We conducted the study in managed forests in western Poland, in five forest districts: Babki, Czerniejewo, Jarocin, Konstantynowo, and Łopuchówko (Fig. 2). We located study plots between 51°59'4.08"N and 52°40'9.36"N and 16°35'28.98"E and 17°37'13.26"E, in two geographical regions: the Greater Poland Lakeland (northern part) and Greater Poland Lowland (southern part). The climatic conditions are similar in the study area with an annual temperature of 8.5 °C and mean annual precipitation of 500–550 mm (BDL 2024).

We aimed to sample a quantitative gradient of invader biomass. To obtain a range, we selected study plots based on invader cover, which is straightforward to estimate, and then, after plots were chosen, we quantified aboveground biomass, following established methods (Bury and Dyderski 2024b), and described in more detail below. During initial plot selection we search for control plots (zero individuals of studied invaders ≥ 1.3 m height), medium (< 30% cover), and high (> 50%) cover areas. Therefore, in our plots there could have been P. serotina or R. pseudoacacia individuals shorter than 1.3 m (included in the natural regeneration survey), however, they were rare as the density of studied neophytes depends on the proximity of propagule sources (Dyderski and Jagodziński 2018). When calculating the gradient of invader biomass, we accounted only for individuals taller than 1.3 m, and the biomass of those few individuals in the regeneration layer in control plots was negligible. We stratified our samples into two habitat types: nutrient-rich habitats that are typical of the invasive species in their native range, and nutrient-poor, where invaders had been massively introduced to improve these habitats (Starfinger et al. 2003; Cierjacks et al. 2013). Nutrient-poor sites included Leucobryo-Pinetum W. Mat. (1962) 1973 communities or secondary P. sylvestris forests. In our study, nutrient-rich sites include different subtypes of Galio sylvatici-Carpinetum betuli Oberd. 1957 communities or secondary Quercus spp. forests. Some areas had characteristics of poorer communities or slightly more fertile ones, with species characteristic of Potentillo albae-Quercetum Libb. 1933 or Querco-roboris Pinetum Mat. et Polak. 1955 s.l. We also included two management contexts: stands in the middle of rotation age (medium age) and close to rotation age (mature age), as these age classes differ in light conditions beneath stand canopies. Stands in the middle age and those of close to rotation age differ in structure and growth dynamics (Jiang et al. 2017; Li et al. 2024). Middle-aged stands are characterized by a rapid increase in biomass, while stands close to rotation age are characterized by maximum biomass, but its increment decreases with age (Jiang et al. 2017; Li et al. 2024). Stand age is related to light availability by stem density and canopy closure, as well as in terms of higher species richness of forest specialists, related to a longer time since disturbance (Jagodziński and Oleksyn 2009; Felton et al. 2010; Conradi et al. 2020). In total, we established 160 plots (500 m² per plot), including 32 control plots (8 replications × 2 habitat types × 2 stand age classes), 64 plots with R. pseudoacacia (8 replications × 2 invasion levels × 2 habitat types × 2 stand age classes) and 64 plots with P. serotina (same as R. pseudoacacia) (Fig. 1). Plots representing the same plot variant (invader × invasion level × habitat × stand age) were a minimum of 5 km apart to reduce spatial autocorrelation.

Figure 1.

Scheme of the study design. Photos: S. Bury.

Figure 2.

Distribution of the study plots (n = 160). The background map of forest cover comes from the Forest Data Bank (BDL 2024).

We estimated invader biomass of 102 plots in the autumn of 2021 and 2022, measuring the diameter at breast height (DBH) of all the individuals in the plots following García et al. (2023). The other 58 plots were sampled in autumn 2022 and 2023. For these, we measured the diameter at the breast height only on trees larger than 5 cm, and we counted trees thinner than 5 cm by species. Then, from the database of the 102 plots, we calculated the average DBH of individuals thinner than 5 cm by species (Suppl. material 1: table S1 for mean and SD values). This approach should not affect the validity of the results, and indeed is more detailed than previous studies that either omitted trees with DBH < 5 cm or used a DBH midpoint for smaller trees (Dyderski and Jagodziński 2021a). Then, we used published allometric formulas (Suppl. material 1: tables S2–S3) to calculate the aboveground biomass for individual trees and stands (Brown 1976; Alberti et al. 2005; Forrester et al. 2017; Zasada 2017; Jagodziński et al. 2018, 2019).

