Farmland biodiversity: wildflower-sown islands within arable fields and grassy field margins both promote spider diversity

Agricultural intensification and the concomitant landscape homogenization is leading to a worldwide decline in farmland biodiversity. Non-crop habitats in agroecosystems may counteract the loss of arthropods such as spiders and thus contribute to sustainable agriculture. However, the effectiveness of field margins and set-aside wildflower-sown patches in maintaining spider diversity is not well understood. Here, we investigated the effects of three different non-crop habitats, namely field margins, set-aside wildflower-sown patches under power poles (‘power pole islands’), and grassland fallows on spider diversity as compared to wheat fields in an agricultural landscape in western Germany. Using pitfall trapping and suction sampling, we show that species richness and overall conservation value were higher in non-crop habitats than in wheat fields. Interestingly, field margins and power pole islands differed from long-term grassland fallows only in conservation value, which was significantly higher in grassland fallows. Species assemblages differed considerably between grassland fallows, field margins and power pole islands, and wheat fields, documenting the added value of using different conservation strategies. Small-scale non-crop habitats adjacent to wheat fields were surprisingly effective in promoting spider diversity in an agricultural landscape, with field margins and power pole islands being equally effective. To maximize overall diversity in agricultural landscapes, we propose a combination of larger long-term fallows and smaller non-crop habitats such as field margins or set-aside wildflower-sown patches.


Introduction
Though agriculture is essential for providing food and other resources to an ever-growing human population, intensification of land use is considered to be one of the main causes of global biodiversity loss (Donald et al. 2001;Benton et al. 2002;Seibold et al. 2019). Increasing mechanization and the demand for higher crop yields have led to a simplification of agricultural landscapes by enlarging field sizes and reducing crop types, and to the widespread use of fertilizer and pesticides Tryjanowski et al. 2011). In parallel, the proportion of semi-natural landscape elements, which are important for protecting biodiversity, has decreased in agricultural landscapes (Billeter et al. 2008;Fahrig et al. 2011). The frequent disturbance on cropland, e.g. due to the application of agrochemicals and tillage, has detrimental effects on farmland biodiversity (Thorbek and Bilde 2004;Batáry et al. 2008Batáry et al. , 2012. Thus, intensively used crop fields are unsuitable for many organisms, and agricultural intensification comprises a major driver of biodiversity loss in industrialized countries (Donald et al. 2001;Kleijn et al. 2009).
To counteract the widespread decline of biodiversity in agricultural landscapes, a wide variety of measures has been applied in recent decades, some of which aim to increase landscape heterogeneity, an important driver of species richness (Benton et al. 2003;Fahrig et al. 2011). Especially for habitat specialists, the size and spatial configuration of noncrop habitats are important determinants for the maintenance of viable populations (Öberg et al. 2007;Pluess et al. 2010;Knapp and Řezáč 2015;Šálek et al. 2018). Only heterogeneous landscapes, composed of arable fields and adjacent non-crop habitats, may offer essential features such as hibernation and feeding places, refuges, and dispersal corridors (Pfiffner and Luka 2000;Sunderland and Samu 2000;Östman et al. 2001).
Semi-natural habitats, e.g. long-term grassland fallows, are also important for the conservation of plant, arthropod and bird species in agricultural landscapes (e.g. Benton et al. 2003;Hendrickx et al. 2007;Rischen et al. 2021). They typically show a more complex vegetation structure, more stable conditions, and more variable soil moisture compared with the surrounding crop fields (Plath et al. 2021). Other non-crop habitats such as grassy field margins and set-aside fields, can link isolated fragments of semi-natural habitats and contribute to landscape heterogeneity (Critchley et al. 2004;Kovács-Hostyánszki et al. 2011;Morris et al. 2011). In general, non-crop habitats provide refuges for arthropods during field manipulations (Schmidt-Entling and Döbeli 2009), and contribute to the conservation of generalist predators such as beetles and spiders in agricultural landscapes (Frank et al. 2012;Plath et al. 2021). Overall, the conservation and restoration of non-crop habitats is thus one of the most effective conservation measures in intensively managed agricultural landscapes (Tscharntke et al. 2002;Knapp and Řezáč 2015). Therefore, agri-environmental schemes within the European Union's Common Agricultural Policy (CAP) aim to promote such habitats in agroecosystems (Holland et al. 2017). These include, among others, fields margins and set-aside crop fields subsequently sown with wildflowers (Van Buskirk and Willi 2004;Haaland et al. 2011;Plath et al. 2021).
While various greening measures are applied in agricultural landscapes, best practises for promoting spider diversity are not resolved (Mestre et al. 2018;Ganser et al. 2019). Spiders are one of the most abundant and diverse invertebrate groups in agricultural landscapes, and contribute to ecosystem services by feeding on pests (Symondson et al. 2002;Michalko et al. 2019). In particular, so-called 'agrobionts', i.e. spider species thriving in agricultural landscapes, can be dominant natural enemies (Luczak 1979;Samu and Szinetár 2002). Even for agrobionts non-crop habitats can be important, for example as overwintering sites (Schmidt-Entling and Döbeli 2009). More specialised spider species may be restricted to certain habitat types, such that spider assemblages may differ between non-crop habitats (Nardi et al. 2019). Moreover, landscape and habitat features may influence the functional diversity of spider communities (e.g. body size, occurrence of ecotypes, preference for specific strata; Schirmel et al. 2016;Gallé et al. 2018). For example, body size may be related to land-use intensity, with smaller body size being associated with more disturbed habitats (Schirmel et al. 2016;Plath et al. 2021). In addition, body size and habitat preference can serve as indicators of the extent of ecosystem services provided by spiders in agricultural landscapes (Rusch et al. 2015). For evidence-based conservation recommendations, it is hence necessary to investigate the effectiveness and conservation potential of different non-crop habitats for spiders in agroecosystems (e.g. Košulič et al. 2014). While the impact of field margins on spiders has been addressed in several studies (e.g. Huusela-Veistola 1998; Gallé et al. 2018;Zhang et al. 2020), the effectiveness of set-aside wildflower-sown fields for spider conservation is largely unclear.
We here compare the effectiveness of three non-crop habitat types, namely long-term grassland fallows, field margins, and wildflower-sown patches under power poles ('power pole islands'), in relation to wheat fields for promoting spider diversity in an agricultural landscape in western Germany. More specifically, we compare spider taxonomic species richness and diversity, community structure, and functional diversity between these habitat types. We predicted that, due to the lower management intensity, (1) species richness and diversity of spiders are higher on power pole islands and field margins than on wheat fields, but lower than on grassland fallows with their larger patch size (see Bonte et al. 2002). Since the occurrence of spiders is related to the availability of appropriate environmental niches (Michalko and Birkhofer 2021), we furthermore predicted that (2) wheat fields have a higher proportion of xerophilous and euryoecious spiders compared to non-crop habitats due to their dry and disturbed conditions. Finally, we predicted that (3) grassland fallows exhibit the highest proportion of hygrophilous spider species, while field margins and power pole islands harbour more agrobionts, as edge effects may favour their immigration due to their small area and proximity to wheat fields (Samu and Szinetár 2002;Blitzer et al. 2012). Consequently, we expected differences in community structure between habitat types, allowing us to explore which combination of habitats enhances overall spider diversity the most in an agricultural landscape.

