Spotlight on the invasion of a carabid beetle on an oceanic island over a 105-year period

The flightless beetle Merizodus soledadinus, native to the Falkland Islands and southern South America, was introduced to the sub-Antarctic Kerguelen Islands in the early Twentieth Century. Using available literature data, in addition to collecting more than 2000 new survey (presence/absence) records of M. soledadinus over the 1991–2018 period, we confirmed the best estimate of the introduction date of M. soledadinus to the archipelago, and tracked subsequent changes in its abundance and geographical distribution. The range expansion of this flightless insect was initially slow, but has accelerated over the past 2 decades, in parallel with increased local abundance. Human activities may have facilitated further local colonization by M. soledadinus, which is now widespread in the eastern part of the archipelago. This predatory insect is a major threat to the native invertebrate fauna, in particular to the endemic wingless flies Anatalanta aptera and Calycopteryx moseleyi which can be locally eliminated by the beetle. Our distribution data also suggest an accelerating role of climate change in the range expansion of M. soledadinus, with populations now thriving in low altitude habitats. Considering that no control measures, let alone eradication, are practicable, it is essential to limit any further local range expansion of this aggressively invasive insect through human assistance. This study confirms the crucial importance of long term biosurveillance for the detection and monitoring of non-native species and the timely implementation of control measures.

The contribution of anthropogenic activities to biological invasions is escalating rapidly and unrelentingly 1 , placing the introduction and spread of non-native organisms amongst the most important contemporary ecological and conservation themes 2,3 . Human-assisted biological invasions 4 can be considered as a six-step continuum: (1) entrainment of living/viable specimens or propagules in their native range, (2) transport, (3) introduction (release) in a new area, (4) establishment, i.e. successful completion of the full life cycle in the new area, (5) sustained population increase at the introduction site(s) and (6) further geographic expansion from the introduction site [5][6][7] . In a wide range of taxa, studies have examined how and why non-native species have breached natural environmental barriers to spread ( 8,9 ; reviewed by 10 for insects and arachnids). Studies assessing the level of invasiveness of non-native organisms and the invasibility of (micro)habitats [11][12][13] have also been undertaken. However, empirical studies that document the early stages of biological invasions, i.e. the establishment and proliferation of non-native populations, and the early stages of subsequent local range expansion, are rare in cases of unintentional introductions 14,15 .
For many if not most unintentional insect introductions, it is challenging to properly document the (1) geographical origin, (2) initial site of introduction, (3) means and date of introduction, and (4) subsequent natural (i.e. not further human assisted) spread of the non-native species within the colonized area (but see the example of the gypsy moth Lymantria dispar (Linnaeus, 1758) (Lepidoptera: Erebidae) whose geographic spread is continuously monitored in the USA 16,17 ). Often, the non-native organisms are only observed for the first time after their population densities have increased markedly 18,19 , and/or when they start to have economic impacts, limiting our capacities to better understand lag effects during invasions [19][20][21][22] . In the cases of the introduction of the ladybird Harmonia axyridis (Pallas, 1773) (Coleoptera: Coccinellidae) in Europe in 1964 and the wasp Vespa velutina Lepeletier, 1836 (Hymenoptera: Vespidae) in France in 2005, both their arrival and subsequent Temporal and spatial spread of Merizodus soledadinus in the Kerguelen Islands. After its initial introduction, the beetle remained restricted to the vicinity of the introduction site at Port-Couvreux for several decades 37 . In 1977, localities to the north of Port-Couvreux were colonized (Cap Kersaint, Presqu'Île Bouquet de la Grye) 54 . In 1982, Tréhen and Voisin 50 reported the presence of M. soledadinus at Port Elisabeth (with a most probable establishment date around 1970), and it then colonized coastal habitats along the north-east coast of the Péninsule Courbet (Fig. 2) as far as the Baie des Cascades by 1983 54 ; these localities were most probably colonised in the 1970s.
