Introduction

Introduced mammalian predators like mice (Mus musculus), rats (Rattus spp.), stoats (Mustela erminea), and cats (Felis catus) threaten biodiversity globally (Clout 2002; McCreless et al. 2016; Doherty et al. 2015) through direct predation and competition (Medina et al. 2011; Harper and Bunbury 2015; Sánchez-Bayo and Wyckhuys 2019; Dueñas et al. 2021). The evidence for significant population declines and extinctions of native fauna caused by introduced mammalian predators is vast and extends across taxonomic groups including mammals, birds, reptiles, and invertebrates (Kats and Ferrer 2003; Gibbs 2009; Medina et al. 2011; Dueñas et al. 2021). Island ecosystems are especially at risk as many islands have lower species richness than mainland areas and fewer or no native mammalian predators (Kier et al. 2009; Medina et al. 2011). As a result, species endemic to islands can have limited behavioural and physiological defences against introduced mammalian predators (Angel et al. 2009).

New Zealand, which is made up of three large ‘mainland’ and numerous small ‘offshore’ islands, represents an ideal opportunity to explore the effects of introduced predators, including mice, on populations of native invertebrates that evolved without native mammalian predators (Parkes and Murphy 2003; Murphy et al. 2019). Early Polynesian settlers were responsible for the introduction of kiore (Rattus exulans) and kurī (domestic dogs, Canis familiaris) to New Zealand (Holdaway 1996; Clark 1997). Later, European settlers brought ship rats (R. rattus), Norway rats (R. norvegicus), and house mice (as stowaways on ships; King and Forsyth 2021), while stoats, feral cats, brushtail possums (Trichosurus vulpecula), and hedgehogs (Erinaceus europaeus) were intentionally released for reasons such as establishing a fur trade and as bio-control for other introduced mammals that had already started to become pests (King and Forsyth 2021). Many of these introduced mammalian predators have become prevalent throughout New Zealand and are key threats to the survival of native fauna (Hitchmough et al. 2016, 2021; O’Donnell et al. 2018; Robertson et al. 2021). As a result, suppressing predator populations through aerial poison applications and intense trapping programmes is a major component of New Zealand conservation action (Pryde et al. 2005; Innes et al. 2010; Reardon et al. 2012; Hoare et al. 2013). Predator control initiatives, including large-scale trapping programs and aerial 1080 poison operations, typically prioritise larger introduced predators like stoats, cats, and possums, while overlooking mice, which may play a significant role as predators for invertebrates.

There is ample evidence that invasive mice are particularly harmful to island invertebrates and are implicated in the suppression and local extinction of many invertebrate groups (St Clair 2011). In a review of the impacts of mice on islands in the Southern Ocean, (Angel et al. 2009) found evidence from comparative studies that strongly suggest mice can be highly selective predators and can influence invertebrate community composition, body size, and relative abundance. On sub-Antarctic Marion Island, where on average mice consumed 194 g (dry mass) of invertebrates per day, mice preferentially selected larger invertebrates including adult moths, moth larvae, and weevil larvae (Smith et al. 2002). Researchers have found similar trends on Guillou Island (Le Roux et al. 2002) and Macquarie Island (Copson 1986) with invertebrates making up a considerable portion of mouse diet (spiders, for example, were in 70% of mouse stomachs on Guillou Island). On Antipodes Island, where mice were abundant before their eradication, mice preferentially fed on Amphipoda, Lepidoptera and some species of Coleoptera and there were significantly fewer large macroinvertebrates compared to nearby mouse-free Archway Island (Russell 2012).

At lower elevations on mainland New Zealand, there is evidence that mice are similarly destructive to the invertebrate community. Introduced mammalian predators are key drivers of declines for invertebrate communities in forests, shrublands and grasslands (McGuinness 2001). Watts et al. (2011, 2022) have shown that mammals, and mice in particular, have significant impacts on the ground-dwelling invertebrate community within a lowland North Island forest. Watts et al. (2011) studied how the removal of thirteen mammal species (all introduced herbivores and predators except mice) from Maungatautari Sanctuary (3239 ha, 240–797 m asl) influenced the invertebrate community. The eradication resulted in an increase of adult tree wētā (Hemideina thoracica) from 0.2 to 1.2 wētā caught per pitfall trap (n = 20), while other wētā species increased from effectively 0 to 2 wētā per trap (Watts et al. 2011). In a subsequent 5-year study in a South Island grassland ecosystem (Macraes Conservation Area, eastern Otago, 400-600 m a.s.l.), Norbury et al. (2022) found that mice needed to be excluded from or extremely scarce (< 5% tracking tunnel printing rate based on the density of footprints) before invertebrate communities began to recover. Overall, our understanding of the interactions between mice and invertebrates in New Zealand is primarily confined to off-shore islands and lower elevation areas of the mainland. There is a lack of knowledge regarding whether, or to what degree, the impact of mice extends to alpine regions.

