1. Introduction
Landscape changes caused by human activities are the main cause of the current biodiversity crisis [
1,
2,
3,
4]. The scale and extent of these activities create specific anthropogenic ecosystems [
5]. On the other hand, land-use diversity is an important environmental factor that promotes species richness treated as a measure of local alpha diversity. However, understanding processes that shape the wider measure of species composition among different sites (beta diversity) remains a fundamental challenge for ecologists [
6]. Assessing the relationship between land use and both measures (alpha, beta) of regional diversity can affect our understanding of global biodiversity patterns [
7].
Agricultural landscapes contain areas of various land use and habitats. There is a lack of research on the responses of animals to habitat changes in such landscapes and whether their habitat requirements impact those responses filtering the community. The concept of environmental filtering assumes that the occurrence, function, and survival of a species at a particular site is driven by its tolerance or adaptation to environmental conditions [
8,
9], and various species differ in this tolerance. The measured variation in species composition between habitats (or sites) is the definition of beta diversity [
10]. Assessing beta diversity can be understood as studying variation in species among sites linking biodiversity at local scales (alpha diversity) and the broader regional species pool (gamma diversity) [
11,
12]. Alpha measures how diversified the species are within a site, whereas beta measures how diversified the sites are in species composition within a region [
11,
13,
14]. Beta diversity has become a sophisticated concept [
14,
15,
16].
Beta diversity can be assessed through the extent of species compositional differences between sites to detect the assembly mechanisms [
10]. The division of beta diversity into two components, turnover and nestedness, is recommended. Turnover assumes that species are replaced by different ones comparing one site to the next (i.e., species replacement), whereas nestedness patterns appear when species loss or gain causes species-poor sites to resemble a strict subset of species-rich sites [
10,
17,
18]. It is worth emphasizing that replacement and richness difference or nestedness indices should be enriched with a careful ecological interpretation [
18]. Turnover may result from species endemism at various spatial scales and does not take into account that similar species may be spatially exclusive due to their properties or have specific range boundaries [
19]. Nestedness may result from local extinctions or colonization [
20]. Therefore, spatial variation in the composition of species must take into account the temporal aspect, because the current community is only its temporary version and may be the result of deterministic and/or stochastic processes [
21]. Deterministic processes result from habitat properties (e.g., habitat filtering or competitive interactions for resources) in determining the composition of local communities [
10]. Stochastic processes are local colonization or extinction that are the responses of species to processes outside the studied environment [
21].
In the case of mobile animals such as birds, which respond rapidly to changes in habitats, examining nestedness and turnover separately may be misleading. For example, species nestedness accounted for a small proportion of bird beta diversity in Sri Lanka’s biodiversity hotspot. Bird community assembly was, on the other hand, strongly influenced by climate at a surprisingly small scale [
22]. The filtering effect of vegetation is sometimes independent of species losses and replacements, although often remains a key predictor of their beta diversity [
9]. In line with this phenomenon, vertically developed vegetation affected bird assemblages through niche partitioning from ground to tree canopies in forests, and the largest independent effect concerned canopy height [
23,
24]. Although individual layers of the forest contain different bird assemblages, it is difficult to expect that this type of variability of species in space reflects beta diversity, because it is located in one type of habitat. In contrast, the positive relationship between habitat heterogeneity [
9] and beta diversity is highly detectable in studies investigating the environmental/habitat heterogeneity hypothesis. Since decisions and the range of occurrence of such small taxonomic groups as species may be the result of habitat filtering rather than relationships between species at different sites following the idea of beta diversity, it seems reasonable to use the concept of beta diversity when comparing broader and ecologically related groups, e.g., several species choosing a similar habitat.
Investigating determinants of the diversity of groups is important, as in a biodiversity crisis, not only are specific species extinct, but entire communities associated with specific environments are at risk. A large number of studies explored bird diversity in urban areas [
25], agricultural lands [
26,
27], and forests [
28]. Biodiversity loss has been a disproportionally ongoing problem in open landscapes such as grasslands, agricultural lands, and steppes in Europe in the last few decades [
29]. Such declines are believed to be driven by agricultural intensification and subsequent habitat disturbances [
30] as well as by the abandonment of agricultural land [
31]. The relationships between habitats and community diversity in agricultural landscapes may be critical to the survival of species and biodiversity across a wide range. However, the patterns of population changes were hardly detectable when based on species using agricultural lands without more precise habitat discriminations [
32]. This is the reason why the present article examines the diversity of birds sorted into assemblages connected with the vegetation layers from the ground to the tree canopy [
23,
24]. This research investigated changes in the following bird assemblages: ground/herb dwellers, bush foragers, ecotone birds, and tree foragers. This was in line with the fact that the vertical composition of vegetation located in a mosaic of habitats can be a driver of the birds’ beta diversity patterns [
23,
24].
