Introduction

Along with agricultural intensification, strong and ongoing decline of biodiversity in the Central European landscapes have been documented (Wesche et al. 2012; Meyer et al. 2013; Richner et al. 2015; Seibold et al. 2019), and even where plant richness remained more stable, populations of many species declined, often coupled with a strong species turnover (Jandt et al. 2022). Although edges have declined with increasing field size, agricultural landscapes still include a lot of transition zones between habitats of different land use and thus different quality with respect to biodiversity. Edge effects can expand deep into the adjoining habitats and therefore influence a considerable amount of the landscapes (Ries et al. 2004). Much attention has been paid to edge effects, not least because of their obvious importance for conservation and protected area design, but the patterns are still not fully understood.

Ries and Sisk (2004) developed a predictive model to explain species responses to edges between habitats of different quality. They predict increasing, decreasing or neutral response curves, depending on the distribution of resources in the two adjoining habitats. Starting from an unfavourable habitat to a neighbouring preferred habitat, response curves should be increasing if resources are not divided between the two habitats, while they should show a peak at the edge in case of divided resources (Fig. 1). Although this framework was developed for the abundance of individual organisms, it should also work and was already used for species richness (Wimp and Murphy 2021; Barahona-Segovia et al. 2023).

Fig. 1
figure 1

Theory-based models for possible edge responses (for abundances) between a high and low quality habitat. Prediction and illustration according to Ries and Sisk (2004), simplified

Full size image

In Central Europe, arable fields and grasslands together typically cover around 50% or more of the cultural landscapes (Eurostat 2011, Eurostat 2018), yet the interface of these two open habitats is surprisingly poorly studied. Most studies on species response curves at edges look at the interface of forests or other woody structures and open habitats (Erdős et al. 2013; Vespa et al. 2014; Burst et al. 2017; van der Mescht et al. 2023). Moreover, the published studies lack the spatial resolution to detect fine scale response patterns. Studies often include only one side of the edge, sometimes only the first few meters are sampled, or the samples are taken only from very few rough distances rather than continuously. The few studies that focused on small-scale species responses are on arthropods (Pe’er et al. 2011; Madeira et al. 2016; Rischen et al. 2023). An intensive search via Google Scholar revealed, that plant species response has not been studied with high resolution and in a symmetric fashion at the crop field–grassland interface, not to mention semi-natural grasslands (SNG).

Here, we contribute to the overarching topic of how plant species richness responds at the transition of crop fields and SNG. According to Ries and Sisk (2008), edge effects and their magnitude are determined by habitat preferences of respective species. For overall plant species richness, we assume a monotone increase from the field into the SNG (Fig. 2). Edges of SNG have been shown to host less diverse plant communities than their interior (Schöpke et al. 2023). At arable field edges, conditions for plants are less favourable than in SNG (Batáry et al. 2013), but still better than in the field interior (Wietzke et al. 2020). This is due to competition with crops for light (Kleijn and van der Voort 1997; Fried et al. 2009; Kovács-Hostyánszki et al. 2011; Seifert et al. 2014), water and eventually for nutrients (Fried et al. 2009). Moreover, weed control is weaker there (Wilson and Aebischer 1995; Fried et al. 2009) and the spillover from adjacent habitats is higher (Metcalfe et al. 2019).

Fig. 2
figure 2

Hypothesized responses of species richness at the edge between arable fields and semi natural grasslands for all species, arable specialists (A), generalists (G) and grassland specialists (S)

Full size image

For the often nutrient demanding generalists, we expect a positive response to the edge on both sides, leading to a peak at the transition between the two habitats, and a higher richness in the grassland, as SNG are generally more species-rich (Fig. 2). Richness of grassland specialist species should increase monotonously towards the SNG interior, but the increase will start just at the edge and has a steeper slope (Fig. 2). The unfavourable conditions for many grassland specialists at SNG edges are caused by farming practices of the crop field, namely fertilization and pesticides, and either directly or indirectly by the encroachment of competitive generalists (Labadessa et al. 2017). In turn, as the field offers habitats with far higher quality for the arable specialists, we assume a strong decline into the SNG for this group. However, as even for these habitat specialists the edges at least of fields with conventional pesticide-based cultivation methods are more favourable than the field interior (Kleijn and van der Voort 1997; Seifert et al. 2014; Aguiar et al. 2023), we assume a positive response to the field edge, leading to a peak in species richness just before the boundary (Fig. 2).

