Introduction

There is continued interest in understanding how multiple factors control species distribution, abundance and diversity across environmental gradients at different scales (Götzenberger et al. 2012). With particular regard to Maritime Antarctica, this interest has been growing in recent decades due to its importance for climate change monitoring (Thomazini et al. 2018) and its impacts on high-latitude plant diversity (Amesbury et al. 2017). However, few studies have focused on the factors shaping the assembly of plant communities along gradients in this region (e.g. Leishman and Wild 2001).

At landscape scale, the availability of ice-free areas determines the distribution of terrestrial vegetation in Antarctica (Leishman and Wild 2001). These areas are restricted to coastal regions, rocky slopes, or nunataks and soils at different stages of development (Campbell and Claridge 1987; Bokhorst et al. 2007). The vegetation is mostly composed of bryophytes and lichens, with only two native angiosperms species Deschampsia antarctica Desv. and Colobanthus quitensis (Kunth) Bartl. (Putzke and Pereira 2001; Victoria et al. 2013). Differences in species composition can found within small areas (Victoria et al. 2009; Schmitz et al. 2018) due to high environmental heterogeneity (microtopography, geology, soil properties; Victoria et al. 2013; Benavent-González et al. 2018).

In this study we evaluated the relationship between richness and composition of cryptogamous communities and soil chemical and physical properties along a pedoenvironmental gradient on Elephant Island, Maritime Antarctica. Cryptogamous communities inventory data from 300 plots (20 × 20 cm), distributed among fifteen pedoenvironments, were used to evaluate the effect of soil texture and soil fertility on the species richness and composition. In this context, the following questions were addressed: (1) What is the richness, species composition and coverage pattern along a pedoenvironmental gradient in Maritime Antarctica? (2) Is there variation in species associations along the pedoenvironmental gradient? (3) What are the effects of soil chemical and physical properties on species richness and composition?

Material and methods

Study area

This study was conducted at Stinker Point (61°21′ S, 55°20′ W), located on Elephant Island (Fig. 1). Elephant Island is located north of the South Shetland archipelago, between the Drake Passage and the Weddell Sea (Pereira and Putzke 1994) and has a cold climate with extreme fog and snowfall conditions. Mean annual temperature is −2°C, and mean summer temperature is 1°C (Šťastná 2010; Putzke et al. 2020). Snow storms and fog are frequent at Stinker Point (Putzke et al. 2020).

Fig. 1
figure 1

Location of the study area: Antarctica (a), South Shetland Islands (b), Elephant Island (c), Stinker Point with fifteen sampled pedoenvironments (d)

Full size image

Geologically, the island is composed of Paleozoic metamorphic rocks of the Skotia arc (O’Brien et al. 1979). Ice-free areas are found mainly in coastal areas, on raised marine deposits mixed with glacial drift, forming alternating cliffs with steep glaciers fronts or ice cliffs (Navas et al. 2018).

Stinker Point is one of the largest ice-free areas of Elephant Island (CGE-UAM-UFRJ 2005; Navas et al. 2018) and possesses a rich fauna and flora (Pereira and Putzke 1994). These authors recorded the occurrence of two native angiosperms for the Maritime Antarctic region, Colobanthus quitensis (Kunth.) Bartl. and Deschampsia antarctica Desv., 38 species of moss, seven liverworts, 68 lichens, two species of terrestrial algae and four macroscopic fungi.

Selection of pedoenvironments

We selected fifteen pedoenvironments (defined here as P1 to P15) on different landforms, following the types described by López-Martínez et al. (2012) and Navas et al. (2018), see Table 1, with contrasting soils (Fig. 2).

