The Island of South Georgia is about 170 km long and up to 40 km wide and is located in the South Atlantic, between 54°S–54°55'S and 35°50'W–38°W, about 1000 km north-east of the Antarctic Peninsula. Despite its geographical isolation and relatively harsh sub-Antarctic climate, South Georgia currently hosts species of plants and invertebrates that were introduced by sealing – and later shore-based whaling – industries between the late 18^th and mid-20^th centuries (Convey and Lebouvier 2009; Convey et al. 2011; Black 2022) with new introductions occurring up to the present day (Convey et al. 2010; Tichit et al. 2023). Some introduced species have negative impacts on native communities (Ernsting et al. 1995; Houghton et al. 2019) and the island’s terrestrial ecosystems may be particularly vulnerable to introductions due to the presence of vacant niches that are readily available to new competitive invasive taxa (Convey and Lebouvier 2009; Houghton et al. 2019). To tackle this issue, the Government of South Georgia & the South Sandwich Islands (GSGSSI) has implemented rigorous biosecurity measures and invasive mammals have been successfully eradicated from the Island (GSGSSI 2013; Martin and Richardson 2019). However, some plant species such as the dandelion Taraxacum officinale agg. and the meadowgrass Poa annua and invertebrates such as the carabid beetles Trechisibus antarcticus and Merizodus soledadinus may now be too widespread for any realistic possibility of eradication (GSGSSI 2021).

To assess the ability of established introduced species to colonise deglaciated areas, we surveyed six glacial foreland sites during the austral summer on the north coast of South Georgia (Fig. 1). Three sites, investigated in April 2022, were located in the vicinity of the following tidewater glaciers at low elevation (46 ± 40 m a.s.l.): the Nordenskjöld Glacier between Barff and Greene Peninsulas, the Harker Glacier between Greene and Thatcher Peninsulas and the Lyell Glacier that defines the western limit of Thatcher Peninsula. In January and February 2023, the surroundings of three extant or recently-extinct inland cirque glaciers at higher elevation (378 ± 122 m a.s.l.) were surveyed: Hodges and Col Glaciers on Thatcher Peninsula and an unnamed icecap west of Husvik on Busen Peninsula (locally known as Husvik Glacier). Of these three sites, only Col Glacier currently persists as an ice remnant approximately 30 × 100 m in size. Hodges Glacier rapidly receded after 1970 and was lost in 2008 (Bakke et al. 2021), while Husvik Glacier was lost by the late 1950s (“Map of Falkland Island Dependencies: South Georgia” 1958).

Figure 1.

Maps and overview of the six glacier sites on the north coast of South Georgia. Plant and invertebrate communities at three tidewater glacial forelands (1–3 purple) and three inland deglaciated sites (4–6 blue) were surveyed. Centre-right: example of transects (black dots) along former deglaciation fronts (from 1993 to 2017, light to dark purple) in the vicinity of Nordenskjöld Glacier.

To assess the dynamics of foreland colonisation by terrestrial communities, locations with contrasting times since deglaciation (tsd) were sampled at each foreland. For tidewater glaciers, detailed maps of glacial front changes were available (Cook et al. 2010; South Georgia GIS, accessed February 2022), enabling us to sample locations positioned with precision along former glacial fronts with tsd between five and 30 years. Depending on site accessibility and logistical constraints associated with fieldwork, we were able to sample one to three replicates uniformly distributed and at least 80 m apart on the retreat line for two to four values of tsd at the foreland of each tidewater glacier (Suppl. material 1). As no map or record of deglaciation dynamics was available for the inland glaciers, an area with recent tsd was identified closest to the current (Col Glacier) or last known position (Hodges and Husvik Glaciers, “Map of Falkland Island Dependencies: South Georgia” 1958) of the ice remnant (Suppl. material 1). For comparison, a second area with older tsd was determined at a distance of approximately 200 m in the flow direction of the glacier. The validity of this approach relies on the unverifiable assumption that the retreat rate was similar across the three inland glaciers and over their deglaciation history. For both recent and older tsd, three approximately equidistant replicate locations were sampled.