In the summers of 2021, 2022, and 2023 we counted natural regeneration on four schematically distributed subplots with a 3 m radius (4 × 28.26 m² = 113.04 m²). The centers of the subplots were systematically set at 4.21 m (1/3 of the main plot radius) from the center of the plots in the four cardinal directions (N, E, S, W), using a compass and measuring tape (Fig. 3). Within these subplots we identified and counted all individuals of trees and shrubs < 1.3 m height, similarly to Kerr and Mackintosh (2012) and Mousavi et al. (2012). For each plot, we identified all seedlings germinated in the study year by species, as well as all saplings up to 1.3 m in height. Saplings may have been the product of prior years’ seedlings, or clonal propagation. We treated all saplings growing separately from the soil as single individuals (Radtke et al. 2013).

Figure 3.

Schematic of the distribution of the subplots (four dashed line circles, counts of natural regeneration) within each plot (solid line circle, stand structure measurements). Plot area = 500 m² (r = 12.62 m), subplots area = 4 subplots × 28.26 m² = 113.04 m² (r = 3 m).

All analyses were conducted in R (R Core Team 2023). Because one of the plots with P. serotina in the stand (P. serotina aboveground biomass = 47.11 Mg ha^-1; Table 1) gave a strongly biased result, we excluded it from all analyses. The aboveground biomass of P. serotina trees in this area was about three times higher than the next highest result. Due to differences in species pools in each habitat type, we separately analyzed the relationships with the natural regeneration for both studied invaders on both habitats. We accounted for different stand development phases, including stand age as a covariate in analyses.

We used Canonical Correspondence Analysis (CCA) to compare the effect of invader biomass and stand age. We used the cca() function from the vegan package (Oksanen et al. 2018) to develop CCA. Due to differences in sapling density among species and plots, we log-transformed data using the decostand() function from the vegan package (Oksanen et al. 2018). Furthermore, we added invader aboveground biomass and stand age as constraints. We used the step.cca() function from the vegan package to choose the optimal set of predictors based on Akaike’s Information Criterion. To assess the significance of constraints, we ran a permutation-based ANOVA-like test, implemented in the anova.cca() function. We visualized the results using the ggplot2 (Wickham 2016) and ggrepel (Slowikowski 2024) packages.

Furthermore, we used Threshold Indicator Taxa Analysis, implemented in the TITAN2 package (Baker et al. 2023) to check which sapling species increase or decrease with invader biomass. We log-transformed biomass using log(x+1) transformation, to include also plots with zero invader biomass. For species with purity and reliability ≥ 0.95 we visualized results using the plot_taxa_ridges() function with default settings. We conducted the analysis using only species that occurred in at least three plots. We did not use any additional filters in the settings. The plot_taxa_ridges() function generates graphs on which the x-axis represents an environmental factor. In our case, this is the aboveground biomass of an invasive species (log(x+1) transformation was used). Changes in the abundance of individual taxa were assessed along the gradient of the environmental variable. In our case, this was the number of saplings of individual species on the plot. Ridges are generated for species that have achieved purity and reliability of 95%. These ridges look different for each species. The shape of the ridge tells us where in the environmental gradient a given species achieves the highest probability of occurrence. The top of the ridge indicates the value of the environmental indicator for which the abundance of the species is the highest. The greater the width of the ridge, the greater the discrepancies in the data. Taxa are divided into two groups. Species whose abundance increases along the environmental gradient are marked in red (increasers), while species whose abundance decreases along the environmental gradient are marked in gray/blue (they are called decliners). The z-score value indicates the strength of the impact of a given factor or, in other words, the higher the z-score, the higher the indicator value of a given species. The higher the z-score, the darker the red or gray color (for low values it is light blue). The function also generates black vertical lines on the graph. These are the so-called threshold values, which tell us where on the gradient there is a sharp change (an increase in increasers or a decrease in decliners) in the abundance of a given species.

