Response of wild bee diversity, abundance, and functional traits to vineyard inter‐row management intensity and landscape diversity across Europe

Abstract Agricultural intensification is a major driver of wild bee decline. Vineyards may be inhabited by plant and animal species, especially when the inter‐row space is vegetated with spontaneous vegetation or cover crops. Wild bees depend on floral resources and suitable nesting sites which may be found in vineyard inter‐rows or in viticultural landscapes. Inter‐row vegetation is managed by mulching, tillage, and/or herbicide application and results in habitat degradation when applied intensively. Here, we hypothesize that lower vegetation management intensities, higher floral resources, and landscape diversity affect wild bee diversity and abundance dependent on their functional traits. We sampled wild bees semi‐quantitatively in 63 vineyards representing different vegetation management intensities across Europe in 2016. A proxy for floral resource availability was based on visual flower cover estimations. Management intensity was assessed by vegetation cover (%) twice a year per vineyard. The Shannon Landscape Diversity Index was used as a proxy for landscape diversity within a 750 m radius around each vineyard center point. Wild bee communities were clustered by country. At the country level, between 20 and 64 wild bee species were identified. Increased floral resource availability and extensive vegetation management both affected wild bee diversity and abundance in vineyards strongly positively. Increased landscape diversity had a small positive effect on wild bee diversity but compensated for the negative effect of low floral resource availability by increasing eusocial bee abundance. We conclude that wild bee diversity and abundance in vineyards is efficiently promoted by increasing floral resources and reducing vegetation management frequency. High landscape diversity further compensates for low floral resources in vineyards and increases pollinating insect abundance in viticulture landscapes.

Research demonstrated that wild bees are threatened by intensive agricultural practices (Kremen, Williams, & Thorp, 2002) such as high pesticide application (Woodcock et al., 2017), and/or frequent soil tillage (Williams et al., 2010), which result in reduction of floral resource availability  and contribute to landscape simplification (Senapathi, Goddard, Kunin, & Baldock, 2017).
Vineyards cover about 7.6 million hectares worldwide (OIV, 2018). The commercial grape vine (Vitis vinifera L.) is self-pollinated and wind pollinated, thus pollination by insects only plays a minor role for grape yield (Cabello Saenz, Luis Villota, & Tortosa Tortola, 1994). Bees were rarely observed foraging on grapevine flowers (Vorwohl, 1977), but vineyards can provide habitats for wild bees to increase pollination for insect-pollinated crops, fruit trees, cover crops, and wild plants. Maintaining wild bee diversity is essential for the resilience of pollination services (Bartomeus et al., 2013; and also enhances diversity of associated plants pollinated by wild bees (Biesmeijer et al., 2006). Improving habitats for pollinators simultaneously enhances ecosystem services like biological pest control, soil and water quality protection, or landscape aesthetics (Wratten, Gillespie, Decourtye, Mader, & Desneux, 2012). Establishing and maintaining noncrop flowering areas within the farmland matrix promotes the native plant community, provides habitats for a range of insects, bird and mammals, and thus contributes to biodiversity conservation . Further, a spill-over effect of flower visitation rates in insect-pollinated crops from field margins was observed for wild bees, which increased crop yields in closer proximity to field margins (Woodcock et al., 2016).
As winegrowers experience an increased consumer demand for eco-friendly produced wine (Schütte & Bergmann, 2019), establishing flower-rich habitats for wild bees in vineyards can be used for marketing.
Depending on the vegetation management intensity, vineyard inter-rows are comparable with field margins or wildflower strips in agricultural landscapes, which increase wild bee diversity (Haaland, Naisbit, & Bersier, 2011). Winegrowers manage inter-row vegetation by tillage, mulching, or herbicide application to mitigate potential water and/or nutrient competition between the vines and the interrow vegetation (Pardini, Faiello, Longhi, Mancuso, & Snowball, 2002).
The intensity of this disturbance varies among wine-growing areas across Europe according to local pedological and climatic conditions. Wild bees in vineyards have been shown to benefit from biodiversity-friendly management practices and from mosaics of semi-natural elements within the viticultural landscape (Kehinde & Samways, 2014a, 2014bKratschmer et al., 2018). Further, species characterized by certain traits may respond similarly to a certain vegetation management measure or landscape configuration in wine-growing areas. For example, ground-nesting species could benefit from undisturbed soil conditions for nesting in permanently vegetated inter-rows. Further, larger species may compensate low landscape diversity with their increased activity range and forage in more fragmented landscapes (Zurbuchen et al., 2010). A meta-analysis communities were clustered by country. At the country level, between 20 and 64 wild bee species were identified. Increased floral resource availability and extensive vegetation management both affected wild bee diversity and abundance in vineyards strongly positively. Increased landscape diversity had a small positive effect on wild bee diversity but compensated for the negative effect of low floral resource availability by increasing eusocial bee abundance. We conclude that wild bee diversity and abundance in vineyards is efficiently promoted by increasing floral resources and reducing vegetation management frequency. High landscape diversity further compensates for low floral resources in vineyards and increases pollinating insect abundance in viticulture landscapes.