At each sampled location, plant communities were surveyed along a 30 m transect. All vascular plants present within six adjacent quadrats (5 × 5 m) either side of the transect line were recorded (yielding 12 records of plant presence/absence per transect), while the cumulative number of bryophyte and lichen morpho-species (photographs provided as Suppl. materials) across two quadrats at opposite ends of the transect was recorded. For the tidewater glacier sites, the cover of plants and lichens was measured through a point-contact sampling procedure using a frame with 10 equidistant pins (length = 50 cm), placed every 2 m along the transect (yielding 15 records of plant cover per transect). The vegetation at inland glacier sites was too sparse to achieve representative point-contact sampling; we therefore used a Braun-Blanquet scale to estimate the cover of plants and lichens in each 5 m quadrat (Suppl. material 2; yielding 12 records of plant cover per transect).

Several invertebrate sampling methodologies were applied in order to achieve the most comprehensive description of the communities present. Ground-dwelling (and secondarily flying) arthropods were sampled using pitfall traps (n = 3) consisting of 250-ml beakers half-filled with a water/washing detergent solution and buried to ground-level at the start, middle and end of each transect. Traps were retrieved after being deployed for approximately 48 h. Macro-invertebrates were extracted from approximately 200 ml of substrate obtained at the same three positions along the transect, using Tullgren extractions for 8 h. Micro-invertebrates were sampled on an opportunistic basis from soaked aliquots of the same substrate. Invertebrates sampled using this non-quantitative method were not included in statistical analyses. Finally, ground-dwelling invertebrates under stones and debris were recorded and sampled during hand searches of 8 min and flying insects were captured using sweep nets along a span of 5 m either side of the transect. All sampled invertebrates were rapidly transferred to ethanol for preservation.

All vascular plants were readily identifiable to species level in the field, based on published description and nomenclature (Burton and Croxall 2012b; POWO 2023) with the exception of the native lesser rush, Juncus inconspicuus that was considered to be a synonym of the native greater rush Juncus scheuchzerioides Gaudich (Kirschner 2002). Observations of the hybrid between the native greater and lesser burnet Acaena magellanica × tenera were merged with the data for A. magellanica. The introduced species aggregate Taraxacum officinale may contain several micro-species and so was reported as Taraxacum officinale agg.

While some macroinvertebrates could be identified to species level in the field (Burton and Croxall 2012a), most specimens required detailed assessment of anatomical features under stereo- or light microscopy, with reference to the available literature (Enderlein 1912; Hendel 1937; Gressitt 1970; Convey et al. 1999; Kits 2011). Identification confidence for each taxon was categorised as ‘possible’, ‘probable’ or ‘certain’. All macro-invertebrates and springtails were identified to species-level. All Sminthuridae were reported as Sminthurinus jonesi, since there were no consistent taxonomic features supporting the presence of other species, contrary to what was suggested by Convey et al. (1999). Mites and other micro-invertebrates were categorised into morphotaxa.

All statistical analyses were performed in R (R Core Team 2022). To assess if the observed presence data reflected the true presence of organisms in the glacial foreland communities (Buddle et al. 2005), we plotted species accumulation curves for each transect with the function specaccum from the package ‘vegan’ (Oksanen et al. 2007). Visual inspection of these curves indicated if the encounter rate of new species across samples taken was sufficient to compute representative diversity metrics.

To visualise the taxonomic composition of vascular plant communities across tsd and glacier site, we performed an ordination on a Jaccard dissimilarity matrix of the presence/absence data at the transect level, using non-metric multidimensional scaling (nMDS) with the function metaMDS in ‘vegan’ (Oksanen et al. 2007). The score of each species was displayed on the ordination plot.

The presence data were modelled as a function of tsd and glacier site through an ordination with the function cca from the package ‘vegan’ (Oksanen et al. 2007). ANOVA-like permutation tests (n = 999 permutations) for constrained correspondence analysis with the function anova.cca were used to test whether community composition was significantly constrained by tsd, glacier site and their interaction.