Table 1 Description of the 15 sampling pedoenvironments (P). Where: R= Species Richness; Cover = percentage of vegetation cover in the pedoenvironment
Full size table
Fig. 2
figure 2

Barplots soil properties. For analysis, available: pH (H2O), exchangeable acidity (H + Al), sum of bases (SB), Ca2+, Mg2+, Na, P, K+, organic matter (MO), Cu, Mn, Fe, potential cation exchange capacity (T), percentage of bases saturation (V), remaining phosphorus (P Rem), aluminium saturation (m), Zn, and soil texture as coarse sand, fine sand, clay and silt contents were included

Full size image

Cryptogamous communities survey

On each pedoenvironment 20 plots of 20 × 20 cm were sampled, following the Braun-Blanquet (1932) square methodology, adapted to Antarctic vegetation conditions, to assess the coverage of each species at each plot. In total, 300 plots were sampled in the austral summer of 2017/2018. Bryophytes were identified at species level in each plot using the taxonomic identification keys by Putzke and Pereira (2001), and Ochyra et al. (2008); Redon (1985), Øvstedal and Lewis-Smith (2001), and Olech (2004) were used to identify lichens. Lichens and mosses collected were deposited in the HCB Herbarium of the University of Santa Cruz.

Ecological significance index (ESI)

The importance of the species in each pedoenvironment, the plant community’s classification and species association, were determined using the ecological significance index (ESI), which combines the frequency and coverage of each species in the plot (Lara and Mazimpaka 1998; Marques et al. 2005). These researchers defined the ecological significance index as:

$$ \mathrm{ESI}=F\left(1+C\right) $$
(1)
$$ F=100x/n, $$
(2)

where F is the relative frequency of the species in the area or habitat and is generated by the number of occurrences (x) divided by the total number of samples considered (n), and C is the average coverage of the species in the samples, calculated as:

$$ \mathrm{C}=\sum (ci)/x, $$
(3)

where ci is the class of coverage and x is the number of samplings in which the species occurs (e.g. Schmitz et al. 2018). This index determines the scale of importance of the species in the area, which ranges from 0 to 600, where values above 50 indicate ecological significance (Victoria and Pereira 2007). The species with the highest values and their form of growth define the name of the community, following the classification by Longton (1988). The associations are characterized by dominant species or by restricted occurrence in more specific habitats (Schmitz et al. 2020).

Soil collection

In order to measure the soil properties within each plot, a topsoil sample was taken (at 0–10 cm depth). Soil properties were measured following standard protocols (EMBRAPA 1997). The following parameters were assessed: Soil pH was determined in distilled water (pH H2O), phosphorus available exchangeable (P), potassium (K+), sodium (Na), calcium (Ca2+), magnesium (Mg2+), aluminium (Al3+), potential acidity (H + Al), sum of bases (SB), potential cátion exchange capacity at pH 7.0 (T), percentage of bases saturation (V), aluminium saturation (m), sodium saturation index (ISNa), organic matter (OM), remaining phosphorus (P Rem), copper (Cu), manganese (Mn), iron (Fe), zinc (Zn), total nitrogen (N) and soil texture, classified as coarse sand (Sand_c), fine sand (Sand_t), clay and silt contents.

Data analysis

All analyses were carried out using the R Environment (R Core Team 2018). For soil variables we tested normal distribution with the Shapiro-Wilk test and by evaluating the Q-Q plot; The homogeneity of variances was assessed with Bartlett´s test using the ‘dplyr’ package (Crawley 2013). To compare soil properties (non-normally distributed data), between pedoenvironment sites, we used Kruskal-Wallis’s test followed by a post hoc Dunn’s test performed with the ‘dunn.test’ package (Dinno 2017).

We compared species richness patterns between pedoenvironments using sample-based data to estimate rarefaction and extrapolation curves based on the first Hill number (Chao et al. 2014). Extrapolations were made based on presence/absence of species in the plot data using Hill number order 0 (Colwell et al. 2012), with the purpose of evaluating the sampling sufficiency in each pedoenvironment. Estimates were obtained using the ‘iNEXT’ package (Hsieh et al. 2016). The Hill number was estimated as the mean of 100 replicate bootstrap runs to estimate 95% confidence intervals (e.g. Schmitz et al. 2020). Whenever the 95% confidence intervals did not overlap among assemblages at each pedoenvironment, and species numbers differed significantly at P < 0.05, we considered in the analysis (Colwell et al. 2012).

