Introduction

Two species of the giant knotweeds (Reynoutria, syn. Fallopia, Polygonum), which are members of the Polygonaceae family, are highly invasive and are recognised as noxious weeds in areas of invasion. One species, Fallopia japonica, is native to Japan, Taiwan and northern China (Beerling et al. 1994). The other species, Fallopia sachalinensis, is native to the southern part of Sakhalin Island, the southern Kurile Islands, and the Japanese islands Hokkaido and Honshu (Sukopp and Starfinger 1995). Within their invasion area in Europe and North America both species are most common on sites disturbed by human activities such as industrial areas or roadsides situated in open grassland and forests as well and on sites with natural soil erosion such as river banks (Beerling et al. 1994, Sukopp and Starfinger 1995, Weston et al. 2005). Fallopia sachalinensis additionally invades semi-natural riparian sites (Pyšek and Prach 1993).

Stands of both Reynoutria species are remarkable for their height, leaf area and production values as summarised by Sukopp and Starfinger (1995): Reynoutria sachalinensis reaches a height of 2–4.5 m. The biomass can exceed 200 t/ha equalling an energy content of ∼12,000 MJ/ha, which is above any other herbaceous vegetation. These extreme values are reached by daily height increments of up to 15 cm in early summer. The value of leaf area index is 21.0 which exceeds that of dense forests. The total cover of Reynoutria in areas of invasion is in the range of 95–100%. Native vegetation is mostly replaced. The only species able to persist within Reynoutria stands are spring geophytes (Beerling et al. 1994, Kappes et al., 2007). Outcompeting the native vegetation may be enhanced by allelolechemicals such as the anthraquinone compounds emodin and phycion which were isolated from roots of R. sachalinensis (Inoue et al. 1992).

The most common concern related to invasion are negative interactions between invaders and native species (Crooks 2002). In Europe, Reynoutria often establishes in herb-rich stands dominated by Urtica dioica (Sukopp and Starfinger 1995). When displacing Urtica stands, Reynoutria invasion is related to a decrease in soil pH and changes in habitat structure, including a reduced stem density (Kappes et al., 2007). In contrast to Urtica-dominated stands, Reynoutria stands provide a continuous detritus supply. Reynoutria continuously produces leaf litter as it drops the shaded leaves in the interior of the stands. Parts of the stem litter persist at least to the next growing season (Beerling et al. 1994).

Plant diversity is directly related to diversity of herbivore insects (Siemann 1998). Moreover, indigenous species such as nettles harbour a higher number of phytophagous insects (Davis 1983; Zabel and Tscharntke 1998) than the invasive Reynoutria (Zimmerman and Topp 1991; Beerling and Dawah 1993). Generally, assemblages of aerial insects associated with invasive vegetation considerably differ from those of native vegetation as invaders support lower abundance and species richness (Herrera and Dudley 2003).

Changes in the physical structure, nutrient cycling, productivity and food webs are substantial effects on ecosystems (Vitousek 1990; Crooks 2002; Zedler and Kercher 2004). These ecosystem changes should also affect ground-dwelling species of different trophic levels, although many of these arthropods are not as closely associated with vegetation as are aerial insects (Herrera and Dudley 2003). Among insects, beetles are the taxon with the highest species richness and include numerous ground-dwelling species with diversified feeding habits. Thus, beetles should be good indicators of the effects of plant invasions.

We studied ground-dwelling beetle assemblages and selected six pairs of Urtica dominated stands and invading Reynoutria stands growing in two different locations. One is a semi-natural softwood forest in which disturbance originates from natural processes such as erosion and sedimentation. The other is an artificial embankment in which human activities transformed late successional habitats into ruderal habitats. The goal of our study was to analyse changes in species richness, abundance and rarefaction diversity and in functional groups associated with Reynoutria invasion. We hypothesise that (1) the invader supports a lower species richness of ground-dwelling beetles, that (2) Reynoutria invasion especially affects herbivores and detritivores, and that (3) the impact of Reynoutria invasion on beetle assemblages is the same in the semi-natural softwood forest and the ruderal embankment because both habitats are subjected to disturbance by floodings.

