An insect invasion of Antarctica: the past, present and future distribution of Eretmoptera murphyi (Diptera, Chironomidae) on Signy Island

Rising human activity in Antarctica, combined with continued warming of the polar climate, means the risk of non‐native terrestrial species colonising and establishing in its biodiversity‐ and nutrient‐poor ecosystems is increasing. Of the five non‐native invertebrate species currently in terrestrial Antarctica, the flightless midge Eretmoptera murphyi (Schaeffer, The Museum of the Brooklyn Institute of Arts and Sciences 2:90–94, 1914) is perhaps the most persistent insect invader. Accidentally introduced to Signy Island (60°S) in the 1960s from sub‐Antarctic South Georgia (54°S), E. murphyi has steadily increased its distribution, however, its status has not been reassessed for a decade. Here, we update the distribution of E. murphyi on Signy, specifically assessing whether footpaths to regularly visited research sites represent dispersal corridors. Our findings show that both the abundance and range of E. murphyi have increased significantly since 2009, particularly along paths leading away from the original introduction site, and that the species is now on the cusp of moving into new valley systems. We identify a moderate association with soil/substrate and vegetation types and build Maximum Entropy (MaxEnt) models to predict areas of the island that may be at highest risk of future colonisation. As a detritivore with no competitors or predators, E. murphyi may have a major impact. For example, accelerating nutrient cycling which may have wider impacts on all levels of biodiversity. This study highlights the need for an assessment of current biosecurity protocols applied within the Antarctic Treaty system, as well as the need for systematic regular monitoring of introduced and invasive species in Antarctica.


Introduction
The Antarctic region is the least invaded area in the world, in part thanks to its geographic isolation, harsh climate and limited history of human activity. As a result, few terrestrial species have established naturally in the area south of 60 S latitude since the last glacial maximum (Frenot et al., 2005;Hughes et al., 2015). However, increasing human activity in the region in recent decadesprimarily in the forms of tourism and scientific explorationhas diminished the isolation of Antarctica, particularly within the maritime Antarctic, a region encompassing much of the Antarctic Peninsula and the archipelagos of the Scotia Arc. Most recent arrivals of non-indigenous species in the Antarctic region have occurred because of accidental introductions by humans and, as our footprint has increased (Pertierra et al., 2016), so records of non-indigenous species becoming established have risen, particularly within the South Shetland Islands and northern Antarctic Peninsula (e.g. Greenslade et al., 2012;Volonterio et al., 2013;Hughes et al., 2015;Molina-Montenegro et al., 2012). The few terrestrial introductions that have occurred to date above 60 S latitude have been in, or close to, research stations (i.e. Molina-Montenegro et al., 2010). For example, the introduction of Lycoriella sp. (Diptera, Sciaridae), via imported fresh vegetables, to Casey Station on the eastern continental Antarctic (Anon, 2002). Further colonisation of non-indigenous species is likely, given the continuing increase in activity around Antarctica as new research stations are established, and land-based exploration as well as tourism expands. Establishment is then facilitated by predicted changes that an altered and ameliorated climate may bring (Convey, 2011;McGeoch et al., 2015;Turney et al., 2016). Understanding the biology of established non-indigenous species and monitoring their distributions will be key to understanding how best to control and mitigate this challenge. Only with deeper knowledge of species' physiology, life history, dispersal mechanisms and rate of population expansion, can management and governance processes in the Antarctic be effective in preventing further introductions, as well as minimising the risks of species invasions that have already occurred Hughes & Pertierra, 2016).
The brachypterous midge Eretmoptera murphyi (Chironomidae, Orthocladiinae) (Schaeffer, 1914) is an established nonindigenous species on Signy Island (South Orkney Islands, maritime Antarctic, 60 43'0"S,45 36 0 0"W). Native to sub-Antarctic South Georgia, where it is paleoendemic (Allegrucci et al., 2012), E. murphyi is thought to have been accidentally introduced to Signy during plant transplant experiments conducted in the 1960s (Burn, 1982;Block et al., 1984). Since its initial discovery on Signy in 1980, it has progressively expanded its range from the original introduction site immediately adjacent to the research station (Burn, 1982;Block et al., 1984;Smith, 1996;Dózsa-Farkas & Convey, 1997;Hughes & Worland, 2010). At some sites, the midge is now found at densities exceeding all other native micro-arthropod fauna combined and, consequently, is influencing litter turnover, with implications for the terrestrial ecosystem and native biodiversity (Hughes et al., 2013). The species is currently considered a persistent alien in the absence of direct evidence of impacts on native species, which would lead to it being classified as an invasive alien (Frenot et al., 2005).
