Beetle (Coleoptera) communities inside and outside the pest-resistant fencing of a New Zealand ecosanctuary

ABSTRACT The ecological impacts of invasive mammals are widely documented around the world. In New Zealand, fencing designed to exclude non-native mammals is used for conservation and restoration efforts. The Orokonui Ecosanctuary is a 307-hectare coastal Otago reserve (Dunedin, New Zealand) that is surrounded by mammal-exclusion fencing. The goal of the present study was to investigate how excluding mammals and including (native) animals inside the Orokonui Ecosanctuary fence has influenced ground- and litter-dwelling beetle (Coleoptera) abundance, diversity, and community composition. We hypothesised that beetle abundance, diversity, and community composition would be different whether the beetles sampled were from sites inside the fence or outside the fence. Beetles were extracted from the litter and soil of six sites (three inside, three outside) two times (once in winter, once in summer). The abundance, diversity, species composition, size distribution, and trophic guild distribution of beetles inside and outside the fence and between seasons were compared. Our results suggest that sites inside the fence harbour a greater abundance and diversity of beetles. We found a high abundance of native beetles both inside and outside the Orokonui Ecosanctuary’s fence. Further research may find evidence that the fenced sanctuary is providing a ‘halo effect’ whereby native beetles thriving within the Orokonui Ecosanctuary are spreading out into the surrounding landscape and also outcompeting introduced beetles.


Introduction
The impact of mammals on ecosystems has long been of concern in places where they have been introduced outside of their natural historic range, such as in New Zealand (Coblentz 1978;Clout & Russell 2008;Dolman & Wäber 2008). The flora, fauna, and distinctive ecosystems of New Zealand evolved in the absence of nonvolant terrestrial mammals over many millions of years before human settlement by Māori, approximately 800 years ago (Simberloff 1995;Dobson et al. 1997;Courchamp et al. 2003;Worthy et al. 2006). Birds, including the large-bodied moa, were the primary herbivorous browsers and grazers occupying ecological niches typically dominated by mammals elsewhere in the world ). More than 31 species of terrestrial mammals that are now established in New Zealand were introduced by human settlers both intentionally and accidentally (Veblen & Stewart 1982;Williams & Cameron 2006;Parkes & Murphy 2003). Many of these were deliberately brought to New Zealand from the eighteenth century onwards by European settlers for use as domestic stock, pets, and recreational hunting (Parkes & Murphy 2003). Introduced mammals can modify the habitat structure and functioning of terrestrial ecosystems through particular modes of feeding, trampling, uprooting, and burrowing (Rooney & Waller 2003;Miyashita et al. 2004;Campbell & Long 2009;Holt et al. 2011), which can often greatly differ from the modes that have evolved with the native vertebrate and invertebrate fauna (Atkinson 2006;Towns et al. 2009).
Insectivorous mammals can directly and indirectly affect invertebrate assemblages via predation (Ruscoe et al. 2013;Gish et al. 2017), and herbivorous mammals can indirectly impact invertebrate density and diversity through habitat modification, especially by changing vegetation structure (Baines et al. 1994;Allombert et al. 2005;Ueda et al. 2008). Invertebrates are crucial to ecosystem functioning and in comparison to mammals are generally poorly known, yet constitute a highly diverse fauna (Wilson 1987).
Beetles (Coleoptera) are one of the most species-rich orders among the insects, representing most functional roles (Leschen et al. 2003;Ponomarenko 2003;Jäch & Balke 2008;Forbes et al. 2018). Because of this, taxa within this group have been proposed as important bioindicators of environmental change (Choi et al. 2010;Salomão et al. 2019). For example, soil and ground-dwelling beetles, like dung beetles (Scarabaeidae) and ground beetles (Carabidae), are important components of soil and litter ecosystems (Petersen & Luxton 1982;Pizzolotto et al. 2018;Salomão et al. 2019) that can be used as indicators of ecosystem disturbance, taxonomic diversity, and other aspects of ecosystem health (McGeoch et al. 2002;Vandewalle et al. 2010;Vergara et al. 2021).