Finally, we used Generalized Linear Mixed-Effect Models (GLMMs), using the glmmTMB package (Brooks et al. 2017), with Poisson or negative binomial family distribution, to exactly determine the relationships between the abundance of seedlings and saplings of each species with invader aboveground biomass. To test the invasional meltdown hypothesis, we also evaluated the relationships between the abundance of saplings of invasive species (excluding the dominant invader whose effects we were exploring) and invader aboveground biomass. For saplings we created models for species that occurred in at least 20% of the plots in a given variant (invader and habitat type) (Suppl. material 1: table S4). For seedlings we developed models for species that occurred in at least 10% of the plots in a given variant (invader and habitat type) (Suppl. material 1: table S13). In the Results section we present only statistically significant results (with p < 0.05). We used the DHARMa package (Hartig 2022) to conduct formal zero inflation and dispersion tests for each model. We started from models assuming Poisson distributions, due to the count character of our data. If we did not find problems with overdispersion we tested zero inflation, and in the case of statistically significant zero inflation, we used zero-inflated Poisson distribution. If we found statistically significant overdispersion we used negative binomial distribution, adding zero-inflation when necessary. We used invader aboveground biomass and stand age as fixed continuous effects and forest district and the year of natural regeneration assessment as random intercept, to cover spatial and temporal dependence within our dataset. We used the dredge() function of the MuMIn package (Bartoń 2017) to choose the best model according to comparing Akaike’s Information Criterion of null model AICc₀ with the final model AICc. We presented the results using marginal responses implemented in the ggpredict() function from the ggeffects package (Lüdecke 2018). These responses show mean model prediction for each level of predictor, assuming remaining predictors at a constant (mean) level, and excluding random effects (prediction for global population). We excluded two outlying observations in the model of Cerasus avium saplings density for R. pseudoacacia in rich sites (densities: 304 and 37 ind.) and one in the model of Carpinus betulus saplings density for P. serotina rich sites (density: 1206 ind.). In these plots very high regeneration density resulted from an abundance of propagule pressure in proximity and did not allow for developing models reflecting overall conditions. We used the ggplot2 package (Wickham 2016) to present results on the graphs. In the results, we provide extreme values of sapling density for some species, i.e. zero and a value close to the maximum of the gradient for R. pseudoacacia and P. serotina in individual habitats. All mean values are followed by ± SD, except ± SE in the results of GLMMs.

Species composition depended on invader biomass, both for stands with P. serotina and R. pseudoacacia, and both on the nutrient-rich and nutrient-poor sites. The stand age was statistically significant for R. pseudoacacia on nutrient-poor (p = 0.004 Fig. 5, Table 2) and nutrient-rich sites (p = 0.034). Prunus serotina biomass was positively correlated with the frequency of Sorbus aucuparia, P. serotina, Fagus sylvatica, and Q. robur on nutrient-poor sites, and A. platanoides, Fraxinus excelsior, Q. robur, Prunus cerasifera, Prunus padus, P. serotina, F. alnus, U. minor and Sambucus nigra on nutrient-rich sites. Robinia pseudoacacia biomass was positively correlated with S. aucuparia, R. pseudoacacia, and S. nigra on nutrient-poor sites, and R. pseudoacacia, A. platanoides, Acer campestre, Acer pseudoplatanus, U. minor, F. alnus, P. serotina, P. cerasifera, C. avellana, Euonymus europaeus, and S. nigra on nutrient-rich sites (Fig. 5).

Figure 5.