To investigate the effects of deglaciation time on the richness of communities and the presence or cover of species, we employed multivariate models using Bayesian Inference with the package brms (Bürkner 2017), treating the glacier site and tsd as the main explanatory variables. Tidewater and inland glaciers were modelled separately. As the response of communities may depend on glacier sites, we ran a model with a simple interaction between glacier site and tsd. For tidewater glaciers, we ran a model with a quadratic term for tsd to reflect non-linear responses, as well as a model with both interaction and quadratic terms. We selected the simplest and most informative model using pairwise comparisons of the expected log pointwise predictive density (ELPD) with function loo (Vehtari et al. 2017, Suppl. material 3). The variable tsd was a categorical variable for inland glacier sites (recent or old), but continuous and scaled to zero mean and unit variance for tidewater glacier sites. As the availability of mapped former glacial fronts in the period 1993–2018 varied between tidewater glaciers (Cook et al. 2010), the sampling of tsd was heterogeneous and not synchronised across glaciers, which prevented the use of a categorical variable to model tsd. When the sampling unit (pitfall traps, pin frame or 5 m quadrat) was nested within a transect, transect identity was included as a random effect. Response variables were observed for species richness, plant cover (at species level or higher) and species presence/absence and were modelled using Poisson, zero-inflated binomial and Bernoulli distributions, respectively. Weakly-informative priors determined by a Gaussian distribution (mean µ = 0, standard deviation σ = 10) were used to model the effects of predictors. Random effects were drawn from a Student’s t-distribution (df = 3, mean µ = 0, standard deviation σ = 10). Models were run using four chains for 5000 iterations (including 2500 burn-in iterations). Traces of the sampling behaviour of each predictor were scrutinised (Suppl. material 4) and the R-hat convergence diagnostic (Vehtari et al. 2021) was computed (Suppl. material 5) to verify that the models converged towards reliable predictions. A posterior predictive check was used to compare modelled and observed data and evaluate the quality of the models (Suppl. material 4). The significance of each effect being positive or negative was assessed using Bayesian 95% credible intervals (CI₉₅).

Eighteen native species of vascular plant were found at tidewater glacier sites (Suppl. material 6), representing 78% of the native species known from South Georgia. At inland glaciers, only seven native species were observed (30%). Four introduced species occurred on both types of glacier sites: Poa pratensis, Taraxacum officinale agg., Cerastium fontanum and Poa annua (Suppl. materials 7, 8). At tidewater glaciers, C. fontanum was the second and P. annua was the seventh most frequent species (Fig. 2a). On forelands of inland glaciers, P. annua was the second, C. fontanum the fourth and T. officinale agg. the eighth most frequent species (Fig. 2b).

Figure 2.

Frequency of occurrence across transects of the 10 most common vascular plants around tidewater glaciers (a, n = 21 transects), inland glaciers (b, n = 18 transects) and most frequent invertebrates around tidewater glaciers (c, n = 21 transects). Introduced species are highlighted in red.

Sixteen native species of terrestrial invertebrates were identified with high confidence at tidewater glacier sites (Suppl. material 6), representing 48% of the native species on South Georgia. At inland glacier sites, only five native species were present, representing 15% of known native species. Five introduced species were found at the tidewater glacier sites and none at the inland sites (Suppl. material 7): Merizodus soledadinus (Coleoptera), Hypogastrura viatica (Collembola), Apterothrips secticornis (Thysanoptera), Mycomya sp. (Diptera) and Trichocera regelationis (Diptera). Merizodus soledadinus and H. viatica were the fourth and tenth most frequent invertebrate species, respectively, at tidewater glacier sites (Fig. 2c).

Accumulation curves of vascular plant species were close to saturation for most of the sampled transects (n = 39, Suppl. material 9), indicating that samples were largely representative of the communities present. However, accumulation curves of invertebrate species did not reach a plateau or were not possible to produce when transects were the smallest replication unit. In the following analyses, we thus calculated diversity metrics only for plants, for which an assumption of near-complete detection was reasonable. We investigated drivers of presence for a subset of the ten most frequently encountered invertebrate species, assuming that detection – though likely incomplete – remained equally probable across sampling sites.

The structure of vascular plant communities was significantly constrained by tsd, glacier site and marginally by their interaction in forelands of both tidewater and inland glaciers (Table 1, Fig. 3).

Figure 3.