Our research represents one of the first studies to specifically explore the relationship between introduced mammalian predators, particularly mice, and the large-bodied ground-dwelling invertebrate community in the New Zealand alpine zone. On the South Island of New Zealand, common terrestrial alpine invertebrates include ground beetles, wētā (the New Zealand name for Orthoptera in the families Anostostomatidae and Rhaphidophoridae), spiders, landhoppers (terrestrial amphipods), grasshoppers, centipedes, and harvestmen (O’Donnell et al. 2017). Although mice likely influence a wide range of invertebrate groups, we focus on large-bodied invertebrates because they have been shown to be preferentially targeted in lowland studies (Watts et al. 2011, 2022). This targeting may be due to the fact that they are flightless, long-lived, relatively slow to develop, ground-dwelling, and slow-moving making them especially vulnerable to predation and potentially slower at bouncing back from predation pressure (McGuinness 2001; Stringer & Hitchmough 2012). Consequently, we also expect this group to be the first to respond to predation pressures from introduced mammalian predators in the alpine zone.

We used pitfall traps to compare the composition, abundance, and body size of the large-bodied invertebrate community (which we define as individuals that are greater than or equal to 1 cm in body length) across sites with and without mice and with and without predator control. Our sites consisted of four alpine areas on the South Island of New Zealand and four alpine areas on nearby offshore islands. Although predator control targeted stoats, rats, possums and cats, our study focused on the results for stoats and mice, as they are the resident introduced predators in alpine areas of Fiordland (O’Donnell et al. 2017). Given the known effects of mice on the invertebrate community at lower elevations, we expected mice to reduce invertebrate abundance and body size. Stoats are prolific hunters of invertebrates, but they have a varied diet that also includes mice (Smith et al. 2005; McAulay et al. 2020). Removing stoats from alpine zones may benefit some invertebrate groups, like wētā, which are a key part of the stoat diet (Smith et al. 2005), but it might also release predation pressure on mice or weasels (Mustela nivalis) allowing them to prey upon invertebrates (Peltzer et al. 2019; McAulay and Monks 2023). Consequently, we expected predator control not to alter large-bodied invertebrate community composition, abundance, or body size because the presence of mice would counteract any benefits received from controlling stoats.

Methods

Study area

This study occurred at eight sites above the climatic treeline in Southern New Zealand. Site altitude ranged from 700 to 1600 m above sea level (a.s.l.). Four sites were on the South Island, while the others were on smaller ‘offshore’ islands. All sites had vegetation dominated by tussock grasslands with small sections of scree, boulders, and patches of dense shrub. Tussock grasslands consisted of 50–60% mid-ribbed snow tussock (Chionochloa pallens), narrow-leafed snow tussock (C. rigida), red tussock (C. rubra), fescue tussock (Festuca novae-zelandiae), and needle snow tussock (C. acicularis). Other common plants included mountain buttercup (Ranunculus lyallii), dracophyllum (Dracophyllum rosmarinifolium), common speargrass (Aciphylla squarrosa), mirror bush (Coprosma prostrata) and species of the woody shrub Gaultheria. Scree and boulder sections were 60–80% rock-covered, ranging from small pebbles to scree to large boulders and interspersed with mosses and ferns. Shrub patches were dominated by woody shrubs no more than 1 m tall, including hebes (e.g., Veronica hectorii) and dracophyllum, as well as a variety of non-woody native plants like mountain daisies (Celmisia spp.), mountain buttercup, common speargrass, and various ferns and mosses.