Canonical Correspondence Analysis (CCA) for a nutrient-poor sites with P. serotina (n = 47 plots) b nutrient-rich sites with P. serotina (n = 48 plots) c nutrient-poor sites with R. pseudoacacia (n = 48 plots) d nutrient-rich sites with R. pseudoacacia (n = 48 plots). Species with a frequency > 20% are labeled. Green arrows and green labels represent environmental variables. Red dots = control plots, light blue dots = plots with P. serotina or R. pseudoacacia. Abbreviations: log1p(Biomass) — natural logarithm of invader aboveground biomass.

For P. serotina, we observed similar trends on both nutrient-poor and nutrient-rich sites (Fig. 6a, b, Suppl. material 1: tables S5, S6). The analysis revealed that Q. petraea saplings density declined with increasing P. serotina biomass and the opposite trend for P. serotina saplings. In the stands with R. pseudoacacia, more species revealed any response (Fig. 6c, d, Suppl. material 1: tables S7, S8). The decliners were Pinus sylvestris, Q. petraea, and B. pendula on nutrient-poor sites and Q. petraea on nutrient-rich sites. On the nutrient-poor sites S. nigra, P. padus, A. platanoides, and R. pseudoacacia increased their saplings density with increasing R. pseudoacacia biomass. On the nutrient-rich sites A. platanoides, Q. robur, and R. pseudoacacia increased their saplings density with increasing R. pseudoacacia biomass (Fig. 6c, d, Suppl. material 1: tables S7, S8).

Figure 6.

Results of Threshold Indicator Taxa Analysis (see Methods section for interpretation of the graph) for a nutrient-poor sites with P. serotina (n = 47 plots) b nutrient-rich sites with P. serotina (n = 48 plots) c nutrient-poor sites with R. pseudoacacia (n = 48 plots) d nutrient-rich sites with R. pseudoacacia (n = 48 plots). Grey/blue density estimators represent species responding negatively to invader biomass gradient (decliners) while red color – positively (increasers). We included here only responses for species that were both reliable (reliability ≥ 0.95) and pure (purity ≥ 0.95). For statistics of all species see Suppl. material 1: tables S5–S8.

We used three different types of analyses, and in almost all cases we reached consistent results (Table 7). Among three tested methods we found the most consistent relationships for Q. petraea in all variants and P. serotina in plots with P. serotina. According to the Threshold Indicator Taxa Analysis, on the nutrient-poor sites with P. serotina, B. pendula were correlated negatively with the invader biomass, but according to the GLMMs positively. We found a contrast pattern in the case of R. pseudoacacia regeneration in stands with R. pseudoacacia, as TITAN2 suggested a positive relationship, but the model suggested a negative (but only for the zero-inflation component).

Observational studies on the impact of invasive species on various ecosystems, including forests, should not be considered as a simple causation based on observed correlations. Ecosystems are very complex and each of their elements is simultaneously affected by various factors. Impact assessment should be multidimensional and a systemic approach. In our plots we observed different densities of saplings and seedlings of individual species. We refer to individual hypotheses in the following sections of the discussion. The relationship between natural regeneration density and biomass of P. serotina and R. pseudoacacia can be both positive or negative, and this is in line with recent studies showing that results depend on the environmental context (Sapsford et al. 2020; Catford et al. 2022) and the reference ecosystem used for comparison (Sádlo et al. 2017; Medvecká et al. 2018; Dyderski and Jagodziński 2021b). Our study improves the knowledge about the relationship between different invasive tree biomasses and ecosystem services.

