Moss-inhabiting diatom communities from Ile Amsterdam (TAAF, southern Indian Ocean)

1Department of Botany & Zoology, Faculty of Science, Masaryk University, Kotlarska 2, 61137 Brno, Czech Republic 2Université de Rennes, CNRS, EcoBio (Ecosystèmes, biodiversité, évolution) UMR 6553, F-35000 Rennes, France 3Meise Botanic Garden, Research Department, Nieuwelaan 38, B-1860 Meise, Belgium 4University of Antwerp, Department of Biology, ECOBE, Universiteitsplein 1, B-2610 Wilrijk, Belgium *Corresponding author: barbora.chattova@gmail.com RESEARCH ARTICLE


INTRODUCTION
Diatoms (Bacillariophyceae) are one of the most abundant and diverse algal groups in the (sub-)Antarctic region (Jones 1996; Van de Vijver & Beyens 1999a;Sabbe et al. 2003). Although the majority of diatom species are restricted to aquatic habitats, a large number of diatom taxa is able to survive in non-submerged or even dry habitats, such as dry mosses, bark, litter, lichens, wet rocks, and soils (Van de Vijver et al. 2002a;Pfister et al. 2017;Foets et al. 2020). Diatom communities can be strongly influenced by environmental differences in their habitat. Within the (sub-) Antarctic region, there are several studies on local factors controlling the diatom community structure. The structure of moss-inhabiting diatom communities determined by water availability was demonstrated on several sub-Antarctic localities such as Heard Island ( Van de Vijver et al. 2004b) and the Prince Edward Islands ( Van de Vijver et al. 2008b). The soil diatom communities on Ile de la Possession (Iles Crozet) appeared to be structured by available moisture, phosphate and sulphate content (Van de Vijver et al. 2002b). Furthermore, the moss-inhabiting diatom communities in the Maritime Antarctic Region also appeared to be strongly influenced by water availability (Kopalová et al. 2014). In addition, the endurance and survival strategies of diatoms have been intensively studied in the form of controlled laboratory experiments, focusing mainly on their tolerance to extreme conditions such as freezing and desiccation (Souffreau et al. 2010(Souffreau et al. , 2013aHejduková et al. 2019).
The present study is the second in a series of three papers characterising the non-marine diatom diversity from freshwater, moss, and soil habitats on Ile Amsterdam, one of the most remote islands in the world, located north of the sub-Antarctic islands Iles Crozet and the Prince Edward Islands in the southern Indian Ocean, halfway between South Africa and Australia. The first paper described the ecological preferences of the freshwater diatom flora (Chattová et al. 2014), while a third paper discussing the soil diatom communities on Ile Amsterdam is currently being prepared. These ecological analyses of the Ile Amsterdam diatom communities followed a thorough morphological and taxonomical revision that led to the description of many new species (Van de Vijver et al. 2008a, 2012, 2014bLowe et al. 2013; Van de Vijver & Cox 2013;Chattová et al. 2017Chattová et al. , 2018. Chattová et al. (2014) presented a history of diatom research on Ile Amsterdam, starting in 1999 with a first small study on the diatom flora in a handful of samples around the scientific station, where almost 100 different taxa were observed ( Van de Vijver & Beyens 1999c). Most of these samples were taken from dry moss vegetation, rehydrated by rain.
