1. Wildflower plantings on farms have been shown to attract foraging wild bees, however, whether these added floral resources increase nesting densities of bees remains largely untested.

2. We placed nest boxes containing natural reeds at 20 fruit farms in Michigan. We then compared nesting densities between farms with and without wildflower plantings and analyzed nest provisions to evaluate use of wildflower plantings for brood provisioning.

3. We found significantly greater nesting at farms with wildflower plantings, with only one out of 236 completed nests at a farm without a planting. The majority of nests were completed by Megachile pugnata, with a portion of nests completed by Osmia caerulescens.

4. We found that nesting bees collected pollen from only a subset of the available flowers in the wildflower plantings, with a strong preference for Centaurea maculosa, and Rudbeckia type pollens. While these species were found growing in the plantings, only Rudbeckia type species were seeded in the plantings.

5. This study provides evidence that wildflower plantings (though not only seeded species) are filling a critical resource gap for stem-nesting bees in agricultural landscapes and likely support local populations.

Study locations

Twenty commercial fruit farms in the western region of Michigan were chosen for this study. These were comprised of ten tart cherry (Prunus cerasus L.) orchards and ten highbush blueberry (Vaccinium corymbosum L.) farms within the main production regions of these crops (Fig. 1A). Half of the farms within each focal crop had established wildflower plantings (Fig. 1B) which were seeded in 2011 or 2013 (Table S1). Locations with plantings were paired with a nearby farm without a planting, and paired farms ranged between 1.14 and 12.36 km apart (mean = 4.55 ± 1.15 S.E.). Most paired farms were managed by the same grower, with similar applications of pesticides and weed management strategies (e.g. mowing, herbicides, etc.) that could affect the suitability of the landscape for bees.

All plantings were seeded with a mix of native forbs and grasses optimized for wild bee conservation (Table S1) and designed to bloom after flowering of the two focal crops. They were located in the field margins adjacent to the crop and bordering a wooded edge. The plantings ranged in size from 0.08 to 1.97 hectares (mean = 0.49 hectares ± 0.17 S.E). Unenhanced, regularly mowed field margins that were also adjacent to a wooded edge were selected at the paired unenhanced comparison farms (mean = 0.24 hectares ± 0.04 S.E).

Floral surveys

Once every two weeks, floral surveys were conducted at each farm. The first visit was in the week of June 25, 2018, and surveys continued for 10 weeks (5 total visits per farm). At each visit, we walked a 50 m transect in front of the nest box, roughly down the middle of the wildflower planting, and parallel to the edge of the crop planting and the wood line. Similar transect samples were taken at unenhanced comparison farms, again directly in front of the nest box. Every 5 m along the transect, a 1m x 1m quadrat was placed on the ground. We then visually estimated the proportional cover (between 0 and 1) of each species of flowering forb within the quadrat. Only the flowering head was considered in the estimate, and flowers that were obviously pre- or post-bloom were not included. Flowering plants were identified to the lowest taxonomic level in the field using Newcomb (1977) and the Pl@ntNet app (Joly et al., 2016), usually to species. Any plants that could not be confidently identified in the field were brought back to the lab, pressed, and identified using the keys available in Newcomb (1977).

Nesting density

At each farm, one 45 cm x 30 cm open faced wooden box was mounted three feet off the ground between two metal posts. These were placed along the edge of each wildflower planting or unenhanced field margin, facing the crop and backing up to a wood line (Fig. 1C). Tangle-Trap Sticky Coating (The Scotts Company LLC, Marysville, OH) was sprayed on the posts below the boxes to limit ant activity. A wooden bar running lengthwise 8 cm below the top of the box was used to hang the nesting materials, consisting of 6mm diameter reeds (Phragmites sp.) (Crown Bees, Woodinville, WA). Bundles of 50 reeds each were assembled using zipties and three of these bundles were hung from the crossbar of each nest box. Nest boxes were set up during the week of June 11, 2018 and removed the week of September 4, 2018.

At the same time as floral surveys were conducted (once every two weeks), all reeds with completed nests were collected and replaced with new reeds (Fig. 1D). This included any reeds that had been capped at the end (either by Megachilid bees or Eumenid mason wasps) or were filled with grass (from grass-carrying Isodontia wasps). Though only those reeds completed by bees were used for future data analyses. Additionally, ants were occasionally found to be stripping parts of the reeds for nesting materials, at which time all reeds in the nest boxes were removed and replaced, and additional applications of Tangle-Trap were made to the posts to reduce ant visitation.

All completed nests were brought back to the laboratory and the bee nests (N=236) and wasp nests (N=37) were separated. If we had collected more than one bee nest that day at a farm then one reed per farm, or 10% of collected reeds if there were twenty or more collected, were randomly selected for rearing out. Nests for rearing (N=24) were then put into mesh pollinator exclusion bags which were secured shut with zipties and stored in an unheated barn to overwinter. After the first male bees emerged (June 2019), the rest of the reeds were frozen and then opened to identify adult bees to species using published keys (Arduser, 2009; Sheffield et al., 2011). All other collected bee nests were used for pollen identification (N=212), and therefore placed in the freezer the same day they were collected to halt larval development. Nests used for pollen identification were not used to identify bee species.

Pollen identification from nests

Up to 10 nests were randomly chosen from each farm on each collection date and opened (Fig. 1E). Pollen provisions (bee collected pollen found in each cell) were removed and placed in individually labeled tubes. Each pollen provision was then suspended in 70% ethanol, with the amount of ethanol added varying in proportion to the size of the provision, but typically between 0.5 and 1 ml. The sample was then vortexed until the pollen was homogenized (2-3 minutes), and 20μl of the sample was immediately placed in the middle of a microscope slide, and the ethanol was allowed to evaporate. Pollen grains (Fig. 1F) were then stained with fuchsin gel and the sample was covered with a cover slip. Volumes of pollen species (Folk, 1951) were visually estimated for each sample. We identified the pollen to the lowest taxonomic rank possible using Sawyer (1981) and a reference collection of 254 known pollen species/types collected by bees in Michigan. Reference slides are housed in the Isaacs Lab at Michigan State University, Department of Entomology and primarily from west Michigan where our project farms were located. Some reference images are also available online: http://bit.ly/MSUbeepollen. If the identity of a pollen was uncertain, then it was lumped into a type category. Each pollen type is made up of a group of possible plant species that the pollen could be attributable to (Table 2).