We found different relationships between particular species natural regeneration densities and R. pseudoacacia and P. serotina biomasses (H1, H2). We confirm both the first (H1) and second (H2) hypotheses. The biomass of R. pseudoacacia was correlated with the density of natural regeneration more than P. serotina (H1). We also confirm the second hypothesis, as individual natural regeneration species showed different patterns of density. Some showed a decrease in density with the biomass increase of invaders, others showed opposite trends. Differences between individual species were seen in the number of individuals in each quantity of the invasive species, the shape (more linear or exponential), and the slope of the curves in the models. Similarly, in TITAN2, we found differences in the number of species and shapes of ridges (H2). Some natural regeneration species revealed positive relationships with invader biomass (mostly P. serotina in the stand with P. serotina) and some negative (Q. petraea, P. sylvestris). Those differences between particular species relationship with R. pseudoacacia or P. serotina should be mostly connected with different light or nutrient requirements of particular sapling and seedling species. As the transformations of both studied neophytes changed along their biomasses, also the density of particular natural regeneration species should change more intensively. There are visible trends connected with the ecological niches of particular trees, but they should be interpreted with caution. More shade-tolerant and nitrophilous species increased their abundances with invader biomass increasing, e.g., F. excelsior or Acer spp. In contrast, light-demanding and acidophilous P. sylvestris decreased its abundance or the abundance remained unchanged. Increasing biomass of either P. serotina or R. pseudoacacia led to reduced light availability on the forest floor and higher nutrient content in the soil (Rice et al. 2004; Dyderski and Jagodziński 2019; Engel et al. 2024). Those transformations did not support P. sylvestris regarding natural regeneration growth, in both saplings and seedlings. Lázaro-Lobo et al. (2021) also mentioned that the response of a particular species’ natural regeneration depends on their functional traits, and the competition between invasive tree species with desirable species depends on their niche spacing. There are also visible differences in the number of significant relationships between sapling species densities and R. pseudoacacia or P. serotina biomasses (H1). Focusing only on GLMMs (Table 7), in nutrient-rich habitats 17 species (13 positively and four negatively) responded significantly to R. pseudoacacia increasing biomass and six (four positively and two negatively) to P. serotina increasing biomass. In nutrient-poor habitats, three species (one positively and two negatively) responded to R. pseudoacacia increasing biomass and six (three positively and three negatively) to P. serotina increasing biomass. According to GLMMs in nutrient-rich habitats (R. pseudoacacia: 76%, P. serotina: 67%), the share of the number of species reacting positively to the invader’s biomass in the number of species that responded significantly positively and negatively was higher than in coniferous habitats nutrient-poor habitats (R. pseudoacacia: 33%, P. serotina: 50%; H3).

For some species, we observed some trends similar to those observed by Dyderski and Jagodziński (2020), conducted in protected forests of the nearby Wielkopolska National Park. The density of forest tree species was lower in invaded stands by both R. pseudoacacia and P. serotina than in non-invaded while shrubs and admixed trees increased their density. Our study provided a significant advance from this study, as we included invader abundance, that allows for assessment of various stages of invasion (López-Núñez et al. 2017).

Our observations regarding the negative correlation between invasive trees biomass and the natural regeneration of forest-forming species are in line with the findings of Terwei et al. (2013) and Kowarik et al. (2019). Terwei et al. (2013) showed that in the hardwood floodplain forests, R. pseudoacacia in the stand was positively correlated with the density of R. pseudoacacia seedlings. However, P. serotina in the stands was positively correlated with P. serotina seedlings but negatively with U. minor seedlings. Langmaier and Lapin (2020), in their review of the impact of different invasive plant species on forest regeneration, also discussed the impact of P. serotina and R. pseudoacacia. Based on the works of other authors (e.g., Rahmonov 2009; Maringer et al. 2012; Petrášová et al. 2013; Radtke et al. 2013; Terwei et al. 2013), Langmaier and Lapin (2020) synthesized the negative impact of R. pseudoacacia on species e.g., Q. petraea, Q. robur, P. sylvestris, U. minor. In our studies, we confirmed a negative correlation of Q. petraea and P. sylvestris regeneration density with invader biomass. For Q. robur and U. minor, we obtained less obvious positive responses. Quercus robur was less frequent than Q. petraea in our plots, while for U. minor we obtained significant results only for nutrient-rich sites. However, it should be borne in mind that their work accounted for habitats not only from Central Europe, but also other ecoregions, e.g., Western European deciduous forests, Pannonian mixed forests, or riparian forests.

