River embankments mitigate the loss of grassland biodiversity in agricultural landscapes

Agricultural intensification has resulted in severe declines in the extent and diversity of seminatural habitats in Europe, whereas the extent of secondary habitats has increased considerably. River embankments have become one of the most extensive and widespread secondary habitats in former floodplains. We compared the diversity patterns of secondary dry and wet grasslands on river embankments with those of seminatural dry and wet grasslands in a Hungarian agricultural landscape using the following community descriptors: (a) species diversity, (b) phylogenetic diversity and (c) functional diversity. We also performed trait‐based analyses to evaluate the ecosystem services provided by these secondary grasslands. Both grassland types of the embankments showed significantly higher Shannon diversity compared with their seminatural counterparts. The cover of generalist species (i.e., cosmopolitan species, weeds and nonindigenous plant species) was high in the secondary grasslands. We found significant differences in phylogenetic diversity between the secondary and seminatural grasslands: secondary grasslands showed significantly lower mean nearest taxon distances than the seminatural grasslands. Functional diversity did not differ between the secondary and seminatural grasslands according to the Rao's quadratic entropy. However, we found higher community‐weighted means of specific leaf area, plant height and flowering period in the secondary grasslands, which are related to important ecosystem services (via biomass production and pollination). Well‐planned management actions and restoration activities could help further improve the ecological function and conservation value of secondary grasslands on river embankments, contributing to the maintenance of species diversity and sustaining the functionality of ecosystems in agricultural landscapes.


| INTRODUCTION
Technological advances in the second half of the 20th century have resulted in unprecedentedly rapid agricultural intensification worldwide (Baessler & Klotz, 2006), leading to a global-scale decline of the species and habitat diversity of ecosystems (Benton, Vickery, & Wilson, 2003). Landscapes with a high proportion of natural grasslands have largely been converted into arable fields, built-up areas and forest plantations (Bastian & Bernhardt, 1993;Biró, Bölöni, & Molnár, 2018). For instance, a large proportion of loess grasslands in Europe have been ploughed for cereal production because of their fertile chernozem soils (Deák et al., 2018;Erd} os et al., 2018). Today, this grassland type mainly occurs in small fragments (e.g., on ancient burial mounds, earthen fortifications, road verges or at the margins of arable fields) and its area is still shrinking Molnár, Biró, Bartha, & Fekete, 2012). The area and diversity of the European wet grasslands have also significantly declined in the last 300 years due to inappropriate management, drainage and river regulation (Maltby & Blackwell, 2005;Timmermann, Margóczi, Takács, & Vegelin, 2006). Along regulated rivers, wet grasslands can usually be found in the narrow and frequently disturbed (i.e., periodically flooded) floodplains between the river and the embankments (Varga, Dévai, & Tóthmérész, 2013).
The precise assessment of the ecological function and conservation value of different habitats is not possible based only on species diversity measures, because these methods neglect the functional complementarity and redundancy of species (Díaz & Cabido, 2001;Schleuter, Daufresne, Massol, & Argillier, 2010) and some of the information provided by more complex analysis of species is lost (Cadotte, Cavender-Bares, Tilman, & Oakley, 2009;Cadotte & Davies, 2016). Therefore, we computed not only species diversity but also functional and phylogenetic diversity and performed traitbased analyses to compare the ecological function and conservation value of secondary grasslands on river embankments and seminatural grasslands. We hypothesized that the secondary grasslands of river embankments have the potential to act as refuges for many plant species and provide important ecological functions that play a crucial role in sustaining the functionality of ecosystems within agricultural landscapes.

