Introduction

Agricultural intensification is the single most important driver of biodiversity losses in the industrialized regions of the world (Tilman et al. 2017). In most European countries, farmland biodiversity has sharply decreased or even collapsed during the past decades (Wesche et al. 2012; Meyer et al. 2013; Hallmann et al. 2017; Heldbjerg et al. 2018; Rigal et al. 2023), imperiling important ecosystem functions such as pollination and biological pest control (Potts et al. 2010; Garibaldi et al. 2016). Cultural grasslands play a pivotal role for farmland biodiversity in Europe (Oppermann et al. 2012). Prior to the recent era of agricultural intensification, managed grasslands were home to a considerable part of Central Europe’s flora and fauna (Dierschke and Briemle 2002; Bruchmann and Hobohm 2010; Luick et al. 2012). For example, around a third of the native vascular plant species of Central Europe resided predominantly in mown or grazed agricultural grasslands, if management was not too intense (Leuschner and Ellenberg 2017). In Germany, more than 1000 vascular plant species occur in agricultural grasslands (Korneck and Sukopp 1998), and about 40% of Germany’s endangered plant species occur in grasslands (Briemle 2003, BfN 2014, 2018). Thus, grasslands have the highest plot-level plant diversity in Central Europe (Merunková et al. 2012), highlighting their importance for biodiversity conservation. Grasslands also provide highly valued ecosystem services such as soil carbon storage and water retention, and protection against soil erosion (Guo and Gifford 2002; BfN 2014; Schils et al. 2022).

The breeding of high-yielding dairy cows has steadily increased the nutritional requirements of the livestock, and thus the production of protein- and energy-rich herbage. Therefore, grassland management has changed in the past decades from extensive pastures and hay meadows to highly intensive grasslands with up to 4–7 cuts per year for silage or high stocking rates on pasture (Mielke and Wohlers 2019). In groundwater-influenced areas such as river floodplains and low-lying coastal plains, intensification depended on large-scale drainage to enable machine access (Meisel 1960; Succow 2001). High yield and forage quality require high fertilizer inputs that can exceed 300 kg N ha−1 yr−1 (LWK-NRW 2021).

With Germany being the largest milk producer in the EU, the north-west German alluvial lowlands belong to the regions with highest grassland management intensity in Europe (Estel et al. 2018), where intensification has started already more than 60 years ago. Intensive drainage, high fertilization, re-seeding with ryegrass, and early and frequent cutting have fundamentally altered the grasslands toward species-poor ‘basal communities’ of a few highly productive grasses (Isselstein et al. 2005; Hopkins and Wilkins 2006; Leuschner and Ellenberg 2017). The associated losses in plant diversity have been documented through several re-sampling studies in wet grasslands of northern Germany and elsewhere, covering 3–6 decades of change (Dierschke and Wittig 1991; Schrautzer and Wiebe 1993; Wittig et al. 2007; Prach 2008; Krause et al. 2014). For example, a large-scale resampling study in five floodplain areas in northern Germany indicated a 30% decrease in species numbers per relevé between the 1950s and the 2000s, significant decline in 23 of 30 characteristic wet meadow species and the loss of many grassland specialist taxa (Wesche et al. 2012; Krause et al. 2014). The list of declining species includes many once widespread plants, while a few nitrogen-demanding, productive species have increased. Wittig et al. (2007) found an even larger decrease in average species numbers per plot from 27.0 (1963) to 10.5 (2006) in floodplain grasslands near Bremen (north-western Germany). The decline in grassland α-diversity is accompanied by cross-regional vegetation homogenization, i.e. decreasing β-diversity (Diekmann et al. 2019).

In many regions, management intensification was associated with large-scale conversion of historical grassland to arable land. For example, in the federal states Lower Saxony and Schleswig–Holstein, 34–48% of former grassland area has been converted between 1950 and 2012 (Leuschner et al. 2014a). In this region, species-rich extensively managed grassland has been reduced to a tiny fraction of the remaining cultural grassland: In Schleswig–Holstein, only 2.1% of the agricultural land was categorized in 2021 as high-nature-value grassland (BfN unpubl. data). Until the 1960s, the floodplains of northern Germany were widely covered by pastures and meadows with 2–3 cuts per year, which typically harbored 28–40 plant species per 20–30 m2 (Dierschke 2008), constituting the LRT type 6510 “nutrient-poor lowland hay meadows” within the EU’s Flora, Fauna and Habitat (FHH) Directive (BfN 2023). Most of this extensive grassland has since then been replaced by highly fertilized, frequently mown silage meadows and rotational pastures. In the lowlands of northern Germany, the concerted action of plot-level decreases in α-diversity and widespread loss of high nature value (HNV) grassland habitats has led since the 1950s to decreases in population size by > 95% of formerly common grassland species such as Anthoxanthum odoratum, Cardamine pratensis and Lychnis flos-cuculi (Wesche et al. 2009; Krause et al. 2011). Preserving and restoring grassland biodiversity is thus a priority goal in German conservation and in other European countries (Oppermann et al. 2012; Bosshard 2016).