Taxonomic composition of vascular plant communities across time since deglaciation (tsd in years) at tidewater (a) and inland (b) glacier sites. The two first components of a non-metric multidimensional scaling (NMDS) from an ordination on the presence data at the transect level are mapped. Each circle, rectangle or triangle corresponds to a transect from a given site and deglaciation time (black to red). Small crosses represent the score of the ten most frequent species on the ordination plots.

Around tidewater glaciers, the number of bryophyte morpho-species increased linearly with tsd, while the number of vascular plant species initially increased and then reached a plateau (Fig. 4, Suppl. material 5). Bryophytes seemed to be more abundant at intermediate tsd and there was no clear effect of tsd on the cover of vascular plants, but the cover of lichens was higher in areas exposed for longer (Suppl. materials 5, 10). Compared to Harker Glacier, Lyell Glacier hosted a lower number of vascular plant species and a higher number of bryophyte morphospecies, while Nordensköld Glacier was associated with a higher cover of bryophytes (Suppl. materials 5, 10).

Figure 4.

Effect of time since deglaciation (tsd in years) on the number of vascular plant species and bryophyte morpho-species at tidewater (a, b) and inland (c, d) glacier sites modelled with Bayesian Inference. Transparent points represent the original data. Lines (a, b) or points (c, d) are the estimated mean effects of tsd. Purple, turquoise and yellow areas (a, b, Harker: purple, Lyell: turquoise, Nordenskjöld: yellow) or intervals (c, d, Husvik: purple, Col: turquoise, Hodges: yellow) represent the Bayesian 95% credible intervals, respectively.

At inland glacier sites, there were significantly more vascular plant species in older than in recently-deglaciated areas (Fig. 4, Suppl. material 5), as well as a higher cover of vascular plants and bryophytes (Suppl. materials 5, 10). Compared to Husvik Glacier, Col Glacier hosted a lower number of vascular plant species and reduced cover of vascular plants and of bryophytes (Suppl. materials 5, 10). Hodges Glacier did not differ from Husvik Glacier in terms of community-level metrics.

At tidewater glacier sites, the introduced Cerastium fontanum and Poa annua were most likely to occur across a broad range of intermediate tsd (Fig. 5a, Suppl. materials 5, 10), which was also the case for the native species Deschampsia antarctica and Colobanthus quitensis. The probability of occurrence of the native Phleum alpinum initially increased and then reached a plateau with increasing tsd. The native Festuca contracta, Acaena tenera, A. magellanica, Rostkovia magellanica and Galium antarcticum were more frequent at older deglaciated sites. Similarly, the two lichens Stereocaulon sp. and Pseudocyphellaria sp. and the mosses Syntrichia robusta and Polytrichum sp. were more likely to occur in older deglaciated areas, while the presence of the liverwort Marchantia berteroana was not affected by tsd. The cover of C. fontanum decreased with increasing tsd (Fig. 5b, Suppl. materials 5, 10), while the cover of P. annua showed no evidence of change. Amongst native vascular plants, the cover of C. quitensis showed no evidence of change with tsd, D. antarctica and P. alpinum were most abundant at intermediate tsd and F. contracta had higher cover in older deglaciated areas. Amongst bryophytes and lichens, the cover of Stereocaulon sp., Pseudocyphellaria sp. and Marchantia berteroana showed no evidence of change with tsd, while Polytrichum sp. was most abundant at intermediate tsd and Syntrichia robusta increased in cover with tsd.

Figure 5.

Scaled effect of time since deglaciation (tsd in years) on plant presence (a), vascular plant cover (b) and invertebrate presence (c) at tidewater glaciers modelled using Bayesian Inference. Effects on introduced and native species are represented in red and blue, respectively. Points are the mean effects of tsd on the logit scale. Intervals represent the Bayesian 95% credible intervals. The vertical dotted lines correspond to the null hypothesis (effect is zero).

