Introduction

Biological invasions are one of the major threats to biodiversity (Mack et al. 2000; Davis 2003; Essl et al. 2017). Ecological success of invasive species is a derivate of three groups of factors: species invasiveness, habitat invasibility, and propagule pressure. Species invasiveness compromises life history traits providing advantages over native species, which allow alien species to establish in their secondary range (e.g., Tecco et al. 2010; Pyšek et al. 2014; Gallien et al. 2016). Habitat invasibility describes its vulnerability to colonization by alien species (e.g., Lonsdale 1999; Alpert et al. 2000; Richardson and Pyšek 2006). Propagule pressure reflects quantity, quality, and availability of alien species’ dispersal units (e.g., Lockwood et al. 2005; Křivánek et al. 2006; Vanhellemont et al. 2009). However, invasion success depends not merely on these factors treated separately, but rather on their interactions (e.g., Alpert et al. 2000; Davis et al. 2005; Jagodziński et al. 2015).

Disturbance is a catastrophic event removing part of the ecosystem biomass, caused by biotic or abiotic factors. Examples include biotic insect outbreaks (Jonášová and Prach 2004; Kučerová et al. 2008; Zeppenfeld et al. 2015) and abiotic wildfires (Clarke et al. 2015; Benscoter et al. 2015; Frelich et al. 2017). Human activity may also create disturbance events (e.g., Prach et al. 2001, 2014; Dyderski et al. 2016). Disturbance regimes differ in both frequency and severity, i.e., length of interval between consecutive disturbance events and in proportion of biomass destroyed during single disturbance events (Herben et al. 2016, 2018). Thus, some plant species are adapted to disturbed habitats, which is manifested in their life strategies (Grime 2001; Pierce et al. 2013; Herben et al. 2018), reflecting their functional traits (Baker 1974; Westoby 1998; Prach et al. 2001). For that reason, disturbance is one of the major factors shaping species composition of vegetation in some biomes (Turetsky and Louis 2006; Johnstone et al. 2016; Frelich et al. 2017).

Disturbance dependency is an example of invasion ecology concepts merging these aspects (Hobbs and Huenneke 1992; Chabrerie et al. 2008; Jeschke 2014; Rodríguez et al. 2017). This concept assumes that invasive species reach higher ecological success in disturbed sites (habitat parameter). As intermediate disturbances lead to increase of species richness (Fox 1979), according to the biotic acceptance theory (Stohlgren et al. 2006), this leads to the increase of alien species richness. Moreover, changes in disturbance frequency and intensity (both increase and decrease), as well as change of disturbance type may alter the species composition and create the opportunities for invasive species (Hobbs and Huenneke 1992). For that reason, species with functional traits favoring disturbance have higher spread availability in the secondary range (invasiveness). Moreover, disturbed sites may attract species that are vectors of propagules, e.g., frugivorous birds, increasing propagule availability (propagule pressure). Disturbance is also claimed to be a reason for invasions in agroecosystems (Baker 1974; Groves 2006), riparian landscapes (Pyšek and Prach 1993; Richardson et al. 2007; Dyderski and Jagodziński 2016), or urban ecosystems (Knapp et al. 2010; Jarošík et al. 2011; Obidziński et al. 2016), as well as in natural forests (Chmura and Sierka 2007). Nevertheless, our literature review did not find studies about disturbance impacts on the youngest life stage of invasive trees and shrubs.

The aim of this study was to assess the role of disturbances in shaping forest understory vegetation and responses of invasive tree and shrub species to disturbances in forest plant communities. Based on earlier studies on invasion ecology we hypothesized that (1) disturbances are more important factor shaping species composition of vegetation than fertility gradient, (2) invasive species perform better (i.e., reaching higher biomass and density of natural regeneration) in plant communities with more frequent and more severe disturbances, and (3) invasive species will respond differently to disturbance indices—competitive species responses will be negative and ruderal species responses positive.