Bryophyte communities represent an important part of the sub-Antarctic vegetation. Sphagnum peatlands are absent from all sub-Antarctic islands in the southern Indian Ocean, except for Ile Amsterdam and Ile Saint-Paul where several (endemic) Sphagnum species, forming extensive peatlands above 500 m a.s.l., have been discovered (Flatberg et al. 2011). The moss vegetation in the sub-Antarctic region is clearly inhabited by highly diverse diatom communities. One of the first papers discussing the sub-Antarctic mossinhabiting diatom flora provided an analysis of epiphytic diatom communities from Campbell Island (Hickman & Vitt 1974), reporting the presence of 59, mostly cosmopolitan, diatom species. In recent years, the moss diatom flora of sub-Antarctic islands has been increasingly well studied, with 104 taxa reported from terrestrial mosses of South Georgia (Van de Vijver & Beyens 1997), more than 190 taxa found on aquatic and terrestrial mosses of Ile de la Possession in the Crozet archipelago (Van de Vijver & Beyens 1999b), and 192 taxa reported from aquatic and terrestrial mosses of Heard Island ( Van de Vijver et al. 2004b). A highly diverse diatom flora (170 taxa) was reported in a combined dataset of freshwater samples and aquatic moss samples from Iles Kerguelen (Van de Vijver et al. 2001), another archipelago in the southern Indian Ocean. Moreover, Gremmen et al. (2007) examined the distribution of moss-inhabiting diatoms (130 species) along an altitudinal gradient on Iles Kerguelen. A final paper, describing the diversity and distribution patterns of freshwater and moss-inhabiting diatoms on the Prince Edward Islands, was published a year later and lists 214 diatom taxa (Van de Vijver et al. 2008b). The results of the aforementioned studies together with recent taxonomic revisions of the sub-Antarctic diatom flora (e.g. Van de Vijver et al. 2004a, 2014a, 2014b have uncovered a number of endemic species previously misidentified as cosmopolitan taxa, due to "force-fitting" of species into American or European names (Tyler 1996;Sabbe et al. 2003).
The main objective of the present study is to characterise and discuss the diatom communities living on aquatic and terrestrial mosses on Ile Amsterdam, in order to reveal the principal factors influencing the moss-inhabiting diatom diversity. Furthermore, we aim to compare the moss diatom flora of Ile Amsterdam with the floras of other sub-Antarctic islands to add to our biogeographical knowledge of the sub-Antarctic diatom flora.

Study site
Ile Amsterdam (77°30′E, 37°50′S) (supplementary file 1), with a total surface area of 55 km 2 , has the shape of a small cone culminating at 881 m a.s.l. (Mont de la Dives) and is geologically very young, with its main part appearing during the period 400-200 ka (Giret 1987;Doucet et al. 2003). Permanent waterbodies are restricted to the higher plateau (Caldera, Plateau des Tourbières) in the centre and the westsouthwestern part of the island (Falaises d'Entrecasteaux, Grandes Ravines). Almost all other areas lack (semi-) permanent waterbodies due to the steepness of the slope and the extreme permeability of the substrate and underground, composed of numerous lava tunnels, holes, and fissures. Due to frequent fog, low clouds, and a high relative humidity in the upper areas (from 500 m a.s.l.), extensive peat formations have developed, especially in the volcanic caldera (Heger et al. 2009). On the higher central plateau, the vegetation has a typical sub-Antarctic character, composed of mosses, small ferns (e.g. Austroblechnum penna-marina (Poir.) Gasper & V.A.O.Dittrich), grasses, sedges such as Uncinia brevicaulis (Mack.) Thouars, and several Lycopodium species (Tréhen et al. 1990). Reflecting the isolation of the island, the number of native spermaphyte species is low (17) but 56 species were introduced (Frenot et al. 2001). The lichen flora is still incompletely known but in 2011, Aptroot et al. reported 77 species, including the newly described Caloplaca amsterdamensis Aptroot & Ertz (Aptroot et al. 2011).

Sampling
In total, 148 moss samples were collected during a field campaign in November and December 2007. Both aquatic and terrestrial mosses were sampled spanning a broad range of moisture content (as expressed by the F-values of Jung 1936). The sampling sites are shown in supplementary file 1. Each sample consisted of at least 10 cm 3 of the living part of the moss, without rhizomes. Samples were collected in PVC bottles and fixed with 3% formaldehyde. When possible, the samples were geographically localised using GPS, accompanied by a detailed site description. When possible, pH, conductivity, and water temperature were measured in situ using a WTW 340i Multimeter. For 25 samples, which were all aquatic mosses collected from pools and rivers, water was collected 10 cm below the water surface, filtered in situ, and deep-frozen to be subsequently analysed. , Cl -, Na + , K + , Mg 2+ , and Ca 2+ were analysed at the Laboratory for Ecosystem Management (University of Antwerp, Belgium) using a continuous flow analysis (CFA-SKALAR). Supplementary file 2 list all samples used for further analyses, together with their characteristics.
The moisture content of each sample was determined using the F-classification of Jung (1936). It is a humidity scale based on water content: FI = submerged mosses, FII = free floating mosses, FIII = very wet (water drips from the samples without pressure), FIV = wet (water drips with a slight pressure), FV = quasi-wet (water drips after moderate pressure), FVI = moist (little water produced after high pressure), FVII = quasi-dry (only a few drops of water can be squeezed out), FVIII = dry (contains no water).

