Introduction

Invasive alien plants often have a negative impact on native species and ecosystems (Pyšek et al. 2012b; Vujanović et al. 2022). Usually, plant invasions take place differently across complex environments in different landscapes (Higgins et al. 1996; O’Reilly-Nugent et al. 2016). Since each invasive species has specific ecological requirements, the course of the invasion can be modelled mathematically in invaded regions (Coutts et al. 2011). This allows us to anticipate how the species will spread throughout the region and take action to limit it (Caplat et al. 2012; Szumańska et al. 2021). A spread model may be biased by changing climate factors since the species distribution is shifting due to climate change (Deutsch et al. 2008; Antão et al. 2022). Knowing the climate change trends, future invasion scenarios can be predicted (Guillera-Arroita et al. 2015; Briscoe Runquist et al. 2019; Stewart et al. 2022).

In plant ecology, Species Distribution Modelling (SDM) is the most commonly used prediction method for species spreading (Mkala et al. 2022; Qazi et al. 2022; Farashi and Alizadeh-Noughani 2023). It is widely utilized for the estimation of present or future potential distribution for invasive aliens or native plants (Pearson et al. 2007; Liu et al. 2013; Skálová et al. 2017; Briscoe Runquist et al. 2019; Cho et al. 2022; Chung et al. 2022; Qin et al. 2022; Ali et al. 2023; Jarnevich et al. 2023). There are several types of SDM algorithms. Profile modelling methods such as BIOCLIM have been used since the late 1980s (Booth et al. 2014). Regression-based models, e.g., Generalized Linear Models (Nelder and Wedderburn 1972), represent the second type of SDMs. However, machine learning models, e.g., MaxEnt (Phillips et al. 2006; Merow et al. 2013; Radosavljevic and Anderson 2014), Random Forests (Breiman 2001), and others, are the most popular methods. Aforementioned algorithms may be employed simultaneously (Naimi and Araújo 2016; Serrano-Notivoli et al. 2022) and can identify any species’ environmental boundaries and assess the potential distribution based on climatic, topographic, edaphic, anthropogenic, or other factors (Ali et al. 2023).

In this study, we focused on the spread of Symphyotrichum lanceolatum (Willd.) G. L. Nesom of the Asteraceae family driven by climatic change. It is an alien invasive species in Europe, introduced from North America in the 18th century. It was first cultivated in the gardens of several countries, e.g., in Germany before 1810 (Mears 1978), and in France around 1815 (Tison and de Foucault 2014). In Hungary, it was recorded before 1825 (Priszter 1997), in Belgium in 1835 (Verloove 2006). Later, it naturalized and spread across most of the continent, from the Iberian Peninsula to southernmost Scandinavia, up to central Russia (Pyšek et al. 2017). It is not exactly known when it was introduced to the central European countries, such as the Czech Republic (Pyšek et al. 2002; Pyšek, Danihelka Pyšek et al. 2012a, b), Austria (Essl and Rabitsch 2002), or Slovakia (Fehér 2008). In its native area (North America), S. lanceolatum is tolerant of a wide range of environmental conditions, as it grows primarily in wet or dry prairies, open woods, or shorelines, but it also moves to neglected fields, abandoned pastures, or roadsides. Ecological preference is greatly influenced by ploidy, as there are three fertile cytotypes and hybrids between them (Chmielewski and Semple 2001). In central Europe, it occurs in natural, human-made, or semi-natural habitats in nitrophilous herbaceous fringes of floodplain forests, in nitrophilous vegetation of alluvial meadows, exposed bottoms, and wet ruderal habitats, in willow-poplar forests of lowland rivers, ash-alder alluvial forests, flooded turfs, fresh-water reed vegetation, tall-sedge vegetation in littoral zones of oligotrophic and mesotrophic water bodies, between willow scrub of loamy and sandy river banks (Medvecká et al. 2012), and in ruderal nitrophilous and fringe vegetation (Rendeková et al. 2019). In certain places, it forms continuous colonies and displaces the native flora (Prach and Jedlička 2006; Láníková 2009; Obratov-Petković et al. 2011; Lastrucci et al. 2012), e.g., in floodplain forests, mechanical soil disturbance as a part of forest management significantly increases its abundance in understorey (Šebesta et al. 2021), and this effect is enhanced by the decrease in soil moisture (Mikulová et al. 2020).