Introduced mammals may impact soil-and surface-dwelling beetle communities in island systems. Areas of overgrazing by goats and sheep in shrubland on Crete (Greece), for instance, had significantly reduced ground beetle (Carabidae) species diversity compared to areas of intermediate grazing intensity (Kaltsas et al. 2013). Depending on country and habitat, some ground beetle species may be good grazing indicators, with the abundance of species such as Carabus banoni (Dejean) and Cymindis lineata (Quensel) being greatly reduced or absent at overgrazed sites compared to sites with low grazing intensity (Cole et al. 2006;Kaltsas et al. 2013), and species such as Platyderus jedlickai Maran being dominant at sites with intense grazing (Kaltsas et al. 2013). In Japan, the native sika deer (Cervus nippon Temminck), considered a pest following the extinction of their natural wolf predator (Takatsuki 2009), have significantly altered beetle assemblages in the deciduous and bamboo forests. While the abundance of some beetle taxonomic groups increased or decreased in response to deer foraging, the overall beetle abundance and diversity did not always show significant changes (Ueda et al. 2008;Iida et al. 2016).
In New Zealand, beetle abundance is significantly higher on offshore islands without rodents compared to those where rodents have invaded (Towns et al. 2009). Moreover, endemic ground beetles, Ctenognathus adamsi (Broun), were absent from Korapuki Island until the removal of rabbits and pacific rats (Towns et al. 1997), and are more abundant inside compared to outside of fenced (mammal-excluding) sanctuaries (Watts et al. 2020;Vergara et al. 2021).
Predation and disturbance by introduced mammals may also affect size distribution of beetle populations. For example, large-sized beetles with lower mobility are more sensitive to grazing disturbance (Cole et al. 2006), and some predators may be more likely to feed on (and remove) largersized beetles from a population (Chown & Smith 1993;Marris 2000;Russell et al. 2020).
On the main islands of New Zealand, an approach for protecting and restoring native biodiversity has been the establishment of pest-resistant ring-fenced ecosanctuaries (Bombaci et al. 2018). Ecosanctuaries have been defined as areas greater than 25 ha within which mammalian pests are proactively removed and native species are deliberately introduced to allow for the recovery of these species (Innes et al. 2019). Although a species-focused conservation approach has historically been the focus in New Zealand, mammal exclusion using fences can also help to restore biodiversity in a wider ecosystem context. For example, when possums and rats were near eradicated in central North Island forests to protect populations of North Island Kokako (Callaeas cinerea wilsoni Forster), not only did Kokako populations recover, but improvements in other populations of native biota including insects and plants were observed (Saunders & Norton 2001;Burns et al. 2012).
While the vertebrates and flora have been monitored since the installation of the Orokonui fence, the invertebrate communities have not been monitored or manipulated (via deliberate introductions or removals). Although we cannot 'rewind the tape' to compare invertebrate diversity and abundance before and after the installation of the fence, we can compare the abundance and diversity of terrestrial litter-and soil-dwelling beetles inside (mammals removed, endemic bird/reptile species maintained) and outside (mammals persist, endemic vertebrate species unprotected) the Orokonui Ecosanctuary. To evaluate whether beetle populations have been affected by the installation of pest-resistant fencing that excludes mammals and shelters protected native species, our first objective was to compare the abundance, species diversity, functional diversity, and body size variation of soil-and litter-dwelling beetles inside and outside the fence. Second, since about 90% of beetle species in New Zealand are endemic (Klimaszewski 1997), we compared the abundances of native and non-native beetles inside and outside the fence. Finally, we considered seasonal effects on beetle populations by sampling in both winter and summer.