Here, we studied the vascular plant species richness in a continuous sampling approach along transects running from the interior of arable fields into the interior of dry SNG to test the following hypotheses (cf. Figure 2): (H1) Across the interface of arable fields and SNG, total plant species numbers show a sigmoidal fit, with a higher richness in the grassland. (H2) Grassland specialists also show a sigmoidal fit, but with a steeper slope starting right at the edge. (H3) Arable specialists show a maximum at the field edge and a strong decline in the SNG. (H4) Generalists reach their maximum along the edge of both habitats and show higher species numbers in the grassland. (H5) The patterns of the species response curves are associated to community-level Ellenberg indicator values, reflecting water, nutrient and light conditions.

Methods

Study regions

The study was conducted in two German landscapes: the Uckermark and the Eifel (Fig. 3a). The climate of both regions is temperate, but there is a stronger influence of continentality in the Uckermark than in the more Atlantic climate of the Eifel. In correspondence, the mean annual rainfall is about 500 mm in the Uckermark and 860 mm in the Eifel. The annual mean temperature is 8.9 °C and 8.0 °C, respectively (stations Angermünde and Kall-Sistig 1981–2010; DWD 2021). Our study sites in the Uckermark and Eifel were about 60 and 420 m a. s. l., respectively.

Fig. 3
figure 3

Transect locations in the study regions, Eifel in the West and Uckermark in the North-east of Germany, with yellow dots for organic sites and white dots for conventional sites (a), and scaled representation of the sampling design (b): One transect consists of twelve 2 m x 2 m vegetation plots (grey quadrats), placed radially in a clockwise direction around midpoints with distances of 25 m to capture heterogeneity in two dimensions. The design has its origin in a broader research project (see https://bonn.leibniz-lib.de/de/forschung/projekte/ integrative-analysis-of-the-influence-of-pesticides-and-land-use-on-biodiversity; source of Germany location map from Wikipedia by NordNordWest, used under CC BY-SA 3.0; Sentinel-2 cloudless - https://s2maps.eu by EOX IT Services GmbH; contains modified Copernicus Sentinel data 2016 & 2017)

Full size image

The Uckermark in north-eastern Germany is a young moraine landscape with base-rich permeable soils, namely Retisols, Luvisols and Arenosols (LBGR 2012, nomenclature according to WRB 2006). The Eifel in the west of Germany represents a very old mountain range with some limestone basins (Meyer 2013), and corresponding soils in the centre with grasslands on rendzic Leptosols, and arable fields mostly on Cambisols (GDI 2021; LGBR 2021). In both regions, on sites not suitable for arable farming or forested, semi-natural calcareous and low-nutrient dry grasslands can be found.

In each study region, we selected ten protected (Natura 2000; EC 1992), dry semi-natural grasslands (SNG) directly adjoining arable fields and being large enough for our study design (see below). Six of the adjoining arable fields were managed organically and 14 conventionally (Table A.1). These grasslands, between 1 and 35 ha in size, were grazed as a conservation measure.

Vegetation survey

We studied the vegetation once per site between end of May and July, in 2019, 2020 and 2021. Single sampling may cause failure to detect early season annuals, but we expected only limited problems here as we frequently detected at least remnants of most annuals, such as Cerastium semidecandrum, Draba verna/praecox and Holosteum umbellatum throughout the entire sampling period at all sites. We placed transects orthogonally to the field/grassland border, stretching each 32 m into both habitats. Transects consisted of twelve 2 × 2 m² plots distributed as shown in Fig. 3b. All vascular plant species on the plots were recorded, cover values were estimated using the decimal scale of Londo (1976), which we extended by 0.05 and 0.01 (i.e. 0.5% and 0.1% of cover).