At inland glacier sites, the occurrence of the invasive Taraxacum officinale agg. did not differ between old and recent areas of deglaciation (Fig. 6a, Suppl. materials 5, 10), while the invasive P. annua and C. fontanum were more frequent in older deglaciated locations. The native D. antarctica, A. magellanica and Polystichum mohrioides did not significantly differ between tsd, while all other native plants became more frequent in older deglaciated areas. The cover of all plants (with the possible exception of D. antarctica and P. alpinum) was higher in old deglaciated areas (Fig. 6b, Suppl. materials 5, 10).

Figure 6.

Effect of time since deglaciation (tsd) on vascular plant presence (a) and vascular plant cover (b) at inland glaciers modelled using Bayesian Inference. Effects on introduced and native species are represented in red and blue, respectively. Points are the mean effects of tsd on the logit scale. Intervals represent the Bayesian 95% credible intervals. The vertical dotted lines correspond to the null hypothesis (effect is zero).

At tidewater glacier sites, the native spider Micromaso flavus seemed more frequent at older deglaciated sites, but there was no clear effect of tsd on the presence of other reliably sampled invertebrates (Fig. 5c, Suppl. materials 5, 10).

For both inland and tidewater glaciers, there were notable differences between sites in the presence and cover of species (Suppl. materials 5, 10). The invasive C. fontanum was scarce at Lyell Glacier and abundant at Nordenskjöld Glacier, where the introduced springtail Hypogastrura viatica was also more common, while P. annua was more frequent at Harker Glacier. At inland sites, T. officinale agg. was more frequent at Hodges Glacier, while P. annua was less abundant at Col Glacier.

Introduced vascular plants and invertebrates were well represented in the recent stages of community assembly after glacial retreat on South Georgia. Four introduced vascular plants were found on glacial forelands, with Cerastium fontanum and Poa annua being very frequent, while Taraxacum officinale agg. and P. pratensis were rarely observed. Around tidewater glaciers, C. fontanum and P. annua occurred across a broad range of tsd and C. fontanum was more abundant in recently rather than in older deglaciated sites, indicating that these species are effective pioneers on glacial forelands along South Georgia’s coast. Notably, flowering specimens of C. fontanum were found in areas deglaciated less than five years prior to the survey, approximately 50 m from the terminus of Lyell Glacier. Both species originate from temperate regions of the Northern Hemisphere (POWO 2023, Suppl. material 7), are widespread on South Georgia (Black 2022) and have successfully invaded most islands in the sub-Antarctic (Frenot et al. 2005). Poa annua is invasive on the maritime Antarctic South Shetland Islands (Molina-Montenegro et al. 2012; Hughes et al. 2015) and both species are also early colonisers of newly-deglaciated areas on a glacier foreland on the sub-Antarctic Kerguelen Islands (Frenot et al. 1998). At tidewater glaciers on South Georgia, the rapid colonisation by C. fontanum and P. annua outpaced that of most native vascular plants, with the possible exceptions of Deschampsia antarctica and Colobanthus quitensis, with these invasive plants effectively short-cutting the successional sequence following glacial retreat. At inland sites, C. fontanum and P. annua were less common in more recently-deglaciated areas, but direct comparisons with tidewater glaciers are not possible due to methodological differences. In contrast, the dandelion T. officinale agg. appeared equally capable of colonising old and recently-deglaciated areas inland, which likely results from its seeds being wind-dispersed over large distances.

We also documented the presence of five introduced invertebrate species on recently-deglaciated forelands, indicating an ability to disperse and survive in challenging environments. This capacity to track glacial retreat is particularly remarkable for three of these invertebrates that are flightless and suggests high mobility through passive dispersal (Hågvar et al. 2020) or active locomotion, as previously reported for the carabid beetle Merizodus soledadinus (Convey et al. 2011; Renault 2011; Laparie et al. 2013; Lebouvier et al. 2020). The invasive springtail Hypogastrura viatica was frequent on coastal forelands (in particular at Nordenskjöld Glacier), which underlines the high dispersal capacity of this species that is also introduced on other sub-Antarctic islands and in parts of the South Shetland Islands (Frenot et al. 2005; Greenslade and Convey 2012).