Until recently, grassland conservation in Central Europe has focused on high-value grassland (Oppermann et al. 2012), i.e. the remnants of extensively used, species-rich meadows and pastures, which covered in 2013 about 5% of Germany’s farmland area (BfN 2014). Not surprising, less attention has been paid to the much larger area of intensively managed grassland, where biodiversity is low today. However, the remaining patches of extensively used HNV grassland are often isolated and have declining genetic diversity (Ellstrand and Elam 1993; Honnay and Jacquemyn 2007). This suggests that a successful strategy to halt biodiversity loss in Central European landscapes needs to include intensive grasslands. This policy adjustment is urgent, as the intensive grassland has profited much less from targeted biodiversity measures in the frame of the EU Common Agricultural Policy (CAP) than has arable land. In the latter, flower strips, conservation headlands and fallows have been established since more than three decades, though with mixed success for biodiversity conservation (Kleijn et al. 2006).

One option to promote biodiversity in intensive grassland might be the regulatory framework of the 2023–2027 CAP, and regional regulations such as “Niedersächsischer Weg” in Lower Saxony, which prohibits fertilizer and pesticide application in 3-m wide riparian buffer strips (measured from the ditch shoulder), i.e. grassland directly bordering streams and major drainage ditches (BMEL 2022; LWK-NS 2023). This is an extension of earlier management restrictions along streams, which required smaller buffer strips. Fertilizer- and pesticide-free zones may not only reduce nitrate leaching, but could also provide residual habitats for the endangered grassland flora and fauna (Haddaway et al. 2018; Cole et al. 2020).

The widespread biodiversity collapse that has happened in the high-intensity dairy farming regions of north-western Germany is reinforced by the fact that farmers tend to use almost all of their grassland for production, sparing only tiny patches from intensive management. Can wider riparian buffer strips thus provide a suitable habitat for the endangered biodiversity of wet grassland amidst intensively used farmland?—With respect to plants, this will critically depend on the diversity of the remaining vegetation at grassland field edges. Limited access by machines and cattle together with existing fertilizer restrictions have reduced management intensity at grassland field edges already in the past. This is reflected in higher species richness in grassland patches near ditches and under pasture fences as compared to the field interior (e.g. Vollrath 1970; Konrad and Ruthsatz 1993; Husicka and Vogel 1999). These investigations showed that such micro-habitats are indeed richer in lower-productive and less disturbance-tolerant species. However, the studies were conducted in times when biodiversity impoverishment was not as advanced as it is today. It is unclear, therefore, as to whether field edges and margins can fulfill a function as relict habitats in grasslands that have been managed intensively for many decades, and whether their seed bank still allows restoring more diverse grassland communities, once wider buffer strips are in place.

Here, we study the small-scale heterogeneity of vegetation, seed bank and soil from the grassland interior through the margin to the field edge in 110 intensively managed wet grassland fields in north-western Germany on marsh and moor soils to assess the current extent of phytodiversity impoverishment and to examine the importance of field edges for plant diversity under the conditions of high-input dairy cattle farming. We hypothesized that (1) soil nutrient availability is lower at the field edge than in the field interior, (2) plant diversity of both vegetation and seed bank decreases from the grassland edge to the interior, and edges bear potential for grassland biodiversity restoration, even though γ-diversity is low, and (3) the effect of soil substrate (marsh vs. moor soil) is small compared to the dominant management effect. A restoration would be assessed as successful, if four or more HNV grassland species were present.

Materials and methods

Study region

The study was conducted in the north-western German lowlands in the federal states of Lower Saxony and Schleswig–Holstein close to the North Sea (Fig. S1). Annual mean precipitation and mean temperature (2008–2018) in the study region are 821 mm and 9.6 °C (DWD 2020). The study sites are located at elevations of − 3 to 8 m a.s.l. The study region covers the landscape units ‘Weser-Ems-Marsch’, ‘Stader Geest’, ‘Elbmarsch’ and ‘Schleswig-Holsteinische Geest’ and comprises the grasslands on marsh and moor (peatland) soils. Marsh grasslands stock on river (five farms) or tidal floodplains (one farm) that are rich in silt and clay, while moor grassland occur on drained bogs (six farms) with high soil carbon content. The study year 2019 was somewhat warmer (10.5 °C) and slightly drier (753 mm) than the long-term average (DWD 2020).