We performed non-metric multidimensional scaling (NMDS) to analyse differences between pedoenvironments in terms of species composition using Euclidean dissimilarities (Clarke 1993). The NMDS was performed using the ‘metaMDS’ function of the ‘vegan’ package (Oksanen et al. 2018). Permutational multivariate analysis of variance (PERMANOVA, 9999 permutations) was used to determine differences in species composition using the ‘adonis’ routine available within the ‘vegan’ package (Oksanen et al. 2018).

Soil variables were summarized using principal component analysis (PCA) to identify a possible pedoenvironmental gradient and to reduce the number of redundant soil properties, following Qian et al. (2014) and Villa et al. (2018); all variables were centred and standardized. We also calculated Pearson correlations among soil properties and the PCA ordination axes. The PCA was performed using the ‘FactoMineR’ package (Husson et al. 2017). To investigate a possible relationship between soil properties and biotic (species) variables, canonical correspondence analysis (CCA) was used. CCA examined similarity or dissimilarity in the floristic composition of plots along the pedoenvironmental gradient. The significance of each soil variable in determining species compositional change was assessed by Monte Carlo randomizations (999 randomizations). The CCAs were performed using the ‘ggord’ and ‘ordiplot’ functions of the ‘vegan’ package (Oksanen et al. 2018). Species coverage distribution was evaluated using species-rank curves, for each study area, by ranking all species from the most to the least abundant (Magurran 2004).

We evaluated the effect of potential predictors on species richness and species composition (e.g. extracting the scores on frequency-weighted NMDS Axis 1; Oksanen et al. 2018; Villa et al. 2018) via linear mixed effect models (LMMs). We have used a nutrient-based PCA (i.e. P, K+, Na, Ca2+, Mg2+, Cu, Mn, Fe, Zn, N) to extract the scores of variability of nutrient-related soil fertility (PCA1f), and after we used a texture-based PCA (clay, silt, sand) to extract the scores of variability of physical properties-related soil texture (PCA1t). To reduce any strong correlations among local environmental conditions (Fig. S2 in the Electronic supplementary material), we used the two axes of the PCA for soil fertility (PCA1f) and texture (PCA1t) variables. Thus, the first PCA axis was considered as a proxy for soil fertility and soil texture gradient across all the tested models (Ali et al. 2016; Villa et al. 2018). We used generalized linear mixed effects models (GLMMs) with Poisson error distribution to investigate the effect of individual soil properties, soil fertility, and soil texture on species richness. Species composition was assessed using LMMs after checking the Shapiro-Wilk test for normality and Q-Q graphs (Crawley 2013). Predictor variables (fixed effects) were plant coverage, soil fertility (PCA1f) and soil texture (PCA1t), defined as the first principal component from PCA, considering all eighteen parameters analysed (see above), as well as further soil parameters such as sand, silt, clay, SB, OM and pH. Plant coverage is defined as the proportion of plant community space occupied in each pedoenvironment (Ji et al. 2009); therefore, plant coverage is considered to be an important predictor with possible effects on plant diversity (Ji et al. 2009; Sanaei et al. 2018). Soil chemistry and texture were used, as well as individual soil parameters, as explanatory variables for modelling, because individual soil parameters may also show a direct effect on species richness and species composition (Villa et al. 2018; Schmitz et al. 2020). For predictor selection, we assessed collinearity between selected predictor variables using Spearman correlation analysis; when two variables were strongly correlated (r ≥ 0.6) the most ecologically-relevant predictors were selected, which were included in separate models (Fig. S2). In all mixed models, pedoenvironments were included as a random factor and the first axis of the soil chemistry and texture PCA was used as a fixed factor (Villa et al. 2018, Schmitz et al. in press). All models were fitted using the package ‘lme4’ (Bates et al. 2014) from the R platform (R Core Team 2018); for illustration, we used the package ‘ggplot2’ (Hadley 2015).

