Introduction

Invasive plant species have broad ecological and economic impacts in both natural and human-altered environments (Higgins et al. 1997; Leung et al. 2002; Vilà et al. 2011), and are among the main global threats to biodiversity (Vitousek et al. 1996; Pauchard and Shea 2006; van Kleunen et al. 2015). In particular, they alter the structure and diversity of native plant communities (Vitousek et al. 1996; Vilà et al. 2011; Benesperi et al. 2012), affect the productivity of native plant species (Chambers et al. 2007) and significantly disrupt the trophic structure of ecosystems (Levin et al. 2006; Heleno et al. 2008; Tallamy et al. 2010; Schirmel et al. 2016), with prolonged impacts on diversity at higher trophic levels (Spafford et al. 2013; Bezemer et al. 2014; Litt et al. 2014; van Hengstum et al. 2014).

The impacts of plant invasions on arthropod assemblages strongly vary among different taxa (Spafford et al. 2013; Bezemer et al. 2014; Litt et al. 2014; van Hengstum et al. 2014; Buchholz et al. 2015). Specialized herbivores or pollinators, evolutionarily bound to a small number of plant species (Traveset and Richardson 2006; Aizen et al. 2008; Moroń et al. 2009; Burghardt et al. 2010) or parasitoids (Simao et al. 2010), usually respond to plant invasions negatively, by decreasing in diversity or abundance (Degomez and Wagner 2001; Spafford et al. 2013; Litt et al. 2014). On the other hand, non-specialized pollinators (Bezemer et al. 2014), predators (Pearson 2009; Hartley et al. 2010) and detritivores (Standish 2004; Litt et al. 2014) are often unaffected by invasions, or their diversity and abundance may even increase in novel habitats.

Among invasive plants, the ecological consequences of woody invaders are particularly profound, due to their strong effects on native habitats (Richardson 1998; Hierro and Callaway 2003). Alien woody plants, by eliminating native species as a consequence of interspecific competition (Vilà et al. 2011; Benesperi et al. 2012), can decrease the food supply in the forest canopy (Litt et al. 2014; Reif et al. 2016; Hejda et al. 2017), or change the habitat structure and therefore affect the composition of the whole community (Harris et al. 2004; Pawson et al. 2010; Litt et al. 2014; van Hengstum et al. 2014; Buchholz et al. 2015).

Arthropods are among the most diverse groups of animals (Ødegaard 2000) and significantly contribute to trophic interactions (Mooney et al. 2010). Therefore, a deeper understanding of the impact of woody invaders on arthropods is important, as the impacts of invasive plants on organisms at lower trophic levels may have consequences for the functioning of the whole ecosystem (Heleno et al. 2008; Tallamy et al. 2010; Bezemer et al. 2014; Litt et al. 2014; Reif et al. 2016; Schirmel et al. 2016; Hejda et al. 2017). Even though there are some studies that focus on the effects of woody invaders on arthropods (e.g. Bezemer et al. 2014; Litt et al. 2014; van Hengstum et al. 2014; Buchholz et al. 2015, Schirmel et al. 2016), our knowledge is still incomplete, given the enormous diversity of arthropods, and their ecological traits and life history strategies. For example, we can predict that herbivores bound to different layers of vegetation may vary in response to changes in light conditions or stand structure (Harris et al. 2004; Pawson et al. 2010), but the studies available have mostly used coarse groups of arthropod communities, and more detailed relationships remain unclear. In fact, woodland arthropod diversity may be more affected by habitat structure than by variability in plant diversity (Gardner et al. 1995; Highland et al. 2013).

