Pollinator communities are more abundant and diverse in agricultural matrices with more natural habitats, although the reasons for these correlations remain unclear. It is possible that forest fragments and edges provide resources for pollinators in the important early weeks of spring, after which time those insects can then “spill over” into crops such as apple orchards during bloom.

To explore how forest edges may feed and therefore promote flower visitor communities in adjacent agricultural habitats, we sampled springtime pollinators in nine orchards and their adjacent forest edge canopies and understories. We identified pollen consumed by pan-trapped bees and flower flies to assess if bees ate pollen where they were caught and if their diets similarly “spill over” from forest to orchard. We further explored sex differences in habitat usage.

Our spatially replicated sampling found that bee and flower fly abundance peaks first in the forest understory, then in the forest canopy, and finally in the orchard.

Analysis of digestive tracts showed significant usage of forest canopy pollen throughout the spring, especially before apple bloom. Pollinators had often eaten pollen from a different habitat than the one in which they were caught, suggesting frequent movement between habitats. Digestive tract pollen is an underused but powerful avenue for ecological insight.

In Andrena, which are important orchard pollinators and one of the most abundant wild bee taxa in this study, male bees were primarily found in the woods but not the orchards where conspecific females were later active.

Synthesis and policy implications: Forested areas, especially forest canopy trees, provide large amounts of early spring resources that facilitate the build-up and spillover of wild pollinator populations into apple orchards during bloom. Forests also provide critical habitats for male bees, which were rarely found in orchards. Despite their importance for bee reproduction, the needs of male bees are usually not considered in conservation planning. Overall, our data indicate that ensuring there is adequate forest habitat adjacent to orchards can improve the long-term sustainability of pollinator populations that provide essential crop pollination services.

Bee and flower fly collection: This study was completed in the Finger Lakes region of New York, in the northeastern United States. Bees and flower flies (Syrphidae), two major groups of agricultural pollinators (Rader et al., 2016) were collected in nine second-growth deciduous forest patches that were selected for their adjacency to orchards. The most abundant tree species across sites were Acer saccharum, Acer rubrum, Quercus rubra, Fagus sylvatica, and Betula spp. Several sites also had Tilia americana, Populus deltoides, Juglans nigra, Carya ovata, or Carya cordiformis (see percent basal area at each site in Supplemental Figure A1). For simplicity, we will refer to bees and flower flies as “pollinators” hereafter. For full forest sampling methods, see Urban-Mead et al., (2021). Briefly, five paired canopy and understory pan traps in sets of vertically randomized blue, yellow, and white were deployed in canopy-dominant trees in second-growth forests and woodlots beginning in mid-March 2018-2019. A BigShot® slingshot was used to set a line of paracord with which to rig up the canopy traps at 25-30 meters above ground; canopy traps were not visible from the forest floor. Understory traps were ~1-meter aboveground.

In addition to our canopy and understory sampling, we placed identical sets of traps in the branches of five randomly selected apple trees haphazardly spaced across an orchard block <300 meters from the forest edge. Management varied between orchards, whose selection was constrained by forest adjacency. These pan traps were emptied and reset once every 7-10 days, on the same schedule as parallel forest sampling, until after apple orchards stopped blooming (corresponding roughly to the first week of June in both years). Pollinators were transferred directly into Whirlpaks (Nasco Whirlpak, Fort Atkinson, WI) with 95% ethanol in the field. Specimens were identified to species with a variety of taxonomic keys (Bouseman & LaBerge, 1978; Gibbs, 2011; Gibbs et al., 2013; Laberge, 1971; LaBerge, 1973, 1980, 1985). Difficult taxa were supported by expert knowledge (Jason Gibbs assisted with Lasioglossum (Dialictus), and Kaitlin Deutsch and Andrew D. Young with family Syrphidae), and DNA barcoding (for full methods, see (Urban-Mead et al., 2021). All specimens have been accessioned to the Cornell University Insect Collection.

For clarity and consistency, we will use the term “farm” or “site” to refer to a given orchard and its adjacent forest location, while “habitat” refers to the unique sampling location within each replicate site: the canopy, understory, and orchard. The five trees per habitat (canopy, understory, or orchard) were 100-200 feet apart, so are summed or averaged for most analyses due to their small spatial scale relative to pollinator flight distances (Greenleaf et al., 2007).

Digestive tract dissection and pollen slide preparation: Each insect was dissected directly upon removal from the Whirlpak bags while still pliable and soft from ethanol storage. Under a dissecting scope, each bee’s sternal segments were peeled backward towards the thorax with fine forceps, allowing for a second pair of forceps to extract the full digestive tract (crop, ileum, midgut, and rectum) without damage to external characters, as these were required for later species identification (Urban-Mead et al., 2021). The gut contents were placed on a microscope slide and macerated with forceps to release internal pollen and homogenized until evenly distributed across the slide. We removed large tissue fragments and broke up pollen clumps using forceps. Several drops of Calberla’s solution were added until pollen grains turned distinctly pink; a constant volume was not possible due to different pollen volumes. Each slide was covered with a 22x50mm coverslip.

Pollen identified from an individual adult bee is usually collected from its scopal loads following net collection, and represents the current foraging bout’s collection for brood provisioning. However, for bees caught in pan traps or other liquid traps, this scopal pollen can wash off or become contaminated, making it difficult to assign external pollen confidently to a given specimen. In these cases, identification and characterization of digestive tract pollen can provide insight into adult consumption (e.g. Cane et al., 2017; Dobson & Peng, 1997; Käpylä, 1978; Taniguchi, 1956; Urban-Mead et al., 2022). Pollen eaten by an adult bee may differ from that collected for brood provisioning, so gut pollen should only be interpreted in the context of an adult bee diet unless both are simultaneously recorded. We suggest that this method has the benefit of likely representing multiple visited flowers over several hours rather than simply the most recent foraging bout; many bees are constant within a single foraging bout but not between bouts or days (Brosi, 2016). Although poorly studied, healthy bumble bees can take over 13 hours to defecate after a meal (Giacomini et al., 2022), so digestive tract pollen likely represents foraging bouts of at least a day.

Pollen identification: In both years, we collected pollen from blooming plants to create a reference library and maintained detailed phenological records at each site. Only the site name and date were revealed to allow for phenologically informed identification; all other metadata was hidden to avoid bias. We used a compound Olympus BX41 microscope at 400x magnification (40x with 10x ocular). We began each sample at a randomly generated point along the slide’s short edge and moved along the length of the slide categorizing all pollen in the field of view. We categorized up to 300 grains or up to ten unique transects per slide; slides were marked as “no pollen” if there were fewer than 30 grains of pollen (Harmon‐Threatt & Kremen, 2015).

Each grain was identified and then sorted into one of three pollen categories: canopy, Rosaceae, and others. Pollen types in the category “canopy” included: Acer, Populus, Betulaceae, Carya, Fagus, Juglans, Pine-type, Quercus, Fraxinus. The “Rosaceous” category, due to similar morphology, broadly included all apple, peach, apricot, and other related orchard trees. It may also include Rubus and wild strawberry, known to be favored hosts of many Syrphid species, and Amelanchier, all of which were often found within the orchard or along edges. Black Cherry (Prunus serotina) is also rosaceous but was not a dominant tree in our woods, so we do not believe meaningfully impacts our inferences (mean percent of total basal area 3.3% ± 1.9SD). Not all understory species could be identified with confidence; these were combined into a macro category “Understory, Other, and Unknown Pollen.” We took multiple voucher images for each slide; see Supplemental Table A1 for detailed pollen categories and vouchers.