Predator control regimes

The eight sites (five that are known to have mice and three that are mouse-free) were subject to a range of active predator control regimes run by government and community groups (Table 1). Sites with predator control were Homer and Gertrude Valley, Shy Lake, Secretary Island, Resolution Island and Table Hill (Table 1). Standard trapping for stoats, rats and hedgehogs in New Zealand uses small to medium-sized kill traps known as DOC150 and DOC200 traps (Warburton et al. 2008). A local conservation volunteer group runs trapping lines to control mustelids, rats, and hedgehogs at the Homer and Gertrude Valleys (Weston et al. 2018). Secretary Island is rodent-free and has an extensive 108 km trapping line to keep stoat populations at low levels (McMurtrie et al. 2011). Mice are present on Resolution Island, and there has been considerable effort to keep stoat levels low through extensive trapping (Edge et al. 2011; Murphy et al. 2016). Blaikies Hill and Table Hill are located on Stewart Island (Rakiura), which is free from mice and mustelids, but rats and cats are present. Blaikies Hill has no predator control. Predator control at Table Hill consists of trap and bait lines to control rats and feral cats (Dowding and Davis 2007). Note that Table Hill is home to a colony of New Zealand dotterel (Anarhynchus obscurus) however this is a critically endangered population (c 250 total individuals) that breed in herb-fields and exposed hilltops across the larger island of Rakiura, so we expect they have very little influence on the invertebrate community. Shy Lake is situated on the peninsula between Breaksea Sound and the Wet Jacket Arm of Dusky Sound, which received an aerial 1080 (manufactured version of fluoroacetate) predator control treatment in June 2020 (Anon. 2020). Aerial 1080 operations target possums and rats, while stoats are killed via secondary poisoning (Eason et al. 2011).

Table 1 Presence (+) or absence (−) of mice and predator control across the eight alpine field sites visited during the 2021/22 spring and summer field seasons
Full size table

Most of the mice present sites have evidence of recent mouse activity. Mice were present but uncommon in surveys at Homer and Gertrude and Shy Lake in 2021 (tracking rate of 0.08 for both sites), while mice were very prevalent in Borland in 2020 (tracking rate of 0.82; DOC unpubl. data). Mice were fairly prevalent at Resolution Island (tracking rate of 0.36) while they were less common at Lake Roe (tracking rate of 0.10); however these surveys are not as recent, and no later data was available (2017 and 2018 respectively; DOC unpubl. data). All tracking rates are from 24-h surveys and are presented in Online Resource 1. Historically, sites with predator control have low levels of stoats and rats. Rakiura is stoat-free but there is no available trapping data for rats at Table Hill. From 21-day surveys using tracking tunnels, no rats were observed at Shy Lake, Resolution Island or Secretary Island while Homer and Gertrude had a rate of 0.05 of 35 tunnels. For stoats, no individuals were observed at Homer and Gertrude, Shy Lake, or Resolution Island. For Secretary Island, stoats occurred at low levels (an average of 13.5 stoats per hectare). Most recent tracking rates for rats and stoats and the number of traps set for each predator control site be found in Online Resource 1.

Study design

Individual sites were monitored on different dates between November 2021 and February 2022 (Austral summer) due to the large distances between sites and the costs associated with accessing them (see Table 1 for dates). We installed four pitfall grids at each of our eight field sites. Grids consisted of four rows of four pitfall traps spaced 15 m apart giving 16 pitfalls per grid (64 pitfall traps per site). Before visiting each site, we generated 16 potential grid locations by creating random GPS points within suitable alpine habitats based on aerial imagery in QGIS (QGIS Development Team 2022). We then visited each potential grid location in turn, starting at point one, and installed a grid if the area was both accessible and large enough. We made the pitfall traps from 1L clear plastic bottles with a 90 mm diameter opening. We cut off the top quarter of each bottle, inverted it to make a funnel, and trimmed the container and funnel to reduce the overall size of the trap (average trap depth around 90 mm). We derived the trap design from Hohbein and Conway (2018), who recommend using funnels to limit bycatch of non-target species (such as lizards) and suggest that traps be made of clear plastic as trap colour can influence which invertebrates are caught (Buchholz and Möller 2018). Because there was still potential to catch threatened native lizards in our traps, we filled the pitfalls with ~ 2 cm of water rather than a preserving liquid so that any lizards inadvertently caught could be released unharmed. This was possible because traps were checked daily when open.

We paired each pitfall grid with a line of five tracking tunnels (Black Trakka™ 500 × 100 × 100 mm; Gotcha Traps, Warkworth, New Zealand), spaced 50 m apart following the New Zealand Department of Conservation’s recommended distance to optimise surveying for rodents (Gillies and Williams 2013). We placed the line of tracking tunnels 50 m upslope from its paired pitfall grid running parallel to the ridgeline. If that was not possible, due to the location of the grid, the tunnel line was run downslope from the gridline. The tunnels were baited with ~ 5 g of Smooth Peanut Butter (Pams Products Ltd, New Zealand) placed on the ink in the centre of each tracking card. We left the tunnels overnight and checked them daily for three days.

