Experiments testing multiple factors that affect the rate of invasions in forests are scarce. We aimed to assess how the biomass of invasive Prunus serotina changed over eight years and how this change was affected by light availability, tree stand growth, and propagule pressure. The study was conducted in Siemianice Experimental Forest (W Poland), a common garden forest experiment with 14 tree species. We investigated aboveground biomass and density of P. serotina within 53 experimental plots with initial measurements in 2005 and repeated in 2013. We also measured light availability and distance from seed sources. We used generalized additive models to assess the impact of particular predictors on P. serotina biomass in 2013 and its relative change over eight years. The relative biomass increments of P. serotina ranged from 0 to 22,000-fold. The success of P. serotina, expressed as aboveground biomass and biomass increment, varied among different tree species stands, but was greater under conifers. Total biomass of P. serotina depended on light and propagule availability while biomass increment depended on the change in tree stand biomass, a metric corresponding to tree stand maturation. Our study quantified the range of invasion intensity, expressed as biomass increment, in a forest common garden experiment with 14 tree species. Canopy cover was the most important variable to reduce susceptibility to invasion by P. serotina. Even a modest decrease of overstory biomass, e.g. caused by dieback of coniferous species, may be risky in areas with high propagule pressure from invasive tree species. Thus, P. serotina control may include maintaining high canopy closure and supporting natural regeneration of tree species with high leaf area index, which shade the understory.

Allometric equations, biomass, invasion dynamics, light availability, natural regeneration, tree species effect

Although invasive trees and shrubs are a worldwide problem (Richardson and Rejmánek 2011; Rejmánek and Richardson 2013), their impacts and mechanisms of ecological success are insufficiently understood. Their longevity and the long duration of their lag period (Kowarik 1995; Lonsdale 1999), as well as the physically large dimensions of invasive woody species, make research more challenging (Richardson and Rejmánek 2011). Invasive tree and shrub species are often classified as “transformers” (Richardson et al. 2000), due to the ability of these species to modify habitats. Invasive trees modify light conditions, usually decreasing light availability (Knight et al. 2008; Mueller et al. 2016) and increase rate of soil nutrient cycling (Binkley and Valentine 1991; Allison and Vitousek 2004; Reich et al. 2005). Due to this potential, invasive species of trees and shrubs are a serious threat to the environment (Richardson 1998; Richardson and Rejmánek 2011; Hawkins et al. 2015; Aerts et al. 2017). On the other hand, these properties of trees also highlight the potential role of native tree species in shaping ecosystem invasibility, by creating regeneration niches for invaders (Knight et al. 2008; Jagodziński et al. 2018) or closing them, by providing unfavorable conditions (e.g. González-Muñoz et al. 2014; Nyssen et al. 2016; Dyderski and Jagodziński 2018).

There are three crucial elements shaping invasion success: ecosystem invasibility, propagule pressure, and species invasiveness (Lonsdale 1999; Richardson et al. 2000). In the case of forest ecosystems, the level of resources (e.g. light, water, nutrients) is relatively constant compared to other ecosystems, due to the low frequency of disturbances within a single forest patch (Herben et al. 2016). Furthermore, differences in the availability of these resources across sites with similar soil and climate are determined mainly by tree species composition (see previous paragraph) and tree density (Jagodziński and Oleksyn 2009). However, disturbances in forests, regardless of their spatial scale, are usually connected with increased resource availability (Verheyen et al. 2003; Chmura and Sierka 2007). Although forest ecosystems usually return to a dynamic equilibrium state after disturbance (Gunderson 2000; Johnstone et al. 2016), the pulse of resource availability is crucial for success of invasive species (Davis et al. 2000). However, establishment of alien species also requires the presence of propagules, and the capacity of alien species to spread is crucial for determination of invasiveness (Richardson et al. 2000). The quantity of available propagules determines propagule pressure, which may be approximated by the number of mature plants (Vanhellemont et al. 2009) and the distance to the nearest propagule source (Deckers et al. 2005; Pairon et al. 2006a; Jagodziński et al. 2015) or area of alien species cultivation (Pyšek et al. 2009).

