Wild Bees in the Trees: Pollen Analyses to Determine Wild Bee Foraging in Early Spring Canopies

Progress report for GNE18-188

Project Type: Graduate Student
Funds awarded in 2018: $14,912.00
Projected End Date: 08/31/2021
Grant Recipient: Cornell University
Region: Northeast
State: New York
Graduate Student:
Faculty Advisor:
Bryan Danforth
Cornell University
Expand All

Project Information

Project Objectives:

In spring 2017, 2018, and 2019, I employed novel canopy-sampling methods using elevated “bee bowl” traps at 11 sites in NY state. I caught wild bee pollinators in the canopy and understory of woodlots adjacent to pollination-dependent fruit trees. Woodlots were dominated by wind-pollinated tree species, and more than half of captured bees were known apple pollinators (Russo et al. 2015, Blitzer et al. 2016). Canopy bees were highly abundant during tree bloom, and rapidly declined after, supporting the idea that bees were foraging.

Here, I propose three research objectives investigating (1) bee abundance and activity in forest canopies, with analysis of the evolutionary relationships and life history diversity of the forest bees (2) the relationship between forest and orchard habitats for male and female bees at different times of year and (3) a comparison of nutritional resource quality and pesticide exposure for bees visiting forest-canopy and orchard blooms.

(1) I am using taxonomic guides, expert support, and DNA barcoding to identify the wild bee species found in the canopy, understory, and adjacent orchards. Many commonly found forest bees are poorly known and thus hard to identify due to underdeveloped taxonomic resources. This is especially true for groups such as mining bee males; the females of this group of bees are highly effective pollinators, so understanding the behavior and foraging choices of males is crucial. 

(2) I am using microscopy to identify the pollen from guts of trap-caught wild pollinators. Traps demonstrated immense diversity of canopy-foraging bees, but not which trees the bees preferred. Pollen identification through microscopy will provide detailed tree pollen use by the wild pollinating bees. I ask if bees are foraging on wind-pollinated forest canopy trees in early spring, which would support more direct action for forest management for bees. I also hypothesize that pollen preferences vary over space, time, and the gradient of forest cover in the surrounding landscape. Note that this microscopy method is change in approach, adjusted with the goal of answering the same ecological questions but with more detail (discussion of change in methods sections).

(3) I will quantify the pollen quality and quantity of important forest and hedgerow trees with metrics relevant to bee health. To estimate pollen quantity, I will improve estimates of flower abundance for trees and shrubs in an in-development floral resource database for the Finger Lakes region of NY (Iverson et al. unpublished data), then combine these data with per-flower estimates pollen abundance. The only available pollen quality data is missing many local species and provides only protein content (Roulston et al. 2000), while protein:lipid ratio of pollen is most important for pollinator choice (Vaudo et al. 2016a, 2016b). I thus propose to collect and calculate protein:lipid ratios for the common wind- and insect-pollinated forest trees and compare these with apple pollen in adjacent orchards. I will ask if the preferred trees identified in objective (2) also have the highest quality or quantity pollen. I will quantify pesticide residues in the pollen of forest and hedgerow trees. Using pollen from trees collected 100-500m from the orchard, I will analyze pollen samples for 41 compounds commonly applied in our landscape at the Cornell Chemical Ecology Core Facilities, through the McArt lab. I will calculate toxicity estimates based on the literature and compare wild tree pollen toxicity relative to apple pollen.


The purpose of this project is to identify the forest, woodlot, and hedgerow tree species that are most important for supporting fruit tree pollinators. While much research has focused on herbaceous wildflower plantings for pollination services in agricultural crops (Blaauw and Isaacs 2012, 2014, Grass et al. 2016), almost nothing is known about how forest and shrub resources support pollinator populations and effect pollination services. This is a surprising gap in knowledge since forests and forest fragments commonly surround orchards in the Northeastern US, and the resources they provide may be critical for robust pollinator populations. For example, our own data from flower counts taken in spring 2017 found the number of flowers on a single canopy maple or oak tree often exceed 600,000 or even 700,000 blooms/tree. A single male red maple can present up to 1350 flowers/m 3 of the outer 2 meters of canopy, which translates to an estimated 1.35 ml nectar/m 3 (Batra 1985). This exceptional resource likely contributes to findings that natural forest habitat and plant diversity support thriving pollinator populations in apple orchards in the northeast (Park et al. 2015, Kammerer et al. 2016), yet these patterns have never been measured directly, nor is the contribution of specific tree and shrub resources known.