Materials and methods

Species studied

As model species, we used the three most frequent woody alien species invading forest ecosystems in Europe (Wagner et al. 2017): Prunus serotina Ehrh., Quercus rubra L., and Robinia pseudoacacia L. All these species come from the eastern part of North America and were introduced for forest management purposes (Starfinger et al. 2003; Cierjacks et al. 2013; Woziwoda et al. 2014). P. serotina in its native range is a tree but in Europe it is most frequently a tall shrub. Due to high rate of leaf litter decomposition, P. serotina was widely used as a soil improver (Muys et al. 1992; Aerts et al. 2017; Horodecki and Jagodziński 2017). P. serotina is able to wait out unfavorable environmental conditions (insufficient light availability) by slow growth rates of seedlings (Closset-Kopp et al. 2007) and benefit from opportunity windows in resource availability (Chabrerie et al. 2008). This species is mainly dispersed by birds (Pairon et al. 2006). Q. rubra, due to large seed mass, is a species strongly dependent on propagule pressure (Riepšas and Straigyté 2008; Myczko et al. 2014; Jagodziński et al. 2018), but also able to use carbohydrate reserves stored in the acorn (Ziegenhagen and Kausch 1995). R. pseudoacacia is a wind-dispersed, pioneer tree species, that usually colonizes disturbed areas (Cierjacks et al. 2013; Vítková et al. 2017; Sádlo et al. 2017), especially due to its ability to fix nitrogen (Boring and Swank 1984; Rice et al. 2004). Thus, the species studied have differing responses to disturbance. In Poland, P. serotina prefers less-fertile sites and R. pseudoacacia—those more fertile, while Q. rubra—both less-fertile and more fertile sites (Chmura 2004; Halarewicz 2012; Gazda and Szwagrzyk 2016).

Study area

The study was conducted in the forests of Western Poland, within Wielkopolski National Park (52º16′N, 16º48′E). This national park covers an area of 7584 ha and its aim is to preserve a valuable post-glacial landscape, including valleys, moraine hills, and lakes. Climate of the study area is temperate, transitional between oceanic and continental. In the years 1951–2010, mean annual temperature was 8.4 °C and mean annual precipitation was 521 mm (Poznań meteorological station, c.a. 15 km from the study area). Main soil types are luvisols (47%) and brunic and haplic soils (30%), podzols constitute 7% (Nowak et al. 2000). Forests cover 61% of Wielkopolski National Park, and the most frequent plant associations are Galio sylvatici-Carpinetum, Calamagrostio arundinaceae-Quercetum, and Querco roboris-Pinetum. However, some patches are covered by fertile riparian forests (Fraxino-Alnetum, Querco-Ulmetum minoris) or poor coniferous forests (Leucobryo-Pinetum). Nevertheless, large areas of deciduous forests were planted with Pinus sylvestris before the national park was established in 1957 (Nowak et al. 2000). This replacement caused transformation of both vegetation and soil chemistry by nutrient leaching and acidification (Binkley and Giardina 1998; Zerbe and Wirth 2006). Another negative aspect of human impacts was introductions of alien woody species (Purcel 2009; Gazda and Szwagrzyk 2016). This resulted in the presence of pure stands of alien species surrounded by large areas of forest ecosystems with different levels of disturbances and human impacts. Therefore, Wielkopolski National Park seems to be an excellent place to test hypotheses about impacts of disturbances on ecological success of invasive species.