However, bird surveys comparing paired sites, for example, controlled vs. disturbed, in the same habitats and landscape heterogeneity are not sufficient to understand the patterns of bird diversity change driven by this disturbance because these paired sites are in some way selected. It is impossible to determine which bird species do not reach the site because they do not appear there due to the location or heterogeneity. Therefore, birds at reference sites (easily available for the studied bird assemblages, although not exposed to disturbance, i.e., far from the landscape heterogeneity of the paired sites) were analyzed with the expectation that they constituted a wider bird community than the paired sites.
As mentioned, agricultural lands are exposed to various anthropogenic disturbances, such as plant invasions. This research evaluates the taxonomic filtering of bird assemblages along a gradient of plant invasion-altered agricultural lands. The focal invaders were Caucasian hogweeds or
Heracleum sp., which were once introduced to be used as crops. These invaders, spreading unpredictably and reaching four meters in height, were assumed to be independent from the turnover and nestedness patterns in line with the fact that in other research, the largest independent effect concerned canopy height [
23,
24]. It has not been studied whether and how invasive plants affect a bird community in the context of the beta diversity concept. The aim of this article is to fill this knowledge gap.
The selected invaders resemble bushes in physiognomy and reach large sizes, so it was predicted that by degrading the habitat, they may have had a significant negative impact on the birds from all assemblages, but may have created new space for bush foragers or ecotone species using bushes. In line with the idea of beta diversity, the composition of these assemblages can differ on invaded sites when compared to uninvaded ones. Assuming that the more individuals there are in an assemblage, the more species there are in its composition, variability was also assessed in the bird abundances of individual assemblages between the sites. This was intended to investigate the possible problem of smaller differences between the sites in alpha and beta diversity in the more widely represented assemblages.
3. Results
During this research, the author noted 80 bird species actively using habitats on study sites; 72 species (1698 bird individuals) were recorded on n = 28 control sites, 65 species (1272 bird individuals) on Heracleum sites (n = 28), as well as 70 species (2928 bird individuals) on n = 56 reference sites. Taking into account the classification of birds into particular assemblages, there were 427 observed ground/herb dwellers, 365 bush foragers, 362 ecotone birds, and 521 tree foragers on control sites; on Heracleum sites, 232 ground/herb dwellers, 362 bush foragers, 252 ecotone birds, and 397 tree foragers; while on the reference sites, 843 ground/herb dwellers, 598 bush foragers, 546 ecotone birds, and 906 tree foragers. The control sites were dominated by the common starling Sturnus vulgaris—ecotone bird, skylark Alauda arvensis—ground/herb dwelling species, and fieldfare Turdus pilaris—ecotone bird, while the invaded sites hosted a large abundance of common whitethroat Sylvia communis—bush forager, common starling—ecotone bird, and great tit Parus major—tree forager. On the reference sites, the three dominating bird species were the following: skylark—ground/herb dweller, common starling—ecotone species, and great tit—tree forager.
Birds appearing on control or invaded sites were part of a set of particular species from the wider community present on the reference sites (
Figure 1). These reference sites contained habitat specialists, like the ortolan bunting
Emberiza hortulana living in agricultural lands near young forests, northern wheatear
Oenanthe oenanthe requiring groups of rocks in agricultural mosaics, as well as wetland birds, like the sedge warbler
Acrocephalus schoenobaenus and great reed warbler
Acrocephalus arundinaceus. On the reference sites, large woodpeckers typical of mature forests were also present, for example, the Eurasian three-toed woodpecker
Picoides tridactylus and black woodpecker
Dryocopus martius. Although on control and invaded sites many low-demanding species appeared and they were shared with the reference sites, bird species like the rook
Corvus frugilegus or barred warbler
Sylvia nisoria found more appropriate habitats on those paired sites in comparison with the reference ones.