We opened the pitfall traps for three nights, except at Blaikies Hill, where traps were open for two nights because of adverse weather. At each site, we installed the four pitfall grids on the afternoon of the first day and checked all pitfall traps daily. From each trap, we identified invertebrates that were ≥ 1 cm long to the taxonomic family level, except for harvestmen which we were only able to identify to the order level. For spiders and harvestmen, we measured body size as diagonal leg span, a common measurement for this group (e.g., Seyfarth et al. 1982). For all other groups, we measured body size as the length from the front of the head to the tip of the abdomen excluding legs, antennae, and ovipositor. We kept at least three example specimens of each family that we caught and preserved these in 70% ethanol. We released additional invertebrates back into the environment after they were measured and recorded. We did not break the surface tension of the water, so many invertebrates were alive when we sampled the catch the following day (although some did drown overnight). We chose not to keep all invertebrates to limit our study’s impact on the invertebrate community. On the last day, we checked, emptied, and removed all the pitfall traps. For the entire three-day period over which pitfalls and tracking tunnels were active at each site, we recorded hourly temperature and relative humidity with two HOBO loggers (model number MX2301A, Onset Computer Corp., Pocasset, MA). We placed the loggers at ground level in shaded locations within two of the four pitfall grids selected at random.

Statistical analysis

Hill numbers, and an accumulation curve of taxonomic richness at the family level (to assess sampling effort), were generated for the invertebrate community at each site using the hillR package (Li 2018) and the random method in the Vegan package (Oksanen et al. 2020) respectively. The random method calculates a mean accumulation curve and its standard deviation based on random permutations of the data (Collwell et al. 2012). To investigate whether the presence of mice or predator control influences large-bodied invertebrate abundance, we ran a generalised linear mixed model (GLMM) with a negative binomial distribution using the package glmmTMB (Brooks et al. 2017). The model contained counts of invertebrates at all eight sites as the response variable. We combined observations of invertebrates at the pitfall grid level so there were four replicates per field site. We then included the presence of mice and predator control as categorical predictor variables in addition to the daily average maximum temperature and relative humidity as continuous predictor variables. We included the site as a random effect to account for repeated measures. We then ran a similar model with a subset of the data that only contained observations of wētā. This model contained the same structure as the model described above but with a Tweedie distribution using the package glmmTMB.

We also ran a generalised linear model (GLM) with an inverse-Gaussian distribution using the package lme4 (Bates et al. 2015) to examine the influence of mice and predator control on the body size of the largest invertebrates. To achieve this, we truncated the data to only include the largest 50% of invertebrates caught (individuals larger than 1.5 cm) to better discern whether introduced mammalian predators were removing the largest invertebrates from the community. As we were not using kill traps, there was potential to re-sample the same individual across the three-day sampling period, so we used data only from the first night of sampling. This was necessary as the smaller sample size, containing the largest 50% of invertebrates would not allow us to include a random effect. The model contained invertebrate body size as the response variable, with the presence of mice and predator control as separate categorical predictor variables and average daily minimum temperature and relative humidity as covariates.

Lastly, we used a permutational multivariate analysis of variance (PERMANOVA, 9,999 permutations; Anderson, 2001) based on a Bray–Curtis dissimilarity index to examine differences in the invertebrate community in response to the presence of mice and predator control. We also conducted a permutational analysis of multivariate dispersion (PERMDISP, Anderson, 2001, Anderson 2006) to determine if the distribution or spread of groups is significantly different. This analysis was completed using the package Vegan in R version 4.2.1 (Oksanen et al. 2020; R Core Team 2023). We then used non-metric multidimensional scaling (NMDS) with a Bray–Curtis dissimilarity index to generate minimum convex polygons (containing all points within the group) to visualise the relationships between invertebrate communities (Fig. 1).