K E Y W O R D S
Apiformes, ecosystem services, floral resource availability, functional traits, GLMM, Shannon Landscape Diversity Index, vegetation management, viticulture landscapes included only two studies about the effects of vineyard vegetation management on pollinators and concludes that knowledge about the effects of inter-row vegetation management on wild bee diversity is scarce . Further until now, studies about wild bee diversity and functional traits in response to vineyard management and in relation to landscape diversity in different climatic regions (i.e., different European countries) have not yet been carried out.
We hypothesized that vegetation management intensity, floral resource availability, and the surrounding landscape diversity affect wild bee diversity, abundance, and functional traits in vineyard inter-rows across Europe. We expected that inter-row vegetation management effects on bees would be less pronounced in vineyard with higher floral resource availability and in heterogeneous than in simpler landscapes.

| Study sites
This study was conducted in four viticultural areas across Europe (Spain, France, Austria, and Romania) in 2016. The locations of the viticultural areas ( Figure 1)  In total, 63 vineyards were investigated that ranged in age from 5 to 61 years. The distance between the vines (in-row) ranged from 0.75 to 1.9 m, and the inter-row width varied between 1.5 and 3 m. Three different intensities of inter-row vegetation management were studied (Table 1): (a) permanent vegetation cover without any disturbance for at least 5 years (Austria, France, and Romania), (b) temporary vegetation cover in every second inter-row (Austria and Romania) or in every inter-row during the winter season (Spain) by tillage, and (c) bare soil management through frequent soil tillage (Spain and Romania) and/or application of herbicides (Spain, France) in all inter-rows. Tillage depths ranged between 5 and 40 cm across the countries. In each inter-row, the vegetation coverage (%) was estimated twice a year (at the beginning of the vegetation period and 2 months later) in four 1 × 1 m subplots. The averaged vegetation cover per inter-row differed significantly (Kruskal-Wallis test: χ 2 = 38.50; df = 2; p ≤ 0.001) among the management intensities. The Spanish inter-rows with temporary vegetation cover were managed more intensively compared to the temporary vegetated inter-rows in Austria and Romania which resulted in a comparatively lower vegetation cover (Table 1). Mulching was done 1-5 times in permanently and temporary vegetated inter-rows. All studied vineyards-with the exception of seven Spanish vineyards with deficit drip irrigation-were rainfed.