| Study area and sampling
The study sites were located in the eastern part of the Great Hungarian Plain, in the proximity of the Körös and Maros Rivers ( Figure 1).
The climate of this region is moderately warm and dry. The mean annual temperature is 10.2-10.6 C, and the mean annual precipitation is 500-550 mm. The main soil types within the study area are chernozem, alluvial and meadow soils (Dövényi, 2010).
Both the Körös and Maros Rivers flow in a westerly direction and are among the major rivers of the Great Hungarian Plain. The hydrograph of the larger rivers in this landscape usually shows two floods: snowmelt-induced floods occur in early spring and rain-induced floods in early summer (Bátori et al., 2016). We chose a 100-km-long section of the Körös River and a 40-km-long section of the Maros River for vegetation sampling. The embankments along these rivers were established in the 18th and 19th centuries to prevent the adverse effects of flood and to provide land for agriculture (Bátori et al., 2016). The slopes of embankments were sown with seed mixtures of native grasses (e.g., Arrhenatherum elatius, Alopecurus pratensis, Bromus inermis and Lolium perenne) in order to reduce erosion and to produce fodder for livestock. At that time, natural grasslands were widespread in the vicinity of the rivers, and soils originating from these grasslands were also used for the construction of the embankments. The current crest width of embankments usually ranges between 4 and 6 m, their height is about 4.5 m, while the ratio for vertical and horizontal dimension of the slopes is 1:3 or 1:4. Soil organic matter content is higher on riverside slopes than landside slopes.
Our previous study indicated that the vegetation on the northfacing landside slopes of the embankments (hereafter "secondary dry grasslands") is similar to the loess grasslands (hereafter "seminatural dry grasslands"), whereas the vegetation on the north-facing riverside slopes of the embankments (hereafter "secondary wet grasslands") is similar to the mesotrophic wet meadows (hereafter "seminatural wet grasslands"). Seminatural dry grasslands in the studied region are dominated by Festuca rupicola, but other grasses such as Agropyron cristatum, B. inermis, Elymus hispidus and Stipa capillata are also common. The high cover of dicots (e.g., Fragaria viridis, Galium verum, Inula germanica, Salvia nemorosa and Thalictrum minus) is also typical (Borhidi, Kevey, & Lendvai, 2012). These grasslands have a high conservation value due to their unique species pool. They also provide habitats for several endangered plant species such as Adonis volgensis, are also abundant in this habitat type, including Carex melanostachya, Euphorbia lucida, Inula britannica, Iris pseudacorus, Ranunculus repens, Thalictrum lucidum and Viola pumila (Borhidi et al., 2012). Grasslands on embankments are usually managed by machine mowing twice per year, whereas seminatural grasslands have been managed for centuries by various management practices (e.g., mowing and grazing).
In order to obtain representative samples from the study sites, we applied a stratified random sampling approach. The embankments of both rivers were divided into 10 subsections, and both the secondary dry and wet grasslands on the upper two-thirds of embankments were sampled (the lower third was omitted to reduce the effects of periodic floods and therefore habitat heterogeneity) in each subsection using three randomly placed 2 m × 2 m plots in both habitat types (120 plots in total). The age of these grasslands is about 45 years. For comparison, we selected 20 seminatural dry and 20 seminatural wet grassland patches within the study area. We randomly placed three 2 m × 2 m plots in each patch (also 120 plots in total). The percentage cover of each vascular plant species was estimated in May to early June 2017 in all 240 plots (see Supporting Information Table S1). Nomenclature follows The Plant List (http:// www.theplantlist.org).