For all species, habitat preferences were defined according to their phytosociological affinities. We first defined grassland and arable specialists, using information from the German Species List 1.5 (GSL; Jansen and Dengler 2008; for phytosociological units see Table A.2). For species assigned as indifferent or without data, attribution was done according to expert knowledge. All remaining species were assigned as generalists, with the exception of crops, fruit trees and those taxa that could not be identified to species level (for full species list with habitat preference see Table A.3).

Statistical analyses

Data entry was done using TURBOVEG (Hennekens and Schaminée 2001), and all analyses were carried out with R version 4.3.0 (R Core Team 2022). To import the data and harmonize them taxonomically, we used the R package vegdata (v0.9.11.4; Jansen and Dengler 2010) with the implemented GSL. Ellenberg Indicator Values (EIV) were also taken from the GSL. As recommended by Carpenter and Goodenough (2014), we used unweighted mean Ellenberg indicator values (EIV).

According to Ewers and Didham (2006), response curves should be modelled with one continuous function. We fitted various linear models for richness as response and distance to the field border as predictor (Table 1), according to our hypotheses. We used function lm for the null model and models with polynomials (quadratic and cubic), and non-linear models with function nlsLM (package minpack.lm; v1.2-3; Elzhov et al. 2023) with self-start functions (if not implemented in R base, we used the nlraa package; v1.5; Miguez 2022). We then selected the best fitting model according to the Akaike Information Criterion (AIC).

Table 1 Models with description and the used R functions
Full size table

Results

Over all transects and both habitats, we found 344 vascular plant species, 169 in the arable fields and 292 in the semi-natural grasslands. Of these, we classified 111 species as grassland specialists and 69 as arable specialists, while 152 species were assigned as generalists. The rest were counted, but not assigned to species level, e.g. Festuca ovina agg., or not counted but assigned, e.g. Anagallis as a genus. Of the arable specialists, 59% occur also in grasslands, and 21% of the grassland specialists occur in fields, too. Of the generalists, we found 43% in the fields and 92% in the grasslands.

The best fitting model in our analyses for all species was the cubic one, i.e. with a minimum in the arable field and a peak in the grassland (Fig. 4a, for AIC see Table A.4). For generalist species, the cubic model fitted our data slightly better than the bell-shaped model (Fig. 4b, for bell-shaped see Figure A.1). For grassland specialists, the sigmoid model showed the best fit, albeit these increased only in the grassland (Fig. 4c). The bell-shaped model was best for the arable specialists (Fig. 4d). The same best models were selected when using data from organic sites alone, with generalists being the sole exception showing a bell-shaped curve very similar to the cubic one (see Table A.5 and Figure A.1). All groups had higher species richness in organic sites (Fig. 4, Figure A.2).

Fig. 4
figure 4

Number of species per plot over distance to the border between arable field (left) and semi-natural grassland (right) for all species, specialists and generalists, with the respective best fitting model

Full size image

The community-level EIV for light showed a similar response as the richness of all species, while the responses of the EIVs for nutrients and moisture were equally similar, yet inverse (Fig. 5).

Fig. 5
figure 5

Ellenberg Indicator Values for light, moisture and nutrients, calculated as unweighted mean per plot, over distance. Filled and open dots are organic and conventional sites, respectively

Full size image

Discussion

To describe plant species richness responses across the interface of crop fields and semi-natural grasslands, we studied the vegetation continuously into the interior of both habitats. In summary and as expected, we found a monotonous response with higher species numbers in the grasslands, driven mostly by grassland specialists and generalists. The high presence of specialists and generalists in the grasslands compensated the inverse response of arable specialists.