Few studies of invasive trees and shrubs have quantified the density and size of different life stages during the initial stages of invasion. Because seedlings are more vulnerable to environmental conditions (e.g. drought, frosts or herbivory) than adult trees (Niinemets and Valladares 2006), their survival and recruitment into adult life phases is important for shaping future tree stand composition (Baraloto et al. 2005). Moreover, there is a lack of long-term studies which focus on the dynamics of species invasion and provide a quantitative assessment of the process.

Due to different invasion metrics used in fully controlled experiments and those used in field assessments, there is a gap in transferability of results between these two types of studies. Experiments in fully controlled conditions have measured invader success as relative growth rate or biomass (e.g. Grotkopp et al. 2010). In contrast, field studies usually measure alien species presence/absence, cover, or density (e.g. Pauchard et al. 2009; Jarošík et al. 2011; Kowarik et al. 2013; Woziwoda et al. 2014). Thus, field studies that quantify the invasive potential of alien species using the metrics of invasion success which are used in experiments, e.g. biomass, would provide discernment between findings due to ecological processes and those due to differences in measurement techniques. Biomass may provide a better measure of ecological success, reflecting space filled by the species, and is also a proxy of plant fitness (Younginger et al. 2017).

We took advantage of the invasion of Prunus serotina Ehrh. into an experimental forest common garden comprised of different tree species. The wide range of ecological characteristics of the cultivated species, similar soil parent material, climate and site history, and presence of alien species, compromise an ideal experimental design for invasion ecology studies (Knight et al. 2008; Jagodziński et al. 2018). The results would facilitate better recognition of the threats and management vulnerabilities of particular invaders.

Our study addresses the following questions: (1) how did invasion intensity, expressed as total biomass of P. serotina, change over eight years? and (2) how was this change connected with changes in light availability, tree stand maturation and seed source availability?

In 2005 and in September 2013, all specimens of P. serotina were investigated within all 53 experimental plots. We measured the root collar diameter (RCD) and height (H) of each tree and we determined its location within the plot with 0.25 m accuracy. Following Knight et al. (2008), we chose 1 mm RCD as a threshold to distinguish current-year from second-year or older specimens. The two size classes were distinguished due to low survival of seedlings after the first year (Dyderski and Jagodziński under review).

Biomass, as a function of both the dimensions and the density of plants, directly results from of invasive species growth, thus providing an approximation of invasion success. To assess the biomass of P. serotina within the experimental plots we harvested a subsample 59 trees (from a total population of 2339) for biomass estimation. We randomly selected 58 trees to harvest from the database of all measured trees, ordered by increasing RCD and H, plus the tallest tree, which was chosen to ensure coverage of the entire range of dimensions (n = 59; Suppl. material 2, Fig. S1). The selection was based on division of the dataset into equal intervals across increasing RCD and height. Then, we preselected trees with particular dimensions within particular study plots and in field we cut the first encountered tree with a given dimension. Model trees were harvested, oven-dried at 65 °C to constant mass and weighed with an accuracy of 0.001 g. Aboveground biomass (AB) of model trees ranged from 0.014 to 26546.52 g. Based on tree dimensions and masses, we developed allometric equations to estimate aboveground biomass (AB; [g]) of P. serotina within plots. For smaller trees (up to 10 mm RCD, n = 52) AB = 0.3066 ∙ (RCD²H)^1.0165, with RMSE = 0.4 g and R² of the linearized model, i.e. transformed into ln(AB) = a + b ∙ ln(RCD²H) form, was 0.997. For larger trees (10–126 mm RCD, n = 7) AB = 0.3576 ∙ (RCD²H)^0.9445, with RMSE = 7.8 g and the R² of linearized model was 0.999. Models were chosen based on Akaike’s Information Criterion (AIC; Suppl. material 2, Table S2). We applied these models to the results of the inventory from 2013 to obtain information about biomass in 2013.

