BSC samples were collected in June 2014 and 2015 from the plots of the “German Biodiversity Exploratories” project with natural protected forests and managed forest (age-class forest) (M. Fischer et al., 2010). Forest plots were located in the Schorfheide-Chorin Biosphere Reserve in northeastern Germany; the plots differed in the dominant tree species: Scots pine (Pinus sylvestris L.) or European beech (Fagus sylvatica L.). Samples were taken from the disturbed areas where BSCs developed on the litter-free bare soil (for illustration, see Fig. 1). The top millimeters of soil, where BSC had been visually detected as a green cover, were collected on a spatula. After transportation to the lab the upper 2 mm of BSC were separated from the adhering soil underneath with a razor blade before being stored dry in paper bags. In total, 31 BSCs were collected from 13 pine and 18 beech plots, of which 23 were managed and 8 were natural forest plots (Table 1).

Solid 3N-Bolds Basal Medium (1.5 % agar) with vitamins (Starr and Zeikus, 1993) was used for the establishment of enrichment cultures. Several 7–10 mm2 BSC pieces were cleaned with forceps to remove all roots and leaves, in order to avoid the growth of fungi and bacteria, and were placed on the surface of an agar plate under sterile conditions. Plates were incubated at 20 ∘C, 30–35 µmol photons m-2 s-1 (Osram Lumilux Cool White lamps L36W/840) under a light/dark cycle of 16:8 h L:D. The plates were regularly inspected and colonies were identified after 4 to 6 weeks' incubation, using a light microscope (BX51, Olympus) with Nomarski differential interference optics and 1000 × magnification. Photomicrographs were taken with an Olympus UC30 camera attached to the microscope and processed with the cellSens Entry software (Olympus). For direct observation of BSC samples, pieces of BSC were rewetted with tap water, put on a glass slide, and analyzed with the above-mentioned microscope at 400 × magnification. Mucilage of algae was stained with an aqueous solution of methylene blue.

Morphological identification of algae and cyanobacteria was based on the standard syllabus (Ettl and Gärtner, 1995) and more recent taxonomic publications on certain algal groups (Darienko et al., 2010; Kostikov et al., 2002; Mikhailyuk et al., 2015). Phototrophic microorganisms were identified as Cyanobacteria, Chlorophyta, Streptophyta, and some Stramenopiles (Eustigmatophyceae). Diatoms were regularly found in direct observations but were excluded from the analyses as the mentioned enrichment cultivation was not suitable for this group of microalgae (e.g., Schulz et al., 2016).

Since the enrichment cultivation did not provide clear information on the abundance of each identified taxon, we used the total number of algae and cyanobacteria species per sample, also known as species richness, as the measure of alpha diversity. As a measure of beta diversity, the similarity between the plots was shown by presence/absence of individual species, combining the total number and the identity of all algal taxa observed. Furthermore, the identified algae and cyanobacteria were categorized based on their life form (filamentous or coccal), since different life forms differ in their ecological function. The proportion of filamentous algae in the total number of algae was used for statistical analyses.

The natural and managed forest plots were characterized by different silvicultural management intensity. In natural forests, no management was conducted, meaning that fallen trees were left in place and no trees were cut. In managed age-class forests, the forest stands were regularly disturbed by tree cuts, removal of dead trees, and usage of skid trails. To evaluate the effect of management, the silvicultural management index (SMI) was used. This index takes into account the tree species, forest stand density and age, as well as the aboveground living and dead wood biomass (Schall and Ammer, 2013). High stand density is reflected by a high SMI; therefore, natural forests have a lower SMI than managed forests, and a pine stand has a higher SMI than a beech stand (Schall and Ammer, 2013).