Materials and methods

Study sites

We selected six sites in two different locations, a semi-natural softwood forest and a ruderal embankment of a canalised river (Table 1). The semi-natural softwood forest is located in the floodplain of the River Rhine at Düsseldorf–Urdenbach (sites 1–3). The herb layer is dominated by U. dioica stands, into which species such as Angelica silvestris, Calystegia sepium, Cardamine amara, Galium aparine, Filipendula ulmaria, Lycopus europaeus, Ranunculus ficaria and Impatiens species are interspersed. The herb layer is invaded by both Reynoutria sacchalinensis and Reynoutria japonica. The semi-natural softwood forest is readily flooded after heavy summer rains and may be inundated for a couple of days to some weeks during floods of the River Rhine in the winter.

Table 1 Charactersitics of the two locations Urdenbach and Rheindorf
Full size table

The ruderal embankment is characteristic for the severely disturbed floodplain of the lower River Wupper close to Leverkusen–Rheindorf (sites 4–6). The herb layer is dominated by Urtica stands into which a few other species such as C. sepium, G. aparine and Lamium purpureum, and some grasses such as Agropyron repens are interspersed. The herb layer is invaded by R. japonica. At Rheindorf, inundation of the three sites occurs only during winter floods.

Reynoutria has invaded the Urtica-dominated stands during the last decades. As Reynoutria established within the native Urtica-dominated stands, we used a paired design and compared the remaining Urtica stands with the stands of the invader Reynoutria in each of the sites.

Sampling

We used pitfall trapping to study the beetle fauna. We are aware that pitfall traps record activity densities rather than true abundances (Luff 1975; Baars 1979). As a consequence, abundance of hunting predators is overestimated when compared to other feeding guilds such as detritivores. Still, pitfall traps are a standard method in terrestrial ecology (Antvogel and Bonn 2001; Brose 2003). In each stand, five pitfall traps (white cups, diameter 85 mm, 2.5% formalin with detergent added) were placed at a distance of 5 m from each other and from the edge of the vegetation stand. Due to the invasion history, some stands of Urtica and Reynoutria were not continuously distributed. Instead they exhibited small corridors of the remaining Urtica or of the invading Reynoutria species. In these cases a stand comprised several patches to meet the above mentioned criterion of distances between pitfall traps and vegetation edges. Urtica patches covered areas between 100 and >3,000 m²; Reynoutria patches covered areas sized 100–700 m².

The 60 pitfall traps were checked every fortnight. In early July, early August and late September heavy rainfalls resulted in the inundation of many pitfall traps in the semi-natural softwood forest. As beetles were lost, we also omitted the corresponding series of the other location for analyses. Data thus originated from 02 to 30 June, 14 to 28 July, 11 August to 22 September and 06 to 20 October 2005 for both locations.

Systematics of the beetles follow Freude et al. (1964–2004). Feeding guilds and ecological preference groups were assigned according to Koch (1989–1992).

Statistics

We pooled the data of the sampling dates (as described above) to yield higher numbers. Consequently, the analysis is based on five replicates per stand.

Rarefaction analyses were employed to standardise species richness for the number of individuals. We performed separate analyses for the two plant species and the two locations. Rarefaction diversity was calculated with EstimateS, Version 7.5.1 (Colwell 2005). To test if differences between the rarefaction curves were significant, we calculated z-values and one-tailed P-values based on the mean and the standard deviation of the S obs value at the lowest number of individuals (with n = 15 samples).

Additionally, the rank order of species was plotted to visualise the influence of the vegetation on the relative abundance in each location. The threshold was set to a total of ≥20 individuals within each location.