Choosing sampling sites
All sampling was performed at Orokonui, Waitati, in the city of Dunedin (Figure 1). Plant species composition can affect invertebrate composition and distribution (Gardner 1991;Sanderson et al. 1995). Therefore, we identified pairs of sites, inside and outside the fence of the Orokonui Ecosanctuary, based on vegetation similarity using a cluster analysis. We used plant species presence/ absence data collected between 2013 and 2014 from permanently marked 10 m × 10 m sites inside (n = 50) and outside (n = 18) the fence (Tanentzap & Lloyd 2017). The cluster analysis was conducted in R version 3.5.1 (R Core Team 2013). Dissimilarity data were produced using Euclidean distance as the distance measure and were clustered using the 'hclust' function ('average' linkage) from the package 'stats' (R Core Team, 2013). Average linkage dissimilarity data were plotted into a cluster dendrogram which highlighted two pairs of sites that contained plant species composition more similar to each other than to any other sites ( Figure 2). Both of these site pairs were dominated by forest and located at the southern end of Orokonui (Figure 3(b)). Site pairs are referred to as 'Raukaua pair' and 'Hebe/Ulex pair' based on the woody species unique to each pair (Table 1). To contrast the forest pairs, we also selected a pair of sites dominated by grasses  and sedges (graminoids) (referred to as 'Grass pair'). Each of these grass sites was located ∼ 1-3 m from the fence near the northern end of Orokonui and was approximately 2 m × 10 m in extent (Figure 3(a); reduced dimensions due to available area for sampling).
Although we did not measure vertebrate activity at our sites, we did witness evidence of mammal presence at sites outside the fence (e.g. plants that had been heavily browsed, cloven hoof tracks, feral pig faeces, and other ungulate faeces).

Invertebrate sampling
Invertebrates were collected from turf samples in winter (May-July) and summer (November-December) 2018. Within each site, a 15 cm × 15 cm square quadrat was blindly thrown from a central position to select 8 turf sampling points in forest sites, and 6 turf sampling points in grass sites. If the quadrat landed on a steep slope (∼30 o and above) or if there were obstacles (large roots and/or rocks), the quadrat was moved to the closest suitable area for sample collection. A sharp spade was used to cut 4 cm deep around the 15 cm × 15 cm square quadrat. This sample together with the surface litter was placed into a labelled paper bag and taken to Invermay Research Centre, Mosgiel (hereafter, Invermay) on the day of collection. Turf samples were kept at 4°C for 1-2 days before heat-extracting invertebrates in modified Tullgren funnels for 7 days (as described in Crook et al. 2004). Each sample was inverted and placed on 4 cm × 2 cm mesh over a funnel, with a 150-watt bulb suspended approximately 30-35 cm above the samples. Invertebrates escaping from the warming, drying effect of the heat source were collected into containers containing approximately 250 ml of monopropylene glycol placed below the funnels. Samples were then sieved through a fine cloth mesh (capable of catching invertebrates larger than 0.2 mm in length), rinsed in tap water, and stored in 70% ethanol. All adult beetles were then extracted from each turf sample, sorted from the other invertebrates under a dissecting microscope, counted, and separately stored in 70% ethanol (See Supplemental File 1, which also includes data on beetle larvae). The remaining invertebrates are reported separately (Chen 2020).
Beetles were pinned and identified to species level where possible (Table 2). If a specimen could not be identified to species, they were assigned as morphospecies (Oliver & Beattie 1993;Barratt et al. 2003). To assist in identifying some of the weevil species (Curculionidae), we dissected out and examined genitalia, which were then kept with their associated specimen. Beetle species were sorted into one of three categories indicating their trophic guild: 'Carnivore', 'Herbivore', or 'Fungivore/Decomposer' based on the predominant category at the family level as recorded in the literature (Hammond 1997;Klimaszewski & Watt 1997;Majka 2006;Barratt et al. 2009). For example, although some fungivorous rove beetles (Staphylinidae) do exist (Lipkow & Betz 2005), rove beetles are predominantly predators (Bohac 1999); therefore, all rove beetle morphospecies were categorised as carnivores for simplicity.
The body length of each pinned beetle specimen was measured using a ruler from the front of the head (rostrum and mouthparts excluded) to the posterior tip of the abdomen. Body length measurements were rounded to the nearest 0.5 mm for use in size distribution analysis. Beetles were then categorised as small (1-2.5 mm); medium (3-6 mm); or large (>6.5 mm).