In the early 1990s, M. soledadinus expanded further along the Péninsule Courbet and reached Cap Cotter (Fig. 2) 42 . It also colonized an island close to Port-Couvreux (Île du Port), and further locations remote from its original point of introduction, Port-Phonolite and Port-Jeanne d' Arc, the latter the location of a historical whaling station that is regularly visited by both scientists and tourists, and Île Haute in the Golfe du Morbihan (Fig. 2). The species was first observed at the research station Port-aux-Français in 1999. At this site a monitoring programme using pitfall traps twice a month has run since 1996 for the documentation of terrestrial invertebrate communities. The beetle was first recorded in trap samples in June 2000, subsequently becoming common in the collections.  (Fig. 2). By 2007, the insect was present along almost the entire coastline around the Golfe du Morbihan as well as on many islands in this bay, including islets that are rarely visited. Specimens were also reported from several additional sites on the main island of the archipelago including Port-Fleuriais (north of Port-Couvreux) and, on the east coast, in the vicinity of an isolated field hut (Estacade). Of particular note, the beetle was recorded for the first time inland, in vegetation along rivers (e.g. Gave de l' Azorella) and from low altitude fell-fields, a widespread sub-Antarctic habitat populated by cushion plants (Plateau du Larzac, in 2005, at 290 m above sea level (asl); Plateau Central, in 2006, five records between 278 and 358 m asl).
By 2018 (Fig. 3) the beetle's distribution had expanded considerably in the interior of the Péninsule Courbet, in the valley between Port-aux-Français and Port-Elizabeth. On the east coast, the entire Baie Norvégienne was colonised, as were several locations between Estacade and Cap Digby. In the Golfe du Morbihan, the species has Port-aux-Français is the research station and Port-Couvreux is an abandoned farm. The map of the Kerguelen Islands was created by the authors of the present study, and thus, the background map we used for the figure is not a copyrightable subject matter. Geospatial data were incorporated into the map with ArcMap in ArcGis 10.4 (https ://www.esri.com), and this software was used to generate the map. An abundance index was applied to each of the 1164 locations (Fig. 3) where the species was recorded between 2005 and 2018 (after a 10 min active search following a standard protocol: low abundance 1-30 adults, n = 680 observations; medium abundance 31-100 adults, n = 265 observations; high abundance > 100 adults, n = 219 observations). To illustrate an anecdotal impression of the highest abundance category, in one case it took less than 3 min to find 150 individuals under a single stone of c. 200 cm 2 area. The beetle was particularly abundant in the vicinity of Port-aux-Français, on Presqu'île du Prince de Galles, and between Cap Cotter and Cap Digby, with more than 300 individuals being routinely counted in 10 min in 2013. It is notable that instances of high abundance were found throughout the colonized area, on the coast, on the islands of the Golfe du Morbihan, and inland, in both recent and older colonized habitats, suggesting that its population density can increase quickly wherever it becomes established. This suggestion is also supported by monthly trapping data initiated in 2005 on both sides of Isthme Bas, where the number of adults captured on the east coast, colonized in 2011, rapidly reached similar levels to those recorded from the west coast, colonized between 2000 and 2005 (Fig. 4).  Dreux et al. 1992), between 1991and 1995(after Chevrier 1996, and between 2005 and 2007 (this study). The geographic occurrence squares report observations of the presence/absence of the insect; when present, the time ranges of these observations do not necessarily correspond to the establishment date of M. soledadinus at each of the surveyed localities. Observations are plotted on a one kilometer grid; coloured square = presence, grey square = absence. The map of the Kerguelen Islands was generated by the authors in ESRI ArcGis version 10.4 (www.esri.com). The background map we used for the figure is not a copyrightable subject matter. Each observation is plotted according to the number of adults found during a 10-min search: (0) absence, (1) low abundance, 1-30 adults, (2) medium abundance, 31-100 adults, and (3) high abundance, more than 100 adults. The map of the Kerguelen Islands was created by the authors; the background map we used for the figure is not a copyrightable subject matter. Geospatial data were incorporated into the map with ArcMap in ArcGis 10.4 (https ://www.esri.com), and this software was used to generate the map. A. aptera was more often present and abundant than expected when M. soledadinus was absent; conversely, the fly was more often than expected absent or at low abundance when M. soledadinus was present.
As adults of C. moseleyi were not recorded throughout the whole year ( Fig. 5c), we considered the records of M. soledadinus and C. moseleyi only from the highest abundance period of C. moseleyi, i.e. from January to March 2005 (n = 177). Here again, expected frequencies significantly differed from observed frequencies ( Table 2) (χ 2 = 36.007, p < 0.001). Numbers of M. soledadinus negatively affected the abundance of adult C. moseleyi: when the ground beetle was present within the habitat, C. moseleyi was more often absent (0) and less often abundant (abundance index 2 or 3) than predicted.