The main objective of this study was to compare the assemblages of nocturnal Lepidoptera between stands invaded by a widespread invasive tree, the black locust (Robinia pseudoacacia) and forest stands formed by native tree species. Impacts of invasive Robinia pseudoacacia have been documented for various kinds of organisms (Degomez and Wagner 2001; Cierjacks et al. 2013; Buchholz et al. 2015; Rocca et al. 2016; Vítková et al. 2017). Nocturnal Lepidoptera (further called “moths”) are a well-studied group of arthropods in Central Europe, with detailed knowledge of their ecology (Summerville et al. 2004; Pavlikova and Konvicka 2012), known direct links to vegetation structure (Highland et al. 2013) as well as with a high diversity of larval feeding strategies, life-histories and other ecological traits (Strong et al. 1984; Pierce 1995). These moths therefore represent excellent study organisms for testing the effects of plant invasions on groups of species defined by their (ecological) traits. In particular, we asked: (i) Do stands of the invasive R. pseudoacacia differ in habitat structure compared to stands of native trees? (ii) Do these stand types differ in the species richness, abundance and biomass of moths? (iii) Do the moth assemblages associated with these stand types differ in their ecological traits and could the alteration of habitat structure explain these potential differences?

Materials and methods

Focal invasive tree

Invasive black locust (Robinia pseudoacacia) occurs naturally in the southeast of the USA, where it represents an important part of early-successional forests, being eventually replaced by climax species (Boring and Swank 1984; Cierjacks et al. 2013). It was introduced to Europe at the beginning of the Seventeeth Century and has further spread worldwide (Cierjacks et al. 2013). At present, it is considered as one of the most widespread invasive species in Europe (Vítková et al. 2017). It was introduced to the Czech Republic at the beginning of Eighteenth Century (Slavík 1995), and was widely planted in warm areas, particularly on barren rocky slopes, for the stabilization of soil, and for wood and honey production (Vítková et al. 2017). Due to its nitrogen-fixing ability, it enriches habitats with nitrogen and supports the spread of nitrophilous herbs and shrubs (Benesperi et al. 2012; Vítková and Kolbek 2010; Vítková et al. 2017).

Study area and design

The fieldwork for this study was carried out in a forested lowland area of ca. 600 km2 (approximately between 49°56′N and 50°08′N, and 14°09′E and 14°26′E; 200–400 m a.s.l.) in central Bohemia, the Czech Republic, Europe (Fig. 1). This area is predominantly covered by stands of deciduous forests, human settlement, farmland and grassland. The forests are formed by native species of oak (Quercus spp.), hornbeam (Carpinus betulus), ash (Fraxinus excelsior), maples (Acer spp.) and limes (Tilia spp.). In the first half of the 20th century, large parts of this area were planted with R. pseudoacacia, which further spread spontaneously to the surroundings (Nožička 1957).

Fig. 1
figure 1

Map showing locations of the study plots (19 plots in the stands of Robinia pseudoacacia and 20 plots in native stands)

Full size image

We established 20 study plots (100 m × 100 m) in stands of native deciduous trees (dominated by Quercus spp.; henceforth “native stands”) and 19 study plots in stands dominated by the invasive R. pseudoacacia (comprising more than 95% of tree cover; henceforth “R. pseudoacacia stands”) (Fig. 1, Online Resource 1). These study plots were established at least 100 m from the forest edges and the minimum distance between adjacent plots was 500 m (Beck and Linsenmair 2000; Hanzelka and Reif 2015). Mixed stands with both alien and native trees were avoided.

This study focused on moths of the superfamilies Hepialoidea, Cossoidea, Zygaenoidea, Drepanoidea, Lasiocampoidea, Bombycoidea, Geometroidea and Noctuoidea. To sample moths in both stand types, we used 39 modified, automatic, portable light traps (Heaths 1965) with similar specifications as used by Brehm and Axmacher (2006). Moths were attracted to traps with two 8 W UV LED strip lights (total luminous flux of 400 lm in the wavelength range of 400–420 nm, powered by 7.2 Ah/12 V lead batteries) and then euthanized by evaporated chloroform. At each plot, a single portable trap was placed on the ground, approximately in the middle of the plot. All traps were exposed on the same night at the beginning of each month from April to November 2014, from dusk until dawn, when the weather was suitable (i.e. no strong wind and no strong rainfall) and the moon was quarter-sized at maximum. Such traps attract flying insects within a radius of a few tens of metres (Truxa and Fiedler 2012). Therefore, they should have attracted moths occurring almost exclusively within the area of the study plots, without sampling individuals from the surrounding habitats. Such a sampling design allowed us to cover all major phenological phases of moth species richness, throughout the part of the year when the moths were actively flying (e.g. Jonason et al. 2014; Tropek et al. 2014), and also to avoid short-term effects that may possibly affect the light trapping (Yela and Holyoak 1997).

