2.1 Study Area and Sites

We conducted our study in the Palatinate wine-growing region of Germany (49.273280°N, 8.020602°E/49.147516°N, 8.175736°E). The studied vineyards were located in the Upper Rhine Valley east of the Palatinate Forest, an area characterized by warm temperate climate (Beck et al. 2018). The average annual temperature measured during the last 15 years was 11.3°C and the total annual precipitation 690.1 mm with a mean temperature of 11.9°C and a total precipitation of 630.4 mm in 2020 (Agrarmeteorologie Rheinland-Pfalz 2024). We selected winegrowers in the region who grow both an FRG and a classic grape variety in close proximity to each other and who were willing to let us conduct our study on their properties. We sampled 16 pairs of vineyards located along a landscape gradient with varying proportions of SNH in the surrounding landscape (Figure 1). Half of them were managed organically following the European Union Regulation No 2092/91, while the other half were managed conventionally. Each pair included one vineyard with a FRG variety and one with a classic grape variety managed by the same winegrower.

2.2 Landscape and Environmental Variables

The cover of SNH within a 500 m radius of each vineyard was calculated by using ATKIS data (Basis-DLM by GeoBasis-DE/BKG (2013); Table 1, Table S1) with intersection of spatial data in an Oracle database 12c (Oracle, 2017; for details see Kaczmarek, Entling, and Hoffmann 2023). SNH was defined as forests, hedges and shrubs, and grasslands. For each pair of vineyards, we averaged the proportion of SNH cover across the two nearby vineyards, resulting in a total of 16 landscapes.

Information on the number of fungicide applications in 2020 was obtained from the winegrowers (Table 1, Table S1; for details see Kaczmarek et al. 2023). Fungicides were applied in all vineyards, while only three conventional vineyard pairs received herbicides and none of the studied vineyards was treated with insecticides. To measure vegetation variables in inter-rows, we conducted three surveys during the season (late April, early July, early September). At each survey, we recorded the proportion of ground covered by vegetation and the presence of insect-pollinated plant species that had flowers in two centrally located plots per vineyard (Table 1). Two subplots, each measuring 1 m² (2 m × 0.5 m), were surveyed in adjacent inter-rows, covering a total of 4 m² per vineyard. Vegetation cover was measured visually in tens from 0% to 100%. For each vineyard, we calculated the mean vegetation cover of the three surveys and the total number of insect-pollinated plant species to use in further analyzes (Table S1).

2.3 Sampling and Identification of Bees

In the center of each vineyard, we placed a yellow pan trap (MAXIMILIAN, Neustadt an der Weinstraße, Germany) about 10 cm above ground between two vines. Traps were operated for three consecutive days every month from April to September 2020. We filled the traps with approximately 1.25 L of water with one drop of soap to reduce surface tension. After 3 days of exposure, we transferred collected material to 70% ethanol and stored the samples in undiluted ethanol. The bees contained in the samples were pinned onto insect needles for subsequent species identification.

We used the identification keys of Amiet et al. (2001, 2004, 2007, 2010), Amiet, Müller, and Neumeyer (2014), Amiet, Müller, and Praz (2017), Pauly (2015, 2021), Schmid-Egger and Scheuchl (1997), along with additional literature of Scheuchl and Willner (2016) and Westrich (2018), to identify the bees to species level. Some species have been grouped into species complexes as they cannot be safely distinguished based on morphology: Bombus hortorum agg. (B. hortorum and Bombus ruderatus), Bombus terrestris agg. (Bombus cryptarum, Bombus lucorum, Bombus magnus, and B. terrestris), Halictus simplex agg. (Halictus eurygnathus, Halictus langobardicus, and H. simplex), and Halictus tumulorum agg. (Halictus confusus and H. tumulorum). Difficult identifications have been verified using barcoding and metabarcoding data. Hylaeus cf. hyalinatus could not be clearly identified to species level due to missing body parts and no assignable result from barcoding. We obtained information on the conservation status from the German national Red List (Westrich et al. 2011) and the Wildbienen-Kataster (Scheuchl, Schwenninger, and Kuhlmann 2018) and information on the feeding, nesting, and social behavior from Westrich (2018).