What characteristics of introduced plants and invertebrates make them capable of infiltrating the dynamics of colonisation on glacial forelands? The isolation and harsh environment of sub-Antarctic islands generate environmental filters that may provide opportunities for invaders with a mixture of typical invasive traits that guarantee high resource acquisition and efficient dispersal (Laparie et al. 2013; Liao et al. 2021) and pioneer traits such as low plant height and resistance to abiotic stress (Laparie et al. 2012; Mathakutha et al. 2019; Bazzichetto et al. 2021). Both P. annua and C. fontanum are small annual plants with shallow roots (although P. annua can adopt a perennial life cycle and develop into swards), investing heavily in rapid growth and early reproduction and lack the vegetative and longer-lived tissues typical of the perennial native species that allow multi-year survival under stressful abiotic conditions (Frenot and Gloaguen 1994; Frenot et al. 1998; Chwedorzewska et al. 2015; Johner 2020). It is possible that ruderal traits act as pre-adaptations that provide a colonisation advantage as communities develop around receding glaciers.

Although native plants seem to co-occur with the two invasive pioneers C. fontanum and P. annua, these invasive plants may have impacts on native communities in glacial forelands. During competition experiments in the field in the South Shetland Islands, P. annua reduced the biomass and photosynthetic abilities of C. quitensis and D. antarctica (Molina-Montenegro et al. 2012). On South Georgia, C. quitensis and D. antarctica might also be the native plants most likely to experience direct competition with C. fontanum and P. annua, given their similar early position in the succession on glacial forelands. Competitive interactions between invasive and native taxa are likely to change with ongoing environmental changes and climate warming, possibly to the advantage of introduced species that generally originate from more temperate regions (Molina-Montenegro et al. 2019; March-Salas and Pertierra 2020; Convey and Hughes 2022; Daly et al. 2023). However, our data also show that C. fontanum declines in abundance, while P. annua remains equally common in post-glacial plant communities as they are eventually colonised by native perennial plants, suggesting that native plants can be stronger competitors than some invasive annuals. Despite this, the two invasive plants remain a component of plant communities at least 30 years after deglaciation both in inland and coastal forelands and these taxa are now so widespread on South Georgia that large-scale control is not feasible (Black 2022).

Current and future impacts of introduced species will likely vary between glacier locations. For instance, we found that Nordenskjöld Glacier had higher plant cover, but was also more invaded by introduced plants and invertebrates than the two other tidewater sites. Whether these local differences are due to contrasting topography, microclimate, soil quality or disturbance by macrofauna or human activities is still to be investigated, but they deserve consideration when assessing impacts of biological invasions on glacier-associated communities and developing conservation strategies.

Introduced invertebrates may also have impacts on the terrestrial ecosystems of sub-Antarctic islands (Convey et al. 2010; Houghton et al. 2019). In our study, they were restricted to coastal sites, suggesting that inland sites might provide refuges for native taxa (but see Lebouvier et al. (2020) who note that M. soledadinus is now invading inland and higher altitude locations on the Kerguelen Islands). However, with very limited representative survey data available away from coastal locations on South Georgia, it is also unclear whether inland areas would be suitable for many/most native taxa given their high degree of isolation, inhospitable conditions and low diversity of habitats. The presence of the predatory M. soledadinus is of great concern for the native invertebrate diversity as it can locally drive prey species to extinction, impact the life cycle of co-existing species and fundamentally reshape invertebrate communities (Convey et al. 2011; Lebouvier et al. 2020). The introduced springtail, H. viatica, might act as a food source to the introduced predator, as well as competing directly or indirectly with native species, such as Cryptopygus antarcticus (Convey et al. 1999). Our results suggest that communities in newly-deglaciated areas may not be exempt from the negative effects of invasive invertebrates that may alter the trajectory of invertebrate community succession compared to when they are absent.

The early expansion of introduced species likely modifies soil characteristics, provides biomass and generates biotic interactions in newly-deglaciated areas (Badenhausser et al. 2022), which might have consequences for the entire succession process following glacial retreat. In other systems, biological invasions can alter the speed (Gallego-Tévar et al. 2020) and trajectory of primary successions (Flory and Clay 2010), but little is known in the context of glacier-associated communities. Moreover, the impacts of invasive species in glacial forelands may not be exclusively negative (Walther et al. 2009), as some native taxa might benefit from a modified succession. Our study highlights the need for future research to understand if and how introduced taxa can alter the trajectory and speed of colonisation dynamics following glacial retreat.

