Introduction

Alien tree species influence ecosystem functioning by impacts on services and biodiversity (Lövei 1997; Vitousek et al. 1997; Mack et al. 2000). Impacts of alien tree species on ecosystems and human activity vary across species life-history traits and introduction history (Vilà et al. 2011; Potgieter et al. 2017; Castro-Díez et al. 2019). However, some of these impacts reveal consistent patterns, thus, alien species causing similar impacts and driven by similar anthropogenic and biological factors might be assessed jointly, considered as the same invasion syndromes (Novoa et al. 2020). Therefore, assessment of sample species representing a particular invasion syndrome allows inferences about groups of species impacts on biodiversity and ecosystem functioning, as well as to adopt similar management strategies.

Alien tree species are known for their significant influence on biodiversity and ecosystem functioning because of their high biomass and longevity (Richardson and Rejmánek 2011; Rejmánek and Richardson 2013; Dickie et al. 2014; Castro-Díez et al. 2019). Thanks to species-specific modification of ecosystem properties (Peterken 2001; Mueller et al. 2012; Castro-Díez et al. 2019), they may severely influence dependent organisms, e.g. mycorrhizae (Dickie et al. 2017; Rożek et al. 2020), bacteria (Stanek and Stefanowicz 2019), fauna (Karolewski et al. 2020; Mueller et al. 2016) and plants (Taylor et al. 2016; Vítková et al. 2017; Gentili et al. 2019).

One of the taxonomic groups of plants threatened by alien tree species are bryophytes. Due to their dependence on air humidity and affiliation to particular substrata (Fritz et al. 2009; Kriebitzsch et al. 2013; Wierzcholska et al. 2018), bryophytes are sensitive indicators of ecosystem alteration (Vanderpoorten et al. 2004; Rydin 2008; Jagodziński et al. 2018). Moreover, bryophytes comprise a significant proportion of understorey biomass in temperate and boreal coniferous forests (Wirth et al. 1999; Muukkonen et al. 2006; Woziwoda et al. 2014). The low number of specialized species and high species richness has caused most bryological studies to be conducted in well-preserved ecosystems, especially in old-growth forests (e.g. Snäll et al. 2004; Király and Ódor 2010; Mežaka et al. 2012). Fewer studies have covered managed forests (e.g. Gustafsson and Hallingbäck 1988; Király et al. 2013; Wierzcholska et al. 2018) and post-industrial areas (e.g. Engelmann and Weaks 1985; Jagodziński et al. 2018).

In contrast to numerous studies that assessed impacts of alien tree species on vascular plants, studies focused on bryophytes are scarce and lead to ambiguous conclusions (e.g. Taylor et al. 2016; Sitzia et al. 2018; Slabejová et al. 2019). Moreover, most of them were focused on epiphytes, indicating, for example, high species richness of bryophytes on the bark of Robinia pseudoacacia (Jagodziński et al. 2018; Fudali and Szymanowski 2019) or Q. rubra (Woziwoda et al. 2017). Recent studies revealed decreasing bryophyte species richness and cover in stands of Prunus serotina (Verheyen et al. 2007; Halarewicz and Pruchniewicz 2015; Vegini et al. 2020) or Pinus radiata (Pharo and Lindenmayer 2009). Effects of alien coniferous tree plantations on bryophytes depend on both the life history traits of invasive tree species studied and the properties of the reference ecosystem, and might be both positive and negative (Quine and Humphrey 2010). Therefore, we aimed to compare the impacts of three alien tree species with North American origin (Table 1), which are the most widespread in European woodlands (Wagner et al. 2017) and represent different invasion syndromes (sensu Novoa et al. 2020) due to their functional differences. We hypothesized that invaded forests will impact terricolous bryophyte species composition by decreasing species richness and cover. We also assumed that the relationships will be co-influenced by soil fertility and light availability, due to their importance in earlier studies (Jagodziński et al. 2018).

Table 1 Life-history traits and invasion history of alien tree species studied
Full size table

Material and methods

Study area

We established a set of study plots in the Wielkopolski National Park (WNP; W Poland; 52°16′ N, 16°48′ E; 7,584 ha). Due to the presence of numerous alien woody taxa (158 taxa; Purcel 2009), WNP is a good place to study spread and impact of invasive species (Dyderski and Jagodziński 2018, 2019a). The post-glacial landscape, characterized by high geomorphological diversity, determines the occurrence of numerous forest ecosystem types in WNP. The climate in WNP is temperate, with a mean annual temperature of 8.4°C and a mean annual precipitation of 521 mm, for the years 1951–2010 (measurements from Poznań meteorological station, ca 15 km from WNP).