In the study region, grassland farming depends on the maintenance of intensive drainage to maintain site accessibility (Behre 1991; Succow 2001). Therefore, most grasslands are framed by drainage ditches, which are regularly excavated, and the extracted material deposited on the ditch flanks. In the study sites, only few fields were bordered by hedges, tree rows, woodlands, grass verges, croplands or roads, or combinations thereof. Ditches bordering grassland paddocks face management restrictions according to federal and state law to reduce leaching of agricultural chemicals out of the paddock and safeguard aquatic life (BMEL 2022). A 1 m-wide buffer strip was required from 2006 onwards; the margin was extended to 4 m (or 1 m in case of precision fertilization) in 2017 (BMJ 2017). Since 2023, a 5-m wide buffer strip without fertilization and pesticide use has become mandatory in case of non-precision fertilization (BMEL 2022; LWK-NS 2023). This may increase the habitat value of buffer strips in future. Exceptions are made in areas with very high stream density and for periodically dry ditches.

Grassland management

The 12 study farms have high-input dairy-cattle husbandry combined with silage production on intensively managed meadows with 4–7 cuts per year. Animal husbandry regimes varied from zero grazing, i.e. all cattle is kept indoors, and grasslands are mown for silage, to rotational grazing on up to 30% of the grassland area (seasonal grazing at a maximum duration of 266 days yr−1), while the remaining grassland area was mown. To cover the variation in dairy-cattle farm types in the study region, farm size varied considerably in our sample (32–225 ha), as do cattle stocks (82–1200 animals, or 32–710 lactating cows). To address variation in soil types, half of the farms were on marsh soil, the other ones on peat soil, respectively. The study sites were not affected by management restrictions of the EU’s Flora-Fauna-Habitat (FFH) Directive or other national nature conservation programs, and agri-environmental schemes were implemented only on very small areas. For management data, we did a personal inquiry of the farmers in 2019, collecting information about grassland management on the study sites.

Plot establishment and soil analyses

We studied 55 marsh and moor sites, respectively (110 in total). In many grasslands, regular ditch maintenance resulted in the deposition of excavated soil in the grassland margin and locally also interior sections, forming ‘beds’ with 20–50 cm higher surface that alternate with depressions. In several sites, minor longitudinal trenches to improve drainage were present also in the field interior. This relief structure is typically parallel to the longer side of the field. In some fields, elevated beds and depressions alternated randomly in the field.

To examine the assumed management intensity gradient from the field interior to the edge, and to account for within-site elevation differences, we placed 2-m wide longitudinal plots of 100-m length (area 200 m2) at three distances to the ditch: (1) directly on the grassland edge (covering the ditch slope), where fertilization is legally prohibited (termed hereafter ‘edge’), (2) at 2–8 m distance to the ditch in the field’s fertilized but marginal area (‘margin’), and (3) in the field interior at least 15 m distant to the margin plot (‘center’). The latter was often situated on a ‘bed’, even though depressions were also covered in a number of center plots. We kept a minimum distance of 5 m to field head areas to minimize edge effects. Field access tracks were excluded by avoiding them or, if necessary, interrupting plots.

Soil samples were collected in May–June 2019 in each of the three 200-m2 plots at 72 grassland sites (36 on marsh and moor soils, respectively), i.e. 216 soil samples in total. Each sample consisted of 1 L of soil that was collected at 0–10 cm depth in 40 randomly placed locations in the 200-m2 plot. Prior to chemical analysis, samples were dried at 30 °C for 4 days until weight constancy. The soil material was milled and sieved through a 2-mm sieve, and analyzed for plant-available phosphorus (P) using the calcium-acetate-lactate (CAL) extraction method (Schüller 1969); soil pH was measured in the fresh soil in H2O.

Vegetation survey and seed bank analysis

Plant species composition was analyzed between April and June 2019 in all 110 sites in each of the three 200-m2 plots (edge, margin, center plots). We chose this plot size to compensate for the generally low plant species diversity. All vascular plant species were recorded, excluding juveniles of woody plants. Due to the large size of the plots, we only noted species presence but refrained from estimating cover, which would have been too imprecise. Plant species names follow Hand et al. (2020).