2.4 Data Analysis

We used R v.4.2.3 (R Core Team 2023) and RStudio v.2023.03.0 (RStudio Team 2023) with the R packages car (Fox and Weisberg 2019), lme4 (Bates et al. 2015), glmmTMB (Brooks et al. 2017), blmeco (Korner-Nievergelt 2015), MuMIn (Bartoń 2020), vegan (Oksanen et al. 2020), indicspecies (Cáceres and Legendre 2009), ggplot2 (Wickham 2016), ggpubr (Kassambara 2020), and dplyr (Wickham et al. 2022) for analyzing data and creating figures.

We investigated differences in the environmental variables between organic and conventional management and between FRG and classic grape varieties by using generalized linear mixed model regressions (GLMM). Poisson distribution and logarithmic link function (log link) was used for spraying events and insect-pollinated plants, beta-binomial distribution and log link for proportions of vegetation cover. Management and grape variety were defined as fixed factors and the pair of vineyards as random factor. We investigated the effects of SNH, management, grape variety, their interaction, vegetation cover, and insect-pollinated plants on the abundance (number of individuals) and richness (number of species) of bees by using a GLMM with negative binomial distribution and log link and a GLMM with Poisson distribution and log link, respectively. Management, grape variety, and SNH were defined as fixed factors and the pair of vineyards as random factor. Accordingly, we investigated effects on the abundance and richness of ground- and above-ground-nesting bees using a GLMM with negative binomial distribution and log link for abundance of ground-nesting bees and GLMM with Poisson distribution and log link for richness of ground-nesting bees and abundance and richness of above-ground-nesting bees. We rescaled and centered continuous variables and excluded the honey bee Apis mellifera from all analyzes because its occurrence depends on the location of beehives. We selected the best fitting family for the GLMM by comparing the AICc of a range of potential families. By using a backward elimination method, the best fitting model was selected based on the lowest AICc, for which we used type II ANOVA to test the effects with a significance level of p < 0.05.

We investigated species composition of bees based on species abundances using redundancy analysis (RDA) with SNH, management, grape variety, vegetation cover, and insect-pollinated plants as explanatory variables. To reduce the influence of highly abundant species, we used Hellinger standardization to transform species abundance data to relative values. The best fitting model was selected using a backward elimination method. Furthermore, we identified insect-pollinated plant species related to either organic or conventional management using presence-absence data in a species indicator analysis.

3.1 Environmental Variables

With two more spraying events compared to conventionally managed vineyards (mean = 6.1, SD = ±3.4), organically managed vineyards (mean = 8.1, SD = ±4.8) had a significantly higher number of fungicide applications (Tables 1 and 2, Figure 2). FRG varieties received significantly fewer applications (mean = 3.8, SD = ±2.9) than classic varieties (mean = 10.4, SD = ±2.2). Vegetation cover was 43% higher in conventionally managed vineyards (mean = 77.3%, SD = ±15.5%) compared to organically managed ones (mean = 53.8%, SD = ±14.2%, Table 2, Figure 2), but the difference was not significant. The number of insect-pollinated plants was not affected by either management type or grape variety. Based on our species indicator analysis, insect-pollinated plant species that were more present in organic vineyards were Chenopodium album agg., Convolvulus arvensis, Fagopyrum esculentum, Malva sylvestris, Raphanus raphanistrum agg., and Trifolium incarnatum, while Bellis perennis and Taraxacum spp. were more present in conventional vineyards (Tables S2 and S3).

3.2 Bee Diversity

We sampled 2015 bees of 14 genera and 85 species (see Appendix S1 for full species list; Tables S4 and S5). The most diverse genera were Andrena (26 species) and Lasioglossum (22). We further found species of the genera Apis (one), Bombus (four), Ceratina (one), Colletes (one), Eucera (one), Halictus (seven), Hylaeus (five), Megachile (one), Nomada (eight), Osmia (five), Sphecodes (one), and Stelis (two). The most abundant genera were Lasioglossum (60.7%) and Andrena (26.7%). The most abundant species were Lasioglossum malachurum (21%), Lasioglossum glabriusculum (12.4%), Lasioglossum lineare (10.8%), Andrena dorsata (7.8%), and Lasioglossum morio (5%).

With 49% more bees on average, organically managed vineyards (mean = 73 bees, SD = ±41) had a significantly higher bee abundance compared to conventionally managed vineyards (mean = 49 bees, SD = ±28; Table 3, Figure 3C,D), while the richness tended to be 15% higher under organic management (Figure 3D). Bee abundance and richness increased significantly with increasing vegetation cover (Figure 3B, Figure 4). SNH, the grape variety, and the number of insect-pollinated plant species did not have any significant effect on the abundance or richness of bees (Figure 3A,C,D). The abundance of ground-nesting bees tended to be 48% higher in organic vineyards and the richness increased significantly with increasing vegetation cover (Table 4, Figure S1). The abundance and richness of above-ground-nesting bees increased with the proportion of SNH in the surrounding landscape (Table 4, Figure 5). The species composition of bees was influenced by SNH, vegetation cover, and management (Table 5, Figure 6).