The variance in the number of species of all birds significantly differed between the sites (Bartlett’s test = 6.876;
p = 0.032), and there were also differences between the means (Welch ANOVA: F = 27.761;
p < 0.001). There were significant differences found in the species richness of all birds between the
Heracleum and control sites (Tukey’s HSD,
p < 0.001), control vs. reference sites (
p = 0.004), and
Heracleum vs. reference sites (
p < 0.001)—
Figure 2. The variance in the abundance of all birds did not differ between sites (Bartlett’s test = 3.031;
p = 0.219), although the means were different (One-way ANOVA: F = 5.638;
p = 0.005). This resulted from the differences between the
Heracleum and control sites (Tukey’s HSD,
p = 0.003), as controls did not differ from references (
p = 0.133) and there were no differences between the
Heracleum sites and references (
p = 0.133)—
Figure 2.
In the case of the abundance of ground/herb birds, there were no differences in the variances between the sites (Bartlett’s test = 4.487,
p = 0.106), but differences in the means between them were found (One-way ANOVA: F = 5.952,
p = 0.003), resulting from differences between the
Heracleum and control sites (Tukey’s HSD,
p = 0.009) and reference and
Heracleum sites (Tukey’s HSD,
p = 0.005). There were no differences between the controls and references (Tukey’s HSD,
p = 0.995)—
Figure 3. No differences in variance were found in the number of ecotone birds between sites (Bartlett’s test = 0.277,
p = 0.871), similarly as in their means (One-way ANOVA: F = 1.529,
p = 0.222). In this assemblage, there were no differences between the control and
Heracleum sites (
p = 0.239), and no differences were found between the remaining sites (controls vs. references:
p = 0.298;
Heracleum vs. references:
p = 0.926). In the case of a number of bush birds, differences in variance between sites were significant (Bartlett’s test = 11.309,
p = 0.003), although no differences were found between the means (Welch ANOVA: F = 2.479,
p = 0.091), which resulted from the lack of differences between the control and
Heracleum sites (
p = 0.989) and no differences between the remaining sites (controls vs. references:
p = 0.161;
Heracleum vs. references:
p = 0.180). There were significant differences in the variances of the numbers of tree birds between sites (Bartlett’s test = 8.767,
p = 0.012) and between the means (Welch ANOVA: F = 4.304,
p = 0.017). There were no differences between the mean abundances of tree birds at the control and
Heracleum sites (
p = 0.057) and no differences between the remaining sites (controls vs. references:
p = 0.308;
Heracleum vs. references:
p = 0.449)—
Figure 3.
In the case of the number of ground/herb species, there were no differences in the variances between the sites (Bartlett’s test = 1.165,
p = 0.558), but differences in the means between them were found (One-way ANOVA: F = 10.56,
p < 0.001), resulting from differences between the
Heracleum and control sites (Tukey’s HSD,
p < 0.001) and references and
Heracleum (Tukey’s HSD,
p < 0.001) but no differences between controls and references (Tukey’s HSD,
p = 0.455)—
Figure 4. No differences in variances were found in the number of ecotone species between sites (Bartlett’s test = 3.627,
p = 0.163), but the differences in the means were significant (One-way ANOVA: F = 6.876,
p < 0.001) and resulted from differences between the control and
Heracleum sites (
p < 0.001), although no differences were found between the remaining sites (controls vs. references:
p = 0.058;
Heracleum vs. references:
p = 0.126). In the case of a number of bush species, differences in variance between sites were significant (Bartlett’s test = 18.982,
p < 0.001), as were differences between the means (Welch ANOVA: F = 3.8,
p = 0.027), which resulted from differences between the controls and
Heracleum sites (
p = 0.045) and no differences between the remaining sites (controls vs. references:
p = 0.086;
Heracleum vs. references:
p = 0.991). There were significant differences in the variances of the numbers of tree bird species between sites (Bartlett’s test = 7.371,
p = 0.025) and between the means (Welch ANOVA: F = 3.609,
p = 0.032). There were differences between the mean numbers of tree species between the controls and
Heracleum sites (
p < 0.038), although there were no differences between the remaining sites (controls vs. references:
p = 0.552;
Heracleum vs. references:
p = 0.325)—
Figure 4.
The PCoA ordination revealed divergence in the bird compositions among the invaded, control, and reference sites (One-way ANOVA: F = 4.929;
p = 0.009), consisting of a gradual shift from the references to the controls and to the
Heracleum sites (
Figure 5). The differences between the reference and control sites were not significant (Tukey’s HSD,
p = 0.717), with the same as between the controls and
Heracleum sites (
p = 0.109), and there were statistically significant differences between the references and
Heracleum sites (
p = 0.011). This pattern points to the presence of some exclusive species in the invasion-altered sites that were either not present or less frequent in uninvaded sites, in this case, the mentioned rook and barred warbler (see above), and was due to the absence of, e.g., some species using agricultural lands (
Figure 1). There was also found segregation between sites in ground/herb dwellers and bush foragers, indicating dissimilar species compositions for these groups, due to the distinctive ground/herb dwellers’ composition on control sites and that of bush foragers on
Heracleum sites (
Figure 6).