To assess the best models (GLMMs and LMMs), a multi model inference approach was applied (Burnham and Anderson 2002) with the ‘dredge’ function from the ‘MuMIn’ package (Barton 2015), which returns all possible combinations of the explanatory variables included in the global model. To determine which of these variables were the most decisive in explaining changes in species richness and species composition, we used an information theory approach based on the Akaike information criterion (AIC) with a correction for finite sample sizes (AICc) and AIC weights (Burnham and Anderson 2002). The model with the lowest AICc was considered to be the best, but all models that differed by fewer than four units from the best model were considered as equally good models (Burnham et al. 2011).

Results

General soil attributes

The soil properties differed among all pedoenvironments studied, forming a marked edaphic gradient (Fig. 2). The soils ranged from highly acidic to alkaline (pH 4.19 to 8.18). Available phosphorus values ranged from 17.44 to 870.44 (mg/dm3), of which only P2 and P4 showed no ornithogenic influence, with lower levels of P. The amounts of soil organic matter were low except for P2 (16.94 dag/kg), P4 (35.39 dag/kg), and P5 (35.39 dag/kg), while the total OM were high for Maritime Antarctica standards. The soil was dominated by the sand fraction, sandy loamy as a predominant textural class, but silt and clay levels were high in P5, P7, P10, P12, and P13 (Table S2 from ESM). The soil texture of P4 could not be measured due to the very high amount of organic matter.

Species composition and species richness pattern

We identified 39 species in the fifteen pedoenvironments studied; 21 bryophytes (19 mosses and two liverworts), fifteen lichen species, two angiosperms (D. antarctica and C. quitensis) and a macroscopic alga (Prasiola crispa, Table A.1 from ESM). The families with the highest species richness were the Bryaceae for the bryophytes and the Parmeliaceae for lichens, both with four species. A distinct richness pattern was observed among the pedoenvironments (Fig. S1 in the Electronic supplementary material), where P1 presented the smallest number, with only two species, and P15 had the highest richness, with nineteen species (Table 1). Coverage also varied significantly among pedoenvironments (Tab. 1, Fig. S1B).

The NMDS revealed that species composition varied significantly between pedoenvironments (PERMANOVA F14,299 = 33.70, P < 0.001), but with a marked overlap among most groups (Fig. 3). Sanionia uncinata was dominant, being present in 139 of the 300 sampled plots; other species present in a large number of plots were Sanionia georgicouncinata (82), Cystocoleus niger (72), and Usnea antarctica (71) (Table S1). On the other hand, the occurrence of some species was restricted to some pedoenvironments, as can be observed in the dendrogram cluster analysis (Fig. 4). For example, the moss species Andreaea gainii only occurred in P5, Syntrichia magellanica only occurred in P7, Hypnum revolutum was recorded only in P6, Ceratodon grossiretis in P9, Brachythecium sp. only in P14, and, finally, Leptogium puberulum, Physconia muscigena and Placopsis contortuplicata only occurred in P15.

Fig. 3
figure 3

Non-metric multidimensional scaling (NMDS) based on species composition from different pedoenvironment sites

Full size image
Fig. 4
figure 4

Distribution of 39 species within 300 samples plots installed along a pedoenvironmental gradient by a two way cluster dendrogram based on Euclidian distance. See full names of the species in Table S1 in the Electronic supplementary material

Full size image

Ecological significance index

Differences in the ecological significance of species (ESI) were observed among pedoenvironments. Based on the coverage of the dominant species, six plant communities with eleven distinct associations were identified: the moss cushion community, with three associations, Bryum-Sanionia (P1), Bryum-Brachythecium (P7 and P8), and Bryum-Hennediella (P13); the moss turf community, with two associations: Chorisodonthium-Sanionia (P2) and Chorisodonthium-fruticose lichens (P11); the moss carpet community with four associations: Sanionia-Polytrichastrum (P3), Warnstorfia-Sanionia (P4 and P5), Sanionia spp. (P6), Sanionia-Brachythecium (P14); the cryptogamic-phanerogamic community with a Sanionia-Deschampsia association (P9); the fruticose lichens community (P15) and associated with Andreaea regularis (P10); and the musciculous-lichens community (P12) (Table 1).