4.1 Effects of Semi-Natural Habitats

The landscape context is known to be an important determinant of biodiversity patterns (Tscharntke et al. 2021). Surprisingly, we did not find a significant positive effect of SNH on the total abundance and richness of bees, which contradicts our first hypothesis, where we expected bee diversity to increase with increasing proportions of SNH in the surrounding. This finding deviates from the strong positive effects of landscape structure on invertebrate biodiversity in vineyards reported in previous studies (Kolb et al. 2020; Barbaro et al. 2021; Kaczmarek, Entling, and Hoffmann 2023). The diversity of bees in our study predominantly comprised ground-nesting species from the genera Lasioglossum and Andrena (Westrich 2018). Since they build their nests in bare ground areas also found in and near vineyards, they may not rely on nesting structures provided by SNH within their habitat range. Similarly, Kaczmarek et al. (2023) found rather small effects of SNH on orthopterans, some of which complete the entire life cycle within vineyards and do not depend on SNH in the surrounding landscape. Nonetheless, SNH supply abundant floral resources especially early in the season, which could also support the presence of some ground-nesting species (Bertrand et al. 2019; Eckerter et al. 2022). However, SNH were significantly increasing the abundance and richness of above-ground-nesting bees, similar to the positive effect of increasing cover of SNH on cavity-nesting bees in vineyards observed by Uzman et al. (2020) and Wersebeckmann et al. (2023). Above-ground-nesting bees depend on cavities in woody structures or stone walls to build nests for breeding and thus rely on such structures in the surrounding area of vineyards (Westrich 2018). In addition to SNH such as forests, solitary trees and gardens as well as vineyard fallows with wild flower strips can benefit wild bees by providing foraging and nesting habitat (Kratschmer et al. 2018; Krahner et al. 2024). Except for Osmia bicornis, the abundance of cavity-nesters, including the genera Osmia, Hylaeus, and Ceratina, was low in our study, which may reflect a lack of suitable nesting opportunities within and near vineyards in our region. The diversity of above-ground-nesting bees in viticultural areas can thus be promoted by a more diverse landscape with suitable SNH such as hedgerows providing nesting opportunities. In addition, structurally rich environments containing SNH can increase overall biodiversity, including beneficial organisms such as parasitoids and predators, and thereby improve natural pest control (Holland et al. 2017; Rösch et al. 2023).

4.2 Effects of Local Vineyard Management

In other cropping systems, organic management clearly benefits biodiversity compared to conventional management and bees are among the most strongly favored organisms in organic arable crops, likely due to the higher abundance of flowering insect-pollinated weeds (Bengtsson, Ahnström, and Weibull 2005). In viticulture, effects of organic management on biodiversity are not as clear, and both positive and negative effects have been reported (Döring et al. 2019; Kolb et al. 2020; Paiola et al. 2020; Ostandie et al. 2021; Beaumelle et al. 2023; Kaczmarek, Entling, and Hoffmann 2023). Consistent with our second hypothesis, where we expected a higher diversity of wild bees in organically managed vineyards, the abundance of bees was higher in organic vineyards, and the richness tended to increase. A major difference between the two management systems is the use of mainly synthetic fungicides in conventional management, while organic viticulture relies solely on inorganic compounds, mostly copper and sulfur (Pedneault and Provost 2016). While insecticide use is low in our study region, fungicides, including copper, and sulfur, can still affect non-target organisms (Nash, Hoffmann, and Thomson 2010; Biondi et al. 2012; Vogelweith and Thiéry 2018). Kaczmarek, Entling, and Hoffmann (2023) reported reduced arthropod biomass in organic compared to conventional viticulture. To find opposing effects on bees may indicate that bees are not as affected by fungicides and that other factors determine their presence, such as different ground management practices and floral resource availability. Consequently, the higher abundance of bees and the trend of increasing bee richness in organic vineyards may be due to differences in vegetation and tillage practices compared to conventionally managed vineyards that affect plant communities and vegetation cover in the inter-rows of the vineyards (Kratschmer et al. 2020). Negative effects of organic management on arthropod biomass, as reported by Kaczmarek, Entling, and Hoffmann (2023), may be more related to taxa that are more exposed or more susceptible to fungicides.

