Introduction

Rapidly growing human activities associated with disturbances of the landscape are reflected by increasing interest in urban floras, with special regard to species of alien origin, which, on average, comprise ~28% of plant species in cities worldwide (Aronson et al. 2014). Cities represent a specific environment for plants, both in terms of ecological factors and vegetation with novel species assemblages (Gilbert 1989; Hobbs et al. 2006). From the viewpoint of plant invasions, human settlements serve as immigration foci from which alien species spread into surrounding landscapes (Chytrý et al. 2005, 2008b; Celesti-Grapow et al. 2006; Lososová et al. 2012b); this usually happens not randomly but along linear corridors such as rails, roads, rivers or canals (Asmus and Rapson 2014, Ricotta et al. 2014). Alien species are more closely associated with the urban environment than native species (e.g. Williams et al. 2015). This is due to opportunities for introduction and transportation of propagules (Sukopp and Werner 1983; Lockwood et al. 2005; Malavasi et al. 2014), habitat heterogeneity (Pyšek 1995; Deutschewitz et al. 2003), adaptation to high levels of disturbance (Celesti-Grapow and Blasi 1998; Davis et al. 2000; Chytrý et al. 2008a), and higher temperature demands that are met due to the phenomenon of ‘urban heat islands’ (McKinney 2006; Lososová et al. 2012b). For these reasons, the urban environment harbours a high number of plant species distributed in a wide range of habitats, and represents a unique “unnatural phenomenon” (Sukopp and Werner 1983). Due to greater habitat heterogeneity and enrichment by alien species, cities are generally richer in species than surrounding landscapes (Haeupler 1974; Kühn et al. 2004), and these differences have become increasingly more pronounced over time (Chocholoušková and Pyšek 2003; Pyšek et al. 2004a). In central Europe, aliens in the floras of big cities make up ~40% of the total number of taxa (e.g., Pyšek 1998; Ricotta et al. 2009; Lososová et al. 2012a), and in the flora of the Czech Republic, for which there is detailed information available, more than half of 1454 alien taxa (representing about one third of the total flora) are confined to human settlements, i.e. cities and villages (Pyšek et al. 2012).

In the last decades habitat alteration and destruction, together with human-mediated introductions of alien plants, have resulted in dramatic changes in the structure of European urban floras. Typical of these trends is a gradually increasing proportion of neophytes (alien species introduced to central Europe after 1500 AD) in the total flora, while archaeophytes (pre-AD 1500 aliens; see Pyšek et al. 2004b for definitions) and native species proportions have decreased (Kowarik 1995) or remained stable. Some studies (e.g., Pyšek et al. 2003; Kühn and Klotz 2006) explain such a pronounced turnover of species during a relatively short time by the rapid growth of cities, which creates novel habitats, i.e. industrial areas and waste dumps, suitable for only a few species adapted to new, and markedly different, conditions. Such habitats are under a strong human influence and mostly invaded by neophytes that are more urbanophilous compared to archaeophytes that prefer habitats less affected by humans (Hill et al. 2002). Also, the presence of neophytes has been shown to depend more closely on propagule pressure than that of archaeophytes, which is generally high in human settlements (Chytrý et al. 2008a). This implies that neophytes are a greater threat to the native flora because they, unlike archaeophytes, have not yet occupied all suitable habitats (Chytrý et al. 2008a). The city of Plzeň, western Bohemia, Czech Republic, can be used to illustrate these trends; here the proportion of neophytes in the city flora increased from 73 to 177 species between 1880–1910 and the 1990s (Chocholoušková and Pyšek 2003). Similar trends are reported from other cities in Europe (e.g., La Sorte et al. 2008; Lososová et al. 2012b; Ricotta et al. 2012). Consequently, understanding the behaviour of neophytes, as relatively recent newcomers and likely invaders of the future, in the urban environment is of crucial importance. This is especially relevant for this group of alien species because many neophytes are able to spread to considerable distances from parental plants, which allows them to rapidly colonize large geographical areas. This subset of taxa is called invasive neophytes (see Richardson et al. 2000; Pyšek et al. 2004b for definitions).