Additional information on the impact of M. soledadinus on A. aptera and C. moseleyi was provided by trapping results from a coastal site on Île Guillou. Trap records of the two native flies were consistent over the austral summers 1994-1997, before the establishment of the beetle at this location, with lower but still consistent numbers of adult C. moseleyi (1-3 per trap per day). When M. soledadinus was first trapped in July 1998 at this site, both flies were still present. The beetle was then regularly trapped until 2003 (0.1-0.7 individuals per trap per day in 13 of the 54 trapping sessions). Over this period a drastic decrease in records of A. aptera and C. moseleyi became apparent (Fig. 6). After a pause in trapping, when it was resumed in 2006, M. soledadinus was recorded in every trapping session and at higher abundance than previously (1-11 individuals per trap per day, with a maximum value of 46 in December 2006). However, the numbers of A. aptera caught were extremely low, and C. moseleyi was no longer recorded.

Discussion
The geographical expansion of distributions of non-native species is a key topic in invasion science and has considerable significance to the management of invading populations 69 . Human-assisted introduction and spread of non-native organisms typically involves multiple introduction points, and the large spatial scales involved often impede in-field monitoring of the distribution and abundance of the species involved 70 . In the present study, we were able to take advantage of a more tractable invasion study system, that of the accidentally introduced carabid www.nature.com/scientificreports/ beetle M. soledadinus on the Kerguelen Islands archipelago, which allowed us to report a combination of spatial and temporal in-field monitoring of the processes of geographical distribution expansion of this non-native insect. We additionally highlight how quickly the population of this insect builds up when establishing in a new habitat, and how much the geographic expansion of this invasive predator could contribute to the depletion of native insect communities in this UNESCO World Heritage Area.

Historical documentation of the invasion.
In February 1939 when M. soledadinus was first observed, already locally abundant, on the Kerguelen Islands 37 , it was restricted to the surroundings of the abandoned farm buildings of Port-Couvreux, suggesting a recent introduction 37 . Jeannel 37 first hypothesized that the activities of American sealers provided the probable source of the introduction in the archipelago, in particular the ship 'Hillsborough' which landed at Port-Couvreux in 1799. However, he later revised this assumption, taking the view that the species would have had a considerably larger distribution range on the archipelago if it had been introduced around 1800 44 . Hence, he suggested that individuals could have been introduced at Port-Couvreux when the farm buildings (piggery, sheepfold) of this locality were enlarged in 1927-1928 44,68 .
For the purpose of this study, we identified the very limited number of vessels recorded to have landed at Port-Couvreux, and their previous itinerary (including landings in other harbours / regions) 67 before the first specimens of M. soledadinus were observed. Taking into account the narrow austral native distribution of the species (southern South America, Falkland Islands) 41,46,62-64 , we conclude that Jeannel's revised introduction assumption also cannot be supported. The sheep farming attempt that took place in 1927 used animals loaded during a call at Durban (South Africa) by the ship 'Lozère' , which was transporting material originally from Le Havre (France). This vessel did not visit anywhere in the region of the beetle's natural distribution. Based on our review of the available shipping records, we identified the 'Jacques' , a vessel belonging to René Bossière who, with his brother, established the farm of Port Couvreux, as the most likely introduction source 67 Table 1. Frequency distribution of the abundance of adult Anatalanta aptera (Aa) and Merizodus soledadinus (Ms) in three habitats (seashore, inland, carrion). Codes for abundance according to the number of adults found during a 10 min active search: 0 = absent; 1 = low abundance, 1-30 adults; 2 = medium abundance, 31-100 adults; 3 = high abundance, > 100 adults. When information on the abundance of the insects was not available, occurrence is reported as absence/presence only. Expected frequencies in italics. Observed frequencies significantly differed from expected frequencies along the seashore (χ 2 = 38.877, p < 0.001), inland (χ 2 = 54.217, p < 0.001) and under carrion (χ 2 = 15.560, p < 0.001). Temporal and spatial spread. After its initial introduction at Port-Couvreux in 1913, the species persisted without expanding its range, at least until 1939, when Jeannel 37 found it at high density at the introduction site but also actively searched for and failed to find the species at multiple sites in the archipelago. A lag phase is commonly reported in studies of invasion processes, with its duration varying across taxa and with the characteristics of the introduction sites. For instance, Kiritani and Yamamura 72 reported a mean lag phase of 11.8 years for the 35 non-native insects they considered, in line with the predicted lag range of 4.4-23.2 years of Morimoto et al. 73 . Similarly, the gypsy moth Lymantria dispar took about 20 years to spread only 500 m from its initial introduction point in the USA 74 , before its subsequent rapid expansion commenced. Thus, even if recent models suggest that it is not possible to make accurate predictions of the duration of lag phases, in particular for introductions occurring in coastal areas 19 as in the case for M. soledadinus, the apparent lag time for this species is consistent with the existing insect invasion literature. The lag time could in part result