Sample preparation and diatom counting
Sample contents were randomised before preparation and prepared for light microscopy analysis following the method described in van der Werff (1955): small subsamples (± 2 cm 2 ) of moss fragments, with an addition of distilled water, were cleaned by adding 37% H 2 O 2 and heated to 80°C for about 1 h. The reaction was completed by addition of an excessive amount of saturated KMnO 4 . Following digestion and centrifugation (three times 10 min at 3700 rpm), cleaned material was diluted with distilled water to avoid excessive concentrations of diatom valves on the slides. Cleaned diatom valves were mounted in Naphrax ® . Samples and slides are stored at the BR-collection (Belgium). In each sample, 400 diatom valves were identified and enumerated on random transects at 1000× magnification (oil-immersion), using an Olympus BX53 microscope, equipped with Differential Interference Contrast (Nomarski) optics and the Olympus UC30 Imaging System. Slides were scanned after counting 400 valves in order to find rare species. For identification of Antarctic and sub-Antarctic species, the following publications were consulted: Van de Vijver et al. (2002a, 2002b, 2004a, 2008a, 2012, 2014b; Van de Vijver & Cox (2013); Lowe et al. (2013); Wetzel et al. (2015).

Data analysis
For a pairwise comparison of the diatom flora of Ile Amsterdam with those of the other sub-Antarctic islands, the Community Coefficient of Sørensen (1948) was used with the following formula: 2c/(a+b+2c), where a and b are the number of species observed exclusively in each of the two sites and c is the number of species shared between these sites. The Ile Amsterdam moss inhabiting diatom flora was compared with terrestrial and aquatic moss diatom flora from Heard Island ( Van de Vijver et al. 2004b), Ile de la Possession (Van de Vijver & Beyens 1999b), Prince Edward Islands ( Van de Vijver et al. 2008b), and with aquatic moss diatom flora from Iles Kerguelen ( Van de Vijver et al. 2001). Datasets were made taxonomically consistent prior to the similarity analysis.
To evaluate the extent to which our sampling effort represented the moss-inhabiting diatom flora of Ile Amsterdam, the incidence-based species richness estimator (ICE, Chao et al. 2000) and the mean Chao2 richness estimator (Chao 1984) were calculated, both using the EstimateS program v.9.1.0. (Colwell 2013).
Non-metric Multidimensional Scaling (NMDS) was performed to reduce the multidimensional species data matrix into two dimensions best reflecting site dissimilarities given by diatom species composition. NMDS was based on Bray-Curtis (Sørensen) dissimilarity calculated on square root transformed species data.
Classical agglomerative clustering was used to group diatom assemblages according to the same Bray-Curtis distances using the unweighted per group average algorithm (UPGMA). Resulting groups were projected onto the NMDS ordination as well as best indicator species of the (combinations of) site groups. Indicator species were identified using the extended Indicator Species Analysis. The indicator value (IV) of an indicator species is a number calculated as the product of two quantities, called A and B, ranging from 0 to 1. Quantity A gives the probability of a site being a member of the site-group combination when the species was found at that site. Quantity B informs how frequently (and hence how easily) the species is found at sites of the site-group combination under study (De Cáceres et al. 2010). Unless otherwise indicated, statistical analyses were performed in the R statistical environment (R Core Team 2017; RStudio Team 2015) using the packages vegan v.2.4-6 (Oksanen et al. 2018) and ggplot2 (Wickham 2016).

Species composition and diversity
The analysis of 148 samples revealed the presence of 122 diatom taxa belonging to 38 genera. Three additional taxa were observed outside the counts bringing the total number of moss-inhabiting diatoms to 125 taxa. Twenty-one samples contained (almost) no diatoms, even after scanning entire slides, and have therefore been removed from further analysis. The distribution of the species numbers per sample (supplementary file 3A) showed that most samples contained between 11 and 15 taxa per sample. The average number (and standard deviation) of taxa per sample was 12 ± 7, with a median of 11. The asymptotic flattening towards the end of the species accumulation curve (supplementary file 3B) indicates that a large part of the moss-inhabiting diatom   ). On the other hand, 108 taxa (more than 87% of all counted taxa) had a total relative abundance of less than 1%. Table 1 provides an alphabetical list of all observed species together with their biogeographical distribution. The genera Pinnularia (28 taxa), Eunotia (nine taxa), Nitzschia (eight taxa), and Humidophila (seven taxa) were the most species-rich genera.