To assess potential links between BSC organisms and environmental parameters, the species' richness, presence/absence of individual algal species, and proportion of filamentous algae were related to the following environmental parameters: dominant tree species (pine or beech), silvicultural management intensity (SMI), pH, and water content of the bulk soil (Table 1, for all 31 samples). Additionally, for a subset of 19 BSC samples, data on total C, N and P content and organic and inorganic P compounds, for labile, moderately labile and stable P, were included. Element data were presented in detail by Baumann et al. (2017), and are thus not presented in this paper.

All statistical analyses were done using the R version 3.3.0 (R Development Core Team, 2009) statistical software. Analysis of variance (ANOVA) was conducted to reveal the effect of environmental parameters on algal and cyanobacteria richness, and proportion of filamentous species; the best predictors for their variance were selected by backward elimination stepwise regression analysis based on the BIC (Bayesian information criterion) using the “step” command in R. The correlation between environmental parameters was determined by Pearson correlation (“cor” and “cor.test” commands in R).

To reveal correlations of single environmental parameters with the presence or absence of individual algal species, PerMANOVA (with the “adonis” function in R, Anderson, 2001) was applied using the Bray–Curtis dissimilarity index (Bray and Curtis, 1957), including a permutation test with 1000 permutations. The adonis function allows application of non-Euclidean distance metrics and handles both categorical and continuous predictors. For analysis of co-correlation of environmental factors, Pearson correlation was used. To test significant differences of environmental factors between tree species, an unpaired two-tailed t-test was performed. Differences with a p-value below or equal to 0.05 were taken as significant.

In total 51 different algae species and one cyanobacterium were detected in enrichment cultures of all 31 BSC samples. Stichococcus bacillaris was the most ubiquitous taxon, observed in 27 out of 31 samples, followed by Coccomyxa simplex and Klebsormidium cf. subtile in 26 and 23 out of 31 samples, respectively. All other algal species were detected in less than 50 % of the BSC samples; 22 algal species were observed exclusively in one sample (Fig. 2). The richness of algae (total species number) at each plot ranged from 3 to 14 species with a mean of 8 and a standard deviation of 2.6 (a complete species list is provided in Supplement Table S1).

The phylum Chlorophyta made up 81 % of all detected algal species, followed by Streptophyta (11 %) and Stramenopiles (6 %). Cyanobacteria were rare in these BSCs: only one species, Microcoleus vaginatus, was observed in only one sample.

The identified algal species were differentiated according to their life form (Fig. 3). Five species with strong filaments (Klebsormidium cf. flaccidum, K. cf. subtile, K. cf. nitens, Xanthonema cf. exile, Microcoleus vaginatus) and two species with short or easily disintegrating filaments (Interfilum paradoxum, Stichococcus bacillaris) were found. In each BSC at least two different filamentous taxa were detected, indicating their importance for the BSC formation. Genus Klebsormidium seemed to be highly important for BSCs in forest since it was registered in every BSC sample (Table S1).

The gravimetric water content of the bulk soil was negatively correlated with the SMI; the pH was neither correlated with the water content, nor with the SMI nor with the dominant tree species (Table 2). The N content was positively correlated with the C content, and N as well as C content were independent of the SMI and pH. Total P and the proportion of inorganic P were independent of the C and N content, as well as of pH and SMI (Table 2).

The richness of algal species and the proportion of filamentous algae in BSCs only correlated with SMI, water content and proportion of inorganic P (Table 3). The remaining tested parameters (C and N content, total P, proportion of organic P, pH, dominant tree species, and soil horizon) were excluded by stepwise model simplification based on the BIC. This means that these factors had no measurable effect on the algal species richness or on the proportion of filamentous algae. The SMI was positively correlated with the species richness, meaning that a higher SMI resulted in a higher species richness (Fig. 4); especially the proportion of coccal algae was increased. BSCs with higher algal richness tended to have lower proportions of inorganic P.

The presence/absence of individual algal species in BSCs significantly correlated with the dominant tree species (15 % explained variance) and with the soil water content (10 % explained variance). The SMI and proportion of inorganic P explained each 5 % of the variance, but this was not significant (Table 3). Therefore, we concluded that the dominant tree species and the soil water content affect the composition of algal species in BSCs.