Study design

In the study area, we established 21 block sites (nine for Q. rubra, six for P. serotina, and six for R. pseudoacacia), each with a central monoculture stand of invasive species housing a total of 378 plots (squares 100 m2 in area). For study design details, see Dyderski and Jagodziński (2018). In the case of P. serotina, occurring only as an admixture in tree stands, these are tree stands with high densities of fruiting trees. In the central part of each monoculture, we established a pair of study plots. Along each of the four sides (N, S, E, and W) of the central pair, we established two pairs of plots: one nearly outside the stand—at the invasion edge, and the second pair 30 m from the invasion edge. The invasion edge was defined as the border between crowns of the invasive species monoculture and the surrounding trees. Thus, each study block compromised 18 study plots. The study design did not generate pseudoreplications, which may be expected due to short distances between plots in blocks, because of the high spatial heterogeneity. This was manifested in high variability of natural regeneration within plots, light availability, and microsite occurrences. Because plots were distributed systematically, six of them were located within forest vegetation paths and were therefore excluded from analyses, so that the final number of plots was n = 372. The plots were divided into nine categories according to dominant tree species: Acer platanoides-Tilia cordata-Carpinus betulus, Fagus sylvatica, Pinus sylvestris-Acer pseudoplatanus, Pinus sylvestris-Acer platanoides-Tilia cordata-Carpinus betulus, Pinus sylvestris-Prunus serotina, Pinus sylvestris-Quercus petraea, Quercus petraea, Quercus rubra, and Robinia pseudoacacia (Online Resource 1). The variability of light availability, different tree stand types, and microsite presence, allowed us to assume that the experimental design did not produce spatial clusters connected with block arrangement, which would have needed to be accounted for in the statistical design, e.g., by application of mixed models (Roberts et al. 2017).

During July of 2015-2017, we measured root collar diameters and heights of all trees of the study species on each plot shorter than 0.5 m (in total 39664 plants). Using allometric equations developed on sample trees (356 of P. serotina, 133 of Q. rubra and 158 of R. pseudoacacia) harvested in a 5-m buffer zone around the study plots (Online Resource 2; Dyderski and Jagodziński 2018), we estimated their total above- and belowground biomass. In addition to natural regeneration of the study species, we assessed species composition of understory vegetation, recording species identities and abundances using the Braun-Blanquet method. We also measured canopy openness index (diffuse non-interceptance; DIFN) using an LAI-2200 plant canopy analyzer (Li-Cor Inc., Lincoln, NE, USA). For each plot, we recorded four series of ten samples. The measurements were conducted in August 2016, as our previous studies indicated that this month has maximal canopy foliage development (Knight et al. 2008). In March 2017, we collected four samples of leaf litter from circular plots (0.16 m2). The samples were dried in the oven at 65 °C to constant mass and weighed using a balance with an accuracy of 1 g. Litter pH was assessed using an electronic pH-meter in distilled water solution after 24 h. We used these variables as approximations of nutrient cycling rates and soil acidity in forest ecosystems.

Data analysis

Using data about vegetation species composition for each study plot, we calculated community-weighted means (CWM) of disturbance indices proposed by Herben et al. (2016). These indices reflect disturbance severity, expressed as fraction of community biomass being destroyed, and disturbance frequency, expressed as logarithm of mean time between disturbance events. These indices were developed based on expert assessment of vegetation typological units (phytosociological classes) and calibrated using a dataset of vegetation samples. The indices are correlated with functional traits reflecting disturbance responses—height, seed mass, and clonality (Herben et al. 2016). Because disturbances in forest ecosystems may affect only the understory or the whole community, we decided to analyze both types of disturbance measures.

To assess the factors shaping vegetation species composition, we performed Detrended Correspondence Analysis (DCA) using the vegan::decorana() function (Oksanen et al. 2016). This type of multivariate analysis was chosen due to long environmental gradients (> 3 SD units). As an unconstrained analysis, DCA reveals dispersion of points in ordination space, reflecting particular species abundances within study plots. To determine factors correlated with the gradients revealed, we used the vegan::envfit() function (Oksanen et al. 2016) for passive projection of environmental factors: DIFN, litter mass, disturbance indices, and CWMs of specific leaf area (SLA), canopy height, and seed mass—functional traits describing plant life strategies, according to LHS concept (leaf-height-seed; Westoby 1998). LHS traits were obtained from the LEDA database (Kleyer et al. 2008). We also used proportion of species representing phytosociological classes (Ratyńska et al. 2010) and life strategies (Klotz et al. 2002). CWM computation was done using the FD::functcomp() function (Laliberté et al. 2014).