Soil moisture
The assemblages of many soil-and litter-dwelling invertebrates, including beetles, can be influenced by the moisture levels of the soil (Kelly 2000; Antvogel & Bonn 2008; Brygadyrenko 2014; Tsafack Table 1. Plant species exclusive to each pair of forest sites. (a) The Raukaua pair included sites #1 (inside the fence) and #75 (outside the fence). (b) The Hebe/Ulex pair included sites #9 (inside the fence) and #70 (outside the fence). Site numbers are highlighted in Figure 2. Site data from Tanentzap & Lloyd (2017 Table 2. List of beetle species and morphospecies identified for this study, including their '# in' and '# out' (the number of individuals that were found inside and outside the fence throughout the entire study), and 'status' (Native, Exotic, or Unknown). before a genus name has been assigned by BIPB in the absence of a confirmed reference specimen. To investigate this, approximately 250 ml of soil was collected from directly beneath the turf samples in the field, hand-sorted to remove macroinvertebrates (which were identified and then returned to the site) and stored in plastic Ziplock bags before being transported to Invermay. To calculate soil moisture, soil was weighed on a digital scale on the same day it was collected in the field (wet weight) and then placed in drying ovens at 65°C for 28 h, after which the soil was re-weighed (dry weight). For each sample, we calculated the proportion of moisture in the soil as: Wet weight − Dry weight Wet weight

Data analyses
Beetle Abundance: Beetle abundance data were analysed using a generalised linear mixed model (GLM) in R version 3.5.1 with the 'glm()' function. The family was specified as 'poisson' because beetle abundances are count data. The chosen model was selected using backwards selection based on having the lowest Akaike information criterion (AIC). The response variable was beetle abundance, and the three predictor variables were 'fence' (inside, outside), 'season' (winter, summer), and soil moisture (percent moisture). Interaction terms and site pair (e.g. vegetation type) were discarded during model selection due to lack of significance and the models being a poorer fit based on AIC. Species, Size, and Trophic Diversity: Beetle data from the samples per site collected within a season were combined, and the soil moisture data were averaged accordingly.
Species richness was calculated as the total number of different species and morphospecies per site per season. Shannon-Wiener diversity indices (H) of the species richness data were calculated using the 'diversity()' function from the 'vegan' package (Oksanen et al. 2019) in R. These entropic diversity indices which have non-linear relationships with species richness, were then transformed into a linear relationship ('effective Shannon diversity' eH, known as Hill numbers) by using the function 'exp()' (which computes exponential values). By transforming the entropic diversity indices (H) into effective diversity numbers (eH), the values of diversity are more intuitive (Jost 2006).
To get the beetle functional diversity values, two functional traits were included: trophic guild and body length category (small: 1-2.5 mm; medium: 3-6 mm; or large: >6.5 mm). Functional diversity values can be described as a measurement of the diversity and abundance of functional groups present in a population where a low functional richness indicates that there may be unused resources and niches in the ecosystem (Mason et al. 2005;Schleuter et al. 2010). 'Rao's quadratic entropy' values were calculated for each site per season in R using the function 'dbFD()' from the package 'vegan' (Oksanen et al. 2019). Rao's quadratic entropy measures functional diversity, but unlike most other measures of functional diversity, it considers within-group differences and species abundance instead of just the number of functional groups (Rao & Nayak 1985;Botta-Dukát 2005).
To compare similarities in presence/absence of beetle species across the sites and seasons, cluster analysis was performed on beetle species assemblages. Dissimilarity data were produced using Euclidean distance as the distance measure using the 'dist()' function and then clustered using the 'hclust()' function ('complete' linkage) from the package 'stats' (R Core Team, 2013). In this case 'complete' linkage was used over 'average' linkage as it produced tighter clusters that were easier to interpret when plotted as a dendrogram.
To determine whether mammal exclusion at Orokonui might affect the distribution of beetle sizes (Cole et al. 2006;Russell et al. 2020), beetle length data (mm, continuous) were analysed using a generalised linear model (GLM) in R using the 'glm()' function with the family specified as 'gamma' as the data were right-skewed. The model was selected using backwards selection based on having the lowest AIC. The model included three predictor variables which were 'fence' (inside, outside), 'native status' (native, exotic, unknown), and 'trophic guild' (carnivore, herbivore, fungivore). Season and soil moisture were dropped from the model during model selection due lack of significance and contributing to higher AIC.