In total, 51 microalgal species and one cyanobacterium were identified in all sampled BSCs (Fig. 2), which is a similar or slightly lower species richness compared to the other reports on BSCs from temperate regions at open sites (Langhans et al., 2009; Schulz et al., 2016), but similar or higher compared to the previous reports on algae from forest bulk soil (Khaybullina et al., 2010; Novakovskaya and Patova, 2008; Starks et al., 1981). Nevertheless, the given number most probably underestimates the real algal richness, since our results are based on the enrichment cultivation followed by morphological identification. Enrichment cultivation promotes the growth of only culturable algae, which represent only a small part of all phototrophic microorganisms in BSCs (Langhans et al., 2009). A recent paper, comparing metagenomic data of a polar BSC with data based on enrichment cultivation and morphological identification of the algae, showed that only about 10 % of the metagenomic data could be confirmed by morphological identification (Rippin et al., 2018). Furthermore, it is not always possible to distinguish dormant from currently active microalgae. However, direct observation of a BSC sample under the microscope gives at least a first hint of the dominant active organisms. With this approach we could confirm that all filamentous algae were abundant and vital in the BSC samples. The morphological identification of algae has known challenges: for example, sibling species have similar characteristics but are genetically distant (Potter et al., 1997). To overcome these limitations, researchers proposed combining molecular and morphological methods of identification, since molecular techniques alone can also fail to detect some taxa, as a result of unsuccessful DNA extraction, inappropriate primers, etc. (Büdel et al., 2009; Garcia-Pichel et al., 2001).

All observed algal species are known to be terrestrial taxa; most of them were already reported from other BSCs (Büdel et al., 2016, and references therein; Ettl and Gärtner, 1995). Chlorophyceae were the most abundant phylum, which is typical for temperate regions (Büdel et al., 2016). Especially most of the unicellular taxa belong to the Chlorophyta (genera such as Chlamydomonas, Chloromonas, Chlorococcum, and Tetracystis). A high richness of Chlorophyta is characteristic of humid habitats and typical for forest soils (Hoffmann, 1989).

Cyanobacteria were represented by only one species. While they are often reported as predominant species in BSCs of arid regions such as Israel and drylands of the USA (Garcia-Pichel et al., 2001; Kidron et al., 2010), cyanobacteria are less abundant in temperate regions (Gypser et al., 2016; Langhans et al., 2009; Pluis, 1994) and even rare in acidic soils, which corresponds to the forest plots of our Schorfheide-Chorin study site (Hoffmann et al., 2007; Lukešová, 2001; Lukešová and Hoffmann, 1996). It seems that cyanobacteria play only a minor role in forest ecosystems, with consequences for the taxa's ecological traits. For example, the ability for nitrogen fixation in phototrophic organisms was only reported for cyanobacteria and never observed in eukaryotic algae. In forest ecosystems, litter and other decomposable biomass might have provided sufficient mineral nitrogen compounds, which could have led to the absence of nitrogen-fixing organisms in these systems in contrast to nitrogen-poor habitats such as dunes or deserts where cyanobacteria are dominant (Langhans et al., 2009; Schulz et al., 2016).