We performed one-way ANOVAs for assessment of the effect of location on the beetle assemblages of the Urtica-dominated stands. Also, we employed one-way ANOVAs to test the effect of vegetation on the beetle assemblages for the two locations separately. In a further step, we performed a nested ANOVA. This method was chosen because of the hierarchical structure of the survey: each pair of vegetation stands was nested in site, and three sites were nested in each of the two different locations. The nested ANOVA was used to simultaneously test the influence of location, site and vegetation on the number of species and individuals. The nested ANOVA was run including the five subsamples per stand to account for within-stand variability. The variances of the abundances were stabilised using log(x + 0.5) transformation in the ANOVAs. The ANOVAs were performed with the programme package SPSS 11.0.

To further characterise the assemblages, we calculated the qualitative Sørensen indices and the quantitative Morisita–Horn indices (using EstimateS 7.5.1) for each combination of pitfall traps and entered the similarity matrices into an analysis of similarity (ANOSIM, programme package Primer 5.0). For testing, we classified four groups. Each group comprised the 15 pitfall traps per vegetation stand of each location. Relations between beetle assemblages were visualised in a detrended correspondence analysis (DCA). For the DCA, we pooled the five replicates of each stand. We only regarded those 96 species that had an abundance of ≥5 individuals. Abundances were log(x + 1)-transformed. Detrending was performed using second-order polynominals. The analysis focused on inter-species distances, using Hill’s scaling to account for the long moisture gradient (Lepš and Šmilauer 2003). The DCA was performed using CANOCO for Windows 4.0.

Results

Differences between locations

In the two locations, a total of 13,244 beetles from 218 species were recorded. We collected 6,189 and 7,055 individuals in the semi-natural softwood forest and in the ruderal embankment, respectively.

Before dealing with the effects of the invasive Reynoutria, we compared the beetle assemblages from the Urtica-dominated stands of the two locations. The two locations differed in species assemblages. For example, some hygrophilus species, such as Patrobus atrorufus, Carabus granulatus and Agonum micans were trapped in high numbers in the semi-natural softwood forest (Table 2). The ruderal embankment of the canalised river was characterised by species such as Drusilla canaliculata, Pterostichus melanarius, Barypeithes pellucidus and Phosphuga atrata (Table 2).

Table 2 Species with a total of ≥40 individuals
Full size table

A one-way ANOVA showed that the pitfall traps on the ruderal embankment caught significantly less carabid individuals (F = 38.1, P < 0.001) and staphylinid species (F = 27.8, P < 0.001), but more staphylinid individuals (F = 85.5, P < 0.001) than those in the semi-natural softwood forest. Also, there were less detritivorous individuals (F = 19.4, P < 0.001) but more herbivorous individuals (F = 28.3, P < 0.001) in the pitfall traps from the ruderal embankment.

Differences between Urtica and Reynoutria stands

Irrespective of location, both the total number of species and the total number of individuals were higher in the Urtica-dominated stands than in the Reynoutria stands (Table 2). A one-way ANOVA revealed that differences between vegetation stands were more pronounced on the ruderal embankment than in the semi-natural softwood forest. In both locations, there were less staphylinid individuals per pitfall trap in Reynoutria stands than in Urtica-dominated stands (softwood forest: F = 5.2, P = 0.031; ruderal embankment: F = 10.1, P = 0.004). On the ruderal embankment, there were also less beetle species (F = 6.2, P = 0.019) and less staphylinid species (F = 8.3, P = 0.008) per pitfall trap in Reynoutria stands than in Urtica-dominated stands.

Rarefaction analyses showed that overall beetle diversity was reduced in Reynoutria stands in both locations (Fig. 1, Table 2). However, the negative effect of Reynoutria stands was approximately twice as high in the softwood forest as on the ruderal embankment. At 2,941 individuals, the difference between the means was −18 ± 3 in the softwood forest (z = −6.2, P < 0.001), whereas the difference between the means was −8 ± 3 on the ruderal embankment (z = −2.9, P = 0.002).