| Data analysis
To evaluate the ecological functions and conservation value of secondary grasslands on river embankments, we compared them with the seminatural grasslands using the following metrics: species diversity, diagnostic species, phylogenetic diversity, functional diversity and functional trait distributions.
We calculated the Shannon diversity for each plot and the phi (Φ) coefficient of association (Chytrý, Tichý, Holt, & Botta-Dukát, 2002) between species and habitat (i.e., secondary dry grasslands vs. seminatural dry grasslands and secondary wet grasslands vs. seminatural wet grasslands). We considered a species diagnostic if it had 0.2 or higher phi value in a particular grassland type (Fisher exact test; p < .01). If a species proved to be diagnostic for more than one grassland type, it was considered diagnostic species to the grassland in which it had a higher phi value. For the further evaluation of these diagnostic species, we classified them into three groups based on their habitat preferences (Borhidi, 1995). The three groups were (a) dry grassland specialists, (b) wet grassland specialists and (c) generalist species (i.e., cosmopolitan species, weeds and nonindigenous species).
For the analysis of phylogenetic diversity, phylogenetic trees were created based on a dated, ultrametric phylogenetic tree of European plants (Durka & Michalski, 2012) and the genera occurring in the studied habitats. The cover values of species from the same genus were summarized. Polytomies were retained as they were represented in the original tree. Phylogenetic diversity of the grasslands was compared using the mean pairwise distance (MPD) and mean nearest taxon distance (MNTD) indices. MPD calculates the MPDs among different taxa, whereas MNTD calculates the mean phylogenetic distance to the closest relative for particular taxa. These two indices capture different aspects of phylogenetic diversity. MPD is generally thought to be more sensitive to tree-wide patterns of phylogenetic diversity, whereas MNTD is more sensitive to the patterns of the tips of the phylogeny (Kembel et al., 2010). These two indices are suitable measures of phylogenetic diversity and are less confounded by species richness than other phylogenetic diversity indices. Therefore, they are appropriate measures in studies where species diversity was also calculated (Barak et al., 2017).
As a measure of functional diversity, we calculated plot-level Rao's quadratic entropy using leaf-height-seed traits (specific leaf We applied separate models to compare the dry (secondary dry grasslands vs. seminatural dry grasslands) and wet (secondary wet grasslands vs. seminatural wet grasslands) grassland types. We set sampling location (i.e., subsection) as random factor in the models.
Seed mass and flowering period traits were log-transformed.
The calculations of phi values were conducted with the JUICE 7.0.25 programme (Tichý, 2002). Diversity indices and linear models were computed in R environment (R Core Team, 2018). Shannon diversity values were calculated with the "diversity" function of the vegan package (Oksanen et al., 2019). We used the "cophenetic," "ses.
Although the number of diagnostic species was higher in the secondary grasslands, the proportion of habitat-specific diagnostic species was higher in the seminatural grasslands. The proportion of dry grassland specialists was 20% in the secondary dry grasslands and 75% in the seminatural dry grasslands. Conversely, the proportion of generalist species was 76% in the secondary dry grasslands and 25% in the seminatural dry grasslands (Supporting Information Table S2).
The proportion of wet grassland specialists was lower in the secondary wet grasslands (29%) than in the seminatural wet grasslands (67%) and secondary wet grasslands had more generalist species (64%) than seminatural wet grasslands (33%).
Secondary grasslands showed similar MPDs compared with the seminatural grasslands. In contrast, the difference was significant for MNTDs; secondary grasslands showed significantly lower MNTDs than seminatural ones (Table 2; Figure 2). There was no difference between the Rao's index for secondary and seminatural grasslands (Table 2). However, the CWMs of SLA were significantly higher in secondary grasslands (Table 3 and Figure 3). There was no difference in the CWMs of plant height between the secondary and seminatural dry grasslands. In contrast, the CWMs of plant height indicated that secondary wet grasslands had potentially higher vegetation than seminatural wet grasslands. We did not find any significant difference for seed mass CWMs. However, the CWMs of the flowering period were significantly higher in the secondary grasslands of embankments in both comparisons (Table 3 and Figure 3). The number of insect pollinated plants was significantly higher (p < .001) in the secondary wet grasslands than in the seminatural wet grasslands, but we did not find any significant difference in the case of dry grasslands (p = .780).

T A B L E 1
Details of the nine traits used for the functional diversity analysis

| Evaluation of ecological function and conservation value
According to the diagnostic species analyses, generalist species play an especially important role in the grasslands of embankments.
Cosmopolitan species, nonindigenous species and weeds could initially colonize these fresh surfaces during the construction of the embankments simultaneously with the sown grasses and other species as founders; therefore, the higher abundances of generalist species on the embankments can be a legacy of this founder effect (Egler, 1954;Grime, 1998). The floodplains of the rivers are densely covered by invasive species and weeds, as the rivers can effectively F I G U R E 2 Shannon diversity and phylogenetic diversity (MNTD) values for the grassland habitats of embankments and the landscape. Statistically significant differences were marked with asterisks. MNTD, mean nearest taxon distance. *p < .05; **p < .01; ***p < .001