In addition, we sampled the soil diaspore bank in the 200-m2 plots in the second half of December 2018. In each center and edge plot, a pooled soil sample was extracted from the topsoil (0–10 cm) with a corer of 1.5 cm diameter that was inserted about 50 times at random positions in the plot, until 1 L soil had been collected (144 1 L-samples in total). The samples were taken in close vicinity of the samples collected for soil chemical analysis. The margin plots were omitted to reduce the very labor-intensive seedling identification procedure. The samples were stored at 3 °C for about six weeks for seed stratification. The subsequent germination experiment was conducted by adopting the protocol of Ter Heerdt et al. (1996). Each soil sample was soaked in water and then washed through a sieve of 0.2 mm pore size to remove fine soil particles. The remaining mixture of sand, organic material and diaspores was then spread out evenly in a thin layer on individual trays of 28 cm × 45 cm size that were filled with 1–2 cm of sterilized and roughly sieved potting soil and a thin layer of sterile sand on top (2–3 mm). To ensure optimal germination conditions, the diaspore-containing suspension was spread out as thinly as possible. The trays were watered daily and placed in a greenhouse for eight weeks at temperatures of about 20–30 °C, depending on sunshine intensity. To test for contamination with seeds that did not originate from the sampled soil, two control trays with the same soil substrate were placed randomly on the tables and kept under the same conditions. Sciarids and other pest organisms were controlled by regularly adding nematodes (Steinernema feltiae). After four weeks, the identification of seedlings started. They were identified, counted and removed or, if species identification was not immediately possible, transplanted into pots to allow further growth. After another 4 weeks, most seedlings had been counted and removed. Watering of the trays was then stopped for 1 week for stimulating further germination by applying drought stress. Tiny seedlings, which could not be identified, were counted as ‘mono-’ or ‘dicots’. Subsequently, the parched sample layers were crumbled and watered again. Newly emerging seedlings were identified and counted again. After a total of 15 weeks, the experiment was terminated, as only very few seedlings emerged at this time. This may underestimate total species richness, as some species likely will emerge only in autumn or in the following spring. Species identification mostly followed Hanf (1999) and Jäger et al. (2017), while 1.9% of the seedlings (821 monocots and 332 dicots) could not be identified, because they died off, were too tiny and generative traits were necessary for determination.

The forage value of aboveground biomass was assessed through its crude protein content which was detected in biomass samples taken in July/August 2019 in a 0.5 m × 0.5 m frame by cutting all aboveground biomass ca. 3 cm above the soil surface. Samples were frozen, transported to the laboratory, dried at 70 °C for about 48 h, and crude protein content and crude fibre content (% d.w.) measured by optical determination with a Foss NIRsystems 6500 scanning monochromator (Foss, Silver Spring, MD, USA) after grinding biomass and sieving the powder through 4-mm and 1-mm sieves.

Statistical analyses

Plant species richness was analyzed for the 200-m2 plots (α-diversity), across the 110 sites (β-diversity), and for all sites (γ-diversity). Plant species recorded in the plots were assigned to the functional groups graminoids, non-legume forbs, and legumes, and grouped according to their conservation interest in (1) HNV species for grassland according to the German Federal Agency for Nature Conservation listing (BfN 2020), (2) species characterizing the habitat type ‘nutrient-poor lowland hay meadows’ (no. 6510) in the EU’s FFH habitat type list, and (3) species not listed under these categories (no conservation value). HNV indicator species are plant taxa associated with habitats of higher plant diversity, i.e. hosting rare and specialized species that thus function as indicators for sites with high conservation value. Grassland with up to three HNV species is not classified as ‘valuable’; the presence of 4–5 HNV species ranks sites as ‘moderately valuable’; 6–7 species indicate ‘high’, and > 8 species ‘very high’ conservation value (BfN 2020). FFH indicator species were assigned according to the list defined by the nature conservation agency of the federal state of Lower Saxony (NLWKN 2011). The indicator value of the species for site nitrogen availability (N) and soil moisture regime (F) was assessed by means of the Ellenberg scores for nitrogen (N) and moisture (F) (Ellenberg et al. 1992), and plot-level mean N and F scores were computed by averaging over the N and F scores of the species present. The species’ mowing tolerance was expressed following Dierschke and Briemle (2002), and simple plot means were computed by averaging over the species’ scores without considering cover.

Statistical analyses and data visualization were conducted with the software R 4.0.4 (R Core Team 2021). Linear mixed-effects models were built to analyse the fixed and interaction effects of soil substrate (moor vs. marsh), position in the field (interior, margin, edge) and management (mown and grazed and mown vs. grazed) on the soil pH and soil P content using the package ‘nlme’ (Pinheiro et al. 2022). The sampling plot was defined as the nested sampling unit located within farm, grassland site and position (n = 204) and treated as a random effect. Normality of the residuals was checked by visual inspection of QQ-plots. Variance homogeneity was evaluated with Bartlett’s test and plots of residuals vs. fitted values, and residuals vs. predictor values (Zuur et al. 2009). Separate variances were allowed per position in the models.