We used ordinary least squares regression to check the strength of relationships between the measures of invasive species ecological success (density and biomass of natural regeneration) and disturbance indices. Prior to analyses, we log-transformed densities and biomasses, due to log-normal distributions. Because high numbers of observations increase statistical significance of low differences between treatments as expressed by p values, for our case of n = 1116 (372 plots × 3 years), we resigned from interpretation of p values, which could be biased by reporting statistically significant results with little or no biological meaning (the so-called ‘p value hacking’), according to the American Statistical Association (Wasserstein and Lazar 2016). Instead, we focused on correlation coefficients and effect sizes. To evaluate the latter, we compared 95% confidence intervals (CI) of predicted measures of invasive species ecological success. This was done by inputting the 0.025, 0.5, and 0.975 quantiles of disturbance indices into the regression equation and visualizing the range of the values obtained. The wider the CI of predicted values, the higher the effect of the considered factor on the dependent variable. Although only R. pseudoacacia has assigned disturbance indices values, in regressions, we used disturbance indices calculated excluding the study species, to avoid circular reasoning (Thomsen et al. 2016). All analyses were conducted using R software (R Core Team 2017).

Results

Impact of disturbance on understory vegetation

DCA (Fig. 1, Table 1) revealed two main gradients in differentiation of vegetation species composition. The first gradient (across DCA1 axis) was identified as availability of light and soil nutrients: from poor P. sylvestris sites with high light availability at the left to fertile oak-hornbeam and R. pseudoacacia forests at the right. This gradient also reflected functional traits connected with competition—life strategy ‘c’ (competitors) and high SLA CWM. The differentiation along DCA1 axis was reflected both at the tree stand and understory species composition levels, which was reflected by opposite proportions of Vaccinio-Piceetea and Querco-Fagetea species cover. The second distinct gradient (along DCA2 axis) described disturbance intensity. Low values of DCA2 indicated the presence of stress-tolerant species and ancient forest indicator species, whereas high values—presence of ‘r’ strategy species—were an indicator of frequent and severe disturbances. This was also correlated with proportion of meadow species (Molinio-Arrhenatheretea). The range of whole community disturbance frequency index ranged from − 1.986 to − 1.107, with an average of − 1.761 ± 0.004, which indicated a mean turnover interval of c.a. 58 years. The analogous index for the herb layer ranged from − 1.252 to − 0.526, with an average of − 0.781 ± 0.002, which indicated a mean turnover interval of c.a. 16 years. Whole community disturbance severity index ranged from 0.224 to 0.447, with an average of 0.296 ± 0.001 and herb layer severity index ranged from 0.042 to 0.381, with an average of 0.172 ± 0.002.

Fig. 1
figure 1

DCA results conducted on understory vegetation of sample plots in three growing seasons (n = 1116). Labels represent passively projected environmental variables (Table 1), colors represent tree stand types: Ap-Tc-Cb, Acer platanoides-Tilia cordata-Carpinus betulus; Fs, Fagus sylvatica; Ps-Ap, Pinus sylvestris-Acer pseudoplatanus; Ps-Ap-Tc-Cb, Pinus sylvestris-Acer platanoides-Tilia cordata-Carpinus betulus; Ps-Pse, Pinus sylvestris-Prunus serotina; Ps-Qp, Pinus sylvestris-Quercus petraea; Qp, Quercus petraea, Qr, Quercus rubra; Rp, Robinia pseudoacacia (Table S1)

Full size image
Table 1 Parameters of environmental variables fitted to the DCA analysis results
Full size table