To see if functional assemblage differs across sites, contingency tables with beetle trophic guild data (herbivore, carnivore, fungivore) against 'fence' (inside, outside) for all pairs combined were analysed using a Pearson's Chi-square test of independence using the 'chisq.test()' function. This analysis was also performed on subsets of the data for each of the three pairs (Raukaua, Hebe/ Ulex, and Grass) separately.

Beetle abundance
There was a sum of 276 beetles counted in all 88 samples. The mean number of beetles per sample was 3.2, with a minimum of 0 and a maximum of 17. Thirteen samples contained no beetles, and 25 samples contained one beetle. Beetle abundance was higher inside the fence compared to outside the fence (SE = 0.12, z = −2.98, df = 1, P = 0.003); higher in winter compared to summer (SE = 0.13, z = 4.85, df = 1, P < 0.0001); and increased with increasing soil moisture (SE = 1.33, z = 3.52, df = 1, P = 0.0004). The inside site of the Grass pair in winter showed the highest beetle abundance despite having almost the lowest soil moisture compared to all other sites (Figure 4).

Species and functional diversity
Of the 276 beetles found in the samples, 261 were sufficiently intact for identification (see Supplemental File 2 for example photographs of all species and morphospecies). We identified 17 species and 46 morphospecies. Of the 63 species and morphospecies, 31 were found exclusively inside the fence, 15 were found exclusively outside the fence, and 17 were found both inside and outside the fence (Table 2). Fourteen species and 25 morphospecies were identified as native to New Zealand. Three species were exotic, while the origin of the remaining 21 morphospecies could not be determined (Table 3). The majority of morphospecies whose origins could not be identified were rove beetles (Staphylinidae). Weevils and rove beetles were the most abundant ( Table 2). The three exotic beetle species were found only in winter and at two sites: Hebe/Ulexinside (Aridius bafasciata (Reitter), n = 3; Coccinella 11-punctata (Linnaeus), n = 1), and Grass-outside (Listronotus bonariensis (Kuschel), n = 3).
Effective Shannon diversity and Rao's quadratic entropy values varied across sites and seasons, except for the grass sites where diversity was higher inside the fence for both seasons (Table 3).
Based on the cluster analyses, the beetle species composition for the two forest sites inside the fence were more similar to one another compared to the two forest sites outside the fence. The grass site outside the fence was an outlier in terms of beetle species composition with the grass site inside the fence being more similar to the forest sites than it was to the grass site outside of the fence (Figure 5).

Trophic diversity
We found no evidence to suggest that beetles belonging to a particular trophic guild varied overall inside vs outside the fence (X 2 = 3.70, df = 2, p = 0.16). When data were analysed for each pair of sites separately, disproportionately more herbivorous beetles were observed inside the fence for the Hebe/Ulex subset (X 2 = 6.15, df = 2, p = 0.046), and more herbivores and fungivores were observed inside the fence for the Grass pair (X 2 = 9.17, df = 2, p = 0.01). The distribution of trophic guilds was relatively similar inside and outside the fence for the Raukaua pair (X 2 = 4.03, df = 2, p = 0.13). Notably, the number of herbivorous beetles was always lower outside the fence compared to inside the fence (Table 4).

Discussion
The purpose of the pest-resistant fence of the Orokonui Ecosanctuary is to exclude non-native pest mammals, as well as enclose and hence protect threatened native New Zealand flora and fauna, which can lead to recovery of the enclosed ecosystem and surrounding landscape (Bogisch et al. 2016;Kitchin et al. 2017;Tanentzap & Lloyd 2017). The present study has shown that soil and litter beetle abundance, diversity, and community structure does vary inside and outside of the ecosanctuary, however, the extent of this difference and the direction of the effect could depend on the vegetation present in the habitat. Overall, beetle abundance was significantly higher inside the fence, and this effect is most evident in the grass compared to the forest sites. In previous studies, using pitfall traps instead of soil/litter samples, beetle abundance was not affected by intensive mammal control, and did not differ inside versus outside the fence (Watts et al. 2014;Vergara et al. 2021). However, Vergara et al. (2021) found higher abundance of rove beetles outside the fence, whereas we found rove beetle abundance was similar inside (n = 43) and outside (n = 44) the fence. On the other hand, pitfall studies found higher abundance of endemic ground beetles (Ctenognathus adamsi) inside compared to outside  predator-proof fencing (Watts et al. 2020;Vergara et al. 2021). While we did not find any C. adamsi, we did find six ground beetles (5 different species) inside the fence and only one ground beetle outside the fence in our study at Orokonui Ecosanctuary. Ground beetles are well studied, readily identified to species level, and are responsive to ecosystem change, so further work looking into ground beetles inside and outside Orokonui could prove insightful (Vandewalle et al. 2010). Furthermore, ground beetles (among other beetles) are often present in the diets of native vertebrates such as kiwi and tuatara, and we know that tuatara translocated into Orokonui feed on ground beetles based on scat analyses (Kitchin et al. 2017).