Comparisons of means were conducted posthoc using Tukey’s HSD test in the package ‘emmeans’ (Lenth et al. 2023) for significant influencing factors. LMMs were also calculated to obtain marginal means of the total number of species and individuals in the seed bank of center and edge plots, and of meadow, mown pasture and pasture plots; the means were compared with F test statistics. A Tukey post-hoc test was used to detect significant differences in plant species richness among the three plot types within a grassland. For identifying environmental and management-related factors determining aboveground species richness, a generalized linear mixed effects model (GLMM) with Poisson distribution was calculated that included the factors plot position in the field (edge, margin, center), substrate (marsh vs. moor), soil pH, management regime (mown vs. grazed, or mown and grazed), and plot-level mean N and F scores as fixed factors. Farm and site were included as random factors to address farm-specific variation in climatic and agronomic factors (cattle stock size and grassland management regime and intensity). The assumption of equal variances between groups was checked with QQ-plots. The influence of each of these factors was expressed in a standardized way through effect sizes based on Hedge’s g, which is the difference between two means divided by the pooled standard deviation. The significance of the influence of predictor variables was examined with an F-test. Finally, to separate specific neighborhood effects from the more general influence of species spill-over from adjacent vegetation types, linear mixed effects models were built to examine the influence of the type of adjacent edge habitat or structure (ditch, fence, road verge, hedge/wood or mixed) on the species richness of the aboveground vegetation and the seed bank. A significance level of p < 0.05 was used throughout the study.

A Principal Component Analysis (PCA) was run to relate the species composition in 216 relevé plots (each 72 center, margin and edge plots) to species richness components (richness of graminoids, non-legume forbs, legumes, HNV grassland species, and all species), plot position in the grassland (edge), and important site and stand characteristics (pH, nitrogen and moisture (F) scores, plant mowing tolerance, forage value of biomass).

The Bray–Curtis Dissimilarity Index, i.e. 1 – the Sørensen Similarity Index Ss (the number of shared species times 2 divided by the sum of species in the two samples), was used to compare the species composition of edge, margin and interior plots.

Results

Small-scale variation in soil chemistry and aboveground vegetation

The pH and CAL-extractable P content in the topsoil were influenced by the position in the grassland, and P content (but not pH) also by the type of grassland management, while no significant interaction (position × management) was detected (Table S1). The type of substrate (marsh vs. moor) influenced neither pH nor P content. While center and margin plots had very similar pH values (means of 5.1 and 4.9), edge plots had on both substrates by 0.5 units lower pH values. The P contents were relatively high with median values of 35–52 mg kg−1. On both substrates, they were by roughly 15 mg kg−1 lower in the edge than in the center and margin (p < 0.05; Table 1). Management significantly influenced soil P contents, with a higher content under grazed (lsmean ± SE: 56.1 ± 3.2 mg kg−1 dry soil) than mown and grazed and mown plots (41.0 ± 2.4 mg kg−1).

Table 1 Soil P concentration and pH as influenced by the position within the study grasslands
Full size table

In total, 148 plant species were found in the aboveground vegetation at the 110 study sites. When excluding the species that occurred exclusively at the unfertilized field edges (mostly the slopes adjacent to ditches), 118 taxa constituted the species pool of intensively managed grassland in the north-western German study region (Table S2). The entire species pool (center, margin and edge) included 37 graminoids, 105 forbs and six legumes; 20 species are listed as high-nature-value (HNV) species, 13 are considered as characteristic for the FFH habitat type ‘wet lowland hay meadows’ (no. 6510). While only 80% of the 148 species (118 taxa) occurred in the grasslands (center and margin), a much larger fraction was present at the edges (95%, i.e. 140 of 148 taxa), and 38 (27%) of the species were restricted to the latter habitat. The most common species with exclusive occurrence at the edge were Epilobium angustifolium, Galium palustre and G. mollugo, Luzula campestris, Lotus corniculatus, Lysimachia nummularia and Potentilla erecta (Table S2).

Median aboveground species richness decreased from the edge (23 per 200 m2) to 15 in the margin and 16 in the center plots (p < 0.001; all sites pooled). Maximum species richness reached 34 in edge plots and 24 in margin and center plots, respectively (Fig. 1a). High-nature-value (HNV) and FFH wet grassland species occurred only at very low numbers in the plots (central and marginal plots: median one species; edge plots: median two FFH and 2–3 HNV species; Fig. 1b, c). The difference between edge plots and the other positions was significant. Overall, only three HNV and three FFH wet grassland species were observed in the center and margin plots, which increased to ten HNV and six FFH species in the edge plots (Table S3). Species richness was not different between marsh and moor soils (Fig. 1).

Fig. 1
figure 1

Aboveground plant species richness (per 200 m2) in the center, margin and edge plots in marsh and moor sites (box-whisker plots with median, 25- and 75-percentiles and minima and maxima): a all vascular species, b grassland high-nature-value species, and c species characteristic for FFH-habitat type 6510 (c) (n = 55 sites for center, margin and edge plots on moor soils, n = 56 for central and margin plots, and n = 57 for edge plots on marsh soils). Significant differences between plot positions are indicated by horizontal lines at the figure top and asterisks (*p < 0.05, **p < 0.01, ***p < 0.001) based on a Tukey post-hoc test

Full size image

Aboveground community composition across different sites was more similar in the center and margin plots than in the edge plots (mean Bray–Curtis Dissimilarity 0.50, 0.53 and 0.60, respectively), indicating a higher β-diversity at the edges.