Eretmoptera murphyi has several physiological traits that have allowed it to succeed in the more extreme maritime Antarctic in comparison to the milder climate of South Georgia. Its larvae have appropriate cold tolerance and can rapidly cold-harden (Block et al., 1984;Everatt et al., 2012;Worland, 2010), as well as an ability to respire in water and withstand ice entrapment . Both eggs and larvae are also desiccation tolerant Bartlett et al., 2018a). The species has a 2-year life cycle and is parthenogeneticas a result, it does not require synchronous adult emergence and subsequent mating (Bartlett et al., 2018b). This contrasts with its closest relative, the sexually reproducing chironomid and only higher insect endemic to Antarctica, Belgica antarctica Orthocladiinae) (Jacobs, 1900, Sugg et al., 1983. This means that E. murphyi can have an extended emergence period across the summer season (Bartlett et al., 2018b), making the most of environmental windows of suitability, and potentially using this flexibility to increase its distribution range via the more mobile, and more-readily wind dispersed, surface-dwelling adults.
Studies of the distribution and abundance of E. murphyi on Signy have only recently begun in earnest, with the benchmark survey of Hughes and Worland (2010), which took place in 2007-2009, c. 40 years after the suspected introduction, and 25 years following its initial discovery (Burn, 1982). Hughes and Worland (2010) provided the first comprehensive assessment of the spread of E. murphyi around its original introduction site and found densities as high as 150 000 larvae m −2 (Lm −2 ), with evidence of the midge being more prevalent along paths leading away from this area. These data suggested that although flightless, adults may be capable of dispersal to some extent. However, it is the mechanical process of human footfall and disturbance of larvae in the soil that could be a key factor facilitating its spreadas has also been noted for nonindigenous plant species in Antarctica (Molina-Montenegro et al., 2012). While Hughes and Worland (2010) did not identify the limit of the species' distribution, their study confirmed that E. murphyi had spread to cover an area of at least 35 000 m 2 , doubling the previous estimate made in 1995(Dozsa-Farkas & Convey, 1997. The survey noted a highly patchy distribution pattern, with E. murphyi typically associated with dead organic matter, rather than live moss or inorganic substrates such as gravel and rock. This is consistent with E. murphyi being a detritivore (Cranston 1985;Hughes et al., 2013). However, as yet, no robust assessment has been made of the availability and extent of suitable habitats for E. murphyi around the island, or how species abundance changes across different substrate types, i.e. evidence of substrate preference.
Species distributions and relationships with associated habitats have typically been analysed using correlative models that use linear and/or logistical regressions to establish the main drivers of species abundance (Wiens et al., 2009;Porfirio et al., 2014). Developments of open source software, using geographical information, now allow ecologists to use known species presence (and absence) data in combination with environmental niche predictors to more accurately model likely changes in distribution (Rushton et al., 2004;Raghavan et al., 2016). Such models can aid understanding of species responses to climate change (e.g. Wang et al., 2018), help assess extinction rates (e.g. Thomas et al., 2004), identify habitat loss scenarios (e.g. Angelieri et al., 2016), and prioritise biological conservation efforts (Pyke et al., 2005). They can also be used to forecast suitable areas for the spread of an invasive species (West et al., 2016;Pertierra et al., 2016) through estimations of probability of presence given a series of biotic and/or abiotic variables such as terrain, habitat classification, temperature or dispersal vectors.
Antarctica is a globally important reservoir of unique ecology and biodiversity, and Signy Island provides the best example of terrestrial biodiversity in the region. As a non-native detritivore in such high abundance, E. murphyi may alter the Signy Island ecosystem, and areas beyond if given the opportunity to disperse. Few studies to date have examined Antarctic invertebrate invasions in depth, with no regular monitoring of long-term abundance and distribution changes of those species that have colonised. In this study, we map the current distribution of E. murphyi across Signy Island and evaluate the role of footpaths as dispersal corridors. We establish its association with preferred environments through physical study of its habitat and, using Maximum Entropy (MaxEnt) species distribution modelling, highlight areas at risk of further colonisation across Signy Island.