Descriptors of soil fertility and texture

High variability in soil chemical properties was observed along the pedoenvironmental gradient (Fig. 5). However, some attributes formed clusters of pedoenvironments (Fig. 5), as can be observed for pH, V, Mn, Fe, and silt, which separated P7, P8, P13, and P15 from the other pedoenvironments. In general, most pedoenvironments presented high values of P Rem, coarse and fine sand (P1, P3, P4, P6, P9, P10 and P14). The first two axes of the PCA explained 54.2% of the variation in soil data (Fig. 5). The first axis explained 29.7% of the data variation and was positively correlated with T (R = 0.95, P < 0.05), exchangeable acidity (Hal, R = 0.89, P < 0.05), Na (R = 0.82, P < 0.05) and clay (R = 0.80, P < 0.05). The second axis of the PCA explained 24.5% of the soil data variation and was positively correlated with Mn (R = 0.89, P < 0.05), V (R = 0.85, P < 0.05), Fe (R = 0.83, P < 0.05) and silt (R = 0.79, P < 0.05), and negatively correlated with P Rem (R = −0.59, P < 0.05) (Figs. S2 and S3).

Fig. 5
figure 5

Principal component analysis (PCA) for the soil parameters of different types of pedoenvironment sites. For analysis, available pH (H2O), exchangeable acidity (Hal), sum of bases (SB), Al3+, Ca2+, Mg2+, Na, available P, K+, N, organic matter (MO), potential cation exchange capacity (T), percentage of bases saturation (V), remaining phosphorus (P Rem), aluminium saturation (m), Zn, sodium saturation index (ISNa), Mn, Cu, Fe, Zn, and soil texture as coarse sand (C_Sand), fine sand (F_Sand), clay and silt contents were included. The level of Pearson correlation of each vector is indicated (cos2)

Full size image

Soil-cryptogamous communities relationship

The first axis of the CCA explained 21.04% of species composition based on differences in soil properties, while the second axis explained 18.2% (Fig. 6). According to the CCA, the moss species Andreae gainii is closely related to organic matter contents (OM), which is in line with the high values found in P5, the only pedoenvironment in which this species was recorded in this study. Similarly, Warnstorfia sarmentosa, which was the dominant species in P4 and P5, showed the highest OM values. H. revolutum, which occurred only in P6, showed a relationship with coarse sand (C_sand), which was at its highest level in this pedoenvironment. The CCA also showed a relationship between Chorisodontium aciphyllum and sodium (Na); although this species appears in several pedoenvironments with varying levels of Na, it was dominant in P2 and P11 (Table S3 from ESM), which showed high values for this attribute (Table S2). The fruticose lichen Cladonia rangiferina was related to the high levels of P found in P10 and P11. Andreaea depressinervis and Cladonia sp. were dominant in P12 and may be related to the high K levels found in this pedoenvironment.

Fig. 6
figure 6

Canonical correspondence analysis (CCA) showing species and plot scores in function of soil properties sampled within different types of pedoenvironments (a), and CCA only with texture properties (b). For analysis, available pH (H2O), exchangeable acidity (Hal), sum of bases (SB), Al3+, Ca2+, Mg2+, Na, available P, K+, N, organic matter (MO), potential cation exchange capacity (T), percentage of bases saturation (V), remaining phosphorus (P Rem), aluminium saturation (m), Zn, sodium saturation index (ISNa), Mn, Cu, Fe, Zn, and soil texture as coarse sand (C_Sand), fine sand (F_Sand), clay and silt contents were included.