Small-scale variation in seed bank diversity and species composition

A total of 59,711 seedlings emerged from the 144 soil samples, corresponding to an average of 415 seedlings per 1-L sample (24–2609) or about 46,100 m−2 surface area. In total, seedlings of 107 plant species could be identified (six taxa only at the genus level), with on average 19 species per 1 L-sample (8–30). The seed banks contained more non-legume forb species (69.2% of all species) than graminoids (22.4%); tree species (6.5%) and legumes (1.9%) contributed only small proportions. However, with respect to germinated individuals, graminoids clearly dominated (82.8% of all individuals), with Juncus bufonius being the by far most abundant species, followed by Poa, Agrostis and Holcus species (Table S4).

In total, 12 plant species (187 individuals) were identified as HNV farmland indicators in the seed bank (including Achillea millefolium, Daucus carota and Ranunculus acris). Yet, HNV species occurred only at 44 of the 72 sites, and at a maximum diversity of four species per site.

Edge plots contained on average by 25% higher species numbers in the seed bank, and more than twice as many viable seeds than center plots (both differences highly significant; Table 2). However, the main difference was a more than three times higher density of Juncus seeds in the edge soil; the difference in other species (on average three additional forb and one graminoid species at the edge) was small. While the grasses Poa trivialis and P. annua, Agrostis stolonifera and Elymus repens, and Stellaria media occurred at higher frequencies in the seed bank of the center, herbs such as Urtica dioica, Persicaria maculosa, Rumex acetosella and Polygonum aviculare (and Holcus lanatus) were more abundant in the non-fertilized edge soils (Table S4). Grazed (and grazed and mown) patches contained somewhat higher species numbers (on average 18.3 vs. 15.8) in the seed bank than mown fields (p = 0.04), while seed numbers were not different (Table S5; Fig. 2). Marsh and moor soils did not differ significantly in seed bank species richness and the number of germinated seeds.

Table 2 Comparison of the seed bank composition in grassland center and edge plots (n = 72)
Full size table
Fig. 2
figure 2

Number of all plant species (a) and grassland high-nature-value species (b) in the seed bank of grassland center and edge on marsh and moor soils (Box-Whisker plots with median, 25- and 75-percentiles and maxima and minima). Numbers are per 1 L of soil (***p < 0.001)

Full size image

Factors determining species richness in intensively managed grasslands

The generalized linear mixed effects model, which included plot position in the field, substrate, soil pH, management type, N and F scores as fixed effects, and farm and site as random terms, revealed position as the main factor determining plant species richness in the aboveground vegetation (Fig. 3). When taking the species richness of the edge plots as a reference, the largest diversity reduction was thus caused by position, i.e. edge vs. field interior (center and margin). The reducing effect of position was larger than that of increased N availability (as inferred from a higher N score of the vegetation). Mowing, and slightly less so mowing plus grazing, showed a reduced species richness when compared to grazing alone. Grasslands on moor soil were more species-rich than grassland on marsh soil, and species richness increased with a rising soil pH. In the PCA analysis, the five factors position in the field, N availability, management type, pH and substrate explained 39.9% of the variation in plant species richness in the aboveground vegetation. In correspondence, the PCA also demonstrated the positive association of species richness with edge plots and the negative influence of N availability (Fig. 4).

Fig. 3
figure 3

Effect size (Hedge’s g) of soil pH, substrate (moor vs. marsh), management type (mown and grazed or mown vs. grazed), mean Ellenberg N score (N fertilization), and location (center and margin vs. edge) on plot-level aboveground species richness (given are estimated means of the covariates and 95% confidence intervals). Estimates were calculated relative to plant species richness in edge plots on marsh soils (*p < 0.05, **p < 0.01, ***p < 0.001)

Full size image
Fig. 4
figure 4

Principal component analysis (PCA) of species richness components (richness of graminoids, non-legume forbs, legumes, HNV and FFH grassland species, and all species in the aboveground vegetation), plot location in the grassland (edge) and important site and stand characteristics (pH, soil nutrients (N) and moisture (F), plant mowing tolerance, forage value of biomass) in relation to the species composition of 216 plots (each 72 center, margin and edge plots); N and F were calculated as unweighted community means of the respective Ellenberg scores. Mowing tolerance after species scores by Dierschke and Briemle (2002). Forage value of biomass according to its crude protein content

Full size image

Linear mixed effects models showed no significant effect of adjacent habitat type (ditch, fence, road verge, hedge/wood or mixed) on aboveground species richness and the number of species in the seed bank (results not shown).