Of the 416 species of wild bees in New York State, over 120 species have been found visiting apple orchards (Russo et al. 2015, Blitzer et al. 2016) . Wild bees emerge after spring thaw immediately requiring pollen and nectar, and forests are among the only available forage resources (Chambers 1968, Russo et al. 2013) . Since the orchards they pollinate will not bloom for several weeks, bees require other resources to stay active until crops bloom, and to amass sufficient resources to reproduce to maintain healthy populations each year (Schellhorn et al. 2015) . While spring ephemerals certainly provide pollen and nectar and are the focus of much research (Motten 1986, Williams and Winfree 2013, Parker et al. 2016, Austen Emily J. et al. 2018) , they pale in comparison to the sheer number of flowers in the canopy. Although their nutritional quality has never been investigated, many scattered literature observations suggest wind-pollinated trees are visited by diverse bee species (Chambers 1968, Raw 1974, Russo and Danforth 2017) .

A healthy wild bee population reduces the need to rent honey bee hives and provides pollination services in poor weather and unpredictable climates (Blitzer et al. 2016) . Biodiverse pollinator populations are insurance against climate change, pests, and diseases, as bees with different traits respond differently to climate pressures (Bartomeus et al. 2013) , so species diversity supports economic security in the form of a successful crop each year (Aizen and Harder 2009, Garibaldi et al. 2013) . It will thus directly inform land management for fruit tree pollination services and thus crop viability to identify the preferred tree species and the quantity and quality of pollen they produce for a maximally diverse bee population.

The value of pollination services in NY State is estimated at $500 million/year for crops including apples, pumpkins, and strawberries (New York State Departments of Environmental Conservation and Agriculture & Markets. 2016) . Pollination is crucial for human diets, including a high proportion of nutrient dense foods (Eilers et al. 2011) . Yet, the average loss rate for honey bee hives in the U.S. in 2014-2015 was 49.0% (Seitz et al. 2015) . Orchardists are struggling to find honey bee hives to rent for pollination services, (personal communication to Scott McArt from Jeff Morris, Glenora Farms, Dundee NY), and rentals have risen to $140/hive on average in recent years (Wheeler et al. 2018). Wild bees increase pollination services regardless of honey bee abundance (Winfree et al. 2007, Garibaldi et al. 2013) , and are important pollinators in New York apple orchards (Blitzer et
al. 2016) . There is also widespread cultural interest and knowledge around managing for wild bees. A recent survey of over 600 orchardists in New York and Pennsylvania found that 93% of respondents highly valued wild pollinators, and sought to actively manage their land for wild bee conservation (Park et al. 2018) . Some orchardists are now choosing to rely entirely on wild bee services (Dunn 2018) .

Wild bees often suffer from nesting habitat and forage loss associated with land-use change and agricultural intensification (Steffan-Dewenter Ingolf et al. 2002, Carvell et al. 2006) . Wildflower plantings, a frequently suggested solution, can increase wild bee abundance, diversity, and pollination services (Blaauw and Isaacs 2014) , yet such strips require expensive establishment and are often overtaken by weeds requiring herbicide management (Landis and Savoie 2018) . In contrast, once established, perennial trees and shrubs provide resources to bees in addition to long-term benefits such as soil stabilization, wind breaks, carbon sequestration, timber, firewood, shade for animals, and even maple syrup or nuts (Farming the Woods 2014, Graham and Nassauer 2017) . Thus, identifying the forest and hedgerow species which provide food to bees could successfully maximize ecological services and increase not just fruit pollination but whole-farm ecological and economic sustainability (Bommarco et al. 2013)(Potts et al. 2016) .