Moth data and traits

The samples from the light traps were stored in paper bags and frozen at − 22 °C. In the laboratory, all moths from the target groups were identified to the species level. To measure their biomass (in terms of dry weight), the moths were dried in an oven at 80 °C for 12 h and then weighed on an analytical scale (van Langevelde et al. 2011). For each species, the evidence of its utilization of R. pseudoacacia as a host plant was determined from the literature (Kulfan 2012).

To analyse the possible effects of R. pseudoacacia resulting from differences in habitat characteristics, a species-traits approach was adopted (Pavlikova and Konvicka 2012). For each species, the following ecological characteristics were recorded (Table 1): (i) general traits, not specific to any particular stage of life cycle (voltinism, overwintering stage, preferred structure of habitat and vegetation layer), (ii) traits specific for the larval stage (length of larval development, diet specialisation and feeding guild) and (iii) traits specific for adults (wing span as a measure of body size and mobility, life span).

Table 1 Presumed relationships of ecological traits of the general, larval and adult stages of selected species of nocturnal Lepidoptera, according to the type of studied forest stands, dominated by either native trees (mainly oaks) or by the invasive Robinia pseudoacacia (RP)
Full size table

Habitat characteristics and environmental variables

To describe differences in the habitat structure of both R. pseudoacacia and native stands, 14 habitat characteristics expected to affect the diversity of arthropods (Strong et al. 1984; Gardner et al. 1995; Hartley 2002; Highland et al. 2013) were recorded in June 2014. At each plot, the age of the forest was recorded. Fallen (FALLEN TREES) and dead (DEAD TREES) trees were counted and other vegetation characteristics were estimated in three equal-sized subplots of 100 × 33 m, and mean values for each parameter were calculated (see Hanzelka and Reif 2016 for details). Specifically, by walking through the whole subplots, we visually estimated the percentage cover of herbs < 0.5 m height (HERB1), herbs > 0.5 m (HERB2), shrubs from 1 to 5 m in height (SHRUB), trees from 5 to 10 m (TREE1), trees > 10 m (TREE2), canopy cover (CANOPY) and clearings (CLEARINGS). Further we estimated the proportion of the number of trees with diameter at breast height (dbh) < 0.2 m (TREES); trees with dbh 0.2–0.5 m (TREEM) and trees with dbh > 0.5 m (TREET). We expressed the light conditions as the presence of a continuous canopy (CANYES/CANNO).

Finally, we considered six environmental variables known to affect arthropod communities (Novotny et al. 2015) expressed as proportions of the following land cover types in the surroundings of study plots: ARABLE—arable land; WATER—water bodies; ROCK; GRASS—grassland; URBAN—urban area; BROAD—broad-leaved forest; CONIF—coniferous forest. The proportions of these land cover types were estimated within a circular buffer of 500-m radius around each plot using ArcGIS version 10.2 (ESRI 2011).

Data analysis

To reduce the complexity of habitat and environmental data without substantial loss of information, the major dimensions of habitat structure and land cover characteristics of the R. pseudoacacia and native stands were determined by principal component analysis (PCA) in Canoco 5.0 (ter Braak and Šmilauer 2012). We ran two separate PCAs: one for vegetation structure and second for land cover characteristics. To determine the number of principal components, we used the screeplot method (Jackson 1993). Based on this criterion, we used the plot scores from the first two principal components of habitat structure (further called “VEG1” and “VEG2”) and land cover characteristics (further called “LAND1” and “LAND2”).