Fig. 1
figure 1

Rarefaction analysis on the beetle fauna (S obs with 95% confidence interval, 50 iterations) of the Urtica-dominated stands and the Reynoutria stands of the semi-natural softwood forest and the ruderal embankment

Full size image

In both locations, we found stand-specific beetle assemblages. Comparing rank orders of beetles from the semi-natural softwood forest visualised that the Urtica stands were preferred by A. micans, Tachinus rufipes, Phosphaenus hemipterus, Acrotichis intermedia and Ocypus melanarius (Fig. 2a). The Reynoutria stands were favoured by Limodromus assimilis, Omalium rivulare, Pterostichus niger, Ancyrophorus flexuosus and Nebria bevicollis. Rank orders of beetles from the ruderal embankment illustrate that the Urtica stand were preferred by Tachinus signatus, A. micans, Poecilus versicolor, C. granulatus, Silpha tristis and Cidnorrhinus quadrimaculatus (Fig. 2b). The Reynoutria stands were favoured by L. assimilis, P. melanarius, B. pellucidus, Ocypus brunnipes, Omias mollinus (Boh. 1834) and Latridius nodifer (Westw. 1839). The significance levels for the species with ≥40 individuals are given in Table 2.

Fig. 2
figure 2

Comparison of the ranks of the most abundant species (≥20 ind.) in the Urtica-dominated stands and the Reynoutria stands of a the semi-natural softwood forest and b the ruderal embankment. Deviations above or below the line indicate a higher prevalence in Urtica-dominated stands or Reynoutria stands, respectively. Strongly deviating species are identified with genus names abbreviated to the first two digits

Full size image

In total, we observed a significanty increased number of detritivores (χ² = 5.7), but decreased numbers of predators (χ² = 111.9) and herbivores (χ² = 20.5) in the Reynoutria stands (Fig. 3). The number of fungivores did not differ between vegetation stands (χ² = 2.4).

Fig. 3
figure 3

Total abundance of the feeding groups in the two vegetation stands. The significance level of the χ²-test is indicated as follows: n.s.P > 0.05; *P ≤ 0.05; ***P ≤ 0.001

Full size image

Combined effects of location and vegetation

In a further step, we analysed the influence of location, site and vegetation in a nested ANOVA. Location and site usually had the most significant effect on the distribution pattern of the total number of individuals, species, feeding guilds and selected taxa (Table 3). However, there were three examples in which vegetation had the strongest influence. The two rove beetles T. rufipes and Atheta fungi preferred Urtica stands. Also, xerophilous ground beetles showed a high activity in the Urtica dominated stands of the ruderal embankment. In this location, xerophilous individuals made up 28% of the total number of ground beetles in the Urtica stands but only 5% in the Reynoutria stands. In the semi-natural softwood forest, the proportion of xerophilous beetles was <2% in both vegetation stands.

Table 3 Nested ANOVA on the influence of location (loc), site and vegetation (veg) on the beetles
Full size table

We found eight examples in which vegetation exhibited the second strongest influence: the species numbers and individuals of all Coleoptera, the three feeding groups of predators, fungivores and detritivores, the two ground beetles L. assimilis and A. micans, and the species numbers of Staphylinidae (Table 3). In seven cases, numbers were higher in Urtica stands. The only exception were detritivorous beetles that occurred in higher numbers in Reynoutria stands. The proportion of detritivorous beetles was about twice as high in the Reynoutria stands than in the Urtica-dominated stands (3 and 1.8%, respectively).

The qualitative Sørensen indices confirmed that species composition differed both between vegetation stands and between locations (ANOSIM, 999 Permutations: all P ≤ 0.002). Similarly, the quantitative Morisita–Horn index confirmed dissimilarities between locations (all P ≤ 0.001) and between the vegetation stands on the ruderal embankment (P = 0.003). However, Morisita–Horn indices could not separate between the vegetation stands of the softwood forest (P = 0.104).