All species

As predicted for edges between habitats with different quality and supplementary resources (Ries and Sisk 2004), richness of all species showed a transitional response. In contrast, Dutoit et al. (2007) found no increase in total plant richness from crop fields to grassland, yet their fields were under an agri-environmental scheme aiming at conserving species richness. Also, studies along the border between forests and grasslands showed results different from ours. Burst et al. (2017) found a maximum in the grassland edge and a decline into the grassland for overall species richness. This was in spite of the grasslands being of high quality, and was mainly caused by a high generalist richness, and by forest and edge specialists occurring at the grassland edge. Erdős et al. (2013) explained the maximum, which in their case occurred at the border between natural forests and grasslands, with intermediate conditions at the edge with respect to light, temperature and soil moisture.

The generally higher richness of plant species at crop field edges rather than in field interiors has repeatedly been documented before. It was explained with a spillover from neighbouring habitats, less effective (Batáry et al. 2012) plus less intense pest and weed management activities, more heterogeneous microhabitat structure, greater proximity to adjoining semi-natural habitats (Gayer et al. 2021), as well as less fertilizers and more light (Gabriel et al. 2006). Regarding all species, the increase is not surprising, as for many species crop fields are habitats of lower quality compared to most adjacent habitats.

Interestingly, our curve for all species reaches a minimum and a maximum after about 20 m into the field and the grassland, respectively, before it turns (up and down, respectively) again. In few cases, the impact of the next habitat border might be visible here. However, at least for the grasslands, we consider the peak to be part of the edge effect, driven by patterns in generalists, which also peak here, as discussed below. Rischen et al. (2023) also found a minimum for the response of carabids species richness in crop fields, and a decrease for spiders in the fallow interior.

Generalists

As expected, the generalist species had higher numbers in the grasslands, but surprisingly, they seemed to carry the peak of all species at about 15 m into the grasslands. This peak may show two edge effects on different spatial scales: Coming from the grassland interior towards the edges, an increase of species richness due to a spillover of additional generalists is visible. Very close to the edge, this increase is counteracted by the detrimental impact of the neighbouring farm practices. Generalists should else show their highest richness at the edge, as shown for generalists in grasslands next to forests (Burst et al. 2017). In our study, we expected a peak at the border on both sides of the habitat edge. For ubiquitous insects at open canopy plantation–grassland interface, van der Mescht et al. (2023) found a similar curve, but with the first sample at 15 m, not allowing for any inferences closer to the edge.

Looking at the entire species pool of the study, most species (152) are categorized as generalists. However, per plot, grassland specialists have higher species numbers, and arable specialist richness is still high in relation to the mere 69 species assigned to this group here. This implies that specialist species occur with higher constancy, while generalists have more scattered occurrences.

Almost all generalist species occurred in the grasslands, while only less than half of them were found in the crop fields. Semi-natural grasslands offer favourable conditions for a wider range of species compared to arable fields, which in turn are unfavourable even for many generalists. When regarding the whole habitat patch and compared to these grasslands, arable fields often have relatively homogenous conditions, with respect to micro-relief, vegetation structure and microclimate.

Although not the focus of this study, we like to briefly highlight the generally higher species richness in organic sites. This has to be seen with caution, as the farm type groups are unbalanced and might be biased regarding their distribution. Still, the pattern is surprising, especially with respect to the interior of the protected grasslands as they are all managed in a similar fashion, independent of the adjoining arable field (for details see Schöpke et al. 2023). However, conventional farming practices, including mineral fertilization and synthetic pesticide application, appear to affect the generalists less than do organic farming practices, especially so when compared to the other species groups (Figure A.2). For nutrient demanding species, the higher fertilization could even be beneficial, which in turn suppresses other species due to competition.