The filamentous alga Klebsormidium was found in nearly all BSCs of our study, whereas species with similar strong filaments (Microcoleus and Xanthonema) were only found occasionally. Filamentous algae can be regarded as key players in BSC communities, because of their BSC-initiating potential by building tight networks among soil particles (Büdel et al., 2016). In some forest BSCs, moss protonema can exert a similar function, due to their filamentous nature (Weber et al., 2016). However, in the forest ecosystems of Schorfheide-Chorin the green algae Klebsormidium seems to be the most important BSC-initiating alga. This genus can tolerate a wide range of environmental factors and has a cosmopolitan distribution in numerous terrestrial habitats (Karsten et al., 2016; Rindi et al., 2011, and references therein). Its presence in other terrestrial habitats, such as natural rocks in lowlands and mountainous areas (Mikhailyuk et al., 2008), caves (Vinogradova and Mikhailyuk, 2009), sand dunes (Schulz et al., 2016), tree barks (Freystein et al., 2008), acidic post-mining sites (Lukešová, 2001), urban walls (Rindi and Guiry, 2004), and building facades (Barberousse et al., 2006), is well documented. As many other terrestrial algae, Klebsormidium is tolerant to light exposure during dehydration (Gray et al., 2007). This is a typical situation, which BSC algae have to cope with, since the increase in light intensity in the morning is often associated with dehydration (Raanan et al., 2016). A recent study in central Europe, however, observed that Klebsormidium is sensitive to increasing light during cellular water loss (Pierangelini et al., 2017). The distribution of Klebsormidium in nearly all BSC samples from Schorfheide-Chorin forest may be explained by a lower solar radiation and lower evaporation rates in forest ecosystems compared with the open habitats (e.g., inland dunes) where besides Klebsormidium other filamentous algae are dominant (Langhans et al., 2009; Pluis, 1994). Also, the forest soil is rather acidic (pH min: 3.23, pH max: 3.86; Table 1), which supports a dominance of Klebsormidium (Škaloud et al., 2014). Thus, the low light availability, low water evaporation, and acidic soil conditions plausibly explain the presence and the dominance of Klebsormidium as a potential BSC-initiating algal taxon in nearly all BSCs from Schorfheide-Chorin forest plots.

Three morphospecies of the genus Klebsormidium were identified in the investigated samples (Fig. 2). All three morphospecies were reported from other aeroterrestrial habitats in central Europe (Glaser et al., 2017; Mikhailyuk et al., 2015). Klebsormidium exhibits morphological features, which can be easily recognized. However, the identification down to species level is difficult due to the high morphological plasticity (Lokhorst, 1996). And still, in times of molecular identification, the debate on species definition in the genus Klebsormidium is ongoing (Mikhailyuk et al., 2015; Rindi et al., 2017). Therefore, the definition of clades within Klebsormidium was and still is a helpful tool to differentiate between morpho- or geno-types (Rindi et al., 2011). Studies comparing these Klebsormidium clades from different localities observed global ubiquity on the one hand, and local endemism on the other (Ryšánek et al., 2014). Clade composition seems to differ depending on the habitat: Klebsormidium cf. flaccidum (B/C clade) was abundant in both closed and open habitats, whereas K. cf. nitens and K. cf. subtile (E clade) were predominantly distributed in forest BSCs (Glaser et al., 2017; Mikhailyuk et al., 2015). In our study, however, BSCs from forests contained more often Klebsormidium cf. subtile and K. cf. nitens than K. cf. flaccidum. In desiccation experiments the recovery rates of these clades were similar (Donner et al., 2017a, b). It is still open which of the environmental factors cause the observed habitat preferences of the different clades. Additional ecophysiological experiments including potential environmental factors, such as light regimes, desiccation frequency, and duration, as well as soil parameters such as pH, in combination with transcriptomic approaches might explain these conspicuous habitat preferences of Klebsormidium clades.

The silvicultural management index (SMI) was used to estimate the forest management intensity. It takes into account the tree species, forest stand age and density. However, intensively managed forest did not necessarily inherit more disturbed sites suitable for the BSC development. In contrast, BSC development is limited in forests with high density (typical for intensively managed forest stands). However, managed forests have a higher risk for complete stand loss, because of either regular clear-cut or strong storms; it is more likely to lose a large part of pine stands with high density compared to natural beech forest.