from Allee effects, which refer to any process whereby any component of individual fitness is correlated with population size ( 76,77 , also see the reviews of 6,78,79 , which describe the different concepts underlying invasion dynamics). Investigations conducted on populations of M. soledadinus support this idea. For instance, while adult beetles have a relatively long lifespan of 1-2 years 79 , the small (3-12) mature egg load per female, and the long developmental period of the juveniles (at least several months 55,61,79 ), may limit individual opportunities to find a mate. Such life history characteristics may have contributed to restricting population growth in the initial years following the species' introduction at Port-Couvreux. Inbreeding depression is a further potential element of the Allee effect 76 , although there is no direct evidence of it playing a role in the establishment of M. soledadinus in the archipelago. However, preliminary studies have found that adults of M. soledadinus obtained from Port-Couvreux exhibited significantly lower levels of heterozygosity than those from native Patagonian populations 80 .
Farming and the human presence at Port-Couvreux ceased in 1931 67 , after which there was little human presence or activity on the Kerguelen Islands until the early 1950s, when the scientific research station Port-aux-Français was established. Thus, initial expansion in distribution of the beetle from Port-Couvreux (Presqu'Île Bouquet de la Grye) did not occur with human assistance, at least until it reached Port-aux-Français (Péninsule Courbet) in the late 1990s. During the austral summer of 1982-1983, confirmation of its presence in the general vicinity of its introduction site (Fig. 2) is consistent with natural dispersal. As well as terrestrial dispersal, some beetles may have directly crossed the inlet separating Presqu'Île Bouquet de la Grye and Plateau Central (Fig. 2) by marine rafting. While the distance from Port-Couvreux to Anse Sablonneuse is 25 km overland, it is only 400 m by direct line crossing the inlet, and experimental data have revealed that M. soledadinus can survive flotation and exposure to saline conditions for several days 81,82 .
Habitat connectedness is a key influence on dispersal performance in insects 83,84 . In particular, a lower landscape permeability in between two patches can restrict geographic expansion in insects 85 . In this context, Table 2. Frequency distribution of the abundance of adult Calycopteryx moseleyi (Cm) and Merizodus soledadinus (Ms) along the seashore. Codes for abundance according to the number of adults found during a 10 min active search: 0 = absent; 1 = low abundance, 1-30 adults; 2 = medium abundance, 31-100 adults; 3 = high abundance, > 100 adults. As information on the abundance of M. soledadinus was not always available, occurrence is reported as absence/presence only. Observed frequencies significantly differed from expected frequencies along the seashore (χ 2 = 36.007, p < 0.001).  89 , there is a possibility that local biodiversity may provide a significant barrier to the spread of non-native organisms on Kerguelen. For instance, at Cap Digby, the presence of a large penguin colony with probable effects on soil composition 90 may limit movement of M. soledadinus along the coast. While roads and rivers are well-known dispersal corridors accelerating the geographic expansion of invasive species 91-93 , here we conclude that the seashore is the most prominent dispersal corridor for M. soledadinus in the Kerguelen archipelago, providing connectivity between areas of habitat. Physiological studies have reported that humidity and water availability are key factors that can quickly impair the survival of adults 94 . However, ponds, waterlogged areas, streams and rivers are frequent in habitats close to the seashore around the archipelago, often combined with the presence of abundant food resources in the form of native and other non-native insects 33,36,95 . Temperature likely represents an additional factor driving the expansion of this species, whose spread rate of ca. www.nature.com/scientificreports/ 3.0 km/year (Île Haute) 57 is far higher than that estimated on the colder South Georgia (0.1 km/year 61 ) where it has also been introduced. However, at the time of the latter study in the 1980s, the species may have still been in the lag phase. Although our data do not allow formal spread rate calculations, assuming that the flightless M. soledadinus invaded the south coast of Péninsule Courbet from the research station of Port-aux-Français from 1999 onwards, and based on our in-field surveys, the spread rate achieved would be between 1.7 and 2.7 km/ year. The spread rate of M. soledadinus at the Kerguelen Islands is thus similar to that of closely related carabid beetles, such as Trechus obtusus Erichson, 1837 (3.0 km/year in Hawaii 96,97 ). The arrival of M. soledadinus in the vicinity of the research station at Port-aux-Français (first observed in 1999) marked a significant milestone in the beetle's expansion in the archipelago. Its presence in an area of considerable human activity created the opportunity for further human assistance in dispersal within the archipelago. A clear example of this is given by Estacade, a location with a field hut that has long been used to support monitoring of penguin populations and, to a lesser extent, by tourists who previously were able to overnight when visiting the penguin colonies. When the beetle was first observed at Estacade in 2005 (after most probable establishment between 1995 and 2000, as the shelter was not used frequently after early 2000), the closest established populations of M. soledadinus were at Cap Digby and Port-aux-Français, both more than 20 km distant from Estacade. It is thus very likely that accidental human transport of small numbers of beetles was responsible for their introduction to Estacade.