Since our data may have suffered from problems of spatial autocorrelation, we applied a method of generalized least squares (GLS) from the package “nlme” (Pinheiro et al. 2017) in all following univariate models. Geographic coordinates of plot centres were used to express the possible spatial effects, and different autocorrelation structures within the residuals (Gaussian, exponential, linear, rational quadratics and spherical) were compared. The parsimony of these models, as well as a model without residual autocorrelation, were assessed using the Akaike Information Criterion, AIC (Zuur et al. 2009). By comparing the AIC values, we selected the most appropriate autocorrelation structure. In all cases, the most parsimonious models (i.e. those with the lowest AIC value) turned out to be the models without spatial effects. Thus, we used linear models without accounting for spatial autocorrelation in further analyses.

To compare the habitat characteristics of native versus R. pseudoacacia stands, linear models were fitted with the principal components of habitat characteristics (VEG1 or VEG2) as respective response variables and the stand type (STAND: native trees or R. pseudoacacia) as the predictor.

To compare the species richness, abundance and biomass between the native and R. pseudoacacia stands, we used the number of moth species in each plot (SPECIES), number of all moth individuals per plot (INDIVIDUALS) and the total dry mass of moths per plot (BIOMASS; in grams) as the respective response variables, all transformed using the natural logarithm.

At first, a full linear model with all main predictors—stand type (STAND: native or R. pseudoacacia), LAND1 and LAND2 (principal components of land cover characteristics), VEG1 and VEG2 (principal components of habitat structure)—was constructed for each of the response variables (i.e. SPECIES, INDIVIDUALS, BIOMASS). Plots of the standardized residuals were checked against each continuous variable for possible polynomial trends. We thus added a quadratic term for VEG2 into the models. Interactions were not included, because there were no meaningful interpretations related to our hypotheses. In the next step, a multi-model inference framework was used (package “MuMIn”, Bartoń 2016) to obtain a minimum adequate set of predictors for each response variable. Due to the small sample size relative to the number of estimated parameters, the candidate models containing all possible predictor combinations were compared by AIC corrected for small sample sizes (AICc—Akaike 1974; Burnham and Anderson 2002). Models with ΔAICc (i.e. the difference between the AICc value of the focal model with the lowest AICc value) < 2 were selected as the best performing models. The predictors that appeared in these best performing models were considered as the minimum adequate set and were used for interpretations. All models were further validated for the assumption of normal distribution of errors, based on a visual inspection of the distribution of standardized residuals (Crawley 2013). All univariate models were fitted in the program R version 3.3.1 (R Core Team 2016).

To test if the stand types differed in the traits of the moths assemblages, redundancy analyses (RDA) were performed in Canoco 5.0 (ter Braak and Šmilauer 2012). However, this method does not take the geographic positions of study plots into account and thus its results may suffer from spatial autocorrelation in the data (Šmilauer and Lepš 2014). Therefore, we combined RDA with principal coordinates of neighbour matrices (PCNM) to account for spatial autocorrelation (Dray et al. 2006; Peres-Neto et al. 2006), following recommendations from the developers of this technique (Šmilauer and Lepš 2014).

In PCNM, the Euclidean distance matrix based on geographical distances of neighbouring sample plots was first calculated. This matrix was then processed by a principal coordinate analysis (PCoA) to obtain the spatial variables represented by respective PCoA axes (Šmilauer and Lepš 2014). Monte Carlo permutation tests (999 runs) were used to test the significance of each axis in the PCoA. From the PCoA output, we extracted the positions of each study plot along the significant PCoA axes (called “PCo scores”) and these scores were further used in all subsequent RDAs as covariate variables capturing the spatial information in the data.