Study design

In 2014 and 2015 we established a set of 189 systematically arranged study plots, with nine plots in each of 21 blocks (Fig. 1). This system was designed to assess the natural regeneration of the invasive tree species studied (Dyderski and Jagodziński 2018). We selected stands dominated by these invasive species (or in the case of P. serotina – P. sylvestris dominated stands with understorey dominated by P. serotina), based on WNP management plans and field inspection of potential sites. In the center of each block we established one study plot (200 m2 rectangle), in the middle of a monoculture of the invasive species. Then, we set up four additional plots, towards the N, S, E and W sides of the central plot, at the boundary of the alien species monoculture and another set of four plots, located 30 m away from each of the four additional plots. We excluded three study plots which due to systematic design occurred in non-forest ecosystems (thus final n = 186).

Fig. 1
figure 1

Schematic arrangement of the study plot blocks (21 blocks, each composed of nine plots). Adapted from Dyderski and Jagodziński 2019c

Full size image

We assigned each study plot to one of nine categories representing the main types of forest ecosystems in WNP, based on dominant tree species, vegetation and invasion level (Table 2), similarly to Dyderski and Jagodziński (2020). We decided to divide Pinus sylvestris forests into two groups: poor and plantation, according to different management history (Dyderski and Jagodziński 2020). Poor P. sylvestris forests refer to mesic habitats, mostly growing on brunic arenosols, where this species naturally occurs. By contrast, P. sylvestris plantation refers to forests occupying more fertile soils, usually cambisols and luvisols, which naturally host Quercus–Acer–Tilia forests. We assigned forests with P. serotina individuals with at least 500 ind.·ha−1 > 1.3 m tall as invaded by this species. Forests invaded by Q. rubra or R. pseudoacacia were former plantations, and these species comprised > 25% (mostly > 75%) of basal area. The age of dominant tree species within studied forests ranged from 23 to 139 years, and only ten plots had tree stands younger than 60 years old.

Table 2 Forest types and summary statistics of measured predictors (Dyderski and Jagodziński 2019c, 2020)
Full size table

Data collection

Within each study plot we assessed the cover of each bryophyte species using the Braun-Blanquet alphanumeric scale in July of each study year (2015–2018; Electronic supplementary material 1). This produced 742 observations (i.e. four years and 186 plots). We also visually estimated cover of all bryophytes at 5% intervals. We focused only on terricolous bryophytes, that is, growing on the soil. Taxonomic nomenclature follows Hodgetts et al. (2020). To account for abiotic characteristics we assessed light availability, soil pH and soil C:N. We used diffuse non-interceptance (DIFN) – the fraction of open canopy above the understorey – as a proxy of light availability. We measured DIFN in August 2016 using a LAI-2200 plant canopy analyser (Li-Cor Inc., Lincoln, NE, USA) and eight series of ten measurements per study plot. This measurement accounted for minimal light availability during the maximum development of canopy cover. We also collected six soil subsamples, systematically along plot borders, from 0 to 10 cm depth, and we pooled them into one sample. Then, we measured the pH of soil solution in distilled water after 24 h using an electronic pH-meter. Samples were also analysed for total carbon content and total nitrogen content, using the dry combustion method (PN-ISO 10694:2002) and Kjeldahl’s method (PN-EN 16169:2012), respectively, in an accredited laboratory (Jars Sp. z o.o, Legionowo, Poland). We used C:N ratio, as a synthetic index of soil fertility.