Overall, we found an increase in the cover and diversity of plants along the chronosequences in glacial forelands. This is consistent with the basic process of primary succession following glacial retreat (Jones and Henry 2003; Flø and Hågvar 2013; Vater and Matthews 2015; Glausen and Tanner 2019; Gwiazdowicz et al. 2020; Pothula and Adams 2022) and we can presume that a similar trend would have been found with invertebrate communities if sampling quality had allowed. Although our study did not include sites deglaciated more than 30 years previously, we found evidence that the rate of accumulation of vascular plant species decreased along the chronosequence, which contrasts with proglacial successions across the world (Jones and Henry 2003; Pothula and Adams 2022), eventually reaching a plateau representing a high proportion of the native flora of South Georgia. Contrasting with other regions globally where glacier-associated taxa represent a very small proportion of the overall biodiversity, these differences may relate to the specificity of the flora on sub-Antarctic islands, that typically consists of species with higher dispersal ability and adaptations enabling survival in harsh abiotic conditions (Convey 1996).

There were interspecific differences in the colonisation speed of native plants that may help to disentangle the mechanisms underpinning the deglaciation chronosequence. Deschampsia antarctica and Colobanthus quitensis were the first native species to colonise tidewater glacier sites alongside two genera of lichen (Stereocaulon and Pseudocyphellaria), followed by Phleum alpinum and, subsequently, Acaena magellanica, A. tenera, Festuca contracta, Rostkovia magellanica, Galium antarcticum and three bryophyte taxa. Interestingly, D. antarctica and C. quitensis are the only two native angiosperms in the more extreme maritime Antarctic and are known for their high degree of tolerance to adverse conditions (Cavieres et al. 2016; Clemente-Moreno et al. 2020), which may contribute to their ability to colonise very recently-deglaciated sites on South Georgia. The pattern of primary succession following deglaciation on the sub-Antarctic Kerguélen Islands showed some differences to South Georgia, as cushion-forming Colobanthus species and tussock-forming Poa kerguelensis, but not D. antarctica, were amongst the first colonisers (Frenot et al. 1998). The early succession on South Georgia resembles proglacial communities of glaciers at high latitudes and altitudes where lichens, mosses and some grasses are first to colonise, but later trajectories differ due to the presence of shrubs and trees (Jones and Henry 2003; Nakatsubo et al. 2005; Garibotti et al. 2011; Fickert and Grüninger 2018; Ruka et al. 2023). Regional idiosyncrasies in the successional colonisation of deglaciated areas on sub-Antarctic islands are likely a result of missing species in disharmonious floras which may lead to increased vulnerability of developing native communities to invasive plants with traits largely absent for the native species pool. Even across glaciers of the same type on South Georgia, community composition was largely determined by glacier identity, which underlines the importance of the local microenvironment and the composition of adjacent communities in shaping successions following glacial retreat (Bayle et al. 2023). Native invertebrate communities around tidewater glaciers primarily consisted of mites, springtails, dwarf Linyphiidae spiders, Promecheilidae beetles and winged dipterans, resembling proglacial arthropod communities across alpine, sub-polar and polar ecosystems (Kaufmann 2001; Hodkinson et al. 2004; Franzén and Dieker 2014; Hågvar et al. 2020; Moret et al. 2020; Ruka et al. 2023).

The authors have declared that no competing interests exist.

No ethical statement was reported.

This study was funded by the Darwin PLUS scheme [DPLUS144] from the UK Department for Environment, Food and Rural Affairs.

PT, PC, PB, RN and WD designed the study. PT, PC and WD coordinated logistical support. PT collected and analysed the data. PT and WD drafted the manuscript. PT, PC, PB, RN and WD revised the manuscript. All authors agree to be held accountable for the content of this paper and approve the final version.

Pierre Tichit https://orcid.org/0000-0003-0310-6073

The original datasheets containing all species field and laboratory observations and images of surveyed bryophyte and lichen morphospecies can be downloaded as supplementary materials. Samples are stored at the University of Liverpool.