Fig. 1
figure 1

Map of the eight field sites visited during the 2022 summer season. Black dots represent sites with active predator control, while red diamonds represent sites without predator control. Secretary Island and Stewart Island/Rakiura are mouse free while mice are present at the remaining five sites. The inset is New Zealand and the zoomed in map is Fiordland and Stewart Island. The outline of New Zealand is the NZ coastline and island Topo map generated by Land Information New Zealand accessed through the Land Resource Information System portal (LRIS; https://lris.scinfo.org.nz/ accessed Aug 2023)

Full size image

Results

Overview

Over the entire study, we observed 1362 invertebrates from 8 taxonomic groups, with wolf spiders (Lycosidae), ground beetles (Carabidae), wētā (Anostostomatidae and), and landhoppers (Amphipoda) being the most common (Fig. 2). Table Hill was the most active site, with 334 invertebrate captures, while Lake Roe and Shy Lake were the least active sites, with 86 captures each (Table 2). The average nightly (9:00 pm to 6:00 am) maximum and minimum temperatures ranged from 5.5–19.8 °C and 2.3–8.6 °C respectively across the eight sites (Table 2). For invertebrates ≥ 1 cm long, average body size was 1.5 cm, and the largest invertebrate was 4.3 cm (a carabid beetle of the genus Mecodema from Borland). Among sites, Shannon diversity (q1) ranged from 1.27 to 4.19, with Secretary Island exhibiting the highest diversity. Simpson diversity (q2) varied from 1.10 to 3.70, with Secretary Island and Lake Roe showing the highest dominance. Richness (q0) ranged from 4 to 7 families, with Secretary Island having the most diversity (Table 2). The species accumulation curve (at the family level) showed little change in the number of families per sample after sample 20, suggesting that we did a fairly good job of sampling the community and that additional samples or observations would not substantially increase the number of observed families (the accumulation curve is presented in Online Resource 1). Lastly, we only recorded one mouse track across all eight field sites over the entire summer, which was recorded at Lake Roe.

Fig. 2
figure 2

Invertebrate abundance (total number of individuals caught) measured over three nights across eight field sites, six in Fiordland and two on Rakiura, between November 2021 and February 2022

Full size image
Table 2 Total number of invertebrates captured (n = 1362) and associated weather conditions for the sampling duration at each of the eight sites
Full size table

Total invertebrate abundance

A total of 1362 observations of invertebrates across all eight field sites (i.e., a cumulative total of all invertebrates) were included in this analysis. We found the presence of mice and predator control did not significantly influence total invertebrate abundance (χ2 (1) = 0.85, P = 0.35; χ2 (1) = 0.03 P = 0.85; respectively). Invertebrate abundance was influenced by daily minimum temperature (χ2 (1) = 3 0.86, P = 0.04, model fit = 0.26). For the wētā only model, we found that the sites with mice had significantly fewer wētā compared to sites without mice (χ2 (1) = 09.98, P = 0.001; Fig. 3a). Predator control, on the other hand, did not influence wētā abundance (χ2 (1) = 0.34 P = 0.55; Fig. 3b). Wētā abundance was also significantly influenced by daily minimum temperature (χ2 (1) = 32.00 P =  < 0.001) where observations of wētā increased with rising temperatures.

Fig. 3
figure 3

Generalised linear model results of the influence of a mouse presence and b predator control (focusing on stoats and rats) on wētā abundance. Figures represent the mean and 95% confidence intervals for each factor

Full size image

Invertebrate body size

To determine the effect of mouse presence and predator control on invertebrate body size, we analysed only the largest half of all invertebrates (individuals greater than 1.5 cm in body size) caught on the first night across all eight sites (n = 223/529). The presence of mice significantly influenced invertebrate body size. The invertebrates caught at sites without mice were slightly larger (mean 2.5 cm) when compared to sites with mice (mean 2.3 cm) (χ2 (1) = 7.65, P = 0.006; Fig. 4a). Conversely, invertebrate body size was not significantly influenced by the presence of predator control (χ2 (1) = 0.52, P = 0.46, Fig. 4b, model fit = 0.17). The largest invertebrates, measuring around 3.5 cm in body size at the mice absent sites were identified as being large wētā (Anostostomatidae). Wētā of similar size were not recorded at the mice present sites (Fig. 4a).

Fig. 4
figure 4

Generalised linear model results of the influence of a mouse presence and b predator control (focusing on stoats and rats) on invertebrate body size. We truncated the data to the largest 50% of invertebrates (individuals larger than 1.5 cm) and only included individuals caught on the first night. Figures represent the mean and 95% confidence intervals for each factor

Full size image

Invertebrate community composition

There was no difference in dispersion between sites with and without mice, nor between sites with and without predator control (PERMDISP; Table 3). We found a small but significant difference in community composition between sites with and without mice (R2 = 0.17, P < 0.001; Table 3; Fig. 5). Wētā were less common in sites where mice were present (Fig. 5). Predator control had no significant influence on community composition (R2 = 0.03, P = 0.36; Table 3; Fig. 6).