Data analysis

We analysed data using R software (v. 3.5.3; R Core Team 2019). Because of the low abundance and species richness, we averaged the cover of each species at the study plot level across all four years of data collection. We used the ‘IndVal’ method (Cáceres and Legendre 2009), calculated using mean cover data, to check whether a particular species is more frequent in a particular forest type (hereafter ‘species pool’), understood as all species occurring in a particular forest type. We used the ‘IndVal’ method to obtain information about the strength and significance of association (Cáceres and Legendre 2009). We compared bryophyte species composition in each study plot (using square-root transformed data) using non-metric multidimensional scaling (NMDS), based on Bray-Curtis distances. We used the ‘vegan’ package (Oksanen et al. 2018) to conduct NMDS. Before NMDS we excluded empty observations (i.e. six observations without bryophytes, in NMDS n = 180) To assess the correlation of the main gradients of species composition revealed by NMDS we passively (i.e. without influencing site and species scores) fit soil C:N ratio, soil pH, DIFN, species richness and cover using the ‘vegan::envfit()’ function (Oksanen et al. 2018). The goodness of fit was assessed by a permutation test (n = 999). We assessed differences in beta diversity among forest types using the vegan::betadisper() function (Oksanen et al. 2018). This analysis checked multivariate dispersion by analysis of mean distance to median of group coordinates in multivariate space. Then, we conducted Tukey’s a posteriori tests to assess differences among groups.

We analysed differences in bryophyte cover and species richness using generalized linear mixed-effects models (GLMMs). In models we accounted for random effects describing study design – study plot blocks. To account for plots without bryophytes in GLMMs describing bryophyte species richness and cover, we used zero-inflated models, that is, a compound of the count models and the zero-inflation models. In the count models we assumed Poisson distributions of species richness and beta distribution of bryophyte cover. In the zero-inflation models we assumed binomial distributions. We developed GLMMs using the ‘glmmTMB’ package (Brooks et al. 2017). In each model we assumed the impact of soil C:N ratio, soil pH and DIFN, to check the marginal response of forest type accounting for constant levels of other predictors, that is, marginal means, obtained using the ‘emmeans’ package (Lenth 2019) and marginal responses of continuous variables using the ‘ggeffects’ package (Lüdecke 2018). We presented models resulting from analysis of variance (ANOVA) type II, based on χ2 tests. We ensured a lack of collinearity among variables by checking variance inflation factors, and we tested zero-inflation and overdispersion using formal tests implemented in the ‘DHARMa’ package (Hartig 2020). Then, we reduced variables in each model to decrease the Akaike information criterion, corrected for a small sample size (AICc).

Results

During our study we found 25 bryophyte species in 180 of 186 study plots (Table 3). Among them, the most frequent were Brachythecium salebrosum (occurring in 93.5% of the plots), Hypnum cupressiforme (54.3%), Pohlia nutans (54.3%), Aulacomnium androgynum (40.3%), Dicranella heteromala (39.8%) and Atrichum undulatum (39.2%). Analysis of species frequency and association to particular forest type revealed that non-invaded P. sylvestris plantations differed from all other P. sylvestris forest types by lower frequency and abundance of Pleurozium schreberi, Pseudoscleropodium purum, Dicranum scoparium and D. polysetum (Table 3). Non-invaded poor P. sylvestris forests were characterized by the exclusive occurrence of Brachythecium albicans. Quercus–Acer–Tilia forests differed from other forest types by the presence of indicative Fissidens taxifolia. Non-invaded P. sylvestris plantations and P. serotina invaded poor P. sylvestris forests hosted the highest number of species, while Fagus and Robinia hosted the lowest number of species (Table 3). We found that invaded forests usually hosted fewer bryophytes than non-invaded: only in the case of poor P. sylvestris forests did the invaded forest type host a higher number of species than non-invaded (Table 3).

Table 3 Frequency [%] of bryophyte species in forest types and their indicative value (statistic and p value of association with forest types marked by the frame using the IndVal method)
Full size table

Bryophyte communities in study plots ordered along NMDS1 axis (Fig. 2) from forest types with low soil C:N ratio and light availability (Quercus–Acer–Tilia, Robinia, Fagus, Q. rubra and Q. petraea) to high (P. sylvestris forests, with the exception of non-invaded plantations). However, within these two groups forest types overlapped. NMDS1 gradient was correlated with bryophyte cover, DIFN, and soil C:N ratio. The cluster of P. sylvestris forests (both invaded and non-invaded) was affiliated with Pseudosleropodium purum, Hylocomium splendens, Dicranum scoparium, D. polysetum, and Pleurozium schreberi. Invaded and non-invaded plots overlapped in NMDS space. Analysis of beta diversity revealed that forest types differed in mean distance from the median (d.f. = 8, F = 3.9315, P = 0.0002), with the highest within-group dissimilarity in F. sylvatica forests and the lowest in Q. rubra forests (Fig. 3).