Materials and methods
Signy Island is located in the South Orkney Islands archipelago in the maritime Antarctic, (60 43'S, 45 36'W; Figs. 1 and 2). The island is 6 × 5 km in size, with a maximum elevation of 288 m across a small mountain range that holds its large ice cap. Signy experiences slightly positive temperatures for most of the austral summer (December to March) with monthly averages of +1 to 3 C. During winter, the island is surrounded by sea ice extending from the Antarctic continent and temperatures remain far below freezing, with a winter average of −10 C. Rising temperatures throughout the Antarctic Scotia Arc region mean that the island is de-glaciating and consequently much ice-free land is experiencing primary succession.

Extent of E. murphyi distribution
Distribution surveys were primarily conducted along frequently used paths radiating from the British Antarctic Survey (BAS) research station on Signy. In total, 14 stop-sites were sampled along the High-tide (HT) and Stonechute (SC) paths ( Fig. 2), with soil cores taken at 1 m and 5 m distances perpendicular to the path, every 20 m AE2 m (measured using a tape reel). Where the path broadened away from the sea, samples were taken on either side of it at both 1 m and 5 m, respectively. Otherwise, all samples were taken to the inland side of the path, with the final stop site at the summit of the SC. Fourteen stop sites were also sampled along the Backslope (BS) path (Figs. 1 and 2), following the same procedure, and sampling on both sides of the path, with the final stop site at the summit plateau between Factory and Observation Bluffs. Whole island survey routes targeted potential habitat of E. murphyi (moss patches or banks) at c. 50 m intervals (closer if habitat irregular), taking cores from each site (Fig. 2). These routes followed the frequently used paths from the top of the SC to the huts on the Gourlay Peninsula (GP trail) where regular penguin colony monitoring has been conducted for over 30 years (14 sites sampled) and from the top of the SC to the cemetery flats (CF) crossing, which is the main access route to freshwater lakes and other parts of the north and west of the island (7 sites sampled). Further routes were surveyed from Waterpipe Hut (WH) over Jane Col to Foca Hut (FH) on the west coast of the island (16 sites sampled), and from North Point (NP), where penguin and bird survey work is currently conducted several times a season, over Spindrift Col and down to WH (Fig. 2, 10 sites sampled). These latter routes reflect historically well-used paths around the island. Focus was given to sampling close to the field huts established on the island.
In addition to the survey routes, a grid of points 100 m apart was plotted over the entire area adjacent to the research station, covering an area of 490 000 m 2 , in order to establish the extent of E. murphyi's distribution (Fig. 4). This edge of extent grid incorporated Berntsen Point at its northern-most point, extended several meters beyond the summit of the SC path at its southern-most point, Observation Bluff and Gash Cove to the east, and as far as the gully between Mooring and Knife Points to the west. As previously, soil cores were sampled at each grid point and analysed for presence of E. murphyi.

Sample collection and processing
Soil cores (n = 3) were taken from each designated sample point, using a steel 2.5 cm ø, 10 -cm long soil corer, with all cores  larvae that were visible by eye (instars 2-4, L2-L4) and egg sacs were extracted from the substrate by washing through stacked sieves (2, 0.5, 0.25 mm mesh sizes) and hand-picked from the remaining soil on the 0.25 mm sieve, with the 0.5 mm sieve checked for any further individuals. L1 larvae could not be included in this survey due to their very small size. Prior to washing, any clumps of moss or peat substrates were teased apart with tweezers, freeing any individuals that may have been amongst the fibres.
The constituent parts of each core's upper and lower dominant substrate components were noted for subsequent correlation analysis with E. murphyi abundance. Substrates were divided into 10 sub-types, six for the upper substrates and four for the lower (Table 1). Larval densities were expressed as the mean for each set of three cores. Egg sacs were also included, with their larval potential calculated using known egg hatching success rates (cf. Bartlett et al., 2018a) (Eqn. 1). Densities per metre square were estimated by calculating the surface area of the corer and scaling up (cf. Hughes & Worland, 2010), This gives a multiplier of 1949 to apply to the combined counts of larvae (L) and larvae from egg sacs (E L ) to obtain a count of larvae per m 2 (Lm −2 ).
where E L is the calculation of larvae produced from egg sacs, E g is the mean egg sac count for three cores, E s is the mean eggs per egg sac (48*), and E h is the % eggs that hatch (based on 35% successful hatch rate*) *after Bartlett et al. (2018a).