Discussion

Current state of grassland phytodiversity

The low number of in total 148 higher plant species in the entire sample demonstrates the floristic impoverishment of wet grasslands in north-western Germany that had resulted from decades of intensive management (cf. Schwartze 1992; Schrautzer and Wiebe 1993). The species pool of the grassland interior consisted in our sample of only about 110 species without any red-listed and only 13 HNV taxa. This is considerably less than the 289 taxa that were recorded in 2008 in five floodplain grasslands of northern Germany in a sample of 277 relevés (Wesche et al. 2012). One reason for this result is that we focused explicitly on intensive grassland for dairy production, which was not the case in the latter study. The steep decline in plant species richness is also demonstrated by the large drop in α-diversity from typically 28–40 species per 20–30 m2 in wet grasslands before agricultural intensification (Dierschke 2008) to on average 16–23 species per 200 m2 in recent time. The high frequency of Lolium perenne, Festulolium and other commonly sown forage grasses such as Festuca pratensis, Phleum pratense or Poa pratensis indicates that many fields have been harrowed and resown in the past years. Not only were red-listed species completely absent at our grassland sites, it is also evident that once widespread and common grassland species such as Achillea millefolium, Anthoxanthum odoratum, Bellis perennis, Galium mollugo, Lotus corniculatus, Lychnis flos-cuculi, Ranunculus acris and Rumex acetosa have been reduced to low frequencies in the current vegetation, similar as was reported by Wesche et al. (2012), Leuschner et al. (2014b) and Krause et al. (2014). This demonstrates that high-intensity management now extends over the entire farmland area, with virtually no remnant habitats being left in the study region.

Even though average plot-level diversity increased from a very low 16 species to 23 species per 200 m2 from the grassland interior to the edges, only few additional HNV-grassland species were recorded (including Anthriscus sylvestris, Galium palustre, Lathyrus pratensis, Lotus corniculatus and Vicia sepium). Clearly, field edges do not function as refugial habitats of plant species with conservation interest in our study region, as was still the case 30 to 40 years ago (Konrad and Ruthsatz 1993; Husicka and Vogel 1999).

With a total of 107 taxa, the seed bank had a lower species richness than found in earlier studies in mesic and wet grasslands of Central Europe (Bossuyt and Honnay 2008; Wellstein et al. 2007). Clearly, a comparison among different studies is complicated by largely different sample numbers and a varying size of soil volume examined. Yet, we detected no red-listed species, and the small number of HNV-grassland species (12 in total) is a clear sign of impoverished seed banks, reflecting decades of intensive management (cf. Bekker et al. 1997). A similarly impoverished seed bank was found in a recent survey of mesic grasslands in three regions of Germany (Klaus et al. 2017). As expected, the severe species reduction in the aboveground vegetation is linked to similar impoverishment of the seed bank, which greatly reduces the restoration potential of the degraded grasslands.

Field edges as refugia?

The increase in plot-level species richness from the field interior to the edge in the aboveground vegetation (+ 44%) was also found in the seed bank (+ 25%). The species that were more common, or occurred exclusively, in the seed bank of edges as compared to the grassland center (such as Urtica dioica, Persicaria maculatum and Polygonum aviculare) are widespread forbs that are not listed as HNV species. In fact, the bulk of seeds that germinated from edge soil were Juncus bufonius, which was not found in the aboveground vegetation. Juncus species are a dominant element in the seed bank of most temperate grassland communities, as they produce large quantities of persistent seeds (Bakker and Berendse 1999; Valkó et al. 2010). The abundance of Juncus spp. is also the main reason for the relatively high density of seeds in our samples (on average > 45,000 m−2), despite rather low diversity.