The distribution of S. lanceolatum in Slovakia is limited by several environmental conditions, such as climate, soils, altitude, and heterogeneity of habitats, since the Carpathian Mountains provide a natural barrier to the spread of thermophilic invasive alien species. Climate change is shifting this barrier and invasive species are spreading into the mountains (Kochjarová et al. 2023). This offers a suitable background for researching how climate change affects the spread of S. lanceolatum and to test environmental factors shaping its distribution. In addition, only octoploid populations of this species are known in Slovakia (Murín and Feráková 1978) in contrast to other European countries (Dirkse et al. 2014; Kaplan et al. 2019). This excludes the potential bias in the modelling caused by the presence of other cytotypes with different ecological preferences.

Our research offers responses to the following queries: (a) What is the potential distribution of S. lanceolatum in Slovakia? (b) Are climatic changes responsible for S. lanceolatum’s expansion? (c) What are the possible future invasion scenarios?

Materials and methods

Study area

Slovakia, a European country (47.73–49.61 N, 16.83–22.56 E) with an area of 49,034 km², is located in two climatic zones (Metzger et al. 2005). The mean annual air temperature varies from 0.3°C to 12.2°C, and the annual precipitation totals are recorded from 534 mm to 1621 mm (according to WorldClim datasets, Fick and Hijmans 2017). Climate variations are brought on by surface heterogeneity, which includes lowlands in the east and west and mountains in the centre of the study area. While the mountains, with the exception of basins, have rich vegetation cover, the lowlands are intensively used as agricultural areas. Dense river network not only provide optimal habitats for a number of alien species but also aid in their spread (Medvecká et al. 2012; Májeková et al. 2021). For the data collection, the study area was divided into 1577 5’ × 3’ grid cells based on Central European mapping (Niklfeld 1971).

Data collection

For the study, both living plants and herbarium data were used. The 650 living plants from different populations were sampled by visiting various habitats suitable for the species occurrence: grasslands, open woods, neglected fields, abandoned pastures, riverbanks, or roadsides. The 50 herbarium records were collected from Slovak public herbaria (SAV, SLO, Scheffer’s herbarium). Whereas octoploid S. lanceolatum is difficult to distinguish from related hexaploid species and heptaploid hybrids, flow cytometry was used for species determination (Dirkse et al. 2014). The ploidy of samples was determined from leaves dried in silica gel using a commercial reagent kit, Cystain PI OxProtect (Sysmex, United Kingdom) and a Partec CyFlow Ploidy Analyzer, equipped with a green laser (532 nm) for propidium iodide excitation. At least 5000 nuclei were analysed from each sample at least three times on different days. Samples that exhibited a coefficient of variation (CV) greater than 5% were excluded from the analysis. Bellis perennis (2 C = 3.159 pg; Temsch et al. 2022) was chosen as the established standard for the samples. The ploidy of herbarium specimens was also estimated by flow cytometry according to Viruel et al. (2019), but because of the high CV, 18 herbarium records had to be excluded from the study. The octoploid S. lanceolatum records (2n = 64) have DNA amount 2 C = 4.7–5.2 pg (Dirkse et al. 2014). Using flow cytometry, 395 out of the 650 field samples were confirmed to belong to S. lanceolatum. Among the remaining 32 herbarium specimens, 14 octoploids were found (supplement 1); other ploidies were excluded from the study. For each of the 395 field records, the distance from roads and watercourses was calculated by GIS software (QGIS 3.22.3). The herbarium records were only useful for reconstructing the history of expansion since they lacked coordinates.