Table 3 Influence of the presence of mice and predator control on the large-bodied alpine invertebrate community, as determined by a permutational multivariate analysis of variance (PERMANOVA) and including a test for homogeneity of multivariate dispersion using a permutational analysis of multivariate dispersions (PERMDISP)
Full size table
Fig. 5
figure 5

Non-metric multidimensional scaling (NMDS) ordination of large-bodied terrestrial invertebrate groups as measured at sites with and without mice (k = 3). Red points represent sites where mice are absent and blue points represent sites where mice are present. The red polygon is an ellipsoid hull that encompasses all sites where mice are absent, while the blue polygon encompasses all sites where mice are present. Taxonomic group labels are centered around sites that they are the most associated in the 2D space

Full size image
Fig. 6
figure 6

Non-metric multidimensional scaling (NMDS) ordination of large-bodied terrestrial invertebrate groups as measured at sites with (blue points) and without (orange points) predator control (k = 3). The orange polygon is an ellipsoid hull that encompasses all sites that have no active predator control, while the blue polygon encompasses all sites with predator control. Taxonomic group labels are centered around sites that they are the most associated in the 2D space

Full size image

Discussion

Our study provides new evidence suggesting that mice are influencing the large-bodied invertebrate community in alpine areas of the southern South Island of New Zealand. We found that wētā were less common in sites where mice were present. The largest 50% of invertebrates on the first sampling day were also slightly larger at mouse-free sites than at sites with a history of mice. We did not find the presence of mice to influence invertebrate abundance, but we did find that wētā were more abundant at sites that were mice-free. Lastly, the presence of predator control targeting stoats, rats and or cats had no significant effect on invertebrate community composition, abundance, or body size.

At lower elevations, there is strong international evidence that mice influence invertebrate abundance, body size and community composition (Marris 2000; Angel et al. 2009; St Clair 2011; Houghton et al. 2019). For example, comparisons of invertebrates on sub-Antarctic Marion Island with those on nearby mouse-free Prince Island suggest the presence of mice influenced invertebrate community composition, causing a decrease in the body size of medium to large-bodied invertebrates (Crafford and Scholtz 1987; Crafford 1990; Chown and Smith 1993). Similarly, in the North Island of New Zealand, the invertebrate community within a mammal-proof fence where mice are actively controlled (within Zealandia Ecosanctuary) was found to be significantly different from the community outside the fence where mice are not controlled, and some beetle species were up to 20 times more prevalent within the mouse-free enclosure (Vergara et al. 2021). Watts et al. (2022) removed mice from a 24-ha area within the fenced Maungatautari sanctuary and allowed mice to remain in an adjacent, independently fenced, 17-ha area. After surveying the invertebrate community in each enclosure, the treatments were swapped. Watts et al. found invertebrate abundance and body size increased substantially after removing mice, then decreased again after mice were allowed to build up once more.

The lack of an effect of the presence of mice on the invertebrate community in our study could be a result of the low number of mice that were active during our sampling period. Although mice are known to be consistently absent from our mouse-free sites, mouse abundance can vary greatly at sites where they are present as mouse populations boom and bust with food availability (often associated with Chionochloa or beech mast seeding in the alpine zone). For instance, Wilson & Lee (2010) investigated the impact of tussock mast seeding during the summer of 2005/06 on mouse populations. They discovered that mouse density surged from 4 individuals per hectare in autumn to 39 individuals per hectare by spring of 2006, before declining to 8 individuals per hectare the following autumn (May 2007). By the subsequent spring (November 2007), no mice were captured across 768 trap-nights. This research highlights the boom-bust relationship that mice have with seed masts, where populations boom during the mast when food is available, and then crash as the inflated population strips the environment of resources. The last seed mast event on the South Island was in 2019 (Nichols et al. 2021; O’Malley et al. 2022). As a result, our study took place two years after the mast when mice had become relatively scarce, and the invertebrate community may have had time to recover from the intense predation pressure associated an inflated mouse population driven by the tussock mast. To test our hypothesis that mice are driving the differences that we observed in the invertebrate community, we would need to repeat this study before, during and immediately after a mast year to observe the influence of mice on the invertebrate community when predation pressure is at its peak.