One possible explanation of the higher number and species richness of seeds at grassland edges is that soil excavated from ditches is deposited here (Rasran and Vogt 2018). Another possible factor is the absence of fertilization in riparian buffer strips at ditches that might promote less N-demanding species at the cost of more competitive taxa. The greater abundance of seeds of Agrostis capillaris and Rumex acetosella with lower N demand (Ellenberg indicator values 4 and 2, respectively) might indeed be interpreted as an effect of reduced fertilization on grassland edges. On the other hand, our model that analyzed the influence of different types of bordering habitats on seed banks of grassland edges did not detect a significant habitat type effect. This suggests that ditch edges and their specific vegetation composition, together with maintenance works and reduced fertilization, seem not to be a key factor explaining the richer seed bank. In fact, grassland edges bordering road verges without a ditch also harbored richer seed banks than the interior, which makes it more likely that seed spill-over from adjacent habitats is an important factor causing higher seed numbers and species diversity in the seed bank of grassland edges. Yet, the considerable number of germinated seeds of Callitriche ssp., Epilobium palustre, Ranunculus sceleratus and Rorippa palustre demonstrates that field edges adjacent to ditches are enriched by a number of taxa specific to this habitat. Like our results, Rasran and Vogt (2018) found abundant seeds of various characteristic ditch edge plants such as Ranunculus sceleratus, Rorippa palustris, Sagina procumbens and J. bufonius in their seed bank analysis of northern German grassland ditches. Finally, the lower pH at the field edges could also have contributed to higher seed densities, as the size and longevity of grassland seed banks generally increases with decreasing pH (Basto et al. 2015). Thus, the richer seed bank at field edges is likely caused by more than one factor, including effects of spill-over from adjacent habitats, increased within-site habitat and soil diversity, and possibly reduced fertilization.

When comparing seed bank diversity to the diversity of the aboveground vegetation (107 vs. 148 taxa), it must be kept in mind that the seed bank analysis covered less than 0.5% of the ground area of the vegetation relevés. Thus, examining more soil samples may well show that the seed bank is still richer than the aboveground vegetation (Vandvik et al. 2016).

Soil and management effects

The soil chemical data demonstrate that the grassland edge (but not the margin) has a somewhat lower pH and lower exchangeable P content in the topsoil than the interior. This likely is a consequence of 13 years (2006–2019) of non-application of fertilizers and liming in most buffer strips adjacent to ditches. Less clear is the contribution of the deposited ditch excavates.

The soil influence (marsh vs. moor) on species richness was mixed. While aboveground vegetation was slightly (but significantly) species-richer on moor soil (Fig. 3), which is nutrient-poorer, no difference was detected for the seed bank. Irrespective of the position and the soil type, the P concentrations were apparently so high that the species number would not respond to slight fertility differences (Janssens et al. 1998; Riesch et al. 2018). In addition, the long-term intensive management also contributed to vegetation homogenization, resulting in only small diversity differences between the grassland communities (Wesche et al. 2012).

Grazed plots harboured a higher diversity in the aboveground vegetation than mown or grazed and mown plots. Similarly, grazing increased the number of species in the seed bank compared to mowing, while management regime had no influence on the number of germinated seeds. While both grazing and mowing are intense at the study sites, it appears that grazing has, in relative terms, a positive effect on the vegetation by increasing small-scale heterogeneity of the sward (Klimek et al. 2007), which affects also the seed bank. Ma et al. (2018) found that grazing indeed increases the quantity of transient seeds, while decreasing the number of persistent seeds, which are important for ecosystem restoration (Bakker 1989). Jacquemyn et al. (2011) also found higher seed densities in grazed plots. On the other hand, grazing animals are defecating and urinating also at field edges, which could reduce the de-eutrophicating effect of grassland buffer strips in grazed fields. In fact, a global meta-analysis of grazing effects on the seed bank of grasslands revealed that low-intensity grazing increases seed bank richness, while moderate grazing had no effect and high-intensity grazing decreases richness (Shi et al. 2022). It is thus likely that intensive rotational grazing is rather preventing the establishment of a richer seed bank in riparian buffer strips, than supporting it.

Conclusions

Intensively managed dairy grassland in north-western Germany has a strongly reduced biodiversity, both aboveground and belowground, and on marsh and moor soils. Riparian grassland edges are somewhat richer, but do not function as refugia for species of conservation interest, and they have no potential for successful restoration from the seed bank. Restricting the application of fertilizers and pesticides in riparian buffer strips may help reducing the leaching of nutrients and pesticides from intensive fields and lowering soil nutrient contents to some extent, but it is not sufficient for restoring a richer phytodiversity at grassland edges. To achieve the goal of establishing refugia of grassland biodiversity also in intensively managed grasslands, it is necessary to reduce nutrient input in buffer strips. In addition, there is a need to transfer seeds or green hay from rich grasslands, and to reduce management to one or two cuts per year with a sufficiently long period in-between, so that plants can flower.

Since unfertilized riparian buffer strips may in any case produce forage with lower quality, which is best fed to heifers and non-lactating cows, it might be advisable for the farmer to spare buffer strips from the intensive mowing cycle of the grassland and to harvest the buffer strip hay separately only once or twice in summer. Results from various restoration trials demonstrate that substantial increases in phytodiversity can be achieved with this procedure even in former intensive grassland (Wagner et al. 2021; Valkó et al. 2022). If a variable part of the buffer strips is left uncut in autumn, the benefit for plants, vertebrates and invertebrates will be even greater. Moreover, when the improved buffer strips are arranged strategically in the landscape, they could function not only as refugia, but could also enhance habitat connectivity.