Fig. 2
figure 2

Result of non-metric multidimensional scaling (NMDS; stress=0.1924), points represent site scores for each study plot. Ellipses delimit 95% the confidence area for each forest type. Italicized labels represent species scores and are abbreviations of species names (e.g. Mniuhorn = Mnium hornum). Bold names and arrows represent passively fitted environmental variables: soil pH (pH, R2 = 0.139), soil C:N ratio (C:N, R2 = 0.120), light availability (DIFN, R2 = 0.361), terricolous bryophyte species richness (rich, R2 = 0.496) and total cover of bryophytes (cov, R2 = 0.644)

Full size image
Fig. 3
figure 3

Mean (+SE) distance from group median in multivariate space, indicating multivariate homogeneity of group dispersions for the forest types studied. Same letters denote groups which did not differ at the confidence level α = 0.05 after multiple hypotheses adjustment, according to Tukey’s a posteriori test

Full size image

Bryophyte cover depended on forest type and light availability (Fig. 4, Table 4). We found the highest cover in poor P. sylvestris forests, both uninvaded (18.5 ± 4.2%) and invaded (12.6 ± 2.5%) while the lowest – in Q. petraea, Fagus, and Q. rubra forests (4.0 ± 1.1%, 3.9 ± 1.4% and 3.3 ± 1.0%, respectively). Bryophyte cover also increased with increasing DIFN (Fig. 5). Bryophyte species richness depended on forest type, soil pH, soil C:N ratio and light availability (Fig. 4 and 5, Table 4). We found the highest bryophyte species richness per plot in poor and plantation P. sylvestris forests, both invaded by P. serotina (7.3 ± 0.7 and 6.1 ± 0.6, respectively), while the lowest species richness occurred in F. sylvatica forest (2.9 ± 0.6). Bryophyte species richness increased with increasing DIFN and decreased with increasing soil pH and soil C:N ratio (Fig. 5). However, the effect of DIFN was the highest, while the effect of soil C:N ratio was the lowest.

Fig. 4
figure 4

Marginal mean (+ SE) bryophyte species richness per plot and cover for forest types studied. Same letters denote groups that did not differ at the confidence level α = 0.05 after multiple hypotheses adjustment, according to Tukey’s a posteriori test; for model details see Table 4

Full size image
Table 4 GLMM models of species richness and cover
Full size table
Fig. 5
figure 5

Marginal responses (+ 95% CI) of bryophyte species richness per plot and cover for DIFN, soil C:N ratio and soil pH; for model details see Table 4

Full size image

Discussion

Impact of Prunus serotina on terricolous bryophytes

We found different impacts of P. serotina invasion on bryophyte response: invaded poor P. sylvestris forests had higher terricolous bryophyte species richness per plot, but lover cover, than non-invaded, while invaded P. sylvestris plantations had both higher species richness and cover than non-invaded. This might be connected with different levels of available resources in both types of P. sylvestris forests, making P. sylvestris plantations more susceptible to colonization by both alien and native species (Zerbe and Wirth 2006; Jagodziński et al. 2015). As our model accounted for light availability, our results contradicted previous studies indicating the negative effects of P. serotina on bryophyte species richness (Verheyen et al. 2007; Halarewicz and Pruchniewicz 2015; Vegini et al. 2020). This might be connected with the high capacity of P. serotina for light interception (Urban et al. 2009; Jagodziński et al. 2019), resulting in different biomass allocation patterns by developing a large canopy, relative to its biomass (Dyderski and Jagodziński 2019a).

In P. sylvestris plantations invaded by P. serotina we found more bryophytes typical of acidophilous coniferous forests than in non-invaded, for example Pseudoscleropodium purum, Pleurozium schreberi or Dicranum scoparium. However, at larger scales these species are generally frequent in all P. sylvestris-dominated forests. This might be connected with limiting the growth of graminoids and forest-edge forbs by P. serotina, which usually colonize P. sylvestris plantations (Zerbe and Wirth 2006). In poor P. sylvestris forests addition of P. serotina might create microsites suitable for colonization by the generalist Hypnum cupressiforme. However, lack of difference in soil C:N ratio between invaded and non-invaded poor P. sylvestris forests does not support this suggestion. Different patterns of P. serotina impacts on bryophytes in P. sylvestris plantations and poor forests are an example of the context-dependence of biological invasion (González-Moreno et al. 2014; Dyderski and Jagodziński 2019b; Sapsford et al. 2020). Here, a mismatch between habitat fertility and forest species composition changed the impact of an invasive tree species on bryophytes.