In this study, we assessed the role of different ecosystem types within a city area resulting from landscape transformation due to urbanization and heterogeneity, as well as several parameters describing urban structure in shaping the performance of invasive neophytes. We ask how these factors interact in influencing the cover and identity of invasive neophytes, and determined their relative importance along a gradient from the city centre, defined as the densely urbanized inner city, to the outskirts, defined as rural exurban environment. Although distance to city center is a widely used metric, it is a simple measure that does not consider the organic growth of a city (see McDonell and Hahs 2008 for a review). However, in this study area we have used this measure as there was a clear gradient from old city center through suburban housing to agricultural land (see Fig. 1).The effect of distance from the city centre has been often examined in previous studies (e.g., Klotz 1990; Kowarik 1995; Kühn et al. 2004; Celesti-Grapow et al. 2006) based on which we hypothesized that invasive species richness significantly increases towards the more urbanized, less natural sites, which are located closer to the city centre.

Fig. 1
figure 1

Study area depicted on the ortophoto of Hradec Králové

Full size image

Our main aims were: (i) to investigate the relationships between the species richness and total cover of invasive neophytes and the ecosystem types, (ii) to compare the relative importance of invasive neophytes in higly transformed and low/medium transformed ecosystem types, (iii) to examine how much of the variation in the composition of invasive neophytes is accounted for by spatial structure predictors in the highly transformed and low/medium transformed ecosystem types.

Methods

Study area

The study was carried out in the city of Hradec Králové, eastern Bohemia, with a population of ~100,000, which ranks it among the ten largest cities in the Czech Republic. The city was founded in the early thirteenth century. It is located at 235 m a. s. l., in a warm and slightly dry climatic region with moderate winters, mean annual temperature of 8.0 °C and mean annual rainfall of 600 mm (Tolasz et al. 2007). According to the phytogeographic division of the Czech Republic (Skalický 1988), Hradec Králové belongs to the Thermophyticum; this district is located mostly in lowlands that were first colonized in the Neolithic (Chytrý et al. 2008a), which implies an intensive and long-lasting human impact.

Using the approach of McDonnell and Pickett (1990) and Pyšek (1995), a sampling grid of 200 cells (200 × 200 m each) running from the city centre to its outskirts (2 × 4 km in size) was used. The study area included arable land, old city centre, railway station and adjacent industrial zone, as well as important bio-corridors; they are, besides railways and roads, two rivers (the Orlice river empties into the Labe river in the study area), Chaloupská svodnice stream and Labe’s water-channel (Fig. 1). With the exception of inaccessible private gardens and army barracks, the whole study area was sampled, including managed spaces such as parks and lawns.

Data collection

The field work was carried out from July to September 2004 and we identified a list of 64 invasive neophytes. The species selection was based on the catalogue of alien plants of the Czech Republic and included all invasive neophytes occuring in our territory (Pyšek et al. 2002). Of the species labelled as invasive in the catalogue, we excluded the water macrophyte Elodea canadensis, and three species (Cytisus scoparius, Peucedanum ostruthium, Mimulus guttatus) considered post-invasive in Pyšek et al. (2002) and currently classified as naturalized rather than invasive (Sádlo et al. 2007; Pyšek et al. 2012). Arrhenatherum elatius and Myrrhis odorata were also excluded as their neophyte status is doubtful and they are now regarded as archeophytes in the Czech Republic (Chytrý et al. 2005; Pyšek et al. 2012). Even though some species were labelled as naturalized in the updated catalogue (Pyšek et al. 2012) that used a more conservative approach compared to the previous edition (Pyšek et al. 2002), we included them in this study as they are common in the Czech urban alien flora: Epilobium adenocaulon, Galeobdolon argentatum, Mahonia aquifolium, Matricaria discoidea, Oenothera biennis, Physocarpus opulifolius, Rhus typhina, Rumex thyrsiflorus and Syringa vulgaris. Introduction, region of origin and traits of the species included in our study as well as their taxonomy and nomenclature were taken from Pyšek et al. (2012); a synoptic taxon group was used for Erigeron annuus agg. (see online Appendix 1 for the list of particular species with detailed information on the traits considered).