Impact of disturbance on invasive species ecological success

Analysis of correlations between disturbance indices and measures of success of invasive species natural regeneration (density and biomass) revealed low values of correlation coefficients. However, strength of correlations and size effects varied among species, disturbance indices, and life stages (Fig. 2, Table 2, Online Resource 3). The highest correlation coefficients were found for P. serotina seedling density and total P. serotina density and biomass, and effect of herb layer disturbance frequency index on R. pseudoacacia density and biomass (both seedlings and total). P. serotina density and biomass had the highest correlation coefficients in relationships with whole community frequency index and herb layer severity index, and the lowest—with herb layer frequency index. R. pseudoacacia biomass and density were most closely correlated with herb layer frequency index. Q. rubra did not show any relationships with the disturbance indices studied. Regarding direction of correlation (positive or negative), we found that most of the correlations with r > 0.2 were positive. An exception was response of R. pseudoacacia density to herb layer frequency index. However, response of R. pseudoacacia biomass to the same index was positive. The best fit predictor for P. serotina total density—whole community frequency index—predicted CI of 9.6–670.2 ind. ha−1, with median of 33.5 ind. ha−1, whereas real CI was 0.0–19812.5 ind. ha−1 and a median of 100 ind. ha−1. The best predictor for R. pseudoacacia seedling density—herb layer frequency index—predicted CI of 0.9–26.3 ind. ha−1, with median of 6.2 ind. ha−1, whereas real CI was 0.0–13550.0 ind. ha−1, with median of 0 ind. ha−1.

Fig. 2
figure 2

Effect sizes for invasive species biomass and density prediction based on disturbance indices, modeled using simple regression (Online Resource 3). Points represent biomass or density predicted for median value of each disturbance indices, whiskers—range of 95% CI (i.e., values of 0.025 and 0.975 quantiles). Length of CI reflects relative size effect of each disturbance index within a panel

Full size image
Table 2 Pearson’s correlation coefficients (r) between quantity of invasive species studied (rows, biomasses, and densities of seedlings) and all plants in regeneration layers up to 0.5 m height (total) and disturbance indices (columns; Herben et al. 2016). For regression parameters see Online Resource 3
Full size table

Discussion

Impact of disturbance on vegetation species composition

Our study revealed that disturbance was the second factor characterizing vegetation species composition. Due to the extent of the fertility gradient, determining dominant tree species, this group of factors shaped the first gradient. Different tree species have different soil requirements (Ellenberg 1988; Kozlowski and Pallardy 1997), but their presence also shapes soil properties by altering biogeochemical cycles (e.g., Reich et al. 2005; Hobbie et al. 2006; Mueller et al. 2012). For that reason, the fertility gradient was superior over the disturbance gradient. Our previous study conducted in the same set of study plots (Dyderski and Jagodziński 2018) revealed that soil fertility proxies influenced biomass of natural regeneration of the species studied. Patterns of invasive species abundance along fertility gradient would be related to the biotic acceptance theory, which states that plant communities richer in native species are richer in alien species (Stohlgren et al. 2006). This pattern was also confirmed in relationship between native species richness and invasive species abundance (Knight et al. 2008). Species richness is related with soil fertility (Tilman 1986), which also may limit some alien species (e.g., Pyšek and Prach 1993; Chmura 2004; Halarewicz 2012). However, the importance of soil fertility proxies in our previous study (Dyderski and Jagodziński 2018) was lower than importance of propagule pressure.

The disturbance gradient was connected with the presence of ruderals and species typical to meadows and pastures (Molinio-Arrhenatheretea) and negatively correlated with cover of ancient forest species indicators. The latter is a group of forest specialists, connected with undisturbed, long-lasting forests (Peterken 1974; Hermy et al. 1999; Dyderski et al. 2017). These correlations provide convincing evidence that disturbance indicators elaborated by Herben et al. (2016) allow for good assessment of vegetation response to disturbances.

Impact of disturbance on studied species responses

Our study revealed that although disturbances (represented here by indices) are important factors shaping understory vegetation species composition, they have low importance in shaping ecological success of the invasive species studied. This result may seem to contradict previous studies (e.g., Chmura and Sierka 2007; Chabrerie et al. 2008; Rodríguez et al. 2017). However, the difference is connected with the life stage that we studied—natural regeneration. Seedlings of the invasive woody species reflect only one of the stages of the invasion process (Sebert-Cuvillier et al. 2007). Although this stage is crucial for further growth of invaders and their reproduction, our results show that seedling performance is shaped to only a low degree by disturbance. For that reason, our results do not contradict the previous findings, but show the exception to the general rule. Nevertheless, another study we conducted revealed that the most important predictors of biomasses of the species studied were factors describing propagule pressure (Dyderski and Jagodziński 2018). Invasion success is usually controlled by interaction of propagule pressure and invasibility (Davis et al. 2005). Thus, disturbance regime, as only one aspect of invasibility, cannot be the crucial predictor for ecological success of the species studied.