Impact of Quercus rubra on terricolous bryophytes

We found that Q. rubra forests had smaller terricolous bryophyte species pools than all other forest types except F. sylvatica and R. pseudoacacia forests. In our study Q. rubra forests hosted a lower proportion of bryophytes than in post-industrial forests, where Q. rubra forests had 38 of 82 species recorded as terricolous (Jagodziński et al. 2018). Similarly, in two types of Q. rubra forests Woziwoda et al. (2017) found 38 species of terricolous bryophytes. Despite the lowest mean bryophyte cover, Q. rubra forests hosted a relatively constant set of bryophyte species. This was mainly driven by four species: Brachythecium salebrosum, Hypnum cupressiforme, Dicranella heteromala and Aulacomnium androgynum – two generalists and two specialists occurring mainly near the root collar. The three latter species revealed fidelity to Q. rubra forests, and also occurred with higher frequency in F. sylvatica and Q. petraea forests, which are similar to Q. rubra in terms of soil fertility and community structure (Chmura 2013; Major et al. 2013; Dyderski et al. 2020). This group of relatively constant species drove high species richness, and resulted in low beta diversity, expressed by mean distance to the median (Fig. 3) of terricolous bryophyte assemblages in Q. rubra forests. Therefore, high species richness of epiphytic bryophytes on Q. rubra (Woziwoda et al. 2017; Jagodziński et al. 2018; Fudali and Szymanowski 2019) may affect the terricolous pool by expansion from root collars onto the soil. This speculation needs further experimental confirmation.

Impact of Robinia pseudoacacia on terricolous bryophytes

We found that only F. sylvatica forests had smaller species pools of terricolous bryophytes than R. pseudoacacia forests, as well as low species richness and cover. As R. pseudoacacia forests occupy fertile soils, where natural forests host low diversity of terricolous bryophytes, there are almost no data about R. pseudoacacia impacts. Relatively high (within a studied dataset) species richness per plot is also connected with the presence of generalists – Brachythecium salebrosum and Atrichum undulatum, as well as with generally high soil fertility. This resulted in the lowest beta-diversity (Fig. 3) among forest types studied and low distinctiveness. The level of species richness and cover in R. pseudoacacia forests was similar to Quercus–Acer–Tilia, which might be a reference forest type in terms of soil fertility. However, due to the high number of generalist bryophyte species, their composition in R. pseudoacacia forests is closer to P. sylvestris plantations. This indicates that R. pseudoacacia contributes to the biotic homogenization of terricolous bryophytes assemblages.

Impact of study design and methodology

Both models of species richness and cover revealed the high importance of both light availability and forest types. Therefore, marginal responses revealed in our study were adjusted to standard light availability (Fig. 4). In this way, we separated species-specific effects from light availability effects, which are important drivers of bryophyte occurrence (Rydin 2008). Therefore, accounting for differences in bryophyte cover and biodiversity needs to refer to light availability, as the factor determining both species occurrence and abundance. Light availability is strongly modified by dominant tree species (Peterken 2001; Niinemets 2010; Dyderski and Jagodziński 2019a). Therefore, light limitation seems to be the main mechanism of invasive tree species impacts on bryophyte communities. Another factor affecting study design is the choice of the reference ecosystem. Our study accounted for multiple native forests, which might be replaced by invasive species. This expands our inference regarding predicted impacts of invasive species on bryophytes to multiple types of native ecosystems. However, the application of a ‘treatment-control-removal’ approach would have benefits in understanding invasion legacies (Barney et al. 2015), which are important in bryophyte responses to habitat transformation (Pharo and Lindenmayer 2009).

Conclusions

We found that P. serotina invaded forests hosted more bryophyte species and had higher species richness than non-invaded in P. sylvestris plantations. Quercus rubra and R. pseudoacacia invaded forests hosted fewer bryophyte species but had similar species richness as non-invaded forests. However, these results accounted for adjusted light availability. As most of the previously reported effects of alien tree species on bryophytes did not account for light availability and soil variability, we argue that light limitation by invasive species might be more important than dominant tree species identity in limiting terricolous bryophyte species richness and abundance. This suggestion needs further studies, accounting for varied abundances of invasive species, to assess per capita effects of invaders (Ricciardi et al. 2013; Kumschick et al. 2015). Therefore, conservation of sites with high biodiversity of terricolous bryophytes needs greater insight into modification of canopy cover, as a metric of invasive species impacts.