Different sources were used to obtain predictors. Most of the chosen predictors allow us to distinguish ecosystem types with prevailing level of transformation (adapted and simplified after Kowarik 2011). For each cell in the study area, the length of roads and length of water courses were obtained from Urban Atlas 2006–2008. Other descriptors were not reliable enough in this source so ortophotographs for the year of study (Czech Office for Surveying, Mapping and Cadastre 2004) were manually classified into other ecosystem types: built-up areas and sealed surface, agricultural land, railways and associated land (see online Appendix 2 for the list of particular ecosystem types with their detailed description), and public urban greenery, i.e. a summary descriptor including all types of vegetation in the public space (e.g. spontaneous vegetation, parks and plant cultivations). Their proportions for each cell were computed in ArcGIS Pro 1.1 using tool Summarize Within (ESRI 2015).

Based on the local knowledge of the studied city, we distinguished ten habitat categories (listed and described with other ecosystem types in online Appendix 2); their total number was taken as the measure of ‘habitat richness’ that was used in statistical analyses. In the field we recorded (i) presence/absence of the ten habitat categories in each of the 200 cells, (ii) the abundance of each invasive neophyte, expressed as visually estimated total cover (in m2) of all its populations growing outside cultivation, and also its affiliation to the habitat categories present in a cell, and (iii) whether the species is cultivated (yes or no). The distance from the city centre was measured in cell “units” (200 × 200 m)—each cell that included the city centre or its edge was treated “0”, a cell next to it “n”, the next one “n + 1” etc.

Finally, in the case we had two predictors describing similar features, the more precise predictor was used in statistical analyses, e.g. the length of roads instead of presence/absence of roads.

Statistical analyses

Multivariate constrained ordination was used to compare and jointly test the effects of ecosystem types and other predictors on relative importance of invasive species, using Canoco 5 package (ter Braak and Šmilauer 2012). We included only species occurring in at least three cells; hence the analyses are based on 30 species (highlighted in bold in online Appendix 1).

The relative importance value of individual invasive neophytes was expressed as a percentage of the visually estimated proportion of urban greenery. These values were log-transformed and the linear model of RDA (redundancy analysis) was chosen because a very short compositional gradient was sampled, so a linear response of species along this gradient was assumed (Šmilauer and Lepš 2014, p. 27). We also standardized the relative importance values of the species using case norm in the multivariate analyses, so that they reflect only the changes in the relative proportions of individual invasive neophytes along the gradient (Šmilauer and Lepš 2014, p. 30). In this way, the ordination results complement the regression models predicting total cover and richness of invasive species.

To investigate how the spread of invasive neophytes and different ecosystem types distinguished within the study area are spatially structured and whether this creates an indirect correlation between them (spatial nuisance sensu Peres-Neto and Legendre 2010), we used the spatial eigenfunction approach to represent the spatial variation at multiple scales as a set of additional predictors (Legendre and Legendre 2012, pp. 859–905). In particular, we computed so-called distance-based Moran eigenvector maps (db-MEM) and used a selected subset of significant spatial eigenvectors both as a third group in the variation partitioning analyses (ter Braak and Šmilauer 2012) and as covariates in the regression models.

The importance of low/medium transformed vs highly transformed ecosystem types (see online Appendix 2 for classification) was compared by first performing forward stepwise selection within each group and then comparing the effects of selected predictor subsets using a three-way variation partitioning procedure (Šmilauer and Lepš 2014, pp. 88–91), which included significant spatial eigenvectors (hereafter PCOs) as a third group of predictors together with X and Y coordinates of cells and the distance from the city centre. The independent effects of selected ecosystem type descriptors or of their groups were studied with simple constrained ordinations. We also evaluated the simple (independent) and conditional (dependent) effects of individual variables within each predictor group. The estimated Type I errors for both simple and conditional effects were adjusted by converting them into false discovery rate (FDR; Benjamini and Hochberg 1995) estimates.

To focus on the unique effects of the distance from the city centre, we performed another RDA with this distance as the only explanatory variable. All constrained analyses were accompanied by Monte Carlo permutation tests of the significance. With the exception of analyses using the spatial eigenvectors as explanatory variables or covariates, we used constrained permutation tests where the spatial autocorrelation between adjacent cells is taken into account (ter Braak and Šmilauer 2012, p. 74).