The specific assemblages of the locations and the plant stands were illustrated in a DCA (Fig. 4). The Eigenvalues of the first and the second axis were 0.32 and 0.11, respectively. The first two axes cover 50.2% of the total variance. The Eigenvalues of higher axes were lower than 0.07 and only marginally contributed to the entire model. Figure 4 shows that the gradient described by axis 1 separated the beetle assemblages of the two locations. Axis 2 describes a gradient that separates the beetle assemblages from the two plant species.

Fig. 4
figure 4

Biplot of the DCA of the beetle fauna from the Urtica-dominated stands (U) and the Reynoutria stands (R) of the semi-natural softwood forest (sites 1–3) and the ruderal embankment (sites 4–6). The circles comprise the vegetation stands of each location. The black dots represent the species scores (96 species with ≥5 ind.)

Full size image

Discussion

Differences between locations

Many beetles species are highly specialised and thus are valuable indicator species for habitat quality assessment (Boháč and Fuchs 1991; Boscaini et al. 2000; Bonn and Schröder 2001). As a consequence, the assemblages of the semi-natural softwood forest and those of the ruderal embankment were well separated. The habitat preferences of the dominant species reflect the differences between the locations. The prevalence of hygrophilous species is typical for close to natural wetland conditions (Günther and Assmann 2005). The hygrophilous ground beetle P. atrorufus that dominated the assemblages in the semi-natural softwood forest prefers moist floodplain forests, whereas the myrmecophagous rove beetle D. canaliculata commonly inhabits dry ruderal places (Koch 1989–1992). Thus, the severe transformation of the floodplain along the lower River Wupper is mirrored in the beetle assemblages. The effects of site are due to the small-scale heterogeneity that is typical of floodplain habitats (Robinson et al. 2002). Small-scale habitat preferences have for example been found for soil moisture and ground beetles (Antvogel and Bonn 2001) and the amount of detritus and rove beetles of the genus Stenus (Betz 1998).

Differences between Urtica and Reynoutria stands

Differences between vegetation stands resulted from different aspects of bottom-up effects. Food quality and quantity are known to structure insect assemblages (Strong et al. 1984). Cidnorhinus quadrimaculatus, for example, solely feeds on Urtica (Lohse 1983). The preference of the malacophagous silphid P. atrata for Urtica stands may be because snails are less abundant in Reynoutria stands (Kappes et al., 2007). Also, Reynoutria litter harbours a distinct microfungal assemblage (Beerling et al. 1994). These fungi may be of low palatability for species such as T. signatus and Acrotrichis intermedia that are at least partially fungivorous (Koch 1989–1992).

The effects on species richness, abundance and assemblage similarities were more pronounced on the ruderal embankment. We thus assume that the two vegetation stands strongly differ in physical characteristics especially in disturbed habitats. Microclimate is one factor known to influence the distribution pattern of species. The higher incidence of silvicolous species such as L. assimilis and O. brunnipes and the lower abundance of xerophilous individuals suggest that the microclimate in the Reynoutria stands of the ruderal embankment partially mimics that of forest habitats.

Theories on biodiversity predict a higher number of animal species, and a higher diversity in species-rich vegetation (Siemann 1998; Siemann et al. 1998). Our results on the lower beetle species richness and rarefaction diversity in the monospecific Reynoutria stands are in accordance with these theories. Reynoutria invasion consequently poses a threat to the species-rich Urtica-dominated floodplain habitats. According to Siemann et al. (1998) there is no effect of plant diversity on arthropod abundance. However, we also found negative effects of Reynoutria stands on total beetle abundance and both herbivores and predators.

Detritivore beetles were the only feeding group that had a higher abundance in Reynoutria stands. Similarly, the relative share of the detritivorous diplopods and isopods increases in Reynoutria stands (Kappes et al., 2007). Detritivores seemed to have profited from the increased productivity and the resulting high-detritus supply throughout the year. Our results on detritivore abundance also coincide with those of Levin et al. (2006), who observed a shift from an algae-based food web to a detritus-based food web following codgrass invasion in a coastal wetland. Obviously, highly productive invaders such as Reynoutria alter link strengths in trophic cascades shifting primary producer-based food webs to detritus-based food webs.