There is a growing call for basic research into insect use of wind-pollinated resources, a large portion of which are early-blooming trees (Saunders 2018) . Their pollen-releasing catkins are not showy or colorful, so not traditionally understood as attractive to bees and thus ignored in forest resource calculations. Yet several “wind-pollinated” species are in fact used extensively by bees and partially pollinated by insects, including willow (Peeters and Totland 1999) , several maples (Sullivan 1983, Batra 1985) , linden (Anderson 1976) , and others. Scattered historic data suggest that bees frequently collect from wind-pollinated oak, ash, beech, and birch (Chambers 1968, Raw 1974) . Thus, to understand how to select forest and hedgerow trees to support pollinators, we must re-evaluate the importance of these less-showy resources. Therefore, I propose here to calculate the use patterns and quality of this resource for the broad community of wild bees in order to fill this knowledge gap.


Materials and methods:

As a brief background on the methods for this project, in 2017, 2018, and 2019 I respectively sampled in 6, 11 and 9 different forest-orchard pairs across the Finger Lakes region of New York State. Prior to receiving the SARE grant, I worked with each of the growers to confirm sampling goals and areas and completed two seasons of bee sampling. To set up, I used a “Big Shot” 7-ft slingshot to launch a weight attached to a thin rope. In five trees per forest, I aimed precisely for a branch or branching fork that would allow for sampling at the edge of the canopy tree. Once positioned correctly, I then reeled a bee trap into the canopy, and installed a paired trap at ground level. These traps were “bee bowl” cups in standardized fluorescent yellow, white and blue, each color of which is known to attract a different suite of bee species. Together the three bowls comprise a standard bee sampling method. Five similar tri-color traps were deployed in adjacent orchards in the branches of apple trees. Bees were collected weekly from each trap at all sites and placed directly into small Whirlpak bags with 95% Ethanol for later processing. Sampling began each spring immediately after winter thaw and continuing one week after apples finished blooming across all sites (each year, roughly first week of April-first week of June).

All specimens were pinned and databased using Biota software with unique identifiers and associated metadata (Colwell 1994). Specimens have been accessioned to the Cornell University Insect Collection. Specimens were identified to species or nearest possible morphospecies using online guides and existing taxonomic keys (Laberge 1971, LaBerge 1973, 1980, 1985, Gibbs 2011, Gibbs et al. 2013, discoverlife.org, and Andrena subgeneric keys by Mike Arduser, unpublished). Due to unresolved taxonomy and lack of available keys, specimens of Nomada (n=61), Hylaeus (n= 3), and Sphecodes (n=3) were only determined to the genus level; some individuals were determined to the generic or subgeneric level due to damage in bee bowls (n= 35). Other difficult taxa were resolved in two different ways: with the help of taxonomic experts (Jason Gibbs assisted with Lasioglossum (Dialictus)), or using DNA barcoding for a subset of unidentified Andrena (see below).

We extracted DNA from 130 Andrena bees using a CTAB protocol available on the Danforth lab website (http://www.danforthlab.entomology.cornell.edu/research/resources/). These bees were selected as a subset of the full dataset either (1) to confirm a tentative morphological identification, or (2) because the specimen was damaged or exhibited traits that could not be categorized clearly with the existing keys. DNA was extracted from 3 legs which allowed us to retain the remaining, nearly complete, specimen for further study and as a pinned voucher. We used primers developed by Creedy et al. (2020) for barcoding British bees: BeeF (5’-TWY TCW ACW AAY CAT AAA GAT ATT GG-3’) and BeeR (5’-TAW ACT TCW GGR TGW CCA AAA AAT CA-3’). These bee-specific primers worked consistently across nearly all of our samples and rarely yielded non-target sequences. For PCR, we used 35 cycles of 94°C for 45 seconds, 52°C for 45 seconds, and 72°C for 1min. All PCR fragments were sequenced in both directions. After trimming primers, our fragments were 658bp in length. Sequences were edited and aligned in Geneious (Geneious Prime® 2019.0.4; Build 2018-11-27 02:05; Java Version 11+28 (64 bit), available at https://www.geneious.com).

Objective (1) was first to finalize bee abundance and activity in forest canopies, with analysis of trait and life history diversity of bees active in the forest.