Generalized linear models (hereinafter GLM) were used to evaluate the influence of selected ecosystem type descriptors and habitat richness on two summary characteristics: log-transformed sum of invasive neophytic cover and the diversity of invasive species, using R software (R Development Core Team 2008). Significant predictors were chosen by forward stepwise selection, using a model with assumed Poisson distribution when predicting the count of species and assumed gamma distribution (and log link function) when predicting the total cover. The stepwise selection was based on Bayesian information criterion (BIC) values (Schwarz 1978), stopping the selection when the BIC of the best candidate predictor was larger or equal to BIC value of the present model or it was smaller by less than 0.5% of the present model’s BIC. For predictors selected into the model based on the parsimony criterion BIC, parametric tests of significance were done post hoc using χ 2-statistic or F-statistic based tests, respectively for taxon count or total cover. To compare the (confounding) effect of spatial structure, these two models were selected both without and with important spatial eigenvectors (pre-selected again for each response variable using the BIC criterion).

Results

Number of invasive neophytes and their total cover

Forty-two invasive neophytes were recorded in the study area. The highest species richness was found in road margins (32 species), ruderal sites (28 species), and railway sites (23 species); 21 species occured on water edges and in cultivated areas. The lowest numbers of invasive neophytes were recorded in meadows (5 species) and field margins (5 species). The most common life history was annual (14 species), most species were deliberately introduced (28 species), native to North America (20 species) and from the Asteraceae family (11 species) (see online Appendix 1 for details).

Based on the GLM results, the total number of invasive species in cells significantly decreased with increasing distance from the city centre (~4% decrease with DistCent increasing by one cell, i.e. 200 m), but increased with habitat richness (~12% increase with each additional habitat type), see Table 1a. The model predicting the estimated total cover of invasive neophytes yielded slightly different results with the inclusion of public urban greenery (Table 1b). The total cover of invasive species decreased with increasing proportion of the public urban greenery and the distance from the city centre (~11% decrease with DistCent increasing by one cell), and increased with habitat richness (~44% with each additional habitat type). When including important spatial predictors into the model first, none of those effects was retained (Table 1a, b).

Table 1 Overview of selected predictors and their effects on (a) the number of invasive species per cell (species richness), and on (b) the total cover of invasive neophytes per cell in fitted GLMs. For abbreviations see online Appendix 2
Full size table

Determinants of the relative importance of invasive neophytes

Effects of highly transformed ecosystem types

According to computed RDA, the four selected descriptors together explained ~6% of variation. When evaluating the independent contributions of individual descriptors, the proportion of built-up areas and sealed surface was the most important predictor, explaining 3.5% of data variation, followed by the length of roads (Table 2a, simple effects). When the joint effect was examined, the second best predictor was the presence of ruderal sites (see Table 2a, conditional effects).

Table 2 Simple (independent) and conditional (dependent) effects of (a) highly transformed ecosystem types, and (b) low/medium transformed ecosystem types on the relative importance of invasive neophytes in the study area, based on RDA model
Full size table

Effects of low/medium transformed ecosystem types

The predictors significantly affected (pseudo-F = 2.6, p = 0.005) the relative importance of invasive species and explained ~5% of the variability in this measure. When each descriptor was tested independently (see Table 2b, simple effects), five of the six considered descriptors had a significant effect. Given an obvious dependence in their occurrence, their joint effect was examined and only three descriptors (agricultural land proportion, length of water courses and plant cultivations) were retained (Table 2b, conditional effects). However, the effect of plant cultivation is doubtful because it contributes only a small amount of variability in addition to the two already selected predictors.

Effects of habitat richness and distance from the city centre

Habitat richness and the seven most predictive descriptors of ecosystem types (see Table 2, conditional effects) were retained in the constrained ordination (Fig. 2). They significantly affect (pseudo-F = 3.6, p = 0.005) the relative importance of invasive species and explained 9.3% of the variability. Plotted isolines demonstrate the increase in species richness towards increasing habitat richness. Species in the upper right corner, Bidens frondosa and Impatiens parviflora, are predominantly found close to water courses. Species located in the left part of the diagram (Conyza canadensis, Erigeron annuus agg., Matricaria discoidea, Galinsoga quadriata and G. parviflora), represent the common ruderal alien flora.