To assess propagule pressure, we used a GPS receiver to map fruiting trees up to 50 m from each plot, as the probability of occurrence of P. serotina natural regeneration is highest within this distance (Jagodziński et al. 2015). We also recorded fruiting P. serotina within the study plots. This data allowed calculation of two proxies of propagule pressure: distance to the closest seed tree, which may be more important in cases of longer dispersal distances, and the number of fruiting trees within a 50 m radius (see rationale in study area description). We also investigated tree stand structure of the overstory, by measurements of diameter at breast height of all trees within the plots (Suppl. material 1, Table S1). For each plot, we calculated tree stand biomass using allometric equations specific to each species (Suppl. material 3, Table S3) in 2013 and in 2005. To quantify light availability, we measured the canopy openness index (diffuse non-interceptance; DIFN) each month during the growing season (June–August) of 2005 and 2013 using a LAI-2000 and LAI-2200 plant canopy analyzers (Li-Cor Inc., Lincoln, NE, USA) following the methods used in Knight et al. (2008). We took 20 measurements at six locations near the center of each plot at the height of 0.5 m above ground. DIFN indicates the fraction of open-field diffusive solar radiation reaching the forest floor.

We calculated relative increments of biomass for tree stands and for P. serotina as (B₂₀₁₃-B₂₀₀₅)/B₂₀₀₅, where B₂₀₀₅ – biomass in 2005 and B₂₀₁₃ – in 2013. We used relative increments to account for the effect of initial biomass from population dynamics within the study period. Tree stand biomass change was used as a proxy for tree stand maturation, describing increments of tree quantity in the ecosystem. In the case of negative values, biomass change may reflect quantity of trees destroyed during disturbances. Tree stand maturation is also connected with decrease of light availability (Hamm et al. 2017). Although biomass change mainly expressed tree stand productivity, in our case, when some tree stands were damaged, this variable expressed the quantity of damage (in cases of negative values of biomass). To analyze impacts of tree stand maturation (expressed as the change in tree stand biomass), light availability (DIFN) and propagule pressure (distance to the nearest fruiting tree and number of fruiting trees up to 50 m) on biomass crop and biomass relative change of P. serotina we used generalized additive mixed models (GAMM), describing the additive interactions between the analyzed predictors. As mixed-effects models, GAMM also allowed us to handle study design by including random effects describing study block (Fig. 1). GAMMs were prepared using the mgcv package (Wood 2011). We compared GAMMs using Akaike’s Information Criterion to choose the best fit model. Differences between tree stands in P. serotina biomasses and densities were tested using Kruskal-Wallis tests, due to the small number of samples and lack of normality. All analyses were conducted in R software (R Core Team 2017).

We found the highest intensities of invasion, expressed as relative biomass change, within plots with low P. serotina density in 2005. In these plots, invasion accelerated later, which was indicated by the high proportion of seedlings compared to older plants (Fig. 2). In both 2005 and 2013, we found higher densities of older P. serotina in plots invaded earlier, where biomass of P. serotina was the highest. In both 2005 and 2013 the highest total biomass was in plots with P. sylvestris. Within these plots the presence of older trees, with higher height and diameter increments, could have decreased light availability, causing decreased seedling survival (Knight et al. 2008; Dyderski and Jagodziński under review) or lengthened persistence in a “seedling” stage (Mulligan and Munro 1981; Closset-Kopp et al. 2007). Beneath P. sylvestris, most P. serotina biomass was concentrated in old, often fruiting individuals, responsible for high densities of seedlings, connected with high propagule pressure. However, due to poor seedling recruitment to the shrub layer, the density of older plants was low, similar to the mechanism described by Godefroid et al. (2005). A similar pattern was observed in the plots with B. pendula, L. decidua, and P. nigra. Our study also suggests that disturbance, which led to decreases of tree stand biomass and canopy closure on some plots, acted as a pulse of increased light availability. This window of opportunity allowed rapid P. serotina invasion, consistent with the Fluctuating Resource Availability Theory of Invasibility (Davis et al. 2000). However, the relationship between biomass of P. serotina and light availability was strongly modified by propagule pressure (Fig. 3). The compound effect of propagule pressure and invasibility is known from earlier conceptual studies (Davies et al. 2005). Our other field study (Dyderski and Jagodziński 2018) also confirmed the superior role of propagule pressure over invasibility.