Positive effects of reduced fungicide use in FRG varieties were shown for some groups of non-target organisms, for example, predatory mites, spiders, and herb-dwelling orthopterans (Pennington et al. 2017, 2019; Möth et al. 2021; Reiff et al. 2021, 2023; Kaczmarek et al. 2023). Such positive effects of reduced fungicide use, for example on predatory mites, can further enhance the sustainability of viticulture by promoting natural pest control (Möth et al. 2023). Interestingly, and contrasting to our third hypothesis where we expected a positive effect on bees, we found no effect of reduced fungicide applications in FRG varieties on bees, even though they received less than half as many applications compared to classic grape varieties. Reiff et al. (2023) found that the cultivation of FRG varieties significantly reduced the hazard quotient for applied fungicides based on the toxicity to honeybees. They also reported that organically managed vineyards had a higher hazard quotient due to non-selective compounds such as copper and sulfur (Reiff et al. 2023), yet we found a higher abundance of bees under organic management. This supports the previously mentioned assumption that bees, at least in our study region with no insecticide use, are not as affected by the fungicide management as vine-dwelling organisms due to their mobility. Besides negative effects of fungicides, however, also reduced disturbance and soil compaction along with fewer fungicides applications appear to have no strong effects on bee diversity. Therefore, other factors, such as the inter-row vegetation, may play a more prominent role in determining their presence.

Ground management that allows diverse vegetation cover contributes to biodiversity conservation in vineyards (Kratschmer et al. 2018, 2019; Winter et al. 2018). According to our study results, bees benefit from increased vegetation cover in inter-rows. Since in our region the average vegetation cover was not significantly different between organic and conventional vineyards when vegetation variables were measured, high vegetation cover alone does not seem to account for the higher abundance of bees in organic vineyards. However, ground-nesting bees, including the most common species in our study, which nests in aggregations (L. malachurum), may benefit from bare ground areas because they rely on it for nesting (Potts et al. 2005). Furthermore, it is important to note that high vegetation cover does not necessarily indicate a structurally diverse and resource-rich vegetation in the inter-rows of the vineyards. When surveying the vegetation, we had the impression that conventional vineyards tend to have a high cover of grasses, which are unattractive to bees. On the other hand, although the number of insect-pollinated plants did not differ between management types and did not significantly influence bee diversity, some nectar- and pollen-rich plants (e.g., C. arvensis, F. esculentum, M. sylvestris, and T. incarnatum) were typical of organic vineyards. Some of these species are often included in seed mixtures for the greening of inter-rows. Although flower-rich mixtures are not a general characteristic for organic farming, they are more commonly used in organic than in conventional viticulture in our region. This characteristic may have contributed to the increased abundance of bees, including polylectic species, in organic vineyards, which may have benefited from diverse vegetation cover (Sutter et al. 2017; Winter et al. 2018). Oligolectic species, on the other hand, are dependent on certain plant species within their habitat range and would not be present without the presence of those plants (Westrich 2018). Therefore, an inter-row consisting of both diverse and flower-rich vegetation with high cover, as well as bare ground, may enhance the presence of bees. Differences in vegetation diversity and structure could therefore also account for the differences in species composition in our study. Unfortunately, we did not record detailed information on vegetation management practices, such as tillage and mulching practices and the use of cover crop mixtures, which strongly influence plant community composition in vineyard inter-rows (Guerra et al. 2022). We therefore used our own measured vegetation data as a proxy for this information, including insect-pollinated plant community composition and vegetation cover. These values reflect differences in vegetation management. However, for direct inference about effects of different types of vegetation management, future studies should record the respective information from winegrowers.

Notably, our study identified the presence of Lasioglossum subhirtum, which was recently observed in the region for the first time in about 70 years (Burger 2018). Additionally, we found a high abundance of L. lineare, along with other vulnerable bee species, emphasizing the potential of viticultural areas as important habitats for wild bee conservation, provided that species specific floral and nesting resources are abundant. However, by the use of only yellow pan traps, our study may have underestimated the full diversity of wild bees in the studied vineyards. The use of additional sampling methods, including blue and white colored pan traps and transect walks with hand netting, would likely maximize the abundance and richness of species being recorded (Krahner et al. 2021; Acharya et al. 2022).