Statistical Analysis and Data Visualisation
In order to determine any correlation between substrate type (see Table 1) and the abundance of E. murphyi, two regression methods were employed. Firstly, mean abundance per m 2 (for BS and HT/SC paths) were Log 10 transformed and plotted against the binary presence/absence data for each substrate type and then investigated for significant substrate types associated with E. murphyi abundance using an ordinal logit regression (OLR) (XLStat, verified with SPSS) by categorising the population abundance per m 2 into the following categories: zero, low (1-1000), medium (1001-10 000), and high (10 001-100 000). Whilst this provided a good model on what were the significant relationships, further detail was required, so a general linear model (GLM) was run on the Log 10 transformed data (XLStat) to obtain a true R 2 value for substrate typeabundance relationships. Upper inorganic (Ui) substrate had zero observations, whilst upper vascular (Uv) and upper algae (Ua) had just one each, so were removed from the subsequent analyses. Both methods produced the same significant substrate type abundance relationship, validating one another and the decision to take the significant substrates forward into the maximum entropy model.

MaxEnt Modelling for Whole Island Colonisation Risk
Maximum entropy species distribution modelling (MaxEnt) (Phillips et al., 2006) is a common technique used to model a species' potential distribution based on the known geographic distribution and associated environmental background information. Its application here allows identification of similar areas to those already occupied by E. murphyi, indicating the relative potential for establishment. MaxEnt v3.4.1 (http://biodiversityinformatics.amnh.org/open_source/maxent/) was used to model colonisation risk for areas of the island with input layers created as follows: using a digital elevation model (DEM) of Signy Island, GIS raster layers with a cell size of 5 m 2 were generated for aspect, slope and altitude. Using this topographical information, the programme output identified preferential habitats for E. murphyi presence across the island, based on the locations of its current distribution. This layer of topographical preference for E. murphyi occurrence was then combined with other raster layers to account for human activity around the island; preferred substrate type (i.e. findings from abundancesubstrate-type correlations set out above) and distance from the current distribution of E. murphyi. All raster layers were created in either ArcMap 10.4.1 or QGIS v2.18 Las Palmas (QGIS Development Team 2016. QGIS Geographic Information System. Open Source Geospatial Foundation Project. http://qgis. osgeo.org). Human footprint was classified into nine categories from very high to infrequent/nil (Supporting Information Online Resource 1), based on a subjective assessment of the level of activity in an area in any one season on the island, derived from discussions with, and observations of Signy Island research personnel during the 2016/2017 season. The distance layer was created using Euclidean distance interpolation, radiating from the area of current E. murphyi presence as defined in this study. The substrate type layer was constructed based on the results of the OLR/GLM modelling and presence/absence of the substrates preferred by E. murphyi. Terrain models selected environments of similar aspect, slope and altitude to those where E. murphyi is verified to be present in order to predict further at risk areas. As true absences were also recorded during the whole island and edge of extent surveys, it was not necessary to generate pseudo-absences for the model. The final distribution map was then created in ArcMap 10.4.1, with added landscape features and sites of significance. Areas of the island currently covered with ice or permanent water bodies were discounted as areas at risk and masked out accordingly.

Results
Local distribution and abundance of E. murphyi Figure 3a shows that the distribution of E. murphyi now extends well beyond the immediate area of the research station (RS), and it can be found in densities of up to 12 000 Lm −2 close to the summit of the SC path, 240 m beyond the RS. Abundance lower down the path is equally high, with several points exceeding 10 000 Lm −2 . The highest densities were found on the BS path, reaching 98 241 Lm −2 at 40 m and exceeding 50 000 Lm −2 at 60 m.