In the case of ground/herb species, significant divergence between the sites was detected (One-way ANOVA: F = 3.699,
p = 0.028), resulting from differences between the reference and control sites (Tukey’s HSD,
p = 0.031) and the control and
Heracleum sites (Tukey’s HSD,
p = 0.044), with no differences between the reference and
Heracleum (Tukey’s HSD,
p = 0.756)—
Figure 6. No divergence was found in the case of ecotone species between sites (One-way ANOVA: F = 0.61,
p = 0.545). There were no differences between reference and control sites (Tukey’s HSD,
p = 0.867), reference vs.
Heracleum sites (
p = 0.519), or
Heracleum vs. control sites (
p = 0.866). In the case of bush foragers, divergence was shown between the sites (One-way ANOVA: F = 9.748,
p < 0.001), which resulted from differences between the control and
Heracleum sites (
p < 0.001) and the
Heracleum vs. reference sites (
p = 0.021), although there were no differences between the remaining reference and control sites (
p = 0.409). There was no significant divergence in tree foragers between the sites (One-way ANOVA: F = 1.898,
p = 0.155). There were no differences between the control and
Heracleum sites (
p = 0.306), reference and control sites (
p = 0.979), or
Heracleum vs. reference sites (
p = 0.210)—
Figure 6.
4. Discussion
The conducted research showed that invasion-induced habitat changes impacted both the species richness (i.e., alpha diversity) of individual bird assemblages and the divergence of assemblages between sites (i.e., beta diversity), confirming that plant invasions decrease the diversity of natives at least over small spatial scales [
44]. However, in both cases, the extent of the invasion impact was different. The presence of hogweeds significantly reduced species richness in all studied assemblages compared to uninvaded (control) sites, despite the lacking decreases in bird abundances from the uninvaded to
Heracleum sites, excluding ground/herb dwellers, whose number decreased. In terms of comparison at the level of habitat assemblages and beta diversity, the invaded sites were characterized by a unique composition of bush foragers compared to both uninvaded sites, confirming that anthropogenic activities (in this case, the introduction of plant invaders) create specific ecosystems [
5]. The ground/herb dwellers at the control sites represented a distinct set of birds in comparison with other sites. This result indicates the validity of comparing the bird assemblage with reference sites, as it suggests an increase in the species diversity of ground/herb dwellers at the control site exposed to invasion compared to reference sites with possibly natural habitats. Research on assemblages according to the standard methodology with paired sites with the same habitats and in the same landscape sometimes does not make it possible to compare semi-natural and disturbed sites due to indirect effects related to the location of control sites in the landscape.
Comparative site analyses at the species level did not detect significant differences between reference and control sites or between
Heracleum and reference sites. However, ordination analysis investigating the distribution of various species indicated that at least some habitat specialists were interchanging in space, e.g., reference sites included ground/herb dwellers that used agricultural lands, such as the ortolan bunting, but this bird was replaced by the related common reed bunting
Emberiza schoeniclus on the remaining sites (
Figure 1). This, however, confirmed that agricultural land use has a positive effect on species richness [
45], because all sites were in the agricultural landscape and only the references contained the largest croplands, away from settlements. Another example was the presence of barred warblers mainly at the
Heracleum sites and the preference of the common whitethroat
Sylvia communis for the paired control and
Heracleum sites, with an overall greater diversity of warblers from the genera
Sylvia and
Phylloscopus at reference sites (
Figure 1). This means that although significantly lower bird species richness was found in all assemblages at invaded sites rather than control sites, all site types were characterized by a different set of species. This could have resulted from both deterministic and stochastic processes; in the first case, it would have been due to habitat properties [
10], and in the second case, because of local colonization or the absence of specific species [
21]. It could have been due to the effect of resource competition between similar species, because paired and reference sites were similar in habitats (
Table A1 and
Table A2); therefore, differences in environmental suitability may have been an unlikely reason for species turnover in the study system [
46]. These results imply the important role of niche-based assembly processes in driving bird communities across multiple habitats [
47].