Other species-rich genera include Sellaphora (five taxa) and Hantzschia, Luticola, Psammothidium, Stauroneis (four taxa each). Supplementary file 4 lists all genera arranged according to their species number. Eleven taxa (9%) could only be identified to the genus level. Some of these are most likely new to science and will have to be formally described (mainly in the genera Nitzschia, Navicula, Pinnunavis, and Sellaphora). Forty percent of all observed taxa showed a typical cosmopolitan distribution, 17% had a restricted sub-Antarctic distribution, while another 22% can be considered endemic to Ile Amsterdam (supplementary file 3C). The similarity analysis based on presence/absence data indicates that the moss diatom flora of Ile Amsterdam only showed a limited affinity to other sub-Antarctic islands, with Sørensen index values ranging from 0.24 to 0.48 (table 2). The highest similarity was noted with Ile Saint-Paul, whereas the lowest similarity was observed with Gough Island. The latter is part of the Tristan da Cunha archipelago and located in the southern Atlantic Ocean, at approximately the same latitude as Ile Amsterdam.

Community analysis
The NMDS analysis, based on a cluster dendrogram (supplementary file 5), divided the samples into six main groups ( fig. 1) and Eunotia manguinii. The samples were characterised by a slightly acidic pH, higher specific conductance values, and higher Cland Na + values, possibly the result of sea spray.
The fourth group was exclusively composed of highelevation acidic samples with very low specific conductance, nutrient and ion values. This group was characterised by extremely high abundances of its indicator species Frustulia lebouvieri (IV 0.84) and Eunotia paludosa (IV 0.97  The fifth group contained samples collected mostly in the coastal ravines of the Falaises d'Entrecasteaux (southern part of the island) and Del Cano, often taken from mosses or liverworts growing near or in a waterfall. These samples had higher pH values (7.6 ± 0.9) and a higher conductivity and SO 4 2-, Mg 2+ , Ca 2+ , and K + values. The amount of Cland Na + was also high, reflecting the vicinity of the ocean and the possible effect of sea spray. The samples were almost exclusively dominated by Planothidium subantarcticum (IV 0.97) with Sellaphora cf. seminulum, Nitzschia soratensis E.Morales & M.L.Vis, and Halamphora veneta (Kütz.) Levkov as co-dominant taxa.
The last group comprised of moist and dry samples (moisture categories FIV-FVII), collected mainly from caves in the lava tunnel. The samples were slightly acidic with elevated Na + and Clvalues and were dominated by Angusticopula aff. dickiei (IV 0.99), Ferocia setosa (Grev.) Van Figure 3 shows micrographs of typical taxa in the groups 1-5, while fig. 4 illustrates the cave-inhabiting diatom community of group 6.
The relationship between species richness and environmental variables was analysed with the samples divided into two main groups, according to elevation (fig.  5). The majority of the samples (84) were found on higher elevations (above 500 m a.s.l.), whereas the rest of the samples originated from lower elevations (0-500 m a.s.l.). The species richness clearly increased with increasing pH and conductivity in the samples of the lower elevations, whereas no apparent pattern could be observed in samples of higher elevations.

Species composition and general biogeography
Based on the relative abundance data, the most reported taxon in the moss vegetation of Ile Amsterdam is Eunotia paludosa, a cosmopolitan species frequently reported from sub-Antarctica and typically found in semi-wet terrestrial mosses or in small acidic pools with low conductivity values ( Van de Vijver et al. 2014a). However, within the five most abundant taxa, four taxa (Frustulia lebouvieri, Eunotia manguinii, Planothidium subantarcticum, and Kobayasiella subantarctica) are only present on the southern Indian Ocean   islands and even absent from all other studied Antarctic localities (Zidarova et al. 2016 and references therein). The second most abundant taxon, Frustulia lebouvieri, was so far only found on the Prince Edward Islands, the most northerly of the sub-Antarctic Islands ( Van de Vijver et al. 2008b). It is at the same time the most abundant taxon (30.8% of all counted valves) in the freshwater diatom communities on Ile Amsterdam (Chattová et al. 2014).