| Wild bee sampling and functional traits
Wild bees were sampled by a semiquantitative transect method in the vineyard inter-rows. The transects length ranged between 67 and 133 m in order to adjust to the different width of the inter-rows (1.5-3 m). To consider temporary vegetation cover management, each transect included two neighboring inter-rows. Each vineyard was sampled five times in 2016 for 15 min per sampling event. Sampling dates among the countries were synchronized to grapevine phenology (first budburst, first flower buds, full florescence, pea-sized berries, and beginning of maturation) to adapt to the different climatic zones (Bauer, Regner, & Schildberger, 2013). During the sampling process, each transect was walked slowly and wild bees were collected with an aerial net and later identified in the laboratory.
Functional traits of wild bees (Table 2) were selected according to the possible response to management, floral resource availability, and/or landscape diversity. Information on functional traits was gathered from the literature (Scheuchl & Willner, 2016) or expert's evaluation. As a proxy for the activity range and body size, we measured the intertegular distance (ITD in mm) with a digital microscope (Keyence VHX-5000) of 1-5 specimens from each species and averaged per species. This shortest linear distance between the bee's wings at the dorsal side of the thorax corresponds to the size of wing muscles and to the activity range of a species (Greenleaf, Williams, Winfree, & Kremen, 2007).

| Landscape survey
A 750 m radius around each sampled vineyard center was chosen for the landscape survey to get a minimum distance of 1,500 m between the study sites which covers the foraging distance of many wild bee species (Zurbuchen et al., 2010;Zurbuchen & Müller, 2012). In each landscape circle, the landscape structures following the EUNIS habi-

| Data analysis
Honey bee (Apis mellifera) counts were excluded from the main analysis, because their abundance to a great extent depends on TA B L E 1 Mean ± SD vegetation coverage (%) per vegetation management intensity, method of vegetation management, and number of management events per year in the studied countries The response variables species richness and abundance were aggregated across all sampling dates per vineyard. The predictor variables vegetation cover (proxy for vegetation management intensity) and floral resource availability were averaged per vineyard. Floral resource availability was represented by three classes ("very low," "low," and "medium") after averaging, due to missing observations of the levels "high" and "very high." The SHDI was used as index for landscape diversity because it was least collinear with the other predictors and therefore the best option to model its interactions with management intensity and floral resource availability.
Wild bee traits were summarized by community weighted means (CWM; R package "FD" Laliberté, Legendre, & Shipley, 2015). To evaluate significantly associated wild bee traits in vineyards, a PCA was constructed, including a Hellinger transformation to correct for the "arch effect" (Zuur, Ieno, & Smith, 2007). Further, the CWMs were fitted onto the PCA by vector fitting (with the "envfit" function of the "vegan" package; Oksanen et al., 2017). This function calculates the correlation and associated p-values (α ≤ 0.05) between the ordination of species assemblage per plot and the explanatory variables by random permutations (n = 999; Oksanen, 2015). Finally, generalized linear models (GLMs) were used to analyze the effects of the three predictors on these significant associated traits (i.e., sociality and body size). As response variables, we used the CWMs of the body size and for sociality the number of eusocial and solitary species and their abundances.
Model selection was based on an information theoretic approach (Burnham & Anderson, 2002), and a candidate model set of 10 GLMs was formulated with different combinations of vegetation cover, floral resource availability, and SHDI and their interactions ( Table 3). The country was used as predictor in every model to encompass country-specific effects. Species richness and abundance models were formulated as GLMs with Poisson and ITD as GLMs with Gaussian error distribution. Models were ranked by the second-order Akaike's information criterion (AICc; R package "AICcmodavg" Mazerolle, 2016). The cutoff rate to decide whether a model was the most parsimonious compared to the others was set at ΔAICc ≥ 2 (Motulsky & Christopoulos, 2003). Plots of relevant effects of the most parsimonious models were computed with the R package "effects" (Fox, 2003).  The shortest linear distance measured between a wing tegulae across the dorsal thorax (Cane, 1987) Strongly related to the flying distance of a species (i.e., the distance a female can fly to collect pollen and nectar; and affected by landscape features (Gathmann & Tscharntke, 2002;Greenleaf et al., 2007;Zurbuchen et al., 2010) (2007).