This work has now been published, and is available on the Forest Ecology & Management website (see citation and discussion of results below).

Objective (2) is to use pollen analyses to characterize the relationship between forest and orchard habitats for male and female bees at different times of year and, and to do so I am in the process of identifying the pollen collected by each bee species. This will allow us to understand which tree species in the early spring forest are most important for wild bee pollinators. In the first step towards this answer, an undergraduate mentee (Xavier Carroll) and I dissected each of the bees collected in 2018, which had been collected into 95% Ethanol. Bees were surface sterilized so no contaminant pollen was collected. Dissection was done carefully to remove only the internal guts, but not damage the structure of the bee’s external integument, so that they might still be identified to species under the microscope. All guts were immediately placed into individually labeled tubes and placed in the -80C freezer, to maintain them for future identification via molecular (or microscopy) methods. The original method for this pollen identification was to use pollen meta-barcoding, a molecular approach to identify the full community of pollen in a sample (in this case, in a wild bee’s digestive tract). However, since the time of funding, several new papers have demonstrated discrepancies in the ability of this method to accurately quantify the amount of different pollens, and bias depending on the relative abundances in the original sample. This could make it impossible for me to know if bees were actually using significant amounts of certain species of trees that might amplify in their gut samples, or if there were simply contamination or otherwise negligible amounts of those pollens in their digestive tracts. Thus, upon careful reflection, my advisors and I have adjusted my allocation of funds to ensure that I am accurately & thoroughly able to answer the ecological questions here funded. Using the DNA barcoding instead for the wild bee identification allows for a highly accurate and cutting-edge approach to characterizing this relatively-unknown insect community (methods described above). For pollen, using microscopy instead from meta-barcoding allows for accurate insight into relative resource use. This is very important, as I am seeking to particularly identify resources that have been overlooked in the past and might be dismissed as contamination unless we can clearly demonstrate amount of use.   

(2b) A related question we can answer with these data is the relationship between male and female pollen consumption in the genus Andrena, a common early-spring group abundant in both sexes in our dataset. Male bees are commonly expected to eat no pollen, or only incidentally while they drink nectar. 

Objective (3) investigates the nutritional resource quality and pesticide exposure for bees visiting forest-canopy and orchard blooms. During these field seasons, I have collected branches from 18 common tree species in local woodlots, and collected all pollen that dropped from these blooms after they dehisced in order to obtain a pure sample of each. These samples will be analyzed for protein & lipid contents and screened for pesticide content at the Cornell Pesticide Core Facility this spring. 

For quantity estimates, I am collaborating with Prof. Aaron Iverson to estimate blooms/tree for all common tree species in regional woodlots, while taking into account variables such as tree age and canopy position. I am collecting flowers to calculate the amount of pollen / anther, anthers / flower. These numbers will allow us to understand the bee-relevant diet quality in the context of the availability of those resources on the landscape. 


Research results and discussion:

Background Data & Important Ecological Correlates

Descriptive graphs characterizing the forest composition (measured with variable radius plots using an angle gage, BAF 10) and light levels (measured with a hand-held Densiometer; superimposed bloom windows estimated to represent dates compiled across sites).



OBJECTIVE (1) Forest bee community characteristics

Since last winter’s report, I have published my first peer-reviewed journal article, describing the vibrant and diverse wild bee pollinators found in early spring forest canopies.

citation: K. R. Urban-Mead, P. Muniz, J. Gillung, A. Espinoza*, R. Fordyce, M. Van Dyke1, S. H. McArt1, B. N. Danforth.  “Bees in the trees: Diverse wild bee communities in temperate forest edge canopies” Forest Ecology & Management (2021 In Press).

I summarize the major results of this paper here:

  • Despite equal species richness among strata, woodlots had vertically stratified bee communities
  • We found higher overall diversity and more female bees in the canopy
  • The traits of nesting habitat and sociality were strongly predictive of strata occupancy
  • Canopy cover (leaf-out) was negatively associated with understory but not canopy bee abundance
  • Regional amount of coarse woody debris correlated with understory abundance of some trait groups


OBJECTIVE (2) Gut Pollen Dissections & Pollen Identification for Diet Characterization. 

