The study was conducted in Upper Silesia – the region has a long tradition of coal mining (since the 18^th century). The long-lasting black coal mining activities have resulted in large areas of post-coal mine sites, occupying > 2000 ha (Szczepańska 1987). These mineral material sites built of carboniferous sediments on Pre-Cambrian crystalline rocks have shaped the anthropogenic landscape. The carboniferous mudstone and sandstone complexes are mixed with numerous coal elements. These stone complexes are also overlain by Triassic carbonate formations (Cabała et al. 2004). Plant species colonisation and the development of vegetation communities on coal mine heaps is difficult because the mineral material habitats have extreme abiotic conditions, for example, large variations in daily temperatures (often reaching 50 °C) and humidity, substrate instability, lack of soil, susceptibility to erosion, dusting, thermal and chemical activities. In addition to abiotic parameters, the post-coal mine heap is characterised by extreme biotic conditions, such as soil organic matter deficiency in the substrate and lack of seed bank (Woźniak et al. 2021). These habitat characteristics impact the ability of diaspores to establish and the development of vegetation communities and mosaic of ecosystems on post-industrial sites (Bradshaw 2000; Prach et al. 2013), particularly on mineral oligotrophic coal mine heap sites (Woźniak 2010).

From the list of 112 post-mining sites with available information about age, size, vegetation and reclamation method (Woźniak 2010), we excluded 31 sites differing in size, land-use patterns in the neighbourhood, thermal activity or were artificially shaped. As these factors could significantly alter observed patterns of vegetation assembly, we decided to exclude them, focusing on the most frequent cases. Although this decreased the total variance of abiotic conditions, it allowed us to make conclusions about trends not affected by noise connected with the abovementioned treatments. From the remaining 81 sites, we randomly selected 60 sites proportionally to post-coal mine heap size, age and surrounding land cover. Amongst them, we distinguished five types of land cover and randomly selected plots proportionally to cover class (Jagodziński et al., in prep.). Using the results of this investigation, we randomly selected 80 vegetation patches, proportionally to the cover of each land-use class which forms an area of at least 150 × 150 m. Within each randomly selected patch, we established five plots in a cross design (i.e. one central plot and four subplots at distances of 50 m in the north, south, east and west directions; 400 plots in total; Fig. 1). Each plot was circular with a 3 m radius (28.3 m²). In all plots, we registered vascular plant species and their abundances using the Londo scale (Londo 1976). Alien species status (i.e. casual, naturalised and invasive) and historical-ecological groups (i.e. archaeophytes and neophytes) were determined using the database of alien plants in Poland (Tokarska-Guzik et al. 2012).

Figure 1.

Scheme of study design – distribution of study plots within land use types. Additional plots are in north, south, east and west directions at 50 m from the central plot.

These traits include a broad category of plant life history, leaf morphology and reproductive characteristics (Table 1). Traits data were acquired from LEDA (Kleyer et al. 2008), BIEN (Maitner et al. 2018), Pladias (Chytrý et al. 2021) and BioFlor (Klotz et al. 2002).

The functional approach was based on a set of traits known to have significant ecological implications for plant species competitive ability, dispersal, establishment and stress tolerance on post-coal mine heaps. Specific leaf area (SLA) and leaf dry-matter content (LDMC) serve as proxies of species status on the leaf economic spectrum (Perez-Harguindeguy et al. 2016). High LDMC and low SLA unveil a conservative approach with resistance to the harsh abiotic stress in the mineral material of post-coal mine heaps, while low LDMC and high SLA infer the increased importance of an acquisitive strategy by plant species (Perez-Harguindeguy et al. 2016). Plant height (PH) was used as an approximation of plant competitive competence (Westoby et al. 2002). The seed mass (SM) helps to explain the colonisation and establishment ability of plant species – with low seed mass for species found on post-coal mine heaps of younger age and vice versa (Piekarska-Stachowiak et al. 2014).

We used a random forest algorithm in combination with phylogenetic trait imputation to fill gaps in the trait data and not omit missing data (Penone et al. 2014). To strengthen the predictive power of the model, we used the missForest::misForest() function (Stekhoven 2022) and phylogenetic eigenvectors (Diniz‐Filho et al. 1998) derived from the PVR::PVPdecomp() function (Santos 2018). The variation explained by the first 15 phylogenetic eigenvectors was 59.3% of phylogenetic distances. The Normalised Root Mean Square Error (NRMSE) of imputed traits was 1.011 for continuous predictors and the proportion of falsely classified categorical variables was 0.079. In general, trait imputation has been shown to decrease bias when compared to removing species with missing trait data (Penone et al. 2014).