The species-dependent pattern of invasion success resulted from different life history traits of particular species. One of them is growth dynamics, shown by differences in biomass and biomass increment of overstory trees. At the same age, some of them may exhibit different growth stages, due to different costs of growth, connected with specific stem density (Díaz et al. 2016). Moreover, tree species create different light availability due to different branching architecture and foliage morphology (Breuer et al. 2003; Knight et al. 2008; Pretzsch 2009). Earlier invasion was found in the stands of pioneer tree species, where canopy closure and reopening due to self-thinning occurred earlier (Szymański 1982; Jagodziński and Kałucka 2008; Pretzsch 2009). This pattern is consistent with observations that forests dominated by pioneer species are more invasible (Zerbe and Wirth 2006) or that late-successional tree species limit P. serotina invasion (Godefroid et al. 2005; Chabrerie et al. 2008; Jagodziński et al. 2015). Starfinger et al. (2003) claim that presence of dense thickets of P. serotina in P. sylvestris forests results from planting it as a soil improver in these poor habitats. In our study P. serotina spontaneously colonized experimental plots, without being intentionally planted. This indicates that although introductions increase initial propagule pressure, intentional introduction into poor habitat was not crucial for invasion success of P. serotina in the study site.

Higher total biomass and increment of P. serotina in stands of coniferous species may also result from habitat modification by these species, i.e. higher nutrient leaching and acidification (Binkley and Valentine 1991). This effect of tree species is independent of modifying light availability. In our plots coniferous species had lower soil and litter pH, lower calcium content, and higher acidity, which resulted from different elemental contents of foliage (Reich et al. 2005; Mueller et al. 2012). These factors decrease abundance of understory vegetation (Chmura 2013), which competes with naturally established regeneration of invasive species (Knight et al. 2008; Jagodziński et al. 2018). Thus, coniferous tree stands, especially when planted on fertile sites, are more invasible due to negative tree species effects on understory vegetation cover (Zerbe and Wirth 2006; Jagodziński et al. 2015).

Studies on populations of P. serotina reveal that the proportion of seedlings and specimens in the herb layer (< 50 cm height) is very high. In Rogów Arboretum, within a sample of 20,843 specimens, 76.7% of them were < 50 cm height (Jagodziński et al. 2015). In Compiègne forest, seedlings comprised 72.1% of a sample of 4994 specimens (Closset-Kopp et al. 2007). As presence of P. serotina in the shrub layer limits development of the herb layer (Godefroid et al. 2005; Halarewicz and Żołnierz 2014, but see Chabrerie et al. 2008), it also limits the growth of its own offspring. Therefore, we may observe either low or high densities of young P. serotina under tree stands where the species fruits. However, in both scenarios P. serotina biomass is low. The high density of seedlings and low biomass of P. serotina is connected with shade intolerance of older P. serotina specimens (Burns and Honkala 1990). Therefore, mere consideration of density does not reflect invasion dynamics unless seedlings and older trees are considered separately. In contrast, biomass is a function of both density and dimensions of plants, and thus, it increases with increase of both parameters. For that reason, biomass assessment may yield a more accurate understanding of invasion dynamics than density alone.