In the next step, we fitted three RDA models, where each contained the functional traits as the response variables. The value of a particular functional trait for each plot was quantified as the number of all trapped individuals on a particular plot sharing an identical level of a given trait (e.g. SMALL body size; Table 1). These response variables were centred and standardized in all models. The first model (STAND model) included a single predictor, the stand type, and PCo scores as covariables. The second model (COVARIATE model) included land cover characteristics, LAND1 and LAND2, as predictors and PCo scores as covariables. The third model (STAND│COVARIATE model) included the stand type as a predictor and land cover characteristics and PCo scores as covariables.

In addition to RDA models we used variation partitioning (Peres-Neto et al. 2006) to distinguish the marginal, conditional and shared effects of the three groups of predictors—stand type (native/R. pseudoacacia), habitat structure (VEG1 and VEG2) and environmental variables (significant PCo scores from PCNM and land cover characteristics, LAND1 and LAND2)—on the distribution of the ecological traits of moths. Marginal effects are the effects of a given predictor variable (or a group of variables) without taking the other predictors into account; conditional effects quantify the effects of a given predictor variable after controlling for the effects of other predictors; shared effects are the effects shared between a given predictor variable and the other predictors (Šmilauer and Lepš 2014).

Results

Habitat characteristics of native and invaded forest stands

Native and invaded forest stands differed in habitat characteristics (Fig. 2a). The first PC axis, VEG1 (explaining 55.42% of the variation in habitat characteristics), reflected a gradient from older stands with taller trees and a more developed and continuous canopy to younger, open stands with smaller trees, a more developed shrub layer and a higher number of fallen trees (Fig. 2a). Native stands had lower VEG1 scores than R. pseudoacacia stands (t = − 9.075, p < 0.001). The second axis, VEG2 (18.96%), reflected a gradient from plots with a more developed lower herb layer and small area of clearings to plots with a more developed taller herb layer and larger area of clearings (Fig. 2a), and was not significantly different between the native and the R. pseudoacacia stands (t = − 1.933, p = 0.061).

Fig. 2
figure 2

Principal components analysis (PCA) of a habitat and b land cover characteristics between the 20 plots dominated by native trees and 19 plots dominated by R. pseudoacacia. Polygons indicate convex hulls separately bounding the native and R. pseudoacacia plots

Full size image

In case of the land cover characteristics, the first axis, LAND1 (63.45%), reflected mainly the gradient from a landscape with a large portion of broad-leaved forest to a landscape with a larger cover of urban area (Fig. 2b). The second axis, LAND2 (20.73%), reflected mainly the gradient from a landscape with a large share of coniferous forests to a landscape without coniferous forests.

Moth species richness, abundance and biomass

In total, 18,556 individuals of 384 moth species were captured (Online Resource 2), of which 346 species (mean ± s.e. [range] = 122 ± 27 [81–165]) were trapped in native stands and 304 species (90 ± 16 [61–120]) in R. pseudoacacia stands, with 266 species occurring in both stand types. Seventy eight species were more common in R. pseudoacacia and 164 species were more common in native stands. A total of 18 species that had been previously documented to feed on R. pseudoacacia were recorded, 15 of them in both stand types, two species only in native stands and one species only in R. pseudoacacia stands.

Relationships of moth species richness, abundance and biomass to the characteristics of the forest stands were estimated by linear models with performance assessed by AICc. Stand type was included in all except one of the best performing models (ΔAICc < 2) for all of the response variables (Table 2). Specifically, the species richness, the number of individuals, and the total biomass of captured moths were higher in native stands than in invaded stands (Table 3, Fig. 3). In addition, the best performing models for moth species richness, abundance and biomass also included VEG2 and the quadratic term of VEG2 (Table 2). The highest number of species and highest biomass were recorded in stands with intermediate values of VEG2 (Table 3a, c), i.e. with moderate proportions of clearings and both taller and shorter herbs in the understorey. Moreover, one model for moth abundance contained the effect of VEG1 (Table 3b), with increasing numbers of individuals towards stands with a closed canopy and less-developed shrub layer. Finally, the best performing models for moth species richness and abundance also included the effects of LAND2 (Table 2), with the number of species and individuals increasing towards stands surrounded by a higher coverage of coniferous trees (Table 3a, b). Some of the best performing models (m54 for moth abundance and m54 for moth biomass, see Table 2) contained “masquerading” variables and were thus not used for inference. Such variables are included among the terms of the best performing models, but do not improve the fit sufficiently to offset the penalty for their addition, compared to the more parsimonious models without this variable (Anderson 2008; Arnold 2010).