Though there is little information on mouse diet in alpine areas of New Zealand, Wilson and Lee (2010) found mouse stomach contents to be dominated by arthropods and to switch to seeds during a mass seed event. Studies at lower elevations and offshore islands show that mice consume a range of plant, vertebrate and invertebrate species, but that invertebrates often make up much of their diet (Copson 1986). For example, in a study of the stomach contents of 102 mice near Dunedin Otago, 86% of stomachs contained plant material, and 90% contained invertebrates, where Lepidoptera, Coleoptera and Araneae were commonly consumed (Miller and Webb 2001). On Antipodes Island, mouse stomachs were found to contain plant matter, bird remains, and invertebrates including Lepidoptera larvae, Acarina, adult Coleoptera and Araneae (Russell et al. 2020). There is also evidence of mice influencing native lizards in New Zealand, where skink populations were found to have a negative relationship with mouse abundance in the Eglinton Valley, Fiordland National Park (Monks et al. 2024).

Although we detected likely effects of mouse pressure on the alpine invertebrate fauna, there are alternative explanations for why invertebrate body size may differ between sites with and without mice. For example, there are likely to be natural geographic differences in the species present at the different sites, even within taxonomic groups, and these species may differ in size. Indeed, we caught multiple species of Hemiandrus wētā, a genus whose body size can vary greatly. Hemiandrus focalis was species of wētā common to our field sites that are classified as a medium to large wētā (maximum body size of 24 mm, head width 6.8–9.6 mm, used as a proxy for body size; Salmon 1950), while H. maculifrons is classified as a small wētā (head width of 3.6–4.9 mm; Smith et al. 2013). In our study, the maximum body size for wētā in sites with and without mice was similar (3.0 and 3.3 cm, respectively), as was the body size of spiders (2.2 and 2.3 cm, respectively). This suggests that similarly sized wētā and other species of invertebrates were present at all sites, but larger individuals were less common at sites where mice were present.

Temporal effects could also influence our body-size data from different sites as our sampling occurred over several months. Adults of different species of holometabolous groups, like beetles, can emerge at different points in the season (Harry et al. 2011), and juveniles of hemimetabolous groups, like wētā and spiders, get larger over time as individuals mature and eventually become adults. We sampled our eight sites at different times across the austral summer, therefore it would have been possible to sample all the ‘mice present’ sites when there were more early-instar hemimetabolous invertebrates present. Similarly, mice-present sites could have been sampled before large adult beetles emerged or after they reproduced and died. However, we surveyed mice free and mice present sites fairly evenly throughout the summer, visiting representatives of both site categories in early summer (Nov-Dec) and mid-summer (January). Additionally, large Hemiandrus species, including H. fiordensis (which was common at our field sites), may take up to 3 years to reach sexual maturity (Johns 2001; Van Wyngaarden 1975), and at sites we have visited repeatedly (unpublished data) we observed both juveniles and adults to be present in all summer months, rather than distinct periods with only adults or immatures. Thus, it is likely that a range of differently sized wētā and other invertebrates inhabit our sites throughout the summer. Differences in body sizes observed between sites are unlikely to be driven primarily by the time at which we sampled and the correlation we observed with predator presence is likely to be a real effect.

Our community composition results and wētā only model highlight that mice may be particularly threatening to alpine wētā. Other studies have shown that wētā are consumed by mice at lower elevations, and they are a common part of stoat diet in alpine areas of the South Island (Smith et al. 2005, 2010). For example, McAulay et al. (2020) found that stoat diet had a greater reliance on invertebrates, especially wētā, during the summer when invertebrates were more readily available in the environment. This is concerning as many alpine wētā are long-lived and develop slowly compared to other invertebrates. As noted above, H. fiordensis found in our study can take up to three years to reach sexual maturity, and there is evidence that larger species in other alpine genera (Hemideina, Deinacrida) can live even longer; for example the mountain stone wētā (Hemideina maori) can take three to four years to reach sexual maturity and may live for one to four breeding seasons as an adult (Jamieson et al. 2000; Leisnham et al. 2003). While ground wētā have fairly high fecundity (clutch size of 20–60 eggs; Gwynne 2004), stoat diet analyses demonstrate that an individual can consume upwards of seven wētā in a single night (Purdey et al. 2004; Smith et al. 2008). Given how long it takes for wētā to fully develop into adults, this suggests wētā recruitment could struggle to keep up with predation pressure from introduced mammalian predators. Unfortunately, we don’t understand the level of pressure they can tolerate or whether they can fully recover from intense periods of predation pressure such as what occurs following masting events.