To understand important aspects of the functional community structure, we combined plant trait data with species cover to calculate the community-weighted means (CWMs) and the functional diversity indices. We calculated the CWMs of seed mass, plant height and specific leaf area using the FD::FunctComp() function (Laliberté et al. 2014). These traits influence plant germination and dispersal ability, life form and growth rate. We log-transformed numeric trait data to attain normality before the calculation of CWMs. Using the FD::dbFD() function (Laliberté et al. 2014), we quantified functional diversity indices: functional richness and functional dispersion. These indices show the distribution of plant species traits within the community hyperspace (Laliberté and Legendre 2010). Functional richness (FRich) quantifies the trait space of plant functional types present in a community. Communities with a low functional richness of native plants are expected to be more invasible by competitive alien species (Renault et al. 2022). This implies that niche differentiation within the native community will be low, thereby resulting in trait convergence and competition (Czortek et al. 2021). Functional dispersion (FDis) measures distances between functional traits carried by plant species to the centroid (centre point) in the community hypervolume (Villéger et al. 2008). High functional dispersion delineates strong functional differences between native species in a community – thus suggesting co-occurrence rather than competition (Carroll et al. 2011).

All analyses were performed in R software (version 4.2.1) (R Core Team 2022). Using the base::scale() function, we standardised and scaled explanatory variables before analyses. Such an approach helps to reduce biases linked with uneven ranges amongst these variables and it ensures that the estimated coefficients are all on the same scale, making it easier to compare the effect sizes.

To assess the drivers of alien species richness and cover in post-coal mine heaps, we built a generalised linear mixed effect model (GLMM) and linear mixed effect model (LMM), assuming a Poisson distribution with a log linking function and Gaussian distribution, respectively. In these models, heap age and native community characteristics (i.e. native species richness, native species cover, native CWM SLA, native CWM SM, native CWM PH, native FDis and native FRich) were predictors. In our models, blocks of plots nested within the heap are random variables to account for the spatial dependence of the study design. We used the ‘lme4’ package (Bates et al. 2015) to develop GLMM and LMM, and the ‘lmerTest’ package (Kuznetsova et al. 2017) for the p-values of GLMMs. To extract marginal responses of models, i.e. predicted response excluding random effects and assuming a constant (mean) value of all other predictors, we use the ggeffects::ggpredict() function (Lüdecke 2018).

Prior to model development, we assessed correlations between variables using variance inflation factors (VIF). Hypothesised predictors with high collinearity (VIF > 5) were not included in the global model. The final model for alien species richness and abundance on post-coal mine heaps was: glmer(formula = alien.rich ~ native.rich + native.FRich + native.CWM.SLA + (1 |heap/block)); lmer(formula = alien.abundance ~ native.abundance + native.FRich + native.CWM.H + native.CWM.SM + (1|heap/block)), where alien.rich = alien species richness, native.rich = native species richness, alien.abundance = alien species cover, native.abundance = native species cover, native.FRich = Functional richness of native species, native.CWM.SLA = native community-weighted means of specific leaf area, native.CWM.H = native community-weighted means of plant height, native.CWM.SM = native community-weighted means of seed mass.

To identify models with variables that best predict alien species richness and cover on post-coal mine heaps, we used a model selection in the MuMIn::dredge() function (Bartoń 2022) ranked, based on corrected Akaike Information Criterion, corrected for small sample size (AICc). For each model, we reported the AICc of the global model (i.e. all hypothesised predictors), final model and null (intercept and random effect only) model, to show how the final model differs from them. We ensured that the Poisson GLMM was not biased by overdispersion using the performance::check_overdispersion() function (Lüdecke et al. 2021).

We found that alien plant species accounted for 20.4% of all recorded vascular plants (65 out of 318 taxa) on heaps, with 55% of those being neophytes and the rest being archaeophytes. A higher proportion of native species is well-known from other post-industrial sites (e.g. old Solvay process heaps (Cohn et al. 2001); mining sites in the Czech Republic in central Europe (Prach et al. 2013); and the central German lignite mining district (Tischew et al. 2014)). The moderately high establishment of neophytes in our study is an indication that heaps are still at an early age and have relatively stable plant cover. Post-industrial sites left to spontaneous succession are usually characterised by low frequencies of alien species (Prach and Pyšek 1999; Prach et al. 2013; Tischew et al. 2014). However, in our study site, we found a high frequency of important alien species – Erigeron canadensis, Solidago gigantea and Solidago canadensis. In the Czech Republic, Ballesteros et al. (2021) recorded 129 archaeophytes and 67 neophytes in spontaneously established vegetation and the most invaded successional series were the deforested landscapes. Similarly, Simonová and Lososová (2008) found a high proportion of archaeophytes in a large variety of man-made habitats in the Czech Republic. The high proportion of archaeophytes in their study was due to the inclusion of less urbanised areas that were characterised by the presence of archaeophytes.