We found more herbivorous beetles inside the fence for all three vegetation pairs, and this could be partially caused by herbivorous mammals changing the structure of the beetles' host plants outside the fence. During field work, extensive browsing damage to plants was observed at the outside forest sites. Indeed, Shimazaki & Miyashita (2002) found that herbivorous insects that feed on the same plants as deer are found in lower abundance compared to deer exclusion zones in Japan (Shimazaki & Miyashita 2002). Other studies though have shown an increase in herbivorous insect abundance in regions where they overlap with competing herbivorous mammals (Danell & Huss-Danell 1985;Martinsen et al. 1998). Therefore, the presence/absence of competitors is unlikely to fully explain abundance of herbivorous insects.
There was a higher abundance of carnivorous beetles at grass sites, inside and outside, compared to forest sites and 95% of these were rove beetles from the family Aleocharinae. This could be because these rove beetles are generalist active predators that can utilise a wide range of habitats, and the vegetation present at the grass sites presented fewer opportunities for a higher abundance and diversity of herbivorous beetles (Lassau et al. 2005).
Fungivorous beetles were most abundant at the grass site inside the fence, most of them being featherwing beetles of the family Ptiliidae. It is interesting to note that despite the inside and outside grass sites being just a few metres from one another and divided only by a fence, only one fungivorous beetle from the family Leiodidae was found outside the fence, and 20 (14 of which were featherwing beetles) were found inside the fence. Little is known about the featherwing beetles owing to their small size and cryptic behaviours (Majka & Sörensson 2007).
Species and functional diversity values were generally higher at sites inside the fence, except for the Raukaua sites where species diversity was almost the same at both inside and outside sites, and functional diversity outside the fence was either almost the same or higher compared to inside the fence. This could be because the Raukaua sites were dominated by weevils, which are herbivorous. It is possible that the mostly-herbivorous beetle community assemblages at the Raukaua sites respond differently to vertebrate pressures compared to the beetle assemblages at the other sites due to the different niche requirements of each community (Schratzberger & Warwick 1999).
Six weevils from the genus Etheophanus were notably only found at Raukaua sites (both inside and outside the fence). Little is known about the biology and habitat preferences of Etheophanus other than that they are associated with leaf litter (Davis et al. 2019). Also, although we collected 48 specimens classified into nine morphospecies of Aleocharines (rove beetles), this subfamily was notably absent from Hebe/Ulex sites, but present at both the Raukaua and Grass sites. The Aleocharinae is a poorly studied subfamily of rove beetles in New Zealand, even though it is very speciose with at least 16,200 species worldwide with some subgroups being habitat specialists (Leschen & Newton 2015). In New Zealand, Ulex is an invasive nitrogen-fixing legume that can increase nitrogen concentrations in forest litter and stream reaches enough to alter ecosystem functioning (Magesan et al. 2012;Stewart et al. 2019). Despite being invasive and capable of altering nitrogen levels, Ulex is also known to provide protective habitat for some native New Zealand insects (Harris et al. 2004, Watts et al. 2012. Thus, the presence of Ulex in the Hebe/Ulex sites could be a factor accounting for the absence of Etheophanus or Aleocharinae because of its ability to alter ecosystems. These findings could represent opportunities for future studies focused on these beetle groups, and if the presence/absence of these flora affects their distribution. Of the beetle species/morphospecies that we could identify as either native or exotic in this study, more than 90% were native, a statistic that matches the percentage of native beetle taxa for all of New Zealand as reported by Klimaszewski (1997). Of the 63 described species/morphospecies in our study, three (4.8%) were classified as exotic (Aridius bafasciata (Reitter), Coccinella 11-punctata Linnaeus, Listronotus bonariensis (Kuschel)) and found at an inside site (Hebe/Ulex) and an outside site (Grass) at Orokonui. Similarly, in the fenced sanctuary Zealandia (Wellington, NZ), of the 262 described beetle species caught in pitfall traps over the course of ten years (including the two years before pest mammal eradication), only three (1.1%) were classified as exotic (Sericoderus sp., Hylastes ater (Paykull), Hylurgus ligniperda (Fabr.); Watts et al. 2014). Even in non-fenced areas dominated by European gorse (Ulex europaeus) in a catchment on the Wellington South Coast, over 80% of the beetles collected were classified as native (Crisp et al. 1998). Unfortunately, we were unable to determine the native/non-native status of the rove beetles collected, which represented about a third of the species. This family of beetles is quite diverse, and subject to ongoing taxonomic revision (for examples, see: Park & Carlton 2014;Shen et al. 2020), making species identifications difficult.