The richness of algal species as well as the proportion of coccal algae were positively correlated with the silvicultural management index (SMI; Fig. 4). This means that more algal species were discovered in BSCs from managed than from natural forest ecosystems. This finding agrees with conclusions of high algal richness on disturbed or cultivated soils (Gollerbakh and Shtina, 1969; Hoffmann, 1989). The SMI reflects the effect of management practice on the dominant tree species and the stand density. Most biodiversity exploratory studies on forest-soil microorganisms observed a stronger effect of the dominant tree species than of the SMI on the microbial community (Goldmann et al., 2015; Kaiser et al., 2016; Purahong et al., 2014); only one study on litter decaying fungi and bacteria indicated a significant difference between natural and managed beech forests (Purahong et al., 2015). Kaiser et al. (2016) discussed that the different tree species influence soil bacteria by shifting the pH in soil; hence, tree species was designated as the main predictor for bacterial community composition. However, the bulk soil pH did not differ significantly between beech and pine forest in Schorfheide-Chorin (Table 1); hence, the algae in BSCs were not affected by this abiotic parameter. Therefore, we rejected an effect of the SMI via the pH on the BSC algal species richness in Schorfheide-Chorin.

However, the SMI combines other potential factors, which could explain its positive correlation with the richness of algal species as well as the proportion of coccal algae. Water and light availability might have affected BSC microalgae due to forest stand density and tree species. Forest plots in Schorfheide-Chorin were dominated by either beech or pine trees, which affect the light regime differently: in beech forests the canopy shade changes over the year, with usually higher solar radiation on the ground in winter and spring than in summer, while in pine forests no such light fluctuations occur. Also, the stand density, another parameter of the SMI, could affect the light regime on the ground: higher density would result in less photosynthetic active radiation for photosynthetically active soil microorganisms. The radiation is often coupled with evaporation of soil moisture (Raanan et al., 2016); hence, the stand density could have an indirect effect on the BSC organisms via an altered water regime. Thus, the SMI was expected to affect the algal richness in BSCs via lower light availability and lower evaporation rates. This assumption is well supported by the two-way analysis of water content and SMI. Nevertheless, it should be noted that the water content was measured in the bulk soil, which might differ from the one of BSCs. For future studies on microalgae in BSCs it would be important to examine also the incident light on the ground as well as the BSC water content.

Although the SMI positively affected the algal richness, the presence or absence of individual algal taxa was not correlated with the SMI, but with the main tree species. Broadleaf litter has a higher quality in terms of a more favorable C:N and C:P ratio compared to coniferous litter (Cleveland and Liptzin, 2007; McGroddy et al., 2004). It might have been that the community in the pine forest promoted algal species which could cope with a suboptimal C:N:P ratio. But as mentioned above, both light regime and water availability differ between the two forest types and could also have contributed to the observed differences in the occurrence of algal species.

BSCs have different important ecological functions, such as the enhancement of the nutrient content in the top soil layer (Baumann et al., 2017; Evans and Johansen, 1999). To assess the relationship between BSC community and biogeochemical cycling in BSCs, the content of total C, N, and P and additionally the different P fractions (organic, inorganic, labile, and stable fractions) were correlated with algal richness. Although a correlation between the richness of algae and the total C, N, and P content was not observed, the presence of BSCs clearly led to an increased content of total C, N, and P and in particular a higher proportion of organic P (Baumann et al., 2017). These results indicate that algal species are functionally redundant, and that a BSC community with low species richness still has a functional role in increasing C, N, and P content. A more detailed analysis of the P fractions gave a slightly different picture: the proportion of inorganic P was positively correlated with the proportion of filamentous algae and showed a tendency to a negative correlation with the richness of BSC algae. Soluble inorganic phosphate can be assimilated by organisms, and it originates either from the weathering of P-containing minerals, desorption of mineral-bound phosphates, or the mineralization of organic matter (Mackey and Paytan, 2009). Thus, a low amount of inorganic P could indicate a high uptake rate of BSC organisms, and thus a more closed P cycle due to the higher algal richness (Baumann et al., 2017).