However, even with the likelihood that human assistance has played a role in further spreading the beetle after its arrival in the vicinity of the research station, this is unlikely to explain all instances of local colonisation by the beetle, for instance of several islands of the Golfe du Morbihan that are rarely visited. Rather, as suggested earlier, beetles are likely to have arrived by rafting on vegetation or algae 81,82 or through ornithochory. For instance, Kerguelen shags Phalacrocorax verrucosus use seaweeds that they sometimes transport from one island to another when building their nests. Consistent with this, carrion and skulls of rabbits (Oryctolagus cuniculus) are regularly found on the north-west coast of Île Australia, an island where rabbits do not occur. Carrion typically hosts M. soledadinus individuals which prey on fly larvae developing on the cadavers 38 . The transport and consumption of carrion by scavenging birds (e.g. skuas Stercorarius antarcticus lonnbergi and giant petrels Macronestes giganteus) thus represents another possible mechanism of dispersal for the beetle.
Abundances of M. soledadinus recorded in this study are considerably greater than those reported in the early 1990s 57 . They are also much higher than those reported from South Georgia (maximum = 156 collected per hour 61 ). Even though individuals of M. soledadinus have well developed thermal stress tolerance 98 , and have colonized low altitudes on South Georgia 99 , the harsher climatic conditions of South Georgia as compared with the Kerguelen Islands may reduce the beetle's abundance and the speed of its geographic expansion. Mean annual temperature on King Edward Point during the period 1951-1980 was 2.0 °C and snow cover was more or less permanent from May to October in the coastal area 55 . During the same period , the mean annual temperature was 4.5 °C at Port-aux-Français, and despite regular snowfall (ca. 15 days per month from June to August), no permanent snow cover was observed (Source: Meteo France data).
Earlier studies assessing the physiological capabilities of M. soledadinus reported limited signs of thermal stress in adults permanently exposed to temperatures as high as 20 °C 99,100 , and suggested that warming may further assist the local spread of the species. In the Kerguelen Islands, air temperature increases that occurred during the winter months in the early 1990s resulted in a 20-30 days' reduction of the number of freezing days each year (Source: Meteo France data 33 ). Warming may represent a significant driver of the recent colonization of moderate altitudes by populations of M. soledadinus. The highest altitude that established M. soledadinus was recorded was 110 m above the sea level (asl) in the mid-1990s 36 , while populations were found up to 358 m asl in 2005 (this study, 101 ). Comparing the morphological and biochemical characteristics, and metabolic phenotypes, of adult M. soledadinus sampled along altitudinal transects, Ouisse et al. 101 concluded that the presence of the insects at moderate altitudes resulted from the progressively higher occurrence of thermally suitable habitats.