Beetle community composition was most similar in both inside forest sites compared to both outside forest sites, even though their woody plant composition differed, which points to mammal exclusion influencing beetle community structure more so than plant composition. Our data collected at Orokonui Ecosanctuary were similar to that collected at Zealandia Ecosanctuary, that the pest-resistant fence had a stronger effect on beetle community structure than it did on abundance, species diversity, and native dominance (Watts et al. 2014). The change in predation pressure over time because of intensive mammal control (i.e. fewer mammals resulting in more insectivorous reptiles and birds) inside pest-resistant fenced ecosanctuaries can affect beetle communities (Watts et al. 2014) while the pressures of mammals outside these sanctuaries are also affecting beetle communities in different ways (e.g. Ueda et al. 2008;Kaltsas et al. 2013;Iida et al. 2016).
It is possible that because a diverse array of mammals has been present on the New Zealand mainland for the past 150 years, the beetles we observe inside or outside ecosanctuary fences might already be a non-random mammal-resistant subset of New Zealand's endemic beetle fauna (Watts et al. 2014). This selection could also explain why beetle size distribution did not vary inside vs outside the fence, both in this study and in others (Watts et al. 2014), despite evidence that selective predation by rodents can alter the size distribution of New Zealand beetles on islands (Chown & Smith 1993;Marris 2000;Russell et al. 2020). It may also be that 11 years is not enough time for beetle populations to respond to the removal of mammals (Rufaut & Clearwater 1998;Towns et al. 2009).
We used available presence-absence plant data to select sites for this study, but future studies could incorporate plant species abundance and structure to study their effects on invertebrate communities inside and outside of ecosanctuaries, and if their effects interact with pest mammal control (Angelini et al. 2011;Crawford & Rudgers 2013). Previous studies also have measured vertebrate activity with methods such as tracking tunnels and scat counts that could be incorporated into future studies on invertebrate diversity at Orokonui to further isolate the mammal exclusion effects of its pest-resistant fence (Ruscoe et al. 2001;Blackwell et al. 2002;Forsyth 2005;Wilson et al. 2007;Acevedo et al. 2008;Engeman et al. 2013;Vergara et al. 2021).

Conclusions
Evidence from this study suggests that the control of animal movement using pest-resistant fencing at Orokonui can affect soil-and litter-dwelling beetle abundance, diversity, and community composition. Moreover, we show that the differences in beetle community composition inside and outside the fence can also depend on the vegetation composition of the habitat and the season of sampling. This information could be important for conservation efforts in New Zealand. This study only considered large differences between a few sites from data collected within one year. Drawing conclusions on specific causes and effects are difficult here as the abiotic and biotic factors that can affect insects in ecosystems are complex, and mammal exclusion fences can alter a variety of bottom-up and top-down pressures on beetle communities. This study has documented the dominance and diversity of native species in an around the Orokonui Ecosanctuary and thus provides an important baseline and reference collection for future research on ecological restoration in these reserves.
Beetle specimens from this study can be accessed (by request) from the Otago Museum Collections (https://otagomuseum.nz/collections/collections/).