have not yet reached equilibrium with invasive mammalian predators. Instead most mainland lizard populations are still declining due to predation in situations where conservation management does not tackle the full suite of mammalian predators (Hoare et al. 2007a; Reardon et al. 2012; Nelson et al. 2016; this study) and informs our conceptual model (Fig. 3). It also supports the declining population trend qualifier assigned to grass skinks (the Oligosoma polychroma complex) and many other small-bodied lizard species in Aotearoa that contributes to their current conservation status (Hitchmough et al. 2021). Species for which no mammal-free offshore island populations exist are on a trajectory towards extinction without conservation intervention (Hitchmough et al. 2016; Nelson et al. 2016).

The suite of mammalian species that are currently targeted in landscape scale pest control in New Zealand (e.g. the ‘Predator Free 2050’ strategy Russell et al. 2015), have major impacts on the most visible and popular elements of the native biota in Aotearoa New Zealand (Innes et al. 2010). In particular, iconic forest birds such as mohua, kākāriki (Cyanoramphus spp.), kākā (Nestor meridionalis), kea (N. notabilis), and whio (Hymenolaimus malacorhynchus) are particularly vulnerable to rats and stoats and are likely to benefit from the strategy (Kemp et al. 2018; Moorhouse et al. 2003; O’Donnell and Hoare 2012b; Simpkins et al. 2015; Whitehead et al. 2008; Whitehead et al. 2022). However, taxonomic bias in research and conservation planning (Clark and May 2002) means that our understanding of outcomes of this landscape scale pest control for remainder of the terrestrial biodiversity is lagging behind. Incorporating a wider array of indicator taxa into outcome monitoring is urgently needed to understand whether the strategy is achieving its stated purpose of “protecting our biodiversity”, a goal that is much more encompassing than protection of iconic forest birds alone (Hoare et al. 2010).

Current predator control may be detrimental to lizard populations if controlling the other predators contributes to a mesopredator release of mice and, if successful, the Predator Free 2050 strategy, despite being well-intended, may have adverse effects on New Zealand’s lizard and large terrestrial invertebrate populations. All of New Zealand’s 124 lizard species (Hitchmough et al. 2021) are likely to be vulnerable to mouse predation where they co-occur, as well as many of the larger-bodied invertebrate taxa (Watts et al. 2022; Norbury et al. 2023). The larger-bodied animals among both groups are likely to be most vulnerable (Tingley et al. 2013; Watts et al. 2022) and mainland populations could conceivably go extinct if mesopredator release does come into play. Instead of protecting native biodiversity, the strategy may actually harm more species than it benefits.

Research into effective control of mice at a scale that could protect populations of native taxa that are vulnerable to mouse predation, within landscape scale management of rats, stoats and possums, is urgently needed. Targeting weasels, cats and hedgehogs is also likely to be needed to enable protection of lizards and large terrestrial invertebrate species (e.g. Reardon et al. 2012). The risk of not considering the complexity of predator guilds in our national pest strategy is that we may fail to deliver on the goal of protecting native biodiversity, and may actually harm the plight of many smaller taxa like lizards and invertebrates.