Ecological impacts. Alien insects can severely affect native biodiversity, in particular when they are predators bringing novel ecological function into the invaded habitats and are no longer limited by other predators. As insects, they can also often develop large population densities, further increasing the impacts they can have on prey species. The invasion process of the predaceous M. soledadinus has had major impacts on the native entomofauna, even in the most recently colonized locations, where M. soledadinus rapidly becomes dominant. The native flies A. aptera and C. moseleyi appear to have been largely lost or even driven locally extinct in several locations colonized by M. soledadinus ( 33,39 , this study). We also highlight that C. moseleyi, whose population densities are often much lower than those of A. aptera, is more sensitive to the year-round active M. soledadinus 79 than A. aptera in seashore habitats. This is particularly critical for the endemic C. moseleyi, as algae already represent its secondary trophic niche after its primary resource (the Kerguelen cabbage, Pringlea antiscorbutica 95 ) was wiped out from most locations invaded by rabbits 102  www.nature.com/scientificreports/ arthropod predators, the rove beetle Leptusa atriceps (Waterhouse, 1875) (Coleoptera: Staphylinidae) and the linyphiid spiders Myro kerguelensis (Pickard-Cambridge, 1876) (Araneae, Desidae) and Neomaso antarcticus (Hickman, 1939) (Araneae: Linyphiidae). Direct predation may occur, with late-instar larvae and adults of M. soledadinus predating small spiders. Adult M. kerguelensis may be capable of predating larvae of the ground beetle. In South Georgia 217 Trechisibus antarcticus (Dejean, 1831) (Coleoptera: Carabidae) and 68 spiders were found in one litter sample, all of the latter much smaller than M. soledadinus 55 . Spiders may prey on springtails and be preyed upon by the carabid 55 . At South Georgia, M. soledadinus also has a strong impact on the abundance of the endemic perimylopid beetle Hydromedion sparsutum (Waterhouse, 1875) (Coleoptera: Perimylopidae) 65 . Taken together, our data and the available literature suggest that M. soledadinus can have profound impacts on native insect species, acting as an ecosystem engineer and changing significantly the invaded habitats along the invasion gradient.

Conclusions
Although much less visible than the rabbit, which has completely altered vegetation structure and cover in the areas it has colonized in the Kerguelen Islands 103 , the impact of M. soledadinus on the archipelago's terrestrial ecosystems is considerable. The wide-reaching consequences on biodiversity produced by this predatory species create novel disturbances in the Kerguelen Islands, where the beetle is an invasive ecosystem engineer. Despite its inability to fly, its distribution in the archipelago is continuing to expand. This has been especially apparent since its arrival in the vicinity of the active research station, from where inadvertent human assistance has accelerated its geographic spread. No control measures, let alone eradication, are practicable, so it is urgent and essential to limit as far as possible any further dispersal by human activities. Our study confirms the crucial importance of long-term biosurveillance for the detection and subsequent monitoring of non-native species and the timely implementation of control measures. With the species already having a wide distribution in parts of the archipelago, knowledge of its past and current distribution provide valuable insight into the environmental drivers of its geographic spread, and help identify suitable habitats vulnerable to colonisation by the species. This body of knowledge, which could serve for making predictions on future geographic expansion 104 110 . Merizodus soledadinus has been accidentally introduced to two sub-Antarctic islands or archipelagos, the Kerguelen Islands (first record in 1939 37 ) and South Georgia (first record in 1963 47 ). On Kerguelen, adults have been described as being active at night 53 and are found during the day beneath stones and kelp belts 55 . Anatalanta aptera Eaton 1875 (Diptera: Sphaeroceridae) is a wingless fly that is endemic to the Indian Ocean Province sub-Antarctic islands. It can be found on the Crozet and Kerguelen archipelagos and on Heard and McDonald Islands. On the Kerguelen Islands, it is present from sea level to more than 600 m asl and is active yearround. Larvae and adults are saprophagous and feed on decaying organic matter. Anatalanta aptera is abundant in many habitats, especially in seabird colonies, around carrion and in coastal areas enriched by seaweeds 39,111 .