T A B L E 3
Comparisons of trait distributions between secondary grasslands on river embankments and seminatural grasslands (secondary dry grasslands vs. seminatural dry grasslands and secondary wet grasslands vs. seminatural wet grasslands) disperse their propagules to these areas and the disturbance cycle of floods continuously creates new colonization gaps (Bátori et al., 2016). Reconstruction works, dirt roads and the establishment of different flood regulation facilities also create bare surfaces where these species have the potential to colonize successfully once introduced. In addition, the habitats of these embankments are embedded F I G U R E 3 Community-weighted means of specific leaf area, plant height and flowering period for the grassland habitats of embankments and the landscape. Statistically significant differences were marked with asterisks. *p < .05; **p < .01; ***p < .001 in the matrix of agricultural land with a high perimeter-area ratio.
These circumstances can provide good opportunities for many generalist species to survive on the embankments (cf. Theoharides & Dukes, 2007). Although the high species richness of generalist species significantly increased the Shannon diversity on the embankments; these species did not increase the functional diversity of the grasslands because of their similar functional traits. Therefore, the functional structure of these secondary grasslands is similar to that in the seminatural grasslands of the landscape.
Most studies agree that disturbance has the potential to decrease phylogenetic diversity (Barak et al., 2017;Dinnage, 2009;Turley & Brudvig, 2016). Barak et al. (2017) found that prairies that had been restored using seed sowing had lower phylogenetic diversity compared with natural prairie stands, as the MPD and MNTD of the restored prairies showed significantly more clustered structure than those of the natural ones. Turley and Brudvig (2016) showed that oldfields had significantly lower phylogenetic diversity compared to habitats that had never been cultivated. Similar results were obtained by Helmus et al. (2010), who found that disturbances in lakes resulted in clustering in the phylogenetic structure of the zooplankton community. Based on the theory of environmental filtering and limiting similarity, the phylogenetically clustered structure of the secondary habitats can be expected, as disturbance may weaken the strength of competition (Dinnage, 2009). The MNTD analyses supported this theory, as the values of these indices were lower for the secondary grasslands on embankments than for the seminatural grasslands. It also means that the vulnerability of these secondary grasslands is higher and their resilience is lower against the invasion of alien species (Lososová et al., 2015).
Therefore, our results support the conclusion of recent studies showing that the precise assessment of the ecological function and conservation value of different habitats cannot solely be based on species-based diversity indices, as they are not sensitive to functional redundancy and other functional consequences of species identity (Díaz & Cabido, 2001;Kelemen et al., 2017;Petchey & Gaston, 2006;Schleuter et al., 2010;Tilman et al., 1997). The Shannon diversity indices together with the diagnostic species and the MNTD analyses suggested that the reason for the higher diversity in these secondary grasslands is the higher proportion of generalist species that are functionally and phylogenetically more clustered compared with the species pool of the seminatural grasslands.
Our results on single traits can also help understanding vegetation responses to certain environmental circumstances and potential ecosystem services provided by the vegetation. Species with high SLA can respond rapidly to environmental changes (fast plants) because of their high efficiency of photosynthesis and fast growth (Kelemen et al., 2016;Westoby, 1998 Valkó, Török, Matus, & Tóthmérész, 2012) to ensure the reproduction of most plant species in the long run (Moinardeau, Mesléard, Ramone, & Dutoit, 2019) and to prevent the critical decrease of biomass, which is also important for the protection against erosion.
The embankments of the investigated rivers are situated in agricultural landscapes; therefore, the colonization potential of many grassland specialist species (i.e., dry grassland and wet grassland species) is limited. Consequently, active restoration would be needed to ensure the establishment of these species on the embankments. To increase the number and abundance of valuable dry and wet grassland species in the grasslands of the embankments, hay transfer from seminatural habitats and sowing of regional seed mixtures would provide feasible solutions (Klimkowska et al., 2010;Török et al., 2010).
Secondary grasslands on river embankments have the potential to act as refuge sites for many plant species and may provide important ecological functions in the future. Proper management practices are needed to improve the quality of these secondary habitats.

DATA AVAILABILITY STATEMENT
The authors confirm that the data supporting the findings of this study are available within the article [and/or] its supplementary materials.