Full size image

Effects of soil properties on species richness and species composition

From the comparison of models tested, we found that the predictors of soil fertility, soil texture, and individual soil properties (clay, sand, silt, T, OM and SB) had no significant effect on species richness, generally (Table S4 in the Electronic supplementary material). By contrast, according to our best model, significant effects of soil fertility variability on species composition (LMM: Z = 3.01, P < 0.001) were observed (Fig. 7, Table S4).

Fig. 7
figure 7

NMDS axes 1 (Species composition variability) and the axis 1 of PCA of soil nutrients (PCA1f) relationship according with GLM approach. Colour-filled circles indicate data per treatments. The solid line represent the fitted value (prediction) of the model, and the shaded area the 95% confidence interval of the predicted value of the model

Full size image

Discussion

Our results demonstrate that richness, species composition and coverage of cryptogam communities change along the pedoenvironmental gradient on Elephant Island. In the models tested, we have demonstrated the main effects of different environmental predictors (soil texture and fertility), highlighting how soil fertility has a strong influence on species composition variation along the pedoenvironmental gradient, albeit without affecting species richness. To explain the possible processes that determine changes in species composition, we described the relationship between cryptogam communities and soil properties (mainly fertility) for each pedoenvironment. Also, we produced a detailed description of different types of mosses and lichens associations that determine different cryptogam communities along the pedoenvironmental gradient.

The species richness found in this study was considerably lower than that recorded by Alison and Lewis-Smith (1973) for Elephant Island (80 species, 24 of which were mosses) and also that one reported by Pereira and Putzke (1994) at Stinker Point (115 species, 38), because we selected representative and contrasting pedoenvironments that form the gradient, allowing to test our hypothesis. However, we identified previously unreported moss species (Putzke and Pereira 2001; Pereira and Putzke 2013) for the area: Brachythecium sp., Bryum orbiculatifolium, Hypnum revolutum, Sanionia georgicouncinata and Syntrichia magellanica. Recently-documented climate change in this part of the Maritime Antarctic and Antarctic Peninsula (Turner et al. 2016) has promoted increasing ice-free areas, offering favourable conditions for other species to colonize and develop (Amesbury et al. 2017). On the other hand, we assume that, despite the proximity of the pedoenvironments and the small elevational variation between them, this study showed differences in community diversity and cover, due to small scale differences between pedoenvironments.

We detected marked differences in soil chemical and physical properties among the pedoenvironments. Partially because they have varying degrees of ornithogenic influence, landforms, and histories of ice retreat processes (Michel et al. 2014; Turner et al. 2016). Elephant Island's soils are been forming since the mid-Holocene glacial retreat under wetter, warmer climate condition, compared with continental Antarctica, which favours soil development (Bockheim 2015). Recent studies have investigated the local geomorphological features and mapped eight different types of periglacial landforms at Stinker Point, such as marine platforms (flat-floored valleys, laminated cracking on rock, patterned ground, gelifluction sheets and lobes, and vertical stone fields), till deposits (patterned ground, gelifluction lobes, and vertical stone fields), and slopes (debris talus and cones) (López-Martínez et al. 2006, 2012; Navas et al. 2018).