Our study confirmed the sit-and-wait strategy of P. serotina (Closset-Kopp et al. 2007) over a longer time interval by showing the potential for quick increase of biomass after rapid increase of light availability caused by the windthrow. Persistence of the seedling bank and waiting for the window of opportunity allows this species to colonize forest ecosystems which are usually less invasible due to the low frequency of disturbances (Herben et al. 2016). Our study confirms observations of Chabrerie et al. (2008), who found P. serotina as a passenger rather than a driver of disturbances. A similar pattern of emergence after dieback of the overstory trees was observed for Sorbus aucuparia in mountain P. abies forests (Żywiec et al. 2013). P. serotina in its native range is a species able to dominate secondary succession after disturbances caused by windthrow, logging, or fire (Mulligan and Munro 1981).

Similar to Vanhellemont et al. (2009), we found that colonization of uninvaded areas is a combined effect of long-distance dispersal events in combination with windows of opportunity. However, our study had more emphasis on the role of propagule pressure, which is expressed by distance to the seed source and number of fruiting trees. The roles of the distance from the propagule source and invader density, modified by type of tree stand, was confirmed in our previous study (Jagodziński et al. 2015). Long-distance dispersal, mostly bird-mediated (Deckers et al. 2008; Dylewski et al. 2017), comprises only 20% of seeds and the remaining 80% fall in the vicinity of parent trees (Pairon et al. 2006b). This high input of seeds in the invaded sites is responsible for persistence of the seedling bank every year. The subset of bird-mediated seeds is responsible for colonizing new sites, and in the broader perspective it is responsible for expansion of the invaded range. Propagule pressure is one of the most relevant factors driving invasion potential of alien woody species in Europe (Pyšek et al. 2009), and similar to Vanhellemont et al. (2009), we found that low levels of it limit invasiveness of P. serotina, even in invasible plots (Fig. 3).

Our study, for the first time, quantified the range of an invader’s intensity, expressed as biomass increment, in a forest common garden experiment with 14 tree species. Relative biomass increments of P. serotina ranged from 0 to 22,000-fold in eight years. This highlights the urgent need for monitoring even small populations of P. serotina, as this species has the ability for sudden outbreaks. Noticing even small but stable populations of P. serotina that have not reached the dimensions allowing reproduction will provide time for local eradication, which can lower the high cost of P. serotina control (Reinhardt et al. 2003).

Our results also recommend prioritized risk assessment for P. serotina, as was stated by Carboneras et al. (2018). Success of P. serotina varied among different tree species. Total biomass depended on light and propagule availability, while biomass increment depended on disturbances, described by changes in tree stand biomass. Invasion intensity was driven by disturbance, i.e. dieback of trees, especially P. abies, caused increases in light availability, which increased invasibility of these stands. In conjunction with propagule pressure, disturbance allowed growth and survival of numerous P. serotina seedlings. Our findings suggest that propagule pressure and disturbance intensity, expressed in biomass change, should be considered together to understand the susceptibility of forest stands to invasion and to develop methods of control for invasive trees and shrubs (Brundu and Richardson 2016). Even modest disturbances, including vulnerability to dieback in succession of coniferous species, may be risky in areas with high propagule pressure from invasive trees. Thus, silvicultural methods for P. serotina control may include maintaining high canopy closure and supporting natural regeneration of tree species with high leaf area, which shade the understory. This result is consistent with recommendations of Nyssen et al. (2016), who highlighted the importance of life history traits of the invaded species habitats in shaping invasibility.

This study was partially supported by the Institute of Dendrology, Polish Academy of Sciences, Kórnik, Poland, and by the General Directorate of State Forests, Warsaw, Poland (research project: ‘Environmental and genetic factors affecting productivity of forest ecosystems on forest and post-industrial habitats’). We are grateful to Dr Lee E. Frelich (Department of Forest Resources, University of Minnesota, USA) for valuable suggestions and linguistic revision of the manuscript. We are also thankful to Dr Uwe Starfinger (Julius Kühn-Institut, Quedlinburg, Germany), Prof. Aníbal Pauchard, Prof. Dr Ingolf Kühn, and the anonymous reviewer for thorough and valuable comments on an earlier draft of the manuscript.