Erigeron canadensis, Solidago gigantea, Solidago canadensis, Erigeron annuus and Impatiens parviflora were the most frequent alien plant species in the studied plots on heap sites. Most important were S. gigantea and I. parviflora which had a high mean percentage cover. Solidago gigantea was found mainly in open habitats characterised by high light intensity and heap sites with early-successional communities. The species germinates by seed and rhizomes (Weber and Jakobs 2005). Clonal growth allows S. gigantea to form dense stands, promoting its abundance (Jakobs et al. 2004). Szymura et al. (2018) demonstrated the high competitiveness of S. gigantea in a replacement series experiment and found that S. gigantea outcompetes native grasses. Solidago gigantea in post-agricultural lands reaches the highest cover in sites with low functional richness (Czortek et al. 2020). Our study revealed the opposite pattern, as we focused on the cover of all alien species and we studied ecosystems with a lower level of interspecific competition. Two Erigeron species (E. canadensis and E. annuus) were frequent; however, they reached a low cover in the study plots. Both species are widespread in many ecosystem types on the mineral material of post-coal mine heap sites. This is because both E. canadensis and E. annuus plants produce 10 000–50 000 seeds annually that are wind-dispersed over long distances (Stratton 1989; Dauer et al. 2007; Pacanoski 2017). However, as ruderal species, they are more frequent in the initial phases of heap succession.

We found Impatiens parviflora in forest habitats within gaps in the herbaceous layer and heap sites at the late-successional stage. I. parviflora colonises sites with high native species richness (Chmura and Sierka 2006). Forest management practices, for example, canopy openings (gaps), propagule pressure from I. parviflora in plant communities around forests, increasing light availability and partial understorey disturbance promote invasion of I. parviflora in forests (Eliáš 1999). In general, the invasive ability of S. gigantea and I. parviflora is promoted through their physiological adaptation to water stress (Nolf et al. 2014; Quinet et al. 2015) and, for I. parviflora, through a high level of SLA intraspecific variability (Paź-Dyderska et al. 2020).

Prunus serotina was relatively less frequent on heap sites; however, in plots where it occurred, it had a high cover, thus, giving the species a high mean percent cover. Prunus serotina is a woody plant that encroaches on intermediate stages of succession due to its persistence in the shade and quick growth after disturbance (Closset-Kopp et al. 2007; Vanhellemont et al. 2009; Dyderski and Jagodziński 2019b; Jagodziński et al. 2019; Esch and Kobe 2021). Prunus serotina produces large numbers of seeds per year (Van den Tweel and Eijsackers 1986) with a major quantity of seeds present within 5 m of the parent tree and further dispersal of the seeds is done by frugivorous birds (Pairon et al. 2006; Deckers et al. 2008). As birds perch in a mature tree stand, the regurgitated P. serotina seeds are defecated and emptied as faeces, which then germinate (Jagodziński et al. 2019), thereby creating an efficient establishment of P. serotina seedlings within plots. This mechanism could explain P. serotina dispersal and spread.

The most frequent alien species in the studied spoil heaps were mainly herbaceous plants, self or insect-pollinated and self or wind-and-self dispersed. These are traits associated with the invasiveness of alien plants (Pyšek and Richardson 2007). In the analysis of the invasion success of the Czech alien flora, Pyšek et al. (1995) found that alien species in man-made habitats were mainly pollinated by either self or insects. However, Pyšek et al. (1995) found that animal or wind modes of dispersal of alien species were the most frequent in made-made habitats. In our studied system, most alien species are in the Asteraceae and Poaceae families with ruderal characteristics. This is expected because many of the traits contributing to the evolutionary success of Asteraceae and Poaceae have also encouraged some of the species within these families to be successful invaders (Lenzner et al. 2021).

We found that alien species richness and cover increased with native functional richness in the studied heap sites. Our finding is consistent with the niche-filling hypothesis (Thuiller et al. 2010; Loiola et al. 2018). The theory states that there are available niches left for alien species establishment in a functionally-rich community, while in a functionally-poor community, the niches are fewer and already occupied by native species. Alien species likely benefit from the presence of unoccupied ecological niches; therefore, filling them makes the functional space more saturated (Loiola et al. 2018). Therefore, our results do not support the biotic resistance theory – species-rich communities are more resistant to alien invasion than species-poor ones (e.g. Elton (1958), Bezeng et al. (2015)).