According to the map proposed by Lopez-Martínez et al. (2012) and Navas et al. (2018), most pedoenvironments correspond to platforms in Elephant Island. Conversely, the pedoenvironments P10, P11 and P12 are slopes at the edges of the platform, and P7, P8, P13 and P15 are located on unconsolidated newly-exposed glacial drift (moraines). This may explain the higher variation in soil pH found in these pedoenvironments, which is alkaline in moraines, and highly acid in the platforms (Navas et al. 2018). This striking difference corresponds to the more-recent exposure of the moraines, preserving most of the characteristics of the sedimentary material (Navas et al. 2018). Moraines are also closer to glaciers, while the pedoenvironments adjacent to the coast have much greater acidification due to bird nesting and guano mineralization (Tatur and Barczuk 1985; Simas et al. 2007), and phosphatization as a long-term soil formation process (Beyer et al. 2000; Navas et al. 2017, 2018). Similarly, higher soil organic matter on platforms is related to the proximity of the coast, and longer exposure time, combined with the presence of seabirds, responsible for soil nutrient input (Beyer et al. 2000), and vegetation development (Bockheim and Haus 2014). According to Navas et al. (2018), high P and K values present not only on the coast but also on the intermediate platforms of Elephant Island, are related to bird activity, especially recent and abandoned penguin colonies (Michel et al. 2006; Simas et al. 2007). We infer that this crucial nutrient input may be a key driver of community diversity, forming specific associations of lichens and mosses species.

The lowest fertility values (except P) were recorded in P1, which also had the lowest species richness in this study. The dominant Bryum argenteum (ESI = 540), associated with Sanionia uncinata (ESI = 115), formed a dense carpet of mosses only seen at this location on Stinker Point. High P levels is explained by the presence of active giant petrel nests (Macronectes giganteus) on this area, with phosphatization and strong acidification by guano mineralization, as described by Simas et al. (2007). Conversely, P2 showed higher fertility, high OM, and higher Mg values, but lower levels of P. Located on the central platform, on rock fields, its phytophysiognomy is a dense community of peat bog mosses with a dominance of Chorisodontium acyphillum (ESI = 503.5) associated with S. uncinata (IES = 99), and the presence of the muscicolous lichen Cystocoleus niger (ESI = 195.5) growing on both. This pedoenvironment shows lower cover, higher richness (ten species) and a community of S. uncinata-dominated moss carpet (IES = 258.75) associated with Polytrichastrum alpinum (IES = 115.5) with occasional Deschampsia antarctica grass (IES = 75.25). P4 and P5 are both pedoenvironments located on opposite sides of the drainage line that cuts across Stinker Point, and often waterlogged and with high amounts of OM (~ 35.5 dag/kg). However, chemically, soil attributes of P4, and P5 are very dissimilar. The cover, richness, and species composition are similar, and both are dominated by Warnstorfia sarmentosa in the moss carpet community. In P4, W. sarmentosa is associated with Prasiola crispa algae whereas at P5 the moss S. georgicouncinata is the accompanying species. According to the CCA, W. sarmentosa showed a relationship with organic matter, consistently with Putzke and Pereira (2001), and Ochyra et al. (2008), who described this moss as a typically hydrophilic, growing in saturated wetlands.

The community at P6 was the only moss carpet with exclusive dominance of Sanionia spp. Although ten other species of mosses and lichens were identified at the site, only S. uncinata (ESI = 303.75) and S. georgicouncinata (ESI = 231) had ESI >50. The available P amounts varied widely among the plots, even within the pedoenvironment itself, but were high overall (498.13 ± 313.52 mg/dm3). Increasing values are attributed to the presence of abandoned bird nests at some points, as reported by Francelino et al. (2011) and Moura et al. (2012). Despite its low frequency, P6 was the only pedoenvironment where H. revolutum was present and related to coarse sand, with the highest value. Previous studies have described communities dominated by this moss near drainage lines, or areas of snow accumulation (Barták et al. 2005).

P7 and P8 are close to each another, located on an elevated marine terrace near the beach just above an active penguin rookery. These pedoenvironments have the lowest levels of cover, with similar species richness and composition (Fig A.1A and Fig. 4). Both are cushion moss communities, dominated by Bryum orbiculatifolium associated with Brachythecium austrosalebrosum, with occasional, low frequency presence of Colobanthus quitensis this high plant, according to with the CCA, is related to pH and Mn (ranging from 32 to 36 mg/dm), although Parnikoza et al. (2007) report that it inhabits a large ecological range of habitats. The moss Syntrichia magellanica only appeared in the P7 pedoenvironment (Fig. 4), and maybe related to high levels of Ca2+ and SB, the only edaphic attribute that differed significantly from P8. According to Ochyra et al. (2008), this moss grows in a range of habitats, including acidic or base-rich substrates, in either sandy or silty soils, and on moraines, or close to bird colonies.