Fig. 2
figure 2

Redundancy analysis ordination diagram displaying the relation between the habitat richness, the seven most predictive ecosystem types (see Table 2, conditional effects) and the relative importance of invasive neophytes (nine taxa with the highest variance explained by the predictors are shown). The axes shown explain 7.9% of the total variation. Plotted isolines represent the change of species richness across ordination space as fitted by a loess smoother model. For abbreviations see online Appendix 1, 2

Full size image

The significant effect of distance from the city centre on the relative importance of invasive neophytes is summarized in Fig. 3. This effect is represented by the horizontal axis of the diagram, explaining 1.9% of the total variation (pseudo-F = 4.8, p = 0.006). Nine taxa with the highest variance explained by the city centre distance are shown.

Fig. 3
figure 3

Redundancy analysis ordination diagram displaying the effect of distance from the city centre on the relative importance of invasive neophytes (nine taxa with the highest variance explained by the town centre distance, along the horizontal axis of the diagram). The first, constrained axis explains 2.4% (pseudo-F = 4.8, p = 0.006) and the second (unconstrained, vertical) axis 11.4% of the total variation. Plotted isolines represent the species richness change across ordination space as fitted by a loess smoother model. For abbreviations see online Appendix 1, 2

Full size image

In two partial constrained ordinations, both the descriptors of highly transformed ecosystem types (pseudo-F = 3.8, p = 0.005) and those of the low/medium transformed ecosystem types (pseudo-F = 2.1, p = 0.005) retained significant effects on the composition of invasive neophyte composition after accounting for the distance from the city centre. In addition, we found that the effects of distance from the city centre and proportion of the public urban greenery are strongly interrelated. When taking the distance into the model as a predictor and the greenery as a covariate, the distance explains mere 1.2% (pseudo-F = 3.3, p = 0.001), while the effect of the public urban greenery is non-significant when testing its effect in addition to the distance to city centre.

Comparing the effects of highly transformed and low/medium transformed ecosystem types

Variation partitioning (Table 3) indicates that the major parts of the variation explained by highly and low/medium transformed ecosystem types are spatially structured, while their unique contributions are relatively small (and non-significant for the former group, pseudo-F = 1.2, n.s.), which indicates that a very large part of the variation in the composition of invasive neophytes is accounted for by spatially structured predictors, not corresponding to those we were able to measure, and that the (relatively important) effect of measured predictors is mostly spatially structured.

Table 3 Results of variation partitioning among three groups: highly transformed ecosystem types (HT), low/medium transformed ecosystem types (LMT) and spatial predictors (SP)
Full size table

Discussion

By employing descriptors of different types of ecosystems within a city area as proxies for structured urban landscape we can identify the sites within the city with a high probability of invasive species’ occurence. Both measures of the invasive neophytes’ performance we used, species richness and total cover, respond consistently to these factors; they decrease with increasing distance from the city centre and are supported by habitat richness—more habitats recorded for a cell might be a “prerequisite” for the occurence of more species in the cell. This implies that invasive neophytes in the urban environment perform better in areas that harbour a wide range of habitats—available niches for newcomers. Although the assumption that habitat diversity triggers a high plant species richness in urban areas has been mentioned in the previous studies (e.g. Sukopp and Werner 1983), a rigorous evidence based on purposely collated primary data is lacking. For the total cover of invasive neophytes, we also found that it decreases with increasing proportion of public urban greenery. The peripheries in cities in the Czech Republic are usually less populated, with a lower proportion of built-up areas, and still keep their relatively rural character (Celesti-Grapow et al. 2006; Chytrý et al. 2008a). The higher proportion of greenery at the city margins increases the competition among species, which makes the invasion by alien species and their dominance in local communities more difficult; only a few alien species are able to penetrate into these resident plant communities, establish there and create large stands. For neophytes, the sites close to the city centre seem to be more suitable as these species are well-adapted to the limited water supply and high temperatures, which provides them with an advantage over archaeophytes and native species (Pyšek et al. 1995). Although their populations usually do not reach high covers due to the limited spatial extent of sites in which the vegetation can occur, e.g. trampled or otherwise heavily disturbed sites (Sádlo et al. 2007), their total cover is higher compared to the periphery. However, the species richness and total cover within the city (or the considered descriptors themselves) must be strongly spatially correlated as none of the considered descriptors remained in the model when the predictors describing spatial structuring were included.