Edge of distribution and whole island survey
The 100 × 100 m edge of extent grid (Fig. 4) found evidence of E. murphyi along both currently used and historic paths, with a small number of larvae found on a ridge south-east of the old study site path, over 300 m from the original introduction site. Surveys of other paths around the island found no evidence of E. murphyi presence (see Supporting Information Online Resource 2).  (table 1) differed significantly in frequency across collection sites (p < 0.001, Kruskal-Wallis). This was largely driven by the dominance of moss in the upper part of substrate cores (Um = 43%) and soil in the lower core (Ls = 30%). The most common habitats were those of moss banks: Um/Ls (55%) and Um/Lp (14%), whilst the least observed combinations (excluding Us/Lm and Up/Lm which were totally absent) were those of bare ground/peat (Table 2). Results for the OLR found that only soil and peat in the lower core (Ls and Lp, respectively) had a significant positive association with E. murphyi density (Ls, p = 0.032; Lp, p = 0.024) (Fig. 5). These two substrates explain 10% of density variability through a GLM (R 2 = 0.10, p = 0.011). Additional results from the GLM and associated correlation matrix indicated that Ls was the most significant of the two substrate types (p = 0.003). Further analysis of substrates along footpaths around the island, found suitable habitats (moss banks with soils and peat in the lower core: Um/Ls or Um/Lp,) for E. murphyi colonisation around Cemetery Flats, the Gourlay Peninsula and Three Lakes Valley in particular.

MaxEnt Modelling for Whole Island Colonisation Risk
Results from the MaxEnt model ( Fig. 6; Supporting Information Online Resource 3) indicate that, beyond the area of known distribution, the areas of highest overall risk are Berntsen Point, Observation Bluff, Cemetery Flats and the areas of Mooring and Knife Points, and the trail through Moraine Valley. Mediumhigh-risk areas include parts of the Gourlay Peninsula, Elephant Flats, Waterpipe Hut and Three Lakes Valley. Areas at a medium risk of establishment are largely found in patches around Moss Braes near North Point, parts of Skua Terrace, Jensen Ridge and Foca Cove, and areas close to Cummings Hut. Of the environmental variables taken from the DEM, aspect explained 65% of the known distribution envelope of E. murphyi, with slope 25% and altitude 10%.

Discussion
With the increasing risk of non-native species colonising Antarctic terrestrial habitats, there is an urgent need to regularly assess the rate at which successful invaders can extend their distribution, whilst also assessing their preference for available habitats in order to best forecast other areas that may be at risk. This study represents the most comprehensive survey of the flightless midge E. murphyi on Signy Island to date. This species has extended 85 000m 2 beyond its presumed introduction site in the 1960s, and we report distributions totalling 50 000 m 2 beyond the limit of the last survey in 2009 (Hughes & Worland, 2010) (Fig. 7). Thus, while its spread began slowly, it has clearly accelerated in recent years. The importance of human footfall along routes to regularly visited research sites is further supported as a likely means of dispersal.
This survey doubles the known range of E. murphyi since the previous survey conducted a decade ago. The highest population densities were found on the Backslope (BS) path immediately behind the Research Station (RS) (Fig. 3b), with a steady but small decline after this point with distance up the path. Densities along the high-tide (HT)/Stonechute (SC) path were patchier, but densities of >1000 Lm −2 persisted along this route. There was no discernible difference in the substrate types between the two paths, but the HT/SC path is used by personnel on a more frequent basis than the BS, which may account for the higher abundance further along this path.
The combination of methods used here, targeted surveying (path transects) and systematic surveying (extent grid), illustrates the value of both techniques, with the targeted survey highlighting details of density patterns, and the systematic grid locating outlying populations, as well as habitats where the species currently remains absent. We recommend that both methods are used in future monitoring of E. murphyi on Signy as a minimum requirement in assessing changes in abundance and distribution over the largest area possible. Such combined approaches to non-indigenous species monitoring are also recommended as best practice by Rew et al. (2006) in their comprehensive analysis of survey methods used for the monitoring of invasive plant species in the United States. Furthermore, they recommend that all monitoring of species distribution and abundance coincides with the collection of habitat data, such as we have demonstrated here in characterising substrate type and topography, to 'improve our understanding of factors affecting NIS (sic, non-indigenous species) occurrence, and to produce probability occurrence maps of target species'.