When invasive species arrive in new niches, they can change soil chemicals, transform light conditions, and make the ecosystem more suitable for the other alien species (Crooks 2002; Corenblit et al. 2014; Jagodziński et al. 2024). In the longer term, this may increase the negative impact on biodiversity and other ecosystem services. For the total alien species natural regeneration density, we confirm the invasional meltdown hypothesis (H4) (Simberloff and Holle 1999) only for R. pseudoacacia, both on the nutrient-poor and nutrient-rich sites. Since we examined the relationship between R. pseudoacacia and P. serotina and the total density of all alien species in the regeneration layer, it is impossible to find species-specific patterns related to their biology and ecology. However, for some individual species we obtained significant results. Prunus serotina regeneration density was higher in plots with higher R. pseudoacacia biomass on nutrient-rich sites. However, little is known about the possible interactions between the adult R. pseudoacacia and P. serotina natural regeneration and vice versa. The interesting fact was that invasive P. cerasifera abundances decreased with an increase in R. pseudoacacia aboveground biomass. Due to the similar biology and ecology of this species to P. serotina we would expect rather similar responses. Czortek et al. (2024) proved that the presence of P. cerasifera natural regeneration is favored by higher light availability. However, this species avoids places with a higher number of functional types, which may indicate its lower resistance to competition than P. serotina.

According to the propagule pressure hypothesis, the higher the propagule pressure, the more effective colonization (Lonsdale 1999; Lockwood et al. 2005; Blackburn et al. 2011). This dependence of studied species on propagule pressure was confirmed in numerous previous studies (Vanhellemont et al. 2009; Vítková et al. 2017; Dyderski and Jagodziński 2018). Trees acquire reproductive abilities late. However, they retain these abilities for a very long time. More propagule sources should increase the regeneration capacity. We confirmed this hypothesis for P. serotina on both nutrient-rich and nutrient-poor sites, but the predicted values of sapling and seedling densities were higher on nutrient-poor sites. For R. pseudoacacia, we confirmed this hypothesis for saplings on nutrient-rich sites and seedlings on nutrient-poor sites, but the predicted densities were low. In general, in our plots R. pseudoacacia did not regenerate as effectively as P. serotina. Robinia pseudoacacia spreads generatively mostly by wind and therefore prefers open spaces. Prunus serotina seeds mostly fall directly into the soil and to a lesser degree are spread by birds (Deckers et al. 2008; Vanhellemont et al. 2009; Dylewski et al. 2017). Also, both P. serotina (Starfinger et al. 2003) and R. pseudoacacia (Bouteiller et al. 2023) easily regenerate vegetatively by root suckers.