The uniqueness of the Ile Amsterdam's diatom flora is reflecting one of the most typical features of oceanic islands: the fact that the flora and fauna of these islands are usually disharmonic (Gillespie 2007) with some genera tending to be overrepresented whereas others are missing or only show a reduced diversity. The genus Pinnularia accounts for more than 22% of all recorded taxa whereas other typical speciesrich genera such as Navicula or Nitzschia are less abundantly present on the island. Nineteen genera are represented by only one species and cymbelloid taxa, common in the Arctic region, are even completely absent. Typical terrestrial genera such as Humidophila and Luticola are quite diversified on Ile Amsterdam. Humidophila species are present worldwide, with a large number of endemic taxa restricted to the (sub)-Antarctic region (Le Cohu & Van de Vijver 2002; Van de Vijver et al. 2002c;Kopalová et al. 2015). Recently, two new Humidophila taxa were described from Ile Amsterdam (Chattová et al. 2018): Humidophila amsterdamensis, found in a sample taken from semi-dry mosses (FVI) that were collected from a small cave bog pond, and Humidophila rouhaniana, which was observed in an almost dry moss sample (FVII) collected from the Jardin Météo in the vicinity of the base Martin-de-Viviès. Both Humidophila taxa were frequently reported with other endemic taxa that were typically found in caves and lava tunnels on Ile Amsterdam, such as Halamphora dagmarobbelsiana Van de Vijver & Levkov, Mayamaea cavernicola, Microfissurata australis Van de Vijver & Lange-Bert., and Orthoseira verleyenii. The other genus, Luticola, is one of the dominant components of the terrestrial diatom flora in Antarctic terrestrial habitats. The genus has one of the highest levels of endemism of any diatom genus in Antarctica, both in terms of total number of species (taxon endemism) and the percentage of the entire genus (phylogenetic endemism) (Kociolek et al. 2017). The Luticola records on Ile Amsterdam confirm this high level of endemism, with three recently described taxa L. beyensii Van de Vijver et al., L. ivetana, and L. vancampiana Chattová & Van de Vijver (Chattová et al. 2017) confined to Ile Amsterdam and/or Ile Saint-Paul. The high diversity of the terrestrial genera is most likely linked to the large diversity in aerophilic microhabitats (moss vegetation on the moist rock faces or in caves) on the island.
The moss-inhabiting diatom flora of Ile Amsterdam is more species-rich (125 species) compared to the 104 taxa recorded from the freshwater samples on the same island by Chattová et al. (2014), and to the 60 taxa observed in the terrestrial samples from Ile Saint-Paul (Chattová 2017). The high species richness is confirmed with the results of the incidence-based species richness estimator, clearly demonstrating that the moss sampling effort reflected the true diatom species richness, with the best results among investigated habitats (between 88 and 90%). The expected total number of taxa in the freshwater samples on Ile Amsterdam is 110 (ICE) or 113 (Chao2) and 72 (ICE) or 69 (Chao2) in the terrestrial samples collected on Ile Saint-Paul, suggesting that the counting protocol scored between 87 and 89% of the total taxa present in the freshwater samples and between 83 and 86% in the case of the terrestrial flora Figure 5 -A scatterplot showing the relation between species richness and environmental variables (elevation, pH, moisture, and conductivity), with the samples divided into two groups, according to elevation (0-500 m a.s.l. and above 500 m a.s.l.). Note that the species richness is clearly increasing with increasing pH and conductivity in the samples of the lower elevations.
of Ile Saint-Paul. This higher species richness of the mossinhabiting diatom flora is most likely the result of a higher diversity of moss (micro) habitats but can also be the result of undersampling and underreporting in the other habitats.
When comparing the species composition of the freshwater (Chattová et al. 2014) and the moss-inhabiting diatom flora, clear differences can be observed. The main dissimilarities are reflected in the proportion of terrestrial species. The freshwater dataset contained only two Hantzschia species, whereas our moss dataset included four different Hantzschia taxa. Moreover, Microfissurata australis Van de Vijver & Lange-Bert., Chamaepinnularia soehrensis var. muscicola (J.B.Petersen) Lange-Bert. & Krammer, or Geissleria sp. were exclusively observed in mosses.