Regression analyses identified that, whilst live moss in the upper part of the sampled cores (Um) and soil in the lower part (Ls) were the most abundant substrates, it was the presence of Table 2. Combination of occurrences for each substrate type (see Table 1) in the upper and lower core, with the most positive (in bold) or most negative (italics) associations with E. murphyi abundance (R 2 ) shown along with related significance (*p < 0.01). See Table 1 (Table 1) as values of ordered logit coefficient, plotted in order of least to largest influence on E. murphyi abundance (AESEM). Substrate types testing significant through the OLR (p < 0.01) shown as solid fill circles. All variables >0 plotted. soil and peat in the lower part of the cores that had the strongest correlation to E. murphyi population density. This is consistent with E. murphyi being a detritivore. Hughes and Worland's (2010) data also support the negative association of E. murphyi with both live moss and inorganic substances such as stones and gravel in the lower core identified in this study. However, substrate type explained only 10% of variation in density, making it likely that other factors are primary drivers of population density and distribution. It also remains unclear if the association of E. murphyi with soil and peat is because of preference for this habitat, or because the midge is generating the soil substrate type and aiding decomposition of the existing moss and vegetation. Population densities over 12 000 Lm −2 close to the summit of the SC path, and of 1000 Lm −2 1 m from the path at the summit (Fig. 3b), suggest that this species could be on the verge of entering into new valley systems. This is of concern given the availability of prime habitat leading down to cemetery flats (CF). If able to establish in the CF area, the risk of wider spread will be increased by the seal populations there. Fur seals in particular are very mobile, disrupting the moss bank surface as they move (Favero-Longo et al., 2011), and there is the potential that larvae and eggs may be carried in their fur, particularly the egg sacs that have a sticky gelatinous outer membrane (Bartlett et al., 2018a). Results from the MaxEnt model suggest that combinations of suitable habitat and human footfall are frequent through the ice-free areas of the island, but that primarily the areas of CF and the Gourlay Peninsula (GP) are most at risk of further colonisation, largely because of their proximity to the existing distribution of E. murphyi.
Older and less used paths have no trace of E. murphyi on them at present (Supporting Information Online Resource 2) yet do have suitable habitats as identified by both MaxEnt and alongroute sampling. The large moss and peat banks of Skua Terrace and Moss Braes in particular (Fig. 2) present a large area of potential habitat (Cannone et al., 2017), especially where live moss does not extend too deeply. In addition, if larvae or eggs could survive a short low-tide crossing from CFwhere researchers (and seals) quickly move through shallow sea water and pools to access the northern part of the islandthen much of the area around Waterpipe Hut (WH) and Three Lakes Valley would also become accessible for colonisation, with Elephant Flats and soils directly surrounding WH at particular risk.  (Burns, 1982;Block et al., 1984); 1989 distribution~200 m 2 (Smith, 1996) Within the wider region of the maritime Antarctic, favourable climate and habitats for E. murphyi have been identified in the South Shetland Islands and along the western coast of the Antarctic Peninsula, with islands in Marguerite Bay close to Rothera Research Station at particular risk (Hughes et al., 2013). The indigenous and endemic chironomid B. antarctica, which already occurs across this region, favours similar habitats to E. murphyi and has generally similar physiological capabilities (Edwards & Baust, 1981;Sugg et al., 1983;Convey & Block, 1996;Lee et al., 2006;Benoit et al., 2007;Everatt et al., 2012Everatt et al., , 2015. Therefore, any expansion of E. murphyi into this region could result in impacts on, or even displacement of, B. antarctica, particularly given the possible advantage gained from E. murphyi's more flexible parthenogenetic life history strategy (Bartlett et al., 2018b). Hughes et al. (2013) concluded that E. murphyi was significantly affecting litter turnover on Signy Island. With a biomass at some sites of two to five times that of the estimated indigenous micro-arthropod and micro-invertebrate community (mites, springtails, tardigrades, nematodes and rotifers), a mean E. murphyi population of 21 000 Lm −2 could increase litter turnover by as much as 66.51 g dry mass m −2 per annum. This is almost a full order of magnitude (9.3 times) the consumption rate of the native Signy community (cf. Davis, 1981). In an analogous study, Hänel and Chown (1998) found that another invasive chironomid, Limnophyes minimus, on sub-Antarctic Marion Island, was also increasing litter turnover by 6-10 times the rate of the larvae of the native moth Pringleophaga marioni, superseding it as a major contributor to the island's nutrient cycling. As E. murphyi is the only macro-invertebrate on Signy Island and has no predators or competitors, its impact on the terrestrial ecosystem is likely to be unimpeded. Furthermore, as E. murphyi develops greater population densities (Hänel & Chown, 1998;Hughes & Worland, 2010; this study) and is physically larger than L. minimus (cf. Mercer et al., 2001), it is likely to be one of the most significant introductions to a sub-or maritime Antarctic terrestrial ecosystem yet recorded.