Table 2 Characteristics of the best models (ΔAICc < 2) relating (a) number of species, (b) number of individuals and (c) total dry biomass of moths to particular predictors
Full size table
Table 3 Estimates of the effects of particular predictors from the best models (ΔAICc < 2) on the number of species, number of individuals and total dry biomass of moths from the 20 plots in native forests and 19 plots dominated by the invasive Robinia pseudoacacia
Full size table
Fig. 3
figure 3

Comparisons of moth a species richness, b numbers of captured individuals and c total biomass between the 20 study plots with native trees and 19 plots dominated by R. pseudoacacia. Columns show means and bars show 95% confidence intervals

Full size image

Composition of moth communities

The stand type explained 16.8% of the variation of moth traits (Table 4), and its effect remained significant even after controlling for environmental variables (STAND│COVARIATE models; Table 4). Moth communities in the native stands were characterised by a higher presence of univoltine moths, which are specialists of forest habitats and are associated with the canopy layer, and by habitat generalists (Fig. 4a). On the contrary, forest-steppe moths, which are associated with more open habitats or herb and shrub layers, with more generations per season and with chrysalis as an overwintering stage, were more numerous in the R. pseudoacacia stands. Adult moths in the native stands were larger, more mobile and longer-living, while the stands of R. pseudoacacia predominantly supported moths with a faster life cycle and a higher proportion of short-living adults of smaller body sizes (Fig. 4b). With respect to larval feeding guilds, larvae feeding on herb and shrub litter or on thallus were more common in R. pseudoacacia stands (Fig. 4c). Other feeding guilds showed weaker responses to stand type.

Table 4 Results of redundancy analysis comparing the functional trait composition of nocturnal Lepidoptera between the 20 plots in native forests and 19 plots dominated by the invasive Robinia pseudoacacia
Full size table
Fig. 4
figure 4

Redundancy analysis (RDA) relating the a general, b adult and c larval-stage ecological and biological traits of nocturnal Lepidoptera to the two types of stands studied: 20 plots of native stands and 19 plots dominated by the invasive Robinia pseudoacacia. Spatial autocorrelation and significant environmental variables were included as covariables in all these models. See the “Methods” section for plot characteristics and details on traits

Full size image

Shared and marginal effects of all three groups of predictors (stand type, habitat structure, environmental variables) explained a substantial proportion of the variability in moths functional trait composition (Table 5). While the conditional effects of stand type and habitat characteristics were rather weak relative to the marginal effects (8.7 vs. 21.6% and 4.6 vs. 19.8%, respectively; Table 5), the effects of land cover characteristics, controlled for the spatial positions of plots, explained nearly half of the variability compared to the marginal effects (10.6 versus 22.7%; Table 5).

Table 5 The results of variation partitioning showing the shared, marginal and conditional effects of the groups of environmental variables on the functional trait composition of nocturnal Lepidoptera
Full size table

Discussion

Stands formed by native trees differed significantly in their habitat structure from stands of the invasive R. pseudoacacia. The native stands were mostly formed by taller trees with a closed canopy and with a higher cover of shorter herbs. On the other hand, R. pseudoacacia stands were characterized by a more open canopy, with a higher coverage of shrubs and taller herbs in the understorey. Similar to our study, Buchholz et al. (2015) reported a more developed understorey vegetation in R. pseudoacacia stands compared to stands of the native birch Betula pendula (with a significant effect on the cover of herbs and a marginally significant effect on the cover of shrubs). These effects are probably caused by the nitrogen-fixing ability of R. pseudoacacia enriching the soil in nitrogen (Boring and Swank 1984; Cierjacks et al. 2013; Vítková et al. 2017) and by the more open canopy of its stands, which allows for a better transmission of solar radiation into the understorey and consequently supports the growth of herbs and shrubs in the understorey layer.