In 2018 and 2019 respectively, we caught 499 and 621 Andrena (male bees 59 and 53%), the important pollinators commonly called mining bees. Following gut dissections in both 2018 and 2019, I found that male mining bees were more abundant in the forest than in the orchard, and females were most abundant in the canopy, followed by the orchard (all contrasts p = <0.001). Male and female adult bees caught in apple orchards and canopies were more likely to have pollen meals in their digestive tracts (p = <0.001). This suggests that the canopy catches may indeed be due to foraging in those treetops. In 2019, 158 out of 328 (48%) of male Andrena had pollen in their digestive tracts. 


(2a) So far, we have identified the pollen samples from 2018, with 2019 pollen sample processing underway to be completed before the beginning of the spring field season.


Questions that I will explore with these data for an upcoming publication are:

–What are the differences in pollen diet of bees caught in different strata?

–Does the structure of pollen consumption networks differ among male and female bees caught in different strata?

–Does the structure of pollen consumption networks change across the season? Among years? During mast and non-mast blooming years of canopy-dominant tree species?

–Do relative pollen proportions consumed match or differ from the proportion of pollen available on the landscape?

(2b) Through our comprehensive dissection and quantification of gut pollen, we find that male bees eat a non-negligible quantity of pollen–although significantly less than females, who rely on the protein source for egg production.

So far, the proportion of male and female bees with any pollen at all in their guts is similar. 

When comparing absolute counts up to 300 grains, the amount of pollen present in male and female guts of varied among Andrena subgenera, although the absolute amount was always lower in male than female guts. As far as we know, there is no formal quantification of this phenomenon in the literature to date. 

A surprising finding that warrants further study was that of all male Andrena with pollen in their guts, 25% of those with pollen had gymnosperm-type pollen (that is, coniferous). These distinctive grains were highly abundant, suggesting that bees were either consuming them intentionally or otherwise spending sufficient time in conifers that accidental consumption was common. This pattern warrants further investigation in future research. 


OBJECTIVE (3) Quantity and Quality of Forest Tree Pollen 

I have preliminary counts of the absolute amount of pollen produced by 10 of our most common forest trees. For example, two Red Maple trees had 9,400 and 15,600 flowers, while two canopy red maples respectively had 157,100 and 132,800 flowers. Red maple flowers had an average of 10 anthers per flower, for a range of 94,000 to 1,571,000 anthers per tree of these four counted trees. In a silver maple, two trees had 17,800 and 510,100 flowers per tree, 19 anthers per flower for a range of 338,200 to 9,691,900 anthers per tree, and an average of 2330 pollen grains per anther, for a total estimate of 788,006,000 to 22,582,127,000 (~800 million to ~22 trillion) pollen grains per tree.  I will continue to finalize these number and compare them to apple tree estimates, and then multiply these out by landscape scale abundance of trees to understand landscape scale estimates. 

I have also collected pollen in volume from the common forest trees & apple trees in my study sites, in preparation for quality analysis (protein & lipid) and risk (suite of pesticides, analyzed by Cornell’s core pesticide facility) analysis later this summer. I have sufficient pollen for pesticide analysis for six common wind-pollinated regional trees (alder, quaking aspen, silver maple, white birch, box elder, red oak, and black oak), which requires a very large pollen quantity. I have sufficient pollen for protein:lipid analysis for an additional seven species, for a total of 13 local species. This longer list includes apple and several ornamental and insect-pollinated forest trees for comparison with the wind-pollinated forest species. My goal for this spring’s field work includes finalizing the collections of these volumes. I plan to do bioassays this summer with bulk dry pollen to better understand the circumstances when bees choose to eat wind- versus insect-pollinated plant pollen. 

Research conclusions:

K. R. Urban-Mead, P. Muniz, J. Gillung, A. Espinoza*, R. Fordyce, M. Van Dyke1, S. H. McArt1, B. N. Danforth.  “Bees in the trees: Diverse wild bee communities in temperate forest edge canopies” Forest Ecology & Management (2021 In Press).