The accidental introduction of E. murphyi to Signy Island is not a unique case for Antarctica but does represent perhaps the most successful and best studied invertebrate introduction outside of the sub-Antarctic region to date. Introduced to Signy with E. murphyi was the enchytraeid worm, Christensenidrilus blocki (Block et al., 1984), which remains persistent but far less abundant or widely distributed (Hughes & Worland, 2010). In neither instance was eradication attempted, with initial studies instead focusing on the physiology of E. murphyi (e.g. Block et al., 1984), rather than the risks it might pose as an invasive species. Now the distribution of E. murphyi is so extensive that eradication is not feasible without considerable and unacceptable damage to the larger ecosystem. Within the maritime Antarctic, another dipteran species, T. maculipennis, was also recently accidentally introduced to King George Island in the South Shetland Islandsthis species was first discovered around the Uruguayan Artigas Base in 2006 and previously had only been recorded in the Northern Hemisphere (Volonterio et al., 2013). Eradication of this species from the affected sewage system has been attempted on more than one occasion but has not been successful (Uruguayan Antarctic Institute, 2008; Volonterio et al., 2013); meanwhile, further stations have been affected on the island (Potocka & Krzemi nska, 2018). Whilst invertebrate introductions are increasingly frequent in the Antarctic and sub-Antarctic areas (Chown et al., 2012, b), it remains the case that, to date, only E. murphyi, C. blocki and T. maculipennis have successfully established within the maritime Antarctic region for any duration.
The in-depth monitoring of E. murphyi's spread within the current study has identified a potentially imminent risk to other valley systems on Signy Island. However, such monitoring of non-native invertebrate species is not currently required either as an advisory or a mandated requirement within either the Antarctic Treaty area or the various sub-Antarctic islands. Transfer in soil and associated vegetation is a known means of anthropogenic transfer of E. murphyi, with such an event already having taken place from the species' native South Georgia to Rothera Research Station at 68 S on the Antarctic Peninsula, via the treads of industrial plant equipment . Current measures used to prevent the transfer of biological material into the Antarctic region and between different Antarctic locations, primarily rely on manual scrubbing of boots and equipment, and a boot wash containing the antimicrobial/viral disinfectant Virkon-S (SCAR 2009; IAATO Guidelines, 2018; BAS Biosecurity Regulations 2018). The efficacy of these measures in preventing invertebrate transfer, however, is unknown, and Virkon-S has never been formally tested as an insecticide. We suggest that it would be prudent to re-evaluate existing biosecurity procedures, such as recommended by Hughes and Pertierra (2016) and subsequent ATCM proposal (ATCMXLI-IP31 United Kingdom and Spain, 2018). A further species that would benefit from such regular monitoring would be the invasive predatory beetle Trechisibus antarcticus on South Georgia which, ironically, may have drastically reduced E. murphyi populations in its native range, along with other endemic invertebrate species (Ernsting, 1995;Convey et al., 2011). Its current distribution, and thus the magnitude of its impact on South Georgia's terrestrial invertebrate communities, remains unknown and unchallenged.

Conclusions
Eretmoptera murphyi has significantly expanded its range on Signy Island over the last decade, and this appears to be in direct association with routes of human movement across the island. The species' preferred habitat is moss bank with soil and peat substratum layers, largely on a northerly aspect, although this preference only explains a small proportion of variation in its population density. It is also unclear whether E. murphyi is selecting existing peat substrates or actively driving the production of peaty soils. A pre-adapted physiology to survive maritime Antarctic conditions and life history features that permit a flexible and opportunistic phenology, combined with anthropogenic intervention, has enabled E. murphyi to thrive, and gives likely opportunity for further expansion to new areas on Signy in coming decades. This species potentially has a major ecosystem influence by opening nutrient cycling bottlenecks, and its full (negative or positive) impact on the native ecosystem and communities remains to be evaluated. Thus, there is an urgent need to investigate its role in wider ecosystem processes, as well as ensure appropriate biosecurity protocols are in place to minimise both the rate of population expansion on Signy Island and to prevent colonisation of other areas in the maritime Antarctic.