Data analysis

SDM was implemented to determine the potential distribution of S. lanceolatum in Slovakia. This modelling depends on the selection of optimal (presences) and unfavourable environmental conditions for the studied species (pseudo-absences). The result of the modelling depends on the selection of pseudo-absences (Fernandez et al. 2022; Wang et al. 2023). The best method is to use randomly generated pseudo-absences ‘2°far’ from presences, but for smaller areas it is possible to use 1° distance. The number of pseudo-absences depends on the presences; if the number of presences is below 1000, the number of pseudo-absences should be equal and generated at least in five random runs (Barbet-Massin et al. 2012; Descombes et al. 2022). We set 395 samples determined with flow cytometry as presences. The same number of pseudo-absences distal at least 1° from presences was generated in ten random runs and averaged. Seventeen environmental and geographic characteristics have been assigned to all the data (Table 1).

Table 1 Environmental and geographic variables used as predictors in Symphyotrichum lanceolatum distribution forecasting
Full size table

Nineteen SDMs with different algorithms (Table 2) were calculated to assess the likelihood of S. lanceolatum occurring in the study area under past, current, and future conditions using sdm function of ‘sdm’ package (Naimi and Araújo 2016) in R (version 4. 3. 1, R Core Team 2023). Each model was evaluated with 10-fold cross-validation to obtain the true skill statistic (TSS) and area under the relative operating characteristic curve (AUC) values. For predictions, only models with TSS higher than 70% and AUC higher than 90% were selected (Allouche et al. 2006; Phillips and Dudík 2008).

Table 2 Algorithms used in SDM
Full size table

As predictors, seventeen additional environmental and geographic variables (Table 1) were used at 30-second resolution. Using the ‘ensemble’ function of the same ‘sdm’ package, the predictions from all fitted models were calculated by weighted averaging (Naimi and Araújo 2016). We predicted the potential S. lanceolatum distribution for the past, present, and future based on historical (1970–2000), current (2018–2021), and future data (2040–2061, model EC-Earth3-Veg, scenario SSP585) from WorldClim datasets (Fick and Hijmans 2017). The areas with a probability higher than 0.5 were considered to have potential for species expansion (Skálová et al. 2017).

Spearman correlation coefficients were calculated between the three different predictions and between the predictions and all environmental and geographic variables used as predictors using the ‘raster.cor’ function of ‘ENMTools’ R package (Warren et al. 2021).

Using the ‘ctree’ function of the ‘party’ R package (Hothorn et al. 2006), a non-parametric decision conditional inference tree was constructed to estimate the effects of variables (Table 1) affecting the S. lanceolatum distribution.

Results

Distribution of S. Lanceolatum in Slovakia

The history of species expansion was reconstructed on the basis of herbarium specimens and their ploidy (suppl. 1). The earliest confirmed S. lanceolatum records in Slovakia originate from 1917 (SCH0917, Bratislava, bank of the Danube). Before 2000, records were uncommon and linked with the banks of the rivers (the Danube and its tributaries). Flow cytometry confirmed that many S. lanceolatum herbarium records from this period were incorrectly determined as other Symphyotrichum species (at the time included in the Aster genus). After 2000, S. lanceolatum started to spread from the riverbanks to drier areas. At present, 395 sampled records are located in 186 of 1577 grid cells (Fig. 1). The neglected gardens, human settlements, and field edges are the mostly colonized habitats, representing two-thirds of our records. The remaining share of the habitats consists mainly of the margins of forests, meadows, pastures, and riparian forests, as well as the banks of streams or rivers. As can be seen in the histograms (Fig. 2), the highest frequency of S. lanceolatum occurrence is bound to wet habitats near watercourses and to an anthropogenic environment represented by road density.