Generally, we find:

  • Despite equal species richness among strata, woodlots had vertically stratified bee communities
  • We found higher overall diversity and more female bees in the canopy
  • The traits of nesting habitat and sociality were strongly predictive of strata occupancy
  • Canopy cover (leaf-out) was negatively associated with understory but not canopy bee abundance
  • Regional amount of coarse woody debris correlated with understory abundance of some trait groups

Management recommendations for growers include:

  • Value nearby forest as pollinator habitat, overwintering habitat, and food resources
  • Work with foresters to generate wildlife friendly forest management, particularly prioritizing management for:
    • coarse woody debris, snags, and any other sources of deadwood
    • species who offer resin, honeydew, nectar, and pollen value
    • species diversity at the patch scale
    • landscape scale successional diversity
    • vertical structure
  • Plant and maintain hedgerows


Participation Summary
15 Farmers participating in research

Education & Outreach Activities and Participation Summary

1 Journal articles
2 Published press articles, newsletters
2 Webinars / talks / presentations

Participation Summary

20 Farmers
Education/outreach description:

(POSTPONED to 2021, DUE TO COVID-19) June 2 2020: Schuyler County Conservation Field Days — helping with day-long festival outreach

(POSTPONED to 2021, DUE TO COVID-19) July 25, 2020: “Tent of Knowledge” Catskill Forest Festival – Invited presenter (50 minute presentation)

(POSTPONED to 2021, DUE TO COVID-19) September 25, 2020: Cleveland Pollinator & Native Plant Symposium —Invited Speaker

(POSTPONED to 2021, DUE TO COVID-19) September 22, 2020: Volunteer Pollinator Specialist program and ‘Urban Forestry’ career track student visit in Kent, OH;.


May 6, 2020: Catskill Forest Radio —Invited guest.  https://catskillforest.org/radio-new/

I had fun chatting with Ryan Trapani & John MacNaught at Catskill Forest Radio. I’m sorry that the in-person event at the Catskill Forest Festival’s Tent of Knowledge was canceled, but look forward to joining in 2021.


May 14th, 2020: Oregon Bee Atlas Remote Speaker Series —Invited zoom presentation. (40 min)

As a follow-up to my interview on PolliNation with Andony Melathopolous at the Oregon State Extension Service, Andony and Linc Best invited me as a remote speaker as part of the Oregon Bee Atlas’ weekly speaker series! I stayed up late to give a west coast talk, and I’m glad that I did–144 people tuned in. whew!

The link to the youtube video is here, and the link to the Wild World of Bees full speaker series is here.


Staatsburgh State Historic Site March 2020: Video & social media outreach, cellophane bee aggregation


New York Forest Owner’s Magazine Extension publication Jan 2020: Wild Bees Amidst the Trees

My article “Wild Bees Amidst the Trees” on bee biology and forest management for wild bee conservation in the New York Forest Owner’s Journal.



Project Outcomes

2 New working collaborations
Project outcomes:

I hope that this project will contribute to future sustainability by directly integrating pollinator conservation into agroforestry, silvopasture, and small-farm woodlot and hedgerow management. This will be particularly relevant for apple orchards, the focal commodity of our lab’s research. However, many crops and wild plants benefit from or entirely rely on ecosystem services provided by free-living pollinators. 

Knowledge Gained:

I am eager to continue working with growers, researchers, and community members to answer these exciting questions about the role of woodlots and forests in wild pollinator conservation. I am committed to sharing the knowledge I gain through careful research and collaboration.

In my future career, I hope to work at the intersection of extension and teaching, either through non-profit government or a joint extension/teaching faculty position. This current SARE project is deepening my natural history knowledge in the context of working agricultural lands, and should generate concrete outcomes in terms of being able to recommend woodlot and hedgerow management for pollinator conservation. Given the importance of trees for soil stabilization, wind breaks, carbon sequestration, forage, and numerous other sustainability benefits, I am particularly excited to find myself poised at the cusp of integrating pollinator conservation with these efforts. 

Any opinions, findings, conclusions, or recommendations expressed in this publication are those of the author(s) and do not necessarily reflect the view of the U.S. Department of Agriculture or SARE.