Specialists

The sigmoidal curve for the grassland specialists starts to rise only at the very edge of the arable field, reflecting the highly unfavourable conditions for these specialists when land use changes from grazing to arable farming. Organic fields thereby harbour at least some of them. Besides land use change, the relief causes differing conditions, where grasslands are located on slopes and crop fields on the valley ground beneath. Due to runoff the valley grounds are naturally wetter and richer in nutrients. In line with this, only one fifth of all grassland specialist species of this study occurred in the fields. A very similar response curve for grassland specialists was found by Burst et al. (2017) at the forest-grassland interface. Notably, grassland specialists reach full species richness only after 10 to 15 m into the grassland, showing edge effects and detrimental impact of the farm practices applied in the neighbouring habitat. For organic sites alone, we would expect a less detrimental impact, since organic farming does not use additional synthetic pesticides and mineral fertilisers, what also means less pesticide and fertiliser drift (Aude et al. 2003; Holzschuh et al. 2008; Marshall and Moonen 2002). However, since these sites show the same pattern as the conventional sites, relief again could serve as a further explanation, as it may prevent the pollutants from drifting onto the often higher-lying grasslands. A decreasing species richness towards the border with the crop field was also shown for field margin vegetation (Rundlöf et al. 2010) and small SNG vegetation (Kiviniemi and Eriksson 2002). Winsa et al. (2015) found no such decrease towards the edge of restored SNGs and explained this unexpected result with more shrub and tree encroachment in the interior, higher solar radiation at the edge and succession following abandonment.

Despite the pattern for plant species richness in general, it may seem surprising that specialists for arable fields show highest richness at edges rather than in field interiors. As disturbance tolerant species, they can grow in fields and edges, but profit at edges from less intensive weed control and reduced competition from crops (Wietzke et al. 2020). In consequence, if not neighbouring another field (Wietzke et al. 2020), field edges became refuges for many of today’s often threatened arable specialists, making richly structured landscapes so valuable (Batáry et al. 2017). In our study, more than the half of the arable specialists occurred in grasslands as well, but scattered and with high turnover, as indicated by the relatively low species numbers in the grassland plots. This reflects the somewhat similar conditions of both habitats regarding the availability of light and open soil. In turn, only one fifth of the grassland specialists also occurred in the crop field.

The more heterogeneous conditions of the SNG compared to the fields are also reflected in the specialist groups. Both in our study and more general in the regional species pool, considerably less species are classified as arable specialists than grassland specialists.

Ellenberg indicator values

The effects of farm management practices on the neighbouring grassland are as well reflected in the community level EIV. We found a decrease of EIV for nutrients and increasing EIV for light from the field border into the grassland, showing spillover effects of fertilization into adjacent habitats, plus the dense cover of crops. Further, we saw decreasing EIV for moisture, which can not only be explained by the relief of the grasslands. The few grasslands that were neither located higher than the crop fields nor were situated at a slope show the same decrease for moisture. This could be due to a more nitrophytic and thus denser vegetation in the edges, preventing moisture to evaporate. Moreover, the hidden correlation between EIV for moisture and nutrients shows an effect here. This means, many species with high nutrient demand (or tolerance) coincidently have a high demand on water (Figure A.3). The mentioned patterns were visible for both farming types. Not directly studied here, but probable further impacting species diversity, are pesticide applications.

For the first time, we could disentangle fine-scale patterns of plant species richness across edges between crop fields and semi-natural grasslands far into the two adjacent habitats. Due to the resource distribution at the interface of different land use patches, reflected in ecological indicator values, species groups show different responses according to their habitat preferences. We found edge effects to expand far into adjacent habitats, and that organic farming led to a higher plant diversity, not only in the farmed field, but even in the adjacent habitat, showing that organic farming is a good possibility to support typical arable flora as well as adjoining semi-natural grassland flora. When conserving biodiversity in agricultural landscapes, the interplay of neighbouring land use systems should thus be taken into account, particularly if protected areas are small and scattered. We did not explicitly study buffer zones, but the observed biodiversity patterns across the edges support the idea that even by sheer distance alone buffer strips would foster the quality of neighbouring habitats. Still better, organic farming practices should serve as a standard rather than exception.