Biological invasion dynamics can depend on several factors: environmental conditions, interactions between species, anthropogenic factors, and management (González-Moreno et al. 2014; Sapsford et al. 2020; Catford et al. 2022). To exclude the management context, we established study plots in forest patches without visible impacts of silvicultural treatments e.g., planting or removing trees from plots. We also placed them in similar climatic conditions. We expected a more distinct decline in the density of natural regeneration to invader quantity/biomass on nutrient-poor than on nutrient-rich sites (H3) (Chmura 2004; Halarewicz 2011). In general, in nutrient-rich sites, there are higher densities of native trees understory, so the competition with invasive species is stronger. This is indicated by the number of species that reached a frequency > 20%. Comparing the abundances, P. serotina was more successful on nutrient-poor than nutrient-rich sites. Results for P. serotina are connected with the Empty Niche Hypothesis (Elton 1958; Schmitt 2020). This hypothesis suggests that invasive species can successfully settle new ecosystems by occupying weakly filled or unfilled ecological niches, where native species are less common or absent. In Central Europe, nutrient-poor sites with P. sylvestris have lower richness and abundance of native species compared to nutrient-rich habitats. Prunus serotina was massively introduced on nutrient-poor sites by foresters (Starfinger et al. 2003; Engel et al. 2024; Nyssen et al. 2024). Small competition from native trees facilitated the spread of P. serotina to new stands and led to their dominance in these habitats. This species further spreads easily to nearby stands. Especially, on nutrient-rich sites with R. pseudoacacia, we can observe that many of the native tree species regeneration increased their abundance with invader biomass increasing. More shade-tolerant species showed a positive correlation with invader biomass, except for C. betulus on nutrient-rich sites with P. serotina. This finding contradicts a previous study (Dyderski and Jagodziński 2020) revealing a positive response of this species to P. serotina presence. However, in the cited study, this response regarded P. sylvestris plantations on nutrient-rich sites, thus it cannot be directly compared. Our results balance between supporting the biotic acceptance (Stohlgren et al. 2006) and biotic resistance (Elton 1958; Levine et al. 2003) hypotheses. In the case of stands with P. serotina, both in poor and fertile habitats, the density of its regeneration increased with the biomass of the parental trees, mainly due to the availability of propagules and dispersal mechanisms: barochory and zoochory. Nevertheless, the effect sizes were higher on the poor sites. In the case of R. pseudoacacia stands, we observed slightly better regeneration of this species in nutrient-rich habitats. To sum up, for P. serotina our results are rather in line with the biotic resistance hypothesis (Elton 1958; Levine et al. 2003), while for R. pseudoacacia they rather in line with biotic acceptance (Stohlgren et al. 2006). The more visible relationships between natural regeneration and invader biomass in nutrient-poor habitats depend on specific ecological conditions, specific plant species composition, and soil fertility. Therefore, any transformation by an invasive species is more severe for the species occurring there.

We had to adapt the database to the analysis guidelines. CCA was the least conservative analysis in the case of input data. CCA is also the least sensitive on extremal observations. In models, we excluded one plot for C. betulus saplings and two plots for C. avium saplings (see the rationale in the Materials and Methods section). Thanks to the use of TITAN2, we were able to detect the threshold for particular invasion levels e.g., P. serotina saplings reacted quickly with big abundance on even small quantity of P. serotina in the stands, while Q. petraea as a decliner was more tolerant to P. serotina biomass increasing. For C. avium omitting these records in the models did not change the trend (positive/negative) but reduced the standard error and smoothed the regression curve. In the case of C. betulus, removing the extreme observation changed the trend from positive to negative. The negative trend is consistent with the CCA result. The extreme observation results from the fact that there were adult C. betulus in the vicinity of the plot, acting as a propagule source. According to the guidelines of statistical model development, we should remove this outlier. The model after removing the outlier had a more stable distribution of residuals and a lower standard error of estimates. Even though each of the analyses we use is based on slightly different data structures and responds differently to data variability, the results we obtain are very similar. In general, consistent responses revealed by three different methods suggest that all these tools are useful in the assessment of correlations with invasive species biomass. We also found that TITAN2 resulted in the most conservative approach – for P. serotina and R. pseudoacacia on nutrient-rich sites it revealed relationships only in the cases confirmed by two other methods. For R. pseudoacacia on nutrient-poor sites, it revealed relationships not confirmed by two other methods only for three species.