Based on the light-trapping data, we found significantly lower total species richness, abundance and biomass of nocturnal Lepidoptera in stands dominated by the invasive R. pseudoacacia. This is in accordance with the general pattern of decreasing diversity, abundances or biomass of herbivores caused by invasive plants (Liu and Stiling 2006; Gerber et al. 2008; Spafford et al. 2013; Litt et al. 2014; van Hengstum et al. 2014; Schirmel et al. 2016). Similarly, Degomez and Wagner (2001) found in northern Arizona a nearly 30% loss of species diversity in stands of non-native R. pseudoacacia in contrast to stands of native Robinia species. However, some of the studied groups, e.g., the hyperdiverse Hymenoptera and Diptera, did not exhibit losses in diversity (Degomez and Wagner 2001). Also, other studies comparing the diversity of native and R. pseudoacacia stands did not find differences in species diversity of predators (Buchholz et al. 2015) or saproxylic beetles (Rocca et al. 2016). Based on these findings and on the results of our study, it seems that the response of arthropods to the invasion of R. pseudoacacia depends on their feeding strategy, with prevailing negative effects on herbivores.

The lower species richness and lower number of individuals observed in R. pseudoacacia stands could be explained by the paucity of canopy moths, which are also mostly leaf-chewing herbivores (Degomez and Wagner 2001). Despite three centuries of occurrence in Central Europe (Slavík 1995), the spectrum of species able to feed on R. pseudoacacia remains limited (Kulfan 2012). Higher abundances of canopy dwellers in the native stands could also explain the higher proportion of moths overwintering in the egg stage found in the native stands, including polyphagous forest pests with generally higher population densities (Alford 2000). In contrast to understorey species, canopy species are often strongly associated with spring leaf germination (Van Asch and Visser 2007; Hikisz and Soszyńska-Maj 2015), when leaves are more palatable. Since they are probably not able to feed on the alien R. pseudoacacia, they lack a suitable feeding niche in canopies dominated by this tree. Differences in the number of leaf-chewing moths between R. pseudoacacia stands and native stands could also affect the frequency distribution of adult body sizes, because species restricted to the canopy of native trees tend to be larger in body size (Heleno et al. 2008). Thus the lack of canopy species in R. pseudoacacia stands can also explain lower total biomass of moths in these stands.

The species diversity and the number of individuals were also weakly positively affected by the proportion of coniferous stands in the surroundings of the study plots (represented by LAND2). This is in accordance with the known effects of woody plant diversity in the surroundings of traps on the diversity and abundance of moths (Novotny et al. 2015). Even among forest species specialized on coniferous trees, larger and mobile moths can be found (e.g. larger geometrids, hawkmoths, lappets), dispersing occasionally to deciduous stands and thus increasing the total species diversity and abundance.

Our results also showed that aspects of habitat structure, not accounted for by the distinction between native and invaded stands, had important effects on species diversity, abundance and total biomass. Specifically, sites with an intermediate proportion of clearings and an intermediate proportion of lower and taller herbs in the herb layer (the quadratic term of VEG2) had the highest number of species and individuals and the highest biomass of moths. We also detected the direct effects of vegetation structure on the functional composition of moths. Therefore, vegetation structure plays an important role in moth community assembly.