Predator control and mesopredator release

Predator control did not appear to influence community composition, abundance, or body size for the large-bodied alpine invertebrate community. Current predator control strategies in New Zealand focus on protecting avifauna through large-scale poisoning and trapping programmes (Elliott and Kemp 2016; Innes et al. 2019; Leathwick and Byrom 2023; Nichols et al. 2021; Reardon et al 2012). Work in the alpine zone is an extension of large-scale predator control or species-specific localised control for threatened species, like rock wren (Xenicus gilviventris) and focuses on larger predators like stoats (Little et al. 2017; Weston et al. 2018). Trapping and aerial 1080 poisoning are effective tools to control stoat populations (Dilks et al. 2020; Nichols et al. 2021), which likely benefits the invertebrate community. However, mouse kills from pest control tools such as 1080 are variable, and mouse populations bounce back relatively quickly after predator control operations, making large-scale operations less effective for mice (van Heezik et al. 2020). As a result, fauna susceptible to mouse predation may remain vulnerable and continue to face high predation pressure even when they reside in areas with active predator control. It is likely that invertebrates in our predator control sites still face predation pressure from mice (if not in the study season, then in recent previous seasons), which could explain why we saw no difference between locations with predator control and those without.

Controlling for specific predators, instead of the full suite of introduced mammalian predators, can limit conservation outcomes, or even generate negative outcomes for native species. In New Zealand, suppression of cats and stoats resulted in a boom of rodent populations when there was enough food in the system to support them, which increased predation pressure on native fauna (Norbury et al. 2013; Tompkins et al. 2013; Whitau et al. 2023). Similarly, on Little Barrier Island, lethal control of cats increased the number of predatory rats, causing a decline in the breeding success of the Cook’s petrel (Pterodroma cookii). Globally there are multiple examples of mesopredator release as an unintended consequence of predator control that has led to adverse conservation outcomes for species and habitat (Barton and Roth 2008; Ritchie and Johnson 2009; Colman et al. 2014; Marlow et al. 2015). While we lack sufficient data to confirm whether meso-predator release is occurring in alpine areas of New Zealand, it is important to note that excluding mice and other potential predators from predator control efforts could potentially hinder conservation outcomes. As a result, predator control efforts should be more comprehensive and be cognisant of the trophic interactions between predators (Doherty et al. 2015).

To make predator control more comprehensive and support large-bodied invertebrates in the New Zealand alpine zone, our findings suggest that adaptive management strategies including mouse control are warranted given the mounting evidence that mice are key predators for many endemic species. While it is difficult to monitor mouse populations when they are at low densities (we only detected one mouse print from tracking tunnels across the five sites where mice are known to occur), mice can still influence the invertebrate community when they are scarce in the environment (Norbury et al. 2022). On smaller scales, mice have been successfully eradicated from multiple offshore islands through intensive poisoning programmes using anticoagulant rodenticides (Griffiths et al. 2015; Horn et al. 2019; Martin and Richardson 2019). In these cases, bait programmes were meticulously planned to ensure all rodents had a high chance of being exposed to the bait and to limit exposure to non-target wildlife and potable water supplies (Martin and Richardson 2019). As the size of the target area for control increases, so do the logistical challenges, costs, and risk of failure (Livingstone et al. 2022). Consequently, intensive aerial eradication efforts are currently impractical for the scale of poisoning required to eradicate mice from large mainland areas. Intensive baiting in smaller locations with natural barriers to mouse movement may be a way to support populations of individual threatened invertebrate species or localised invertebrate communities in the alpine zone. However, these efforts require constant upkeep, and mice often re-invade quickly (Gillies et al. 2003; Elliott and Kemp 2016); therefore, investment in new tools that are effective against mice across large areas is essential if we hope to support native species at greater scales.

Conclusion

The interactions between introduced predators and native species are complex. Our study makes initial contributions to understanding the relationship between large-bodied invertebrates and the suite of introduced predators in New Zealand’s alpine areas. Our results suggest that mice influence large-bodied invertebrate body size and community composition, particularly the presence of wētā, and we were able to detect these differences in a year when mouse abundance was low. Further work comparing periods of high and low mouse abundance is required to better understand this dynamic. Predator control efforts, primarily targeting stoats, were not found to influence the invertebrate community despite stoats being known predators of invertebrates. These results suggest that predator control needs to be extended to include mice to benefit large-bodied invertebrates in alpine areas of New Zealand. Our findings also support trends in other studies where controlling for a single predator instead of the full suite of predators can limit conservation outcomes for native fauna. As a result, we need to ensure that our predator control programmes are more comprehensive and account for the interactions between predator species to maximise conservation outcomes.

Competing interests

The authors have no competing interests to disclose.