In the context of current trends in forestry, P. sylvestris is still the main species in nutrient-poor sites areas, while Quercus spp. is in nutrient-rich sites. Therefore, referring to habitats studied here, densities of main forest-forming species were negatively correlated with the biomass of studied invasive species, especially Q. petraea. Prunus serotina also hindered the regeneration of P. sylvestris in the poor sites. We found an increasing density of Quercus robur saplings with increasing R. pseudoacacia biomass in fertile habitats, but negatively correlated with P. serotina biomass in poor habitats. In the context of natural forests and ongoing climate change, the situation looks a bit different. Wide-scale studies predict the retreat of forest-forming tree species from Central Europe, especially P. sylvestris, as a response to climate change (Dyderski et al. 2018; Chakraborty et al. 2021; Wessely et al. 2024). That way, studied neophytes can enhance this negative effect by suppressing the natural regeneration of studied species. Recent management strategies propose in some cases the assimilation of invasive species with native ecosystems (Nyssen et al. 2024), also for P. serotina (Nyssen and Vanhellemont 2016; Engel et al. 2024) and R. pseudoacacia (Sádlo et al. 2017). Such a strategy is recommended especially in fertile habitats that are biotically more resistant to the development of invasion. In assimilating studied invasive species, managers should be hypersensitive to their potential impacts on main forest-forming species. When we want to maintain or increase the number of species such as Q. petraea in the regeneration, we should take into account the results of our research and apply methods that will facilitate their survival. Management in fertile habitats should be adapted to a long-term management plan. If we want to mimic natural processes, the presence of R. pseudoacacia (based on our research) may be helpful, as it promotes species such as F. excelsior, U. minor, or Acer spp. On the other hand, if we want to preserve as large a Quercus population as possible, some human action may be necessary. In the case of P. sylvestris stands, we maintain the fact that if we want successful P. sylvestris regeneration, it is necessary to take into account the observed decreasing density of P. sylvestris regeneration with both invasive species studied biomass increasing and support the regeneration of P. sylvestris. Langmaier and Lapin (2020) summarized that there are studies that indicate that in the case of R. pseudoacacia, its eradication measures or adaptation of silvicultural measures are the most frequent management actions, while in the case of P. serotina – early detection. The latter concerns both decision-makers and the entire society, because it is easier to control biological invasions in the early stages. In turn, silvicultural treatments can be modified to promote selective cutting, and appropriately manage the closure of tree crowns and density with local reduction of invasive species combined with the promotion of species of native origin. Unfortunately, our study revealed the negative correlations between studied invasive species biomass and the Q. petraea natural regeneration in poor sites. Due to changing climatic conditions leading to the retreat of coniferous species, many see the potential of Q. petraea to replace P. sylvestris (Hanewinkel et al. 2013; Dyderski et al. 2025). Since R. pseudoacacia and P. serotina are very common in European forests (Wagner et al. 2017; Campagnaro et al. 2018), and are predicted to expand their range under the changing climate (Puchałka et al. 2021, 2023), we may expect the negative impact on Q. petraea regeneration, that will require particular attention.

Although our study focused on managed forests, certain relationships can be related to natural forests. The areas we searched had the structure of semi-natural forests, managed in a way that imitated natural processes. In the case of protected forests, it is important to monitor the presence and impact of invasive species on natural processes and prevent possible damage they may cause. Eradication of invasive trees is expensive and sometimes counter-productive or even makes the situation worse (Namura-Ochalska and Borowa 2015; Nyssen and Vanhellemont 2016; Nyssen et al. 2024). Our study should be helpful for stakeholders in making decisions about the assimilation or eradication of invasive trees in particular types of stands on particular habitats (Nyssen and Vanhellemont 2016; Sádlo et al. 2017).

The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

No ethical statement was reported.

The study was financed by the National Science Centre, Poland, under project no. 2019/35/B/NZ8/01381 entitled: ‘Impact of invasive tree species on ecosystem services: plant biodiversity, carbon and nitrogen cycling and climate regulation’ and by the Institute of Dendrology, Polish Academy of Sciences.

Conceptualization: SB, MKD; methodology: SB, MKD; investigation: SB, MKD; formal analysis: SB; visualization: SB; writing—original draft preparation: SB; writing—review and editing: MKD; funding acquisition: MKD.

Sebastian Bury https://orcid.org/0009-0006-5380-0521

Marcin K. Dyderski https://orcid.org/0000-0003-4453-2781

All data supporting the results are archived in the figshare repository (Dyderski and Bury 2024) 10.6084/m9.figshare.26809084.