Study area
The study was conducted in an intensively used agricultural landscape dominated by crop fields within the Eifel mountain range in western Germany (50° 14′ N, 7° 21′ E; Fig. 1). The study area belongs to the 'Maifeld' region, characterized by an oceanic climate with a precipitation of ~ 598 mm/year and a mean annual temperature of ~ 10 °C (Agrarmeteorologie Rheinland-Pfalz 2020). To a small extent, non-crop habitats, for example, grassland fallows, field margins, and set-aside areas sown with wildflower mixtures occur. Agricultural lands comprised 72.2%, forests 11.8%, and settlement and traffic areas 13.9% of the Maifeld in 2018 (Statistisches Landesamt Rheinland-Pfalz 2018). To evaluate the effects of land use on spider assemblages, we investigated 1 3 four habitat types with nine replicate plots each (total n = 36; Fig. 1; Table A1, Supplementary Material): (1) wheat fields sown with Triticum aestivum (conventional management with fertilizing, fungicide and herbicide spraying), (2) grassy field margins bordering wheat fields (mown once during sampling), (3) set-aside habitat islands under power poles (2 years old fallows under poles of a power line, sown with a mixture of wildflower seeds as 'greening' measure, 12 × 12 m; hereafter: 'power pole islands'), and (4) grassland fallows (unmanaged long-term grassland fallows with a diverse natural vegetation cover, formerly used as meadows but abandoned decades ago). Limiting factors for our comparisons were the availability of grassland fallows (being rare in the study area) and power pole islands (only found along a power line). Therefore, we 'clustered' our sampling plots somewhat, by always choosing a wheat field close to a grassland fallow, and a field margin close to a power pole island (Fig. 1).