| RE SULTS
In total, 113 species and 719 individuals were sampled in vineyards across Europe (species list: Supporting Information Appendix S1: Supporting Information Appendix S1: Figure S1a) and during full florescence (1.46 ± 1.94; Supporting Information Appendix S1: Figure   S1a) of the vines. The highest mean (± SD) abundances of wild bees (2.84 ± 4.61; Supporting Information Appendix S1: Figure S1b) were also sampled when the first flower buds appeared on the vines. The lowest mean species richness and abundance were sampled at the last sampling date when the grapes started to mature (Supporting Information Appendix S1: Figure S1).
Overall, 65% of all wild bee species were ground nesting and 25% were above-ground nesting. The majority (76%) of wild bee species in vineyards were polylectic and only 14% were oligolectic.
Further, wild bee fauna of vineyards consisted of 26% eusocial species, 60% solitary species, and 4% species with insufficient information on sociality. Parasitic wild bees were dominant in three vineyards (two temporary and one permanently vegetated) and represented 10% of all species. The CWM of body size ranged from 0.9 to 3.0 mm ITD and was significantly related to the Austrian wild bee assemblages. Indeed, the mean (±SD) CWM of ITD was highest in Austria (2.10 ± 0.47 mm), followed by Romania In general, wild bee diversity, abundance and the functional traits that were significantly associated with the PCA's ordination in vineyards, were best explained by models including both floral resource availability and vegetation cover and their interaction (Table 4, Supporting Information Appendix S1: Table S2: GLM results). The average floral resource availability was generally low, but highest in Austrian and Spanish inter-rows and lowest in Romanian inter-rows (Supporting Information Appendix S1: Figure S2).  Except for the significant interaction between landscape diversity and floral resource availability on eusocial wild bee abundance, landscape diversity played a secondary role for eusocial and solitary wild bee species richness and abundance (Supporting Information Appendix S1: Table S2).
Wild bee body size was equally well explained by models that included the country, the landscape diversity, or the interaction of both.
However, the intercept-only model was ranked within the most parsimonious models (Table 4) and the explained deviance of the mentioned models was low (R 2 GLM = 12%-16%; Supporting Information Appendix S1: Table S2) which implies the low explanatory value of the chosen predictors for wild bee body size in vineyards.