The low similarity values based on presence/absence data between Ile Amsterdam and the other islands are not surprising, considering the differences in microhabitat diversity, the rather large distances between the islands, the isolated position of Ile Amsterdam, and the relatively young geological age of the island. The most similar diatom flora (Sørensen Index = 0.48) was found on the neighbouring Ile Saint-Paul (at 80 km distance), followed by Ile de la Possession (Iles Crozet), which is located more than 1000 km to the south ( The low similarity also confirms the highly specific status of the diatom flora on Ile Amsterdam. A large number of taxa are only unequivocally identified on genus level and are listed in table 1 as cf. or sp. This high number of unidentified taxa is the result of the more fine-grained taxonomy that is nowadays applied when characterising the (sub-)Antarctic diatom flora. Those species show some affinity with known taxa but at the same time have sufficient differences to allow separation as independent taxa. In the past, these unidentified species would most probably have been given European or North American names as a result of a too broad interpretation of the original species descriptions. This historic "force-fitting" (Tyler 1996) had severe consequences for our understanding of the ecology, diversity, and biogeographical distribution of diatoms, as it not only stretched the morphological boundaries of many species, but also reinforced the idea that most diatom species have cosmopolitan distributions and are ecologic generalists.
Similarly, recent molecular studies of the diatom flora from the sub-Antarctic islands and the Maritime Antarctic Region indicate that the cosmopolitan nature of many nonmarine (sub-)Antarctic diatoms might be overestimated. Distinct Antarctic lineages have been confirmed within two widely distributed diatom species complexes, Pinnularia borealis and Hantzschia amphioxys (Souffreau et al. 2013b;Pinseel et al. 2019). Pinseel et al. (2020) included several samples of P. borealis from Ile Amsterdam and Ile Saint-Paul in their analysis, revealing multiple distinct P. borealis species present on both islands. Based on the dataset of Pinseel et al. (2020) (their Supplementary Fig.  5), those distinct P. borealis species have only been found in Patagonia, Tasmania, and on other sub-Antarctic islands in the southern Indian Ocean. These data, showing high levels of hidden species-level diversity of P. borealis in sub-Antarctica including Ile Amsterdam, suggest the presence of several endemic P. borealis species in the area. It is therefore likely that the true species-level diversity and endemicity of the Ile Amsterdam's diatom flora remain hidden and underestimated. Further molecular research is needed to shed a more detailed light on the cryptic diversity and biogeography of other presumed cosmopolitan diatom species in (sub-)Antarctica, to unravel the true nature of such cosmopolitan morphospecies.

Moss-inhabiting diatom communities
Both environmental data and diatom distributions indicate that apart from the effect of elevation, conductivity, pH, and moisture are the major factors determining the structure of moss-inhabiting diatom communities on Ile Amsterdam. The effect of nutrients and ions as determining factors remains unclear as only a small part of the samples (25) was analysed for physico-chemical parameters and drawing any conclusions should be done with caution.
No elevated phosphorous values were detected in the samples included in this study, all samples showed low PO 4 3--P values (0.01-0.06 mg/l). This is in contrast to an elevated phosphorous value (1.58 mg/l) reported by Chattová et al. (2014) from one sample (W031) taken from a pool in the middle of a fur seal colony, where the presence of marine mammals together with soil erosion might have caused a higher phosphorous load in that sample. None of the moss samples was taken from areas heavily influenced by animals, which is a possible explanation of the generally quite low nutrient values. The high levels of Cland Na + in the coastal samples of groups 3, 5, and 6 most likely result from sea spray that is continuously deposited in the streams overflowing the moss vegetation, combined with soil erosion and strong winds.