Sampling methods
We collected spiders between the 15th of April and the 7th of July 2020 using two complementary methods, pitfall trapping and suction sampling. For pitfall trapping, we used two PET cups (∅ = 5.6 cm, volume 125 ml, filled with 70% water and 30% monopropylene glycol) per replicate plot as traps (4 habitat types × 9 locations × 2 traps). Pitfall traps were covered by a plastic roof for protection and buried into the ground, 5 m apart. In field margins and in power pole islands, pitfall traps were centrally located, while in wheat fields and grassland fallows they were set up in a distance of 30 m from the edge. Traps were emptied every 2 weeks (i.e. a total of six times), with the number of trapping days ranging between 82 and 83 days per plot. Samples were stored in 70% ethanol for later identification. Suction sampling was performed on the same plots using a leaf blower (Stihl SH 56, Dieburg, Germany) with a polyamide stocking attached to the front of the nozzle. Each plot was sampled twice on dry days in May and June 2020, 2 weeks apart. Each sampling unit consisted of suctioning the vegetation along a 5 m transect (i.e. between both pitfall traps), inserting the nozzle into the vegetation ten times for 7 s each (Brook et al. 2008). Samples were transferred to plastic bags, frozen at − 18 °C, and then stored in 70% ethanol. All adult spiders captured were identified to the species level using taxonomic keys (Roberts 1987(Roberts , 1995Nentwig et al. 2020).

Data analyses
Data from pitfall trapping and suction sampling were pooled for subsequent analyses. For each plot, we calculated the total numbers of species and individuals (although pitfall trapping gives data on activity density only, we here use 'individual number' because of the inclusion of the individuals from suction sampling), the effective number of species, which is a derivate of the Shannon-Wiener index and represents the actual diversity (Jost 2006), and the 'conservation value' according to Görn and Fischer (2011). The latter was determined by assigning a point score to each species according to its German red list status (Blick et al. 2016; Table A2, Supplementary Material): least concern = 1 point, near-threatened = 5 points, risk assumed = 10 points, vulnerable = 15 points. We also estimated the sample coverage for each plot to account for differences in detection probability between the four habitat types. Therefore, we calculated the rarefied species richness for a sample coverage of 90% using the iNEXT package (Hsieh et al. 2020) in R 4.0.5 (R Core Team 2020). The body size of each species was determined by using the mean value available in Nentwig et al. (2020). Afterwards, we calculated community weighted mean (CWM) values by weighing the respective body size by each species' relative abundance (Ricotta and Moretti 2011). Platen et al. (1991) and Platen and Broen (2005) were used to assign each species to an ecotype (hygrophilous, mostly hygrophilous, euryoecious, xerophilous species) and the preferred stratum (ground-dwelling, herb layer, taller vegetation). For all subsequent analyses, we standardized the above data by giving the respective proportion per plot.

Statistical analyses
To test for the effect of habitat type on the numbers of species, individuals, and rarefied species, on the effective number of species, conservation value, CWM body size, Fig. 1 Location of the study area near the city of Koblenz within Germany and of the study sites (enlarged) within the 'Maifeld' and the number of individuals of the most common species (dependent variables), one-way ANOVAs were used followed by Tukey HSD posthoc tests to locate significant differences. To meet ANOVA requirements, the values for the number of individuals were LN-transformed and for the numbers of individuals of the four most common species were LN + 1 transformed prior to analyses. We used Moran's I tests (ape-package in R; Paradis and Schliep 2019) to test for spatial autocorrelation of all dependent variables and their residuals (Dormann et al. 2007), being non-significant throughout. We also calculated Moran's I tests only with the 'clustered' habitat types (group 1: wheat fields and grassland fallows; group 2: field margins and power pole islands). Furthermore, we tested for differences in spider community composition across the four habitat types based on Bray-Curtis indices derived from the relative abundance of species. We used permutational multivariate analyses of variance (PER-MANOVAs, with 999 permutations), visualised by nonmetric multidimensional scaling (NMDS). In addition, PERMANOVAs (999 permutations) were used to test for significant differences in the proportion of species with specific ecotypes or stratum preferences between habitat types. We also performed pairwise PERMANOVAs to locate significant differences among habitat types. Statistical analyses were performed with Statistica 12.0 (Tulsa, StatSoft) or using the vegan package in R 4.0.5 (R Core Team 2020; Oksanen et al. 2020).