The moth assemblages of native forests were only partly formed by forest canopy specialists. Another guild occurring more frequently in the native stands were generalists, without distinctive habitat specialization. This is not consistent with some studies on ubiquitous species (Yoshioka et al. 2010, 2014), showing a higher abundance of generalists in invaded habitats. However this discrepancy may be due to the fact that those studies were conducted in non-forest habitats while our research was performed in forest stands. A majority of generalists in our study were migrants or pests with good dispersal ability (Slade et al. 2013). Such species probably disperse more easily through the more permeable native stands, formed by tall trees and without a well-developed shrub layer, than through stands of R. pseudoacacia, with a dense understorey. The more complex structure of invaded forests may therefore represent a dispersal barrier for insect habitat generalists (Barbaro et al. 2005). Similarly, in contrast to predictions and results showing a higher occurrence of diet specialists in native stands (Liu and Stilling 2006; Burghardt et al. 2010; Litt et al. 2014), we found no difference in preferences for stand type in the herbivore monophages and oligophages. This is probably because the loss of canopy diet specialists in R. pseudoacacia stands is compensated by dietary specialists gained in the better developed understorey.

Many studies on detritivores in invaded habitats have shown that the diversity or abundances of detritivores is higher in invaded stands than in native stands due to the higher amount of ground litter and decaying vegetation in non-native vegetation (Standish 2004; Levin et al. 2006; Litt et al. 2014). Interestingly, we found that moths with larvae feeding on litter leaves of herbs or shrubs were more common in stands of R. pseudoacacia. This may be related to the higher cover of shrubs and taller herbs, dominated by native plant species (Hejda et al. 2017), in the R. pseudoacacia stands.

It is interesting that moths with faster life-cycles (i.e. those having shorter larval development, shorter adult lifespans and more generations per season) occurred more frequently in R. pseudoacacia stands. This may be caused by a warmer and drier microclimate in these stands because leaves of R. pseudoacacia, unlike the leaves of native trees, rotate during strong summer heat to be less exposed to solar radiation (Xu et al. 2009), making them less effective in buffering heat stress in the understorey than native trees.

Overall, moth assemblages in R. pseudoacacia stands were similar to those of open-forests or forest-steppe habitats with better light conditions, but lacked canopy species, while forest and canopy dwellers dominated in native stands.

Conservation implications

The lower moth species richness in stands dominated by the invasive R. pseudoacacia indicates that this habitat does not favour Lepidopteran species richness in central European forests. Moreover, we did not record any moth species of conservation concern (sensu Farkač et al. 2005) in the invaded stands, while several such species were recorded in the native stands. Therefore, we suggest that the further spread of this invasive tree should be prevented and its eradication from sites of conservation concern should be prioritized.

At the same time, we found remarkable differences in the proportions of various ecological groups of moths between the native and invasive stands, which were likely caused by differences in habitat structure. Specifically, the native forests had more closed canopies and a less developed understorey than the studied invasive stands. Due to the higher light availability and well-developed understorey vegetation (Buchholz et al. 2015), stands of R. pseudoacacia resembled open forests, which are among the most threatened and vanishing habitats in Europe (Miklín and Čížek 2014). Therefore, from the perspective of moths restricted to the forest understorey, forest-steppe and open habitats, the conservation potential of the studied native stands with the currently prevailing vegetation structure is limited. The second message from our study for the conservation of moths in central European lowland forests is therefore the need to increase the heterogeneity of the habitat structure and canopy openness of native forests (see also Sebek et al. 2015). Even though the composition of native stands supports the diversity of some functional groups of moth fauna due to long-term adaptations (e.g. canopy feeders), the diversity of moth fauna and its functional guilds may be limited by the large-scale homogeneity of native stands, with closed canopies and relatively homogenous age structure. In this respect, the management of native lowland forests may consider active measures (planned clearings, coppicing, disturbances, grazing, creating small-scale gaps or selective cutting) to promote the patch dynamics of new versus old stands, as well as closed canopies versus more open areas (Merckx et al. 2012; Pavlikova and Konvicka 2012; Sebek et al. 2015).