Fig. 1
figure 1

Study area with grid cells 5’ × 3’ occupied by Symphyotrichum lanceolatum (points)

Full size image
Fig. 2
figure 2

Histograms estimated by the Freedman-Diaconis rule showing frequencies of Symphyotrichum lanceolatum occurrence at different distances from the nearest water bodies and roads

Full size image

Species distribution modelling

Using SDM algorithms, potential distribution maps for the past, present, and future were created. The species might expand throughout 23.6% of the study area in the lowlands and the warmer basins of the Carpathian Mountains, according to the prediction for the past (Fig. 3). Due to changes in climatic conditions in last two decades (an increase in mean air temperature of 0.78–1.40 °C and precipitation totals of 42.6–80.9 mm), the species could expand to 35.9% of territory based on the second prediction for the present (Fig. 4). The prediction for the future assumes increasing air temperature and precipitation totals, which would lead to a shift in expansion to 53.8% of the study area (Fig. 5).

Fig. 3
figure 3

Predicted distribution of Symphyotrichum lanceolatum for the past period 1970–2000

Full size image
Fig. 4
figure 4

Predicted distribution of Symphyotrichum lanceolatum for the current period 2018–2021

Full size image
Fig. 5
figure 5

Predicted distribution of Symphyotrichum lanceolatum for the future period 2040–2061

Full size image

Environmental factors influencing the spread of S. Lanceolatum

The Spearman correlation analysis (Table 3) indicates relationships between environmental and geographic variables and probability of S. lanceolatum distribution (Fig. 6). These relationships are specified in conditional inference tree (Fig. 7). The mean annual air temperature below 8 °C is the primary limiting factor. Also, topographic elements, annual precipitation totals, and a pH of soil less than 5.5 determine the species’ occurrence.

Table 3 Spearman correlation coefficients between the various factors and estimated probability of Symphyotrichum lanceolatum distribution in the past (P1970 − 2000), at present (P2018–2021), and in the future (P2040 − 2061); *** p < 0.001, NS – no significant difference
Full size table
Fig. 6
figure 6

Relationships between selected factors (mean annual air temperature T, altitude, soil ph, annual precipitation totals P) and the probability of Symphyotrichum lanceolatum occurrence in the study area

Full size image
Fig. 7
figure 7

Conditional inference tree displaying the effectiveness rates of the few elements influencing the Symphyotrichum lanceolatum distribution. The bars display the likelihood that Symphyotrichum lanceolatum will be present (dark grey) or absent (light grey) under the conditions indicated by the ovals

Full size image

Discussion

Expansion of S. Lanceolatum as a result of climatic change

It is unclear when S. lanceolatum was brought to Europe, but the species was planted here in the 18th century (Scholler 1787). It probably started to spread in Germany and the Silesian region, from where the expansion continued to Moravia in the Czech Republic (Prach and Jedlička 2006; Řepka et al. 2015; Pyšek et al. 2017), and probably along the river Morava further to Slovakia (Dostál 1950) and to Austria, where its vernacular name is ‘March-Aster’ meaning Moravian aster. Our investigations of herbarium items confirm the assumption that the species spread along the river Morava and from there along the Danube in Slovakia between 1917 and 1951. Whether the species grew in the study area earlier is not yet known. According to the records from the 19th century (Endlicher 1830; Reuss 1853), S. lanceolatum did not grow along the Danube in Slovakia, but along the Danube in Hungary it was found before 1825 (Nees von Esenbeck CG 1833; Priszter 1997).