| D ISCUSS I ON
Wild bee species richness, abundance, and functional traits in vineyard inter-rows strongly increased with higher floral resource availability and extensive inter-row vegetation management. Further, the total wild bee abundance as well as the diversity and abundance of solitary wild bees were significantly positively affected by the interaction of higher floral resources in extensively managed vineyard inter-rows. The surrounding landscape had a limited influence on wild bee species richness, abundance, and most functional traits.
However, it played an important role for eusocial wild bees in compensating for low floral resource availability. Most of the wild bee species and individuals were ground nesting, solitary, and generalists regarding the plants they forage on.
Across the studied vineyards, 5.7% of the almost 2000 European wild species (Nieto et al., 2015) were recorded. The recorded species numbers per country (between 20 and 64) corresponds to other vineyard studies. For example, 25-31 wild bee species were reported in 12 and 10 vineyards, respectively, in South Africa (Kehinde & Samways, 2012, 2014a, 2014b   Supporting Information Appendix S1). The divergent climatic, geographic, and/or floral zones of the studied countries are possible reasons for the different species assemblages (Gusenleitner, Schwarz, & Mazzucco, 2012;Nieto et al., 2015;Ortiz-Sánchez, 2011;Polaszek & Mitroiu, 2013;Tomozei, 2010). It is notable that Spanish vineyards exhibited an unexpected low species richness even though the region in southern Spain is one of the diversity hot spots for wild bees in Europe (Nieto et al., 2015). The overall intensive inter-row management in vineyards and the low landscape diversity in our Spanish study region are the most likely reasons for the low bee diversity. The most abundant species in Spanish vineyards, Andrena tenuistriata, prefers Mediterranean-type shrublands as well as arable land as habitat (Roberts, 2014). The majority of those individuals (64.6%) were present in vineyards with temporary vegetation cover which demonstrates the benefit of less intensive disturbance for this ground-nesting species. Austrian vineyards comprised the highest wild bee diversity which conforms with the generally high wild bee diversity in eastern Austria (Nieto et al., 2015). Further, the landscape diversity was highest in the Austrian wine-growing region and inter-row vegetation treatments included the two least intensive managements.
Furthermore, other pollinators like butterflies also respond positively to suitable nectar resources and larval host plants in winegrowing areas .
Vineyard inter-rows are linear landscape elements and are comparable with flowering strips or field margins which can improve pollinator diversity, abundance, and pollination services for insectpollinated crops (Haaland et al., 2011;Williams et al., 2015). The positive effect of increased floral resource availability in vineyards has to be examined critically because the attraction of wild bees could lead to increased pesticide exposure of these pollinating insects.
However, the effect of pesticides and their active ingredients, which are used in viticulture, on wild bee diversity, abundance, and traits, was not studied and should be addressed in future research.
The strong positive effect of extensive vegetation management agrees with other studies reporting the benefits of extensive agricultural management practices for wild bees in different crop systems (Nicholson et al., 2017;Shuler, Roulston, & Farris, 2005), as well as vineyards (Kehinde & Samways, 2012, 2014a, 2014b. Moreover, a recent meta-analysis confirmed that positive affect of extensive management on overall biodiversity and ecosystem services . Ground-nesting bees benefit from undisturbed soil conditions and can utilize vineyard inter-rows as nesting habitat. Indeed, during field work, nesting activity of Lasioglossum marginatum and L. lineare was occasionally observed. In total, most eusocial (70%) and solitary (70%) species were ground nesting, but nesting types were not significantly associated with the PCA and not analyzed further with GLMs. In general, the high proportion of ground-nesting wild bees is characteristic for agroecosystems because nesting habitats are widely available (e.g., unsealed roads, field verges, bare ground below vine rows). Whereas structures for above-ground nesting wild bees (e.g., old plant material, deadwood elements) are often less abundant (Zurbuchen & Müller, 2012).
Further, floral resources are destroyed by frequent soil tillage or herbicide use in bare soil vineyards which amplifies the negative effect of intensive vegetation management. The combined positive effect of higher floral resource availability and vegetation cover on the total wild bee abundance is associated with the high abundance (79%) of ground-nesting eusocial wild bees. The remaining 21% eusocial (above-ground nesting) individuals were represented by bumblebees.
These species colonize pre-existing cavities below, on or above, the ground for nesting and are much likely to be negatively affected by frequent soil disturbance. The same combined positive effects on solitary wild bees are explained by the high abundance (86%) and species richness (72%) of ground-nesting solitary wild bee species.