The mosses in the region near the Falaises d'Entrecasteaux and Pointe Del Cano were taken from waterbodies showing an almost circumneutral to clearly alkaline pH, whereas the waterbodies in the upper Caldera are strictly (very) acidic. The areas of d'Entrecasteaux and Del Cano (groups 3 and 5) represent a similar habitat of very steep cliffs with small ravines and their diatom communities show some clear similarities. Samples of both groups are dominated by taxa preferring circumneutral to alkaline conditions such as Sellaphora cf. seminulum, Rhopalodia rupestris, and several Nitzschia taxa (Hofmann et al. 2011). However, the groups significantly differ in their dominant (indicator) species, with Platessa oblongella and Psammothidium investians characterising the third group of the slightly acidic to circumneutral samples from Del Cano, while Planothidium subantarcticum almost exclusively dominates the alkaline samples of the fifth group. In contrast, the alkaline moss-inhabiting diatom communities on Ile de la Possession (Van de Vijver & Beyens 1999b) and Iles Kerguelen ( Van de Vijver et al. 2001, Gremmen et al. 2007 A major habitat difference between the Caldera region on Ile Amsterdam and the other sub-Antarctic islands is the dominance of Sphagnum taxa as peat-forming mosses on Ile Amsterdam, while Sphagnum taxa are completely absent from the sub-Antarctic islands (Flatberg et al. 2011). Sphagnum is well known for acidifying its environment (Munson & Gherini 1993;Siegel et al. 2006) and this activity in the isolated Caldera region in the combination with the flat geomorphology of the Caldera most likely created the highly acidic and oligotrophic habitat where the acidic diatom community could develop (Chattová et al. 2014). Those relatively uniform and species-poor high elevation diatom communities in the Caldera region are in strong contrast to the diatom communities from the coastal areas of Del Cano and Falaises d'Entrecasteaux. The explanation can be found in the geological history of the island. Ile Amsterdam has been shaped by two consecutive volcanic episodes resulting in a formation of mostly subalkaline, magnesium-rich tholeiitic basalt bedrock (Doucet et al. 2004). This geochemical composition probably influenced the development of the slightly acidic and subalkaline diatom communities of Del Cano and d'Entrecasteaux. Those areas most likely represent the original habitat for diatom communities on Ile Amsterdam.
The typical terrestrial taxa such as Hantzschia amphioxys complex, Humidophila brekkaensis, Pinnularia borealis complex, or Orthoseira roeseana, forming the driest assemblages of group 1, play only a minor role or are completely absent from Ile de la Possession and Gough Island (Van de Vijver & Beyens 1999b; P. Vinšová unpubl. res.). The only similar feature of the terrestrial diatom flora on Ile Amsterdam with those on Ile de la Possession is the dominance of Humidophila cf. contenta. Differences can be found in the associated diatom flora of these communities, mainly composed of Planothidium aueri (Krasske) Lange-Bert. and Fragilaria cf. vaucheriae. The most similar terrestrial diatom flora was found on dry mosses from South Georgia, with the samples dominated mainly by well-known, cosmopolitan taxa, such as Hantzschia amphioxys complex, Pinnularia borealis complex, Orthoseira roeseana, and Humidophila cf. contenta (Van de Vijver & Beyens 1997).
Moisture content is one of the principal factors determining the diatom assemblages in moss communities in the (sub-)Antarctic region (Van de Vijver & Beyens 1999b; Kopalová et al. 2014) and it is one of the factors co-responsible for the classification of the main groups of moss-inhabiting diatom communities on Ile Amsterdam. A clear difference between diatom communities growing on wet versus dry mosses was found (figs 1, 2). Moisture can therefore be considered as an important factor, even when the wettest samples are not the most species-rich on the island, resulting most likely from the fact that the majority of the wet samples are the most acidic, with lower conductivity values. Moreover, the highly diverse cave samples represent at the same time some of the driest habitats on the island. Future diatom research on Ile Amsterdam should focus on the influence and specificity of the cave and lava tube microhabitat. To accomplish this, a large set of new lava tube and cave samples was recently (austral summer of 2016; B. Van de Vijver pers. comm.) collected on Ile Amsterdam and are currently being investigated.

CONCLUSION
The isolated geographic position and physical features of the island have resulted in the presence of a unique diatom flora, characterised by many species found to be endemic to Ile Amsterdam or sub-Antarctica. This study revealed a high number of diatom taxa unknown to science that will need formal description in the future. Future molecular research could lend another perspective and help to reveal the true extent of the likely underestimated diatom diversity and endemism on Ile Amsterdam. The moss-inhabiting diatom flora on Ile Amsterdam showed a limited affinity to the mossinhabiting diatom flora from several other sub-Antarctic islands, indicating the presence of clear biogeographical patterns in the area. Our findings provide useful information for future investigations, including research on the influence and specificity of the cave and lava tube microhabitat on diatoms, the application of diatoms as proxy in palaeoecological research, and biogeographic or molecular studies.