The gradient from the city centre to the outskirts not only influences the invasive species richness and cover, but also has an important effect on the composition of neophyte assemblages, as illustrated by changes in the relative importance of individual species. This implies that species confined to habitats close to the city centre are different from those typical of the periphery. It is also important to emphasize that both the descriptors of the highly transformed ecosystem types and the low/medium transformed ecosystem types retained explanatory power of the invasive neophytes’ composition even after accounting for the distance from the centre, while the public urban greenery itself did not. Obviously, we cannot distinguish the effect of large distance from the city centre from the effect of high proportion of public urban greenery, unless we include in our studies more towns with different location of urban greenery with respect to their city centres.

As regards the invasive species characteristics, we find similar results as previous studies. Most of the invasive neophytes belong to the plant family Asteraceae, which is in accordance with for example Pyšek et al. (2002) and Weber et al. (2008). Regarding the life-history, type of introduction into the country and origin, most of the recorded species are annuals, deliberately introduced and native to North America and this correspond with the findings of other studies not only from Europe (e.g. Lambdon et al. 2008) but also from Asia (Zerbe et al. 2004; Weber et al. 2008).

The effect of different ecosystem types on plant invasions has been documented for urban areas (Gilbert 1989; Sukopp 2002; Celesti-Grapow et al. 2006; Vilà and Ibáñez 2011) as well as for entire landscapes (Chytrý et al. 2008a; Pyšek et al. 2010; Kowarik 2011). When explaining the changes in relative importance of individual invasive species, agricultural land, length of water courses and plant cultivations were the most important predictors from the group of low/medium transformed ecosystem types, and built-up areas and sealed surface, length of roads, ruderal sites and railways and associated landscape from the group of highly transformed ecosystem types. These findings correspond to the strong effect of factors generally considered as driving invasion in different ecosystem types, such as the disturbance regime, high amount of available nutrients, and propagule pressure, including opportunities for spread by water or human activities (e.g., Davis et al. 2000; Chytrý et al. 2008b; Blumenthal et al. 2009; Pyšek et al. 2010; Foxcroft et al. 2011; Vieira et al. 2014).

Although, the results demonstrate that the effects of low/medium transformed ecosystem types and highly transformed ecosystem types are almost independent, the large amount of variation explained by the spatial predictors, but not shared with any measured variables used, implies that some important characteristics determining the identity and cover of present neophytes were missed from the model. Other explanation could be that the significant predictors might be events, structures or patterns no longer in existence, which cannot be traced in any presently measurable property (“historical dynamics” sensu Legendre and Legendre 2012, p. 878). However, similar results were also acquired by Gulezian and Nyberg (2010), who found almost no effects of urban land-use descriptors on invasive species’ abundance in their study. This, together with the fact that different factors are likely to be important for different species, points to the necessity to study the whole spectrum of characteristics should the invasions in urban areas be predicted effectively. For example, Niggemann et al. (2009) suggest that the human mobility in urban areas is a much better predictor compared to habitats or habitat-related factors. Also, the approach of Hahs and McDonnell (2006) seems to be promising. They attempted to assess commonly used urban descriptors and provided an example of how to select objectively a subset of predictors that could help to understand ecological responses to urbanization.

In conclusion, different invasive species prefer habitats either in the vicinity of the city centre or at its periphery and the structure of the urban landscape needs to be taken into account in efforts to manage alien plant species invasions in urban environments. Our study has also implications for future urban planning and the management of green areas within the city, such as parks and gardens. It is also obvious from the results that the management of the habitats itself is the central issue to control alien species. To prevent greater numbers of invasive species from establishment, low/medium-transformed habitat types should be promoted in urban areas, and the habitats that are most suitable for alien species should be suppressed especially near the city center, where the competition from native plant species is low. Highly transformed habitats such as ruderal sites and road and railway margins should be closely monitored to avoid spread of invasive species into habitats of a more natural character within the city.