P9 supported a phanerogamic community of D. antarctica (IES = 369) associated with S. uncinata, with occasional P. crispa algae. This area was located on a slope near the lake, which is often visited by skuas (Stercorarius antarcticus) for settling and nesting in dry lakes (Quintana and Travaini 2000), thereby promoting the of nutrient’s input. The P10 was near the slope edge of the platform, with a strong ornithogenic influence in the past, and currently located near a giant petrel nests, with the highest P level recorded (870.44 ± 635.89 mg/dm3) and high fertility, creating favourable conditions for vegetation (Tatur and Myrcha 1993). This pedoenvironment showed high variability in species composition, representing a fruticose lichen community, dominated by Sphaerophorus globosus and Usnea aurantiacoatra, associated with the moss Andreaea depressinervis. Other species such as Cladonia rangiferina, THE seaweed P. crispa, S. uncinata, and Chorisodontium aciphyllum also had high ESI values, and are commonly reported with association with S. globosus (Øvstedal and Lewis-Smith 2001; Olech 2004).

In P12, less than 60% of the cover was recorded, due to the rockiness stoniness of the area. A muscicolous lichen community (Cladonia sp., C. niger and Ochrolechia frigida) grows on A. depressinervis (ESI = 300) and S. uncinata (ESI = 165) mosses. This pedoenvironment (P12) is the most acidic, and the highest Al3+, exchangeable acidity (H + Al), and cation exchange capacity at pH 7 (T) values. This pedoenvironment is located adjacent to a current penguin colony and is probably an abandoned penguin rookery at the platform slope. The high Na values are caused by the saline sprays carried by winds (Michel et al. 2006). Conversely, P13 was a recently exposed area close to meltwater lakes following glacier retreat. These lakes are often used by wandering skuas, in the vicinity (Quintana and Travaini 2000), representing initial spots for plant colonization, on rich-nutrient soils. The P14 has a dense moss carpet community (S. uncinata–B. austrosalebrosum association) located in a seasonal waterlogged area, with high coverage by the hydrophilic moss B. austrosalebrosum that grows well on sandy soils along the borders of meltwater channels (Ochyra et al. 2008).

Finally, P15 was the pedoenvironment that show the highest species richness, being located at a small elevation above a moraine, and having pH 6.4, high fertility, and high levels of P (suggesting ornithogenic influence) but no current bird nesting. Although S. uncinta was dominant (IES = 360), this formation was classified as a fruticose lichen community, by the widespread presence of S. globosus, U. antarctica, O. frigida, Psoroma cinnamomeum, C. aculeata and H. lugrubis, as well as other moss and lichen with low frequencies (ESI > 50). The Antarctic endemic H. lugrubis was only found in P11 and P15, and the CCA suggested a relationship with high K+ in both pedo-environments. Regarding Choi et al. (2015) this species, showed that the main factors that influence its distribution and development were substrate and soil moisture.

Conclusion

(1) Cryptogamic communities have differences in diversity and cover of cryptogam communities along a pedoenvironmental gradient on Elephant Island. (2) The models employed showed that only soil fertility had significant effects on species composition, but not on richness; Soil nutrient status is therefore the main significant predictor of species composition of cryptogamic communities in Maritime Antarctica. However, differences in species composition were not as pronounced as expected, showing high sharing between different pedoenvironments. (3) Pedoenvironmental filtering determined differences in the composition of species on Elephant Island (Maritime Antarctica), and small-scale heterogeneity contributed to the formation of typical cryptogamic species associations along the pedoenvironmental gradient. (4) Soil attributes not only drive the patterns of diversity, but also the types of cryptogam communities.