Results
In total, 4942 adult spiders belonging to 112 species and 18 families were collected. Pitfall trapping yielded 4407 spiders belonging to 91 species, and suction sampling yielded 535 spiders belonging to 63 species (Table A2, Supplementary Material). Overall, most individuals belonged to the families Linyphiidae (30% of all individuals), Lycosidae (16%), Gnaphosidae (11%), Thomisidae (8%), Theridiidae (7%), and Salticidae (4%). Five of the above six most common spider families showed the lowest numbers of individuals in wheat fields (Table A3, Supplementary Material), while the number of Linyphiidae was highest in wheat fields. Of the 112 spider species captured, 49 species were exclusively caught by pitfall trapping, 21 exclusively by suction sampling, and 42 species by both methods, with patterns being roughly similar across habitat types (Fig. A1, Supplementary  Material). Both, pitfall trapping and suction sampling identified the highest shares of all species in grassland fallows and the lowest ones in wheat fields (Fig. A2, Supplementary  Material). Only four of the species found are included in the red list of Germany as near threatened (2), risk assumed (1), or vulnerable (1; Table A2, Supplementary Material).
Habitat type significantly affected the number of species (F 3,32 = 8.14, p < 0.001), the number of rarefied species (F 3,32 = 5.45, p < 0.01), the effective number of species (F 3,32 = 4.08, p = 0.015), the conservation value (F 3,32 = 12.78, p < 0.0001), and the CWM body size (F 3,32 = 3.51, p = 0.026), but not the number of individuals (F 3,32 = 1.95, p = 0.142; Fig. 2). Differences between noncrop habitats in the above parameters were non-significant throughout, apart from the conservation value, which was significantly higher for grassland fallows than for field margins or power pole islands. Wheat fields, in contrast, differed significantly from non-crop habitats except from field margins and power pole islands for the number of rarefied species, grassland fallows for the effective species number, and field margins and grassland fallows for CWM body size (Fig. 2).