When comparing habitats on invaded, control, and reference sites, they differed in the area of ruderal habitats and water availability (
Table A1). Paired, i.e., invaded, and control sites, were characterized by a larger ruderal area, and
Heracleum sites additionally contained a significantly larger water area (
Table A1 and
Table A2). This would explain the absence of some species using agricultural lands at paired sites despite their presence on reference sites (stochastic process), as ruderal habitats are rather degraded. Caucasian hogweeds being in the vicinity of water, e.g., rivers, is not surprising because this is the route of spread of these invaders [
33]. However, it is worth noting the lack of differences in the areas of most habitat types between sites (
Table A1). This confirms that the reduction in species diversity during invasion was due to its properties [
33,
35], i.e., it is a deterministic process. Because the bird species richness at the reference sites did not differ from that at the paired sites, only the ordination analysis detected stochastic episodes for some bird species.
The conducted research confirmed that the filtering effect of vegetation is independent of losses and replacements of species [
9], because in terms of species composition, the
Heracleum and control sites were more similar to each other than to the reference sites (
Figure 1). And yet despite this difference between paired sites, significant differences between them were detected in the number of species from all assemblages. It is worth adding that while DCA analysis (
Figure 1) indicated a wider range of at least some species on reference sites than on paired sites, beta diversity analysis showed that invaded sites contained unusual lists of species (
Figure 5). The negative impact of Caucasian hogweeds on species diversity in all bird assemblages confirmed that introduced non-native species are associated with lower taxonomic, functional, and phylogenetic diversity of communities and negatively impacts ecosystem functioning, which is particularly evident in habitats where human disturbance favors non-native species [
48]. The results suggest that even within a generally human-modified landscape (here, including, e.g., control sites), invaded community diversity is always more affected by, and thus has a lower resilience to, disturbance. Thus, restoring and protecting natural habitats within human-modified landscapes is likely to increase the resilience of native species [
48].
On the other hand, it must be stressed that the research approach in this paper was based on correlations and it was difficult to fully assess the mechanisms of the observed relationships between assembling birds and plant invasion. In terms of the generally lower diversity of native species in invaded communities, there were two possible effects: (1) invaders directly affected bird species, e.g., through the enhanced competition or particular habitat disturbance decreasing the number of habitat specialists (possible in ground/herb dwelling species, decreased in numbers and richness) and increasing the number of generalists (possible in ecotone birds, bush foragers, and tree species decreased in richness despite no reduction in the number from uninvaded towards invaded sites) and (2) sites with lower habitat and bird diversity could more likely host invading species, which contrasts with invaded sites being treated with caution, as the habitats there may have had low diversity even without invasion.
The results of this research confirmed that beta diversity provides valuable insights into the mechanisms driving biodiversity changes and their consequences for multiple ecosystem functions [
49]. Differences in alpha diversity between paired sites were demonstrated quite surprisingly because, despite the similarities in species and habitats between them, which proves the properties of Caucasian hogweeds, it does not present the mechanisms of birds’ reaction to invasion. Beta diversity analysis, in turn, indicated that the invaded sites created a bush bird assemblage that was significantly different from the one on control sites, but also on the reference sites. The second detected phenomenon concerned the ground/herb dwellers’ assemblage which was distinct on control sites compared to other sites. These results show the mechanisms of the impact of invasion on birds. Invasive hogweeds morphologically resemble bushes and, therefore, have probably become attractive to these birds, but in a selective way depending on the species, shaping an unusual set of bush birds. In the case of ground/herb dwellers, their unusual grouping compared to other control sites may have resulted from the availability of herbaceous ruderal habitats (
Table A2), and then, the rapid impact of invaders on the birds at
Heracleum sites. These are the mechanisms of the impact of invasion on bird assemblages that suggest the implementation of specific management to preserve natural bird assemblages, i.e., hogweeds occurring in the vicinity of shrubs and herbal habitats and, e.g., grasslands being removed. Focusing on beta diversity is especially important in ecological communities that are subject to large environmental fluctuations and disturbances [
49].
It is worth adding that at the regional scale, species diversity results mostly from changes in species composition among habitat patches, which only occur when multiple habitat areas are preserved. In fragmented landscapes, most of the habitat patches are small and individually host a low diversity. However, these patches often greatly differ in beta diversity and this heterogeneity can compensate for much of the local diversity loss, which works especially well in the case of forested areas [
50]. Therefore, ecotone birds and tree foragers could also develop unusual associations in invaded areas, but in this case, it is not easy to change habitats because the availability of other forest patches is limited and often distant. The lack of an effect of invasion on tree and ecotone birds may have been undetected under beta diversity assumptions, although this effect was indicated by alpha diversity analysis.