One of the possible reasons influencing the low impact of disturbance could be the relatively short gradient of disturbances within our dataset. We assumed this to be a limitation of our approach. However, Hobbs and Huenneke (1992) claim that any change of disturbance regime may create opportunity for invasive species. Herben et al. (2016) elaborated their indices for the whole range of vegetation types, from pleuston communities to the most stable forests. The highest disturbance turnover, typical to pleuston vegetation or crops, cannot possibly occur in forest ecosystems, given the high proportion of biomass destroyed annually. Values of the disturbance frequency indicators for the whole community were higher than proposed for forest plant communities (Herben et al. 2016), whereas values of herb layer disturbance frequency indicators were lower. According to the bimodal distribution of indicator values presented by Herben et al. (2016), the range of values obtained by us cover the range of the first distribution peak around − 2. Similarly, values of severity index from our study covered the peak between 0.1 and 0.4. For that reason, we can assume that our study covered the sustainable proportion of disturbance gradients that occur in forest plant communities. This is also supported by the important role in shaping the species composition of vegetation shown by DCA (Fig. 1).

Different responses of disturbance of particular species studied

Despite the low importance and effect sizes of the patterns studied, we expected differences in responses of the invasive species studied to disturbance indices, connected with species biology. However, these conclusions are limited, due to low effect sizes of the relationships studied. The species most expected to be prone to disturbance, Q. rubra did not exhibit any relationship to the parameters studied, likely as a result of seedling dependence on carbohydrates stored in acorns (Sork 1984; Ziegenhagen and Kausch 1995). However, this effect should be lower for the case of the total regeneration layer. One possible explanation may be the ability of persistence in unsuitable conditions for one or two growing seasons, similar to the Oskar strategy of P. serotina (Closset-Kopp et al. 2007). Nevertheless, P. serotina showed positive responses to disturbances, supporting results of earlier studies (Closset-Kopp et al. 2007; Chabrerie et al. 2008; Jagodziński et al. in prep.).

For R. pseudoacacia, the highest correlation coefficient was found in a negative relationship between its ecological success and understory disturbance frequency index. This species is known for acting as ruderal—occupation of disturbed areas, wind dispersal, and production of large number of seeds (Cierjacks et al. 2013; Vítková et al. 2017). For that reason, we assume that this pattern may be an effect of presence of numerous R. pseudoacacia in plots with low light availability, thus favoring forest specialists (Dzwonko 1993; Hermy et al. 1999). In contrast, plots with abundant open habitat species did not host abundant natural regeneration of R. pseudoacacia.

Our study revealed that although disturbance indices are important factors shaping understory vegetation species composition, it has low importance in shaping ecological success of the invasive species studied. Due to the low effect sizes, we can state only limited conclusions about impacts of disturbance on ecological success of natural regeneration of the invasive species studied. The results suggest that for seedlings (up to 50 cm height) disturbance, a leading factor in biological invasions for numerous taxa, has a small role in this case. Our previous study (Dyderski and Jagodziński 2018) revealed the crucial role of propagule pressure in shaping ecological success of natural regeneration for these species. Thus, we may assume that their ecological success is more connected with stochastic processes in the invader’s seedling populations than with stochastic release from competition caused by disturbances. This possibility would lead to two management implications. Firstly, risk assessments should account not for disturbed sites, but rather for propagule sources. Secondly, the regeneration layer needs monitoring as random processes, not predictable using data about disturbances, may lead to quick release of invaders (Hobbs and Huenneke 1992).