Discussion
Species richness and conservation value of spider assemblages were lower in wheat fields than in non-crop habitats. Similar patterns have been demonstrated in other studies, stressing the importance of non-crop habitats for the conservation and enhancement of spiders in agricultural landscapes (Schmidt and Tscharntke 2005;Mestre et al. 2018;Nardi et al. 2019;Pompozzi et al. 2019). Interestingly though, the three non-crop habitat types were statically indistinguishable except for conservation value, being higher on grassland fallows. The lack of differences in species richness between non-crop habitats is surprising, as grassland fallows are much larger and therefore presumably less disturbed than power pole islands or field margins. This may be due to the long history of intensive agriculture in our study area, resulting in an impoverished fauna even in such remnant sites of natural vegetation. This may also explain the overall low number of threatened species in our study. Thus, landscape-level effects may contribute to the decline of biodiversity in agricultural landscapes (Sirami et al. 2019). Fallows next to arable fields may in general have different effects on ground-dwelling predators considered, as they can serve as a sink or source (Birkhofer et al. 2018). Still, our findings demonstrate the high importance of small-scale non-crop habitats for spider conservation in agricultural landscapes. Compared to wheat fields, such habitats are relatively undisturbed and offer structurally diverse refugia with variable microhabitats that are beneficial for spiders (Clough et al. 2005;Bianchi et al. 2006).
In our study, CWM body size was larger in non-crop habitats than in wheat fields, though differences were significant between wheat fields and power pole islands only. Often, crop fields are dominated by smaller spider species, e.g. Linyphiidae, as larger species are typically less dispersive and not able to balloon . In general, larger species appear to be more strongly affected by agricultural practices such as ploughing (Holland and Reynolds 2003), which is why many species of the relatively large predatory Lycosidae, such as P. prativaga, prefer non-crop habitats (Öberg et al. 2007). Other families (e.g. Lycosidae, Gnaphosidae, Thomisidae) are favoured by litter, and therefore prefer non-crop habitats including grassland fallows (Bell et al. 2001;Clough et al. 2005).
As spider species with specific habitat requirements cannot exist in intensively managed crop fields (Knapp Fig. 3 Non-metric multidimensional scaling analyses (NMDS) for spider assemblages across four habitat types (based on 112 species and 4,942 individuals; n = 36 plots, 2 dimensions, Bray-Curtis distance, tries = 20) Fig. 4 Distribution of ecotypes (a; xerophilous; euryoecious; mostly hygrophilous; hygrophilous) and preferred stratum (b; ground-dwelling; herb layer; taller vegetation) of spiders captured in wheat fields (W), field margins (M), power pole islands (I), and grassland fallows (F). Given are percentages for each habitat type. Data were compiled from Platen et al. (1991) and Platen and Broen (2005). Different letters above bars indicate significant differences among treatments (Pairwise tests after PERMANOVAs; p < 0.05; n = 9 plots each) and Řezáč 2015), these are often dominated by a few species adapted to such environments (Samu and Szinetár 2002). Such agrobionts, including O. apicatus and T. tenuis, may reach high densities in crop fields (Topping and Sunderland 1998;Gallé et al. 2018). This may explain the lack of differences in individual numbers between habitat types in our study. In particular, many Linyphiidae thrive in agricultural landscapes (Pluess et al. 2008). Members of this family can readily recolonise arable fields after disturbance due to aerial dispersal (Pfiffner and Luka 2003;Öberg et al. 2007). However, even dispersive agrobionts may need non-crop habitats, e.g. for hibernation (Thorbek and Bilde 2004;Schmidt and Tscharntke 2005), which thus comprise important source habitats (Michalko and Birkhofer 2021). Accordingly, even small-scale non-crop habitats may reach relatively high spider diversity, as these areas harbour agrobionts as well as more specialized species. Thus, proximity to non-crop habitats may increase spider diversity and thus pest control in adjacent agricultural land (Michalko et al. 2019).
Our results show that crop fields, presumably due to their aridity and periodic disturbance, were mainly dominated by xerophilous and euryoecious species, while in grassland fallows an increased proportion of hygrophilous species was found, likely due to the denser vegetation resulting in moister and cooler conditions. The spider communities from power pole islands and grassland fallows showed a higher proportion of species preferring the herb layer, which may be once again related to the more complex vegetation structure. In general, higher vegetation complexity is associated with higher species richness of spiders (Finch et al. 2008).
The above considerations suggest that species assemblages differ strongly among habitat types. This was indeed confirmed by significant differences except between field margins and power pole islands. These findings reinforce the importance of grassland fallows and small non-crop habitats for spiders in agroecosystems. They are also consistent with Feng et al. (2021) showing that older fallows, such as the long-term grassland fallows studied here, harbour more diverse communities and are particularly valuable for the conservation of specialists. In fragmented agricultural landscapes, several spider species use edge habitats for migration (Blitzer et al. 2012;Gallé et al. 2018;Nardi et al. 2019). Thus, linear field margins but also small power pole islands ('stepping stones') may improve connectivity in highly fragmented landscapes (Schirmel et al. 2016), and thereby contribute to maintaining farmland biodiversity (Bianchi et al. 2006;Hendrickx et al. 2007).
For spider surveys in agricultural landscapes, the efficiency of different sampling methods is often debated (Lang 2000;Sanders and Entling 2011). While commonly used, pitfall traps only capture the activity-density rather than the actual population density of the trapped species (Bali et al. 2019). Suction sampling may therefore give better estimates of true densities, but is largely restricted to vegetation-dwelling arthropods (Sanders and Entling 2011). Essentially, both sampling techniques complement each other by targeting different groups (Schmidt-Entling and Döbeli 2009), which is why we used a combination of both.

Conclusions
In summary, we show that small-scale non-crop habitats, such as field margins and habitat islands within wheat fields, were surprisingly effective in promoting spider diversity in an agricultural landscape. Indeed, field margins and power pole islands were as effective in promoting spider species richness as more extended grassland fallows. Nevertheless, grassland fallows had a higher conservation value and harboured different communities than field margins and power pole islands. Grassland fallows, with their structurally diverse conditions, favoured the occurrence of more hygrophilous species and ones preferring the herb layer. Notwithstanding, our results document that even small-scale non-crop habitats are important contributions to the conservation of biological diversity. To maximize overall diversity and trait composition of spider assemblages in agricultural landscapes, we propose a combination of larger long-term the box) and the median (line in the box). Whiskers represent the 1.5 times the lower or upper interquartile range and outliers are shown as dots. Different letters above bars indicate significant differences among treatments (Tukey HSD posthoc tests after ANOVAs; n = 9 plots each) 1 3 fallows and smaller non-crop habitats (such as field margins or set-aside wildflower-sown patches), as both differ clearly in species assemblages. Whether the small-scale structures involve linear field margins or small rectangular patches, e.g. areas under power poles that cannot be easily farmed anyway, seems to be of subordinate importance, as both did not differ significantly throughout. In addition, it may not be particularly important whether such structures are sown with wildflower (as is the case here for power pole islands) or not (field margins).