The species began to spread intensively in Slovak territory later in the 1970s along the Danube tributaries in the lowlands (Fehér 2007). It continues to expand from the west to the east, but the Eastern Slovak Lowland has not yet been invaded. At present, we recorded its presence in 186 of 1577 grid cells, which represents more than 11% of the territory. Such distribution is comparable to that in the Czech Republic (Chytrý et al. 2021). In general, the distribution captured by field research may not be exact and the true status of the expansion can be often twice as high (Mang et al. 2017). Our SDMs indicate that if there was no warming, current expansion would stop at 23.6% of the territory. This would correspond to the current possible true distribution (twice 11%). But due to warming, the potential area for S. lanceolatum has increased to 35.9%, and the future outlook suggests an increase to 53.85% in 2060. Climate change and season lengthening are thought to be the main factors in the species’ expansion in Serbia, another country in the Danube basin (Obratov-Petković et al. 2009, 2011, 2013). Prediction models in Serbia anticipate that the species will expand to submontane and mountainous regions in the Dinaric Alps and Carpathians due to an increase in temperature and precipitation (Nešić 2017). A comparison of our model with the Serbian model (Nešić 2017) reveals that most of the environmental factors influencing the species’ distribution are the same in Serbia and Slovakia. The main difference between the models is the effect of altitude on expansion. The acidic mountain soils of the Western Carpathians might act as a barrier since S. lanceolatum prefers a pH greater than 4.6 (Chmielewski and Semple 2001). In Serbia, acidic soils are not tied to mountains with higher altitudes (Ličina et al. 2011). Our research revealed that the species was less frequent in soils with pH values below 5.5 and rare below 4.6 (3% of records). Significant effect on the predicted distribution we observed also in the type of vegetation cover. Future expansion may also be limited by forest vegetation because the species requires sunlight and is hindered by tree shadowing (Hardin and Wistendahl 1983).

Habitat preference

While in the past the species in central Europe mainly preferred riparian and floodplain habitats (Meinert et al. 2009; Medvecká et al. 2012), today it expands to ruderal vegetation (Rendeková et al. 2019; Hejda et al. 2021; Tabašević et al. 2021). Our results confirm that the species is mostly located close to rivers and the road network. One explanation may be that octaploid individuals are more capable of tolerating a wider range of habitats than diploids or tetraploids (Chmielewski and Semple 2001). Whereas air temperature, precipitation totals, and the acidity of the soil appear to be the main factors affecting S. lanceolatum spread, human settlement can offer optimal conditions for growth. In cities, the precipitation totals are enhanced (Liu and Niyogi 2019) and the temperature of the environment is higher than in rural areas or the surrounding nature (Liu et al. 2022). Another explanation may be the reproduction strategy. S. lanceolatum reproduces in two ways (Chmielewski and Semple 2001). Vegetative reproduction through stoloniferous rhizomes is mostly carried out near water habitats (Prach and Jedlička 2006). Generative reproduction occurs in different environmental conditions, often in disturbed sites, and is affected by temperature, nutrient concentration, and light (Nešić et al. 2013, 2022). Since S. lanceolatum is a nitrophilous species (Pyšek et al. 2017; Nešić 2017), increasing nitrate concentrations in soils and waters by human activity could contribute to its spread in anthropogenic habitats (Sadayappan et al. 2022). The nitrogen eutrophication resulting from anthropogenic activities is a well-known factor that allows invasive plants to dominate native vegetation (Wedin and Tilman 1996; Davis et al. 2000). The continued spread of S. lanceolatum might harm natural plant communities due to its allelopathic influence on other species (Obratov-Petković et al. 2016; Nešić et al. 2016). Knowing the potential outcome, it is necessary to take preventative steps to avoid the spread of the species. However, mechanical elimination is not very efficient, and herbicides are not advised (Schmiedel et al. 2015).

Conclusions

In the past, the potential distribution of S. lanceolatum was restricted to a small area, and the species preferred natural habitats. According to our model, it could occupy one-third of Slovakia’s land area now. The major cause of this is the increase in air temperature. However, human factors must also be highlighted because neglected gardens and open spaces offer the species an ideal habitat. Our study indicates the necessity for appropriate behaviour toward environment and expansion mitigation since forecasts indicate that expansion will continue over the next 40 years. This prognosis may change because our model does not account for the creation or disappearance of optimal biotopes such as neglected gardens, fields, and pastures by individual human behaviour or competition with other invading species.