Our findings revealed that alien species cover decreased with native species cover on heap sites. Early native colonisers may control the establishment of later-arriving species by occupying niches and ensuring their persistence by creating abundant shade (Perry and Galatowitsch 2006). In our studied system, native species, such as Tussilago farfara, Chamaenerion dodonaei and Calamagrostis epigejos, are perennial early colonisers; therefore they persist for some years on heap sites (Stefanowicz et al. 2015; Kompała-Bąba et al. 2020). These native perennials could reduce the chances of the establishment of alien species with the same ecological requirements (Connell and Slatyer 1977). Therefore, ecologically-similar native species and early colonisers would be expected to capture more resources required by alien species due to niche overlap; thus, further suppressing alien species cover via limiting similarity (Abrams 1983). It has been hypothesised that niche takeover would occur when early- and later-arriving species are ecologically similar (Vannette and Fukami 2014). Our finding is in contrast to Lanta et al. (2022), who recorded an increase in alien species cover with native species cover in temperate lowland forests.

Our results showed that native CWM seed mass and plant height significantly predict alien species cover. Studies on the relationship between vegetation cover and the participation of species with different seed masses have shown that low cover (i.e. more open habitats) favoured the occurrence of species with small seed masses, while species with heavy seeds are successful in shaded habitats (Reader 1993; Kidson and Westoby 2000). A comparison of seedling survival of three temperate forest species differing in seed mass (Prunus serotina, Quercus rubra and Robinia pseudoacacia) confirmed that claim (Dyderski and Jagodziński 2019b). In our study, the native species pool had a low seed mass. Usually, alien species tend to avoid habitats where competitive natives with heavy seed masses were successful (Rees 1995; Turnbull et al. 1999). However, within early-successional communities, we found that alien plant cover increased with native plant seed mass. This suggests that pioneer alien species with low cover were frequent on newly-formed heaps (less than a year), while more competitive aliens with high cover (e.g. Solidago spp., Impatiens parviflora) invade sites where abiotic filtering does not limit the native species seedling establishment (early and mid-successional stage). Thus, seed mass of native species at early and mid-successional stages does not lead to strong competitive advantage over alien species during seedling establishment.

Plant height is an important ecological parameter in spontaneously-vegetated heap sites (Woźniak et al. 2011). Similarly, height controls the competitive ability of plant species (Weiher et al. 1999). In our study, the native species pool is characterised by low plant height. However, we found a positive relationship between alien cover and native plant height. This is because pioneer alien species usually have low cover and more competitive ones have a higher cover. Since our study is in successional communities, it only shows part of the whole gradient, revealing methodological differences in context dependence (Catford et al. 2022). Nevertheless, disturbance in closed habitats where natives have tall heights will continue to promote alien establishment through the creation of gaps. A similar result of increased alien species cover with native CWM plant height was obtained in temperate low-land forests (Lanta et al. 2022).

Our findings showed that alien species establishment was prominent in the early stage of post-coal mine vegetation development, but not on newly-formed heaps. Heap sites at the early developmental stage were characterised by alien species showing ruderal features that benefit from disturbance, for example, Solidago gigantea, Solidago canadensis, Erigeron canadensis and Erigeron annuus. These species reached a high level of ecological success. Therefore, to reduce invasibility, we recommend that the management objectives should be directed to the early stage of spontaneous vegetation formation on heap sites. Similarly, reduced ecological disturbance should be encouraged on heap sites to prevent ruderal colonising species and promote competitive native species.

Monitoring alien species invasion level and establishment on heaps and the surrounding landscape has high importance. Recent findings have shown that landscapes surrounding roads, railways and arable land harbour neophytes (Ballesteros et al. 2021). This affirms that the degree of urbanisation around colonised sites is an important invasion pathway and should be prioritised in alien species management strategies.

To prevent secondary invasion – an increase in the colonisation of non-target alien species after the removal of targeted invasive plants (Pearson et al. 2016), native species addition should be encouraged, specifically at the early successional stage. Our findings showed that plant communities at the early stage of spontaneous vegetation development on heaps are most threatened by alien species; thus, the addition of competitive natives would prevent non-target alien species from exploiting the space created by the removed targeted invader (Hess et al. 2019). Similarly, species addition will not only help restore native species lost from the ecosystem due to mining activities, but can also increase the number of competitors which may act to reduce alien species recruitment, invasion level and ecological success (Bakker and Wilson 2004). A more detailed study on the abundance shifts between the alien and native plant species in the vegetation patches during the developmental stages might give additional insight into the relationship between the role of alien and native plant species in the establishment and functioning mechanisms of the novel ecosystems on post-coal mine heaps mineral habitats.