Calycopteryx moseleyi Eaton 1875 (Diptera: Micropezidae) is another wingless fly endemic to the Indian Ocean Province sub-Antarctic islands. It can be found in the Kerguelen archipelago and on Heard and McDonald Islands. Its larvae feed preferentially on the Kerguelen cabbage Pringlea antiscorbutica, but can also frequently be found under decomposing seaweeds along the seashore and on decaying organic matter in penguin rookeries 94 39 . At each site surveyed, we noted the GPS coordinates of the record, the occurrence (presence/absence) and abundance (abundance index) of M. soledadinus, in addition to recording the occurrence of the two native flies A. aptera and C. moseleyi. For the abundance index, preliminary trials showed that a 10-min search by one person was appropriate to detect the presence of the three insects and make accurate estimates of their densities, even when they were present in low numbers. Thus, a semi-quantitative index was designed based on the number of adults found during the 10-min search: (0) absence, (1) low abundance, 1-30 adults, (2) medium abundance, 31-100 adults, and (3) high abundance, more than 100 adults. The nature of microhabitats hosting the three insects was also recorded, and resulted in the following list: stranded seaweed, stones, carrion, and leaves of the Kerguelen cabbage. Inland sites were subsequently surveyed by making the same observations along transects moving inland perpendicular to the shoreline and along altitudinal transects. Point surveys were completed ca. every 100 m along the former and ca. every 20 m in elevation along the latter in order to accurately define M. soledadinus distribution limits at the edges of colonized areas. Observations were stopped when no M. soledadinus were observed at two consecutive survey sites.
Finally, since 2006, as a part of systematic recording of the distribution of entomofauna at the Kerguelen Islands, further georeferenced field surveys of M. soledadinus have been carried out, paying particular attention to conducting active searches at range edges of the species' known distribution.

Assessment of population dynamics and seasonal fluctuations of Merizodus soledadinus in colonized habitats.
To clarify the beetle's population dynamics after its establishment in a new habitat, we monitored invertebrate communities at two sites on Péninsule Courbet from 2005 to 2018. For this long-term monitoring study, three pitfall traps (Ø 9 cm, h 4 cm) were opened for 5 d every 2-3 weeks at each sampling site. In the first site, on the west coast of Isthme Bas, M. soledadinus established between 2000 and 2005. At the second site, on the east coast of Isthme Bas, M. soledadinus was absent at the start of the study, and became established during the monitoring period. At both sites, the pitfalls were placed in herbfield communities dominated by the deciduous dwarf shrub Acaena magellanica (Vahl 1804) (Equisetopsidae: Rosales). After collection, beetles were stored in 70% ethanol, identified to species level, and counted in the laboratory.
Ecological impact of Merizodus soledadinus on native entomofauna. To assess the ecological impact of M. soledadinus on the native entomofauna, and in particular on the two native flies A. aptera and C. moseleyi, we used the abundance index recorded during the exhaustive geographical survey conducted from December 2004 to March 2006. We took into account the period of activity of the three species based on knowledge from the long-term biosurveillance programme for entomofauna running at Port-aux-Français; this consists of (a) three pitfall traps (Ø 9 cm, h 4 cm) opened for 5 d every 2-3 weeks, with the aim of collecting individuals of M. soledadinus, (b) one baited trap continuously operated targeting A. aptera (insects collected every 5-10 days), and (c) one yellow trap operated for 5 days every 2-3 weeks targeting C. moseleyi. Taking into account the seasonal patterns of activity of these species, we focused on the following pairwise comparisons: (1) M. soledadinus versus A. aptera (at least one of the two species present in the 338 records from coastal habitats, including seashore, under seaweeds, stones), (2) M. soledadinus versus A. aptera (at least one of the two species present in the 379 records from inland habitats, i.e. more than 50 m from the seashore, under stones and carrion), (3) M. soledadinus versus A. aptera (at least one of the two species present in the 155 records from carrion, along the seashore and inland), (iv) M. soledadinus versus C. moseleyi (at least one of the two species present in the 177 records from coastal habitats, including seashore, seaweeds and under stones).
Additional information on the impact of M. soledadinus on A. aptera and C. moseleyi was obtained from trapping results from a coastal site on Île Guillou in Golfe du Morbihan. On this island, three pitfall traps (Ø 9 cm, h 4 cm) were opened monthly for 5 days between January 1994 and July 2003, and then from March 2006 aptera, and C. moseleyi were represented in contingency tables, giving marginal (sum of each column, sum of each line) and grand (total number of individuals) totals. Expected frequencies were first computed from the totals assuming that there were no relationships between cells which would result in similar values between expected and observed frequencies. To assess differences among proportions, Chi-square tests were conducted (whenever classes with low frequencies occurred, frequencies from adjacent classes were pooled) as well as Fisher's exact test (when pooling classes resulted in a 2 × 2 table). The analyses were conducted using Minitab 13 (Minitab Inc., State College, PA.).

Data availability
The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request. www.nature.com/scientificreports/