Even though we found a positive effect of landscape diversity on wild bee species richness, it was low, which could be explained by the superior effect of floral resource availability in the inter-rows. These results disagree with other studies which revealed the essential importance of landscape structures on wild bee communities (Kennedy et al., 2013;Nicholson et al., 2017). Conversely, it demonstrates the necessity for increasing floral resource availability on the landscape F I G U R E 6 Eusocial wild bee abundance in response to (a) floral resource availability and (b) the interaction of landscape diversity and floral resource availability. Solitary wild bee abundance in response to (c) interacting effects of floral resource availability and vegetation cover. Error bars/gray shading: 0.95 confidence intervals scale to increase and maintain wild bee species richness and thus adequate pollination services for insect-pollinated wild plants and crops (Winfree et al., 2018).
Eusocial wild bees were significantly associated with countries (Austria and Romania) where extensive inter-row vegetation management was realized because eusocial species are more susceptible to disturbances than solitary species (Williams et al., 2010). Only eusocial wild bee abundance was affected by the interaction of SHDI and floral resource availability which could be explained by their higher vulnerability to habitat fragmentation (Williams et al., 2010).
Continuous floral resource availability during the vegetation period plays a crucial role for the sexual reproduction of eusocial wild bees because a lack of pollen and nectar can lead to a colony collapse in the reproduction phase during summer (Westphal et al., 2009).
Landscape structures like fallows (Toivonen, Herzon, & Kuussaari, 2016), hedges (Morandin & Kremen, 2013), solitary trees, or edges of woods (Nicholson et al., 2017;Rollin et al., 2013) provide different foraging sites for wild bees. Furthermore, these structures may compensate for negative effects of low to very low floral resource availability on eusocial wild bees that nest in the inter-row space of vineyards . Spanish vineyards possessed similar average floral resource availabilities as Austrian vineyards, which, according to our results, should benefit eusocial species.
However, only one eusocial species was documented in Spanish vineyards. The more intensive vegetation management and low landscape diversity limited eusocial wild bee occurrence. This might decrease pollination provision at the landscape scale because pollination performance mainly depends on wild bee species richness (Winfree et al., 2018) and abundance (Winfree, Fox, Williams, Reilly, & Cariveau, 2015). Even though vines and olives, representing the dominant crops in the Spanish study region, do not rely on insectpollination, but other insect-pollinated wild plants require pollination to guarantee long-term survival. This was reported from central Europe, by Biesmeijer et al. (2006) who showed a parallel decline of wild plants and their pollinators due to insufficient pollination.
We expected that increasing average body size of bee assemblages is related to decreasing landscape diversity, because larger species can forage at greater distances (Greenleaf et al., 2007).
Further, if pollen availability is low it leads to a change in maternal resource allocation to offspring, resulting in smaller adults (Renauld, Hutchinson, Loeb, Poveda, & Connelly, 2016). These effects were not observed since body size was not noteworthy affected by SHDI or by any other predictor. This is likely due to an overlapping effect by the distinct species assemblage in each country: The body size of wild bees was related to the Austrian wild bee assemblages. We explain this by the high abundance and species richness of bumble bees in Austrian vineyards compared to France, Romania, and Spain. On the other hand, a high proportion of the individuals in Spain was represented by two small wild bee species (Andrena tenuistriata, average 1.29 mm ITD and Panurginus albopilosus, average 0.89 mm ITD).
In conclusion, the total wild bee diversity and abundance as well as solitary wild bee diversity and abundance benefitted from the combination of increased floral resource availability and extensive vegetation management intensity in vineyard inter-rows. Consequently, vineyard inter-rows can be important habitats for wild bees in viticultural landscapes. High landscape diversity played an important role in compensating for low floral resources for eusocial wild bees. Therefore, we recommend less intensive vegetation management such as infrequent vegetation disturbance to be implemented in vineyard inter-rows in order to achieve resilient pollination provision for insect-pollinated crops and wild plants in viticultural landscapes. Beside enhancing wild bee diversity and abundance through these measures also honey bees will benefit which is especially important for the pollination of mass flowering crops (Brittain, Williams, et al., 2013). The implementation of pollinator-friendly management ultimately benefits other ecosystem services like for example soil erosion mitigation, surface water runoff reduction, or biological pest control as well as biodiversity conservation . Many of those ecosystem services are relevant for winegrowers and positively affected by extensive inter-row management intensities in vineyards . For example, extensive vegetation management significantly improves soil loss mitigation , which is highly relevant in vineyards that are situated on hilly terrain. Extensive management contributes to sustainable farming contributing to the UN sustainable development goals responsible consumption and production as well as life on land (UN, 2015). Further, biodiversity-friendly vineyard management practices (e.g., organic farming) are increasingly demanded by consumers (Schütte & Bergmann, 2019).

ACK N OWLED G M ENTS
This research is part of the BiodivERsA project VineDivers (www. David Horner for proof reading, as well as the reviewers and editors for helping to improve the manuscript.

CO N FLI C T O F I NTE R E S T S
The authors declare no competing financial interests.