1. Agricultural intensification and abandonment of traditional agricultural practices are main drivers of current insect declines. The resulting loss of feeding and nesting opportunities has led to a decrease in pollinator populations like wild bees. While the restoration of floral resources has been widely implemented in wild bee conservation, nesting resources, particularly for ground-nesting species, are barely considered.

2. We assessed wild bee diversity in a wine-growing area in Germany in 15 study sites along a soil gradient and evaluated whether wild bees were distinctly affected by different vineyard types (vertically oriented, terraced, abandoned), local conditions (e.g. shrub and flower cover), and landscape factors in response to divergent nesting needs (above-ground vs. ground-nesting).

3. We found that wild bees responded more strongly to the availability of nesting sites than to flower resources. While ground-nesting bees were determined by the suitability of soil aspects for nesting irrespective of vineyard management types, above-ground nesting bees profited from vineyard abandonment and shrub encroachment in vineyard fallows and were enhanced by the availability of seminatural habitats (SNH) in the surrounding landscape. In contrast, floral resource availability in managed vineyards had only marginal effects on above-ground-nesting bees.

4. Synthesis and applications:  Life history traits like nesting strategies have long been neglected in wild bee conservation approaches, but proved to be highly relevant, especially for ground-nesting bees. For this, agri-environmental schemes can no longer solely focus on the restoration of floral resources, but should equally address nesting resources. Therefore, management efforts for enhancing wild bees in vineyard landscapes should aim at complementing nesting resources for ground-nesting bees (e.g. exposed bare ground patches) and above-ground-nesting bees (e.g. woody elements, hedges) in addition to floral resources. At the landscape level, conserving heterogeneous landscapes at a mixture of actively managed vineyards and semi-natural and woody elements is significant to maintain diverse bee communities.

In brief, more details are given in the manuscript:

Wild bee sampling was carried out from April to August 2019 in a total of three sampling rounds in three vineyard types: vertically oriented, terraced, and fallow (abandoned). At each sampling round, traps were exposed for three consecutive days per plot (18th–20th April, 28th–30th June, 23rd–26th August). For each plot and sampling round we used a set of three differently coloured pan traps, the inside was painted with blue, white, and yellow UV-reflecting paints and filled with water. Traps were placed in the vineyard-type centre at a distance of 3 m from each other and adjusted at the level of the surrounding vegetation.

Wild bees were identified to species level. For analysis, all samples per site were pooled over the three sampling rounds. We assigned species to be ground-nesting or above-ground-nesting following Westrich (2019). Parasitic species, do not build their nests but rather depend on their hosts and were therefore not considered as having nesting preferences.

Environmental parameters were assessed for each sampling round in parallel with wild bee sampling (April, June, and August 2019). We recorded local vegetation parameters for each plot (two samplings per plot and sampling round). Within each rectangle, we visually estimated the percentage cover of bare ground, herbaceous vegetation, shrubs, and the cover of all blooming-colored plants. We determined all flowering plants to species level and summarized them as the total number of flowering plant species in each rectangle. We measured the height of vegetation and shrubs (five random measurements) within each rectangle. Mean values were calculated and averaged over the three sampling rounds.

To account for differences in the soil composition we defined five categories (hereafter referred to as soil–5) along a gradient from fine soil to coarse gravel particles. We visually estimated the proportion and size of fine (< 2 mm) and coarse soil particles in a rectangle of 0.4 × 0.4 m. Soil 1 and 2 predominantly had fine particles and less than 40 % gravel, while soil 3–5 had a minimum of 40 % gravel that increased in size from soil 3 (< 1 cm), soil 4 (1–4 cm) to soil 5 (> 5 cm).

To analyse the landscape structure, we chose a 150 m radius around each study site. Within each radius, we calculated the area of semi-natural habitats, forests, and vineyards. SNH elements encompassed woody habitats such as vineyard fallows (predominantly) and hedges and herbaceous habitats (meadows or field margins with extensive vegetation). As a reference, we used the digital landscape model for Germany from 2009 (DLM-DE 2009), which we matched and updated with orthophotos (resolution of 40 × 40 cm) provided by the Hessian State Office of Land Management and the Surveying and Geographic Information Office Rhineland-Palatinate. Additionally, we measured the distance from the vineyard centre to the closest SNH and forest.

data processing:

Since landscape factors and soil composition altered alongside the river current, we used Rhine kilometres (a measure of overall river channel length) as a proxy to display altering soil and environmental conditions.

We used Pearson correlation coefficients to describe the alteration of landscape parameters, soil aspect, bee species richness, bee abundance, and nesting traits with the river current (Rhine km).

We choose (generalized) linear mixed effect models (GLMMs) and direct ordination methods (RDA) for analysis.

We provide dataframes as CSV files and metadata for the respective dataframes in text files.