Volume 12 Issue 2
Author:
Lisa Wasko DeVetter | Associate Professor of Horticulture | Washington State University | 360.848.6124 |lisa.devetter@wsu.edu
Highlights:
- Beeflow increased honey bee visitation to ‘Liberty’ flowers in 2022, but not to ‘Duke’ flowers in 2021.
- Return forager counts, a measure of honey bee colony strength, was numerically greater among Beeflow-treated hives.
- Berry weight was consistently greater in Beeflow treated sites in 2021 and 2022, although effects on fruit set were not detected.
- Our data suggests Beeflow is a promising technology to enhance honey bee pollination in blueberry systems. However, overall pollination can be influenced by other landscape factors that influence bee density in a field. Furthermore, fruit set, berry weight, and yield outcomes are impacted by other factors including pruning, nutrient management, and irrigation and may limit yield gains despite improved pollination.
- Additional economic analyses are suggested to determine if the investment in Beeflow technology leads to improved profitability at the farm level.
Background:
Adequate pollination is an essential step to sustain economically viable yields of high-quality blueberries. Highbush blueberries (Vaccinium corymbosum) require insect-mediated pollination, which is largely provided by managed honey bees (Apis mellifera) that are rented by growers for pollination services. While low nectar volumes and floral morphology make blueberry flowers poorly suited for efficient pollination by honey bees (Courcelles et al., 2013), these pollinators are nevertheless essential given they can provide the number of foragers needed to pollinate large blocks of flowers in modern blueberry system. Despite this dependence on honey bees, blueberry growers in northwestern Washington frequently experience poor pollination that can result in small berry sizes and limit yields. Environmental variables such as rainfall, wind, and/or temperatures below 55°F depress honey bee foraging (Winston, 1987) and often coincide with the bloom period in western Washington and exacerbate pollination deficits. Some cultivars are also more challenging to pollinate due to floral morphological features and nectar characteristics, which honey bees primarily feed on when deployed in blueberry fields.
There is a need to optimize honey bee mediated pollination in modern blueberry systems, which has been a key focus of the Washington State University Small Fruit Horticulture program led by DeVetter. New technologies are attempting to leverage emerging knowledge about honey bee behavior and attraction. Beeflow (https://www.Beeflow.com/) is a new technology that has experienced success in California almond (Prunus dulcis) production and the company recently partnered with Driscoll’s. Beeflow’s strategy is to “train” honey bees to forage on target crops, even if they naturally tend to be less attractive to honey bees, and to do this across a wider temperature range. They accomplish this through provision of a supplement that “trains” honey bees to forage on the target crop using odorant mixtures that contain floral scents of the target crop. According to Beeflow, the supplement also supports bee immunity and nutrition. The strategy has scientific merit (Arenas et al., 2007; Balbuena et al., 2012; Farina et al., 2022; Negri et al., 2017) and preliminary results have been promising. According to Stine (2019), Beeflow’s technology has led to a seven-fold increase in bee activity in California’s Central Valley at temperatures below 55 °F.
The initial success of the Beeflow technology warrants further investigation. The objective of this study was to assess the impacts of Beeflow technology on honey bee activity and crop pollination in highbush blueberry systems in northwestern Washington.
Approach:
Assessments of Beeflow technology were done over two years (2021 and 2022). In 2021, we established a paired design experiment with the main treatment factor being the presence or absence of Beeflow technology. A hive-applied feed was the primary variable that was different between treated and untreated sites with the feed applied approximately two weeks before placement in treated blueberry fields and repeated every two weeks by the beekeeper or Beeflow team. Treated sites also had beehives placed in a distributed configuration with pallets of beehives adjacent to the field edge. Untreated sites did not manipulate placement location of beehives.
Treatments in 2021 were replicated across three field sites each (n=6 sites total; 4 ‘Duke’ and 2 ‘Last Call’) all stocked at 4-6 hives/acre of honey bees. Unfortunately, the ‘Last Call’ pair was dropped because the Beeflow treatment could not be successfully implemented with our beekeeper cooperator. Remaining sites were all conventionally managed, had the same grower within a pair, and were located in Skagit and Whatcom counties of northwest Washington. Hives were placed in the study sites at approximately 5-10% bloom. To maintain independence, each site was a minimum of 2 miles apart and all had the same beekeeper (Belleville Bees) to minimize beekeeper variation. All plants were established and approximately the same age (6 years or older).
Honey bee foraging (i.e., flower visitation) data were collected weekly on suboptimal and optimal pollination days (temps >55 °F) to evaluate Beeflow’s effect on honey bee activity across environmental conditions. Honey bee foraging was assessed within three, 328-ft-long transects per field with transects starting 30 ft from the field edge and located approximately in the geometric center of the field in order to get an estimate of average pollination conditions. Honey bees foraging data were recorded by slowly walking down a transect (i.e., row) and counting the number of honey bees feeding from open blueberry flowers on a half-bush basis.
Honey bee colony strength was also measured using a non-invasive return forager method whereby the number of honey bees returning to their colony are counted within a 1-minute period (Sagili and Burgett, 2011). Return forager counts were measured from four randomly selected hives per field at the beginning and end of bloom. All honey bee data were collected between 10 am – 4 pm. Each field was visited a minimum of three times, with some fields being visited four times during the bloom window.
Within each transect, three bushes were labeled with flagging tape and flowers were counted from one labeled and representative branch at mid-canopy level per bush. Flowers on labeled branches were counted twice during bloom. Flower counts were later used to determine percent fruit set [i.e., (# blue fruits ÷ # flowers) x 100] from each branch. Average berry weight was measured from fully ripe blueberries collected from our labeled branches shortly before commercial harvest.
In 2022, another paired design experiment was established and conducted in northwest Washington. However, ‘Liberty’ was used instead of ‘Duke’ and one site was only 0.6 miles away from another field site instead of 2 miles. All data were collected using the same protocols implemented in 2021 with a few exceptions. In 2022, three branches were labeled per bush and assigned one of three sub-treatments: 1) Bagged = pollinators excluded/negative control; 2) Hand pollinated = full pollination/positive control; and 3) Open pollinated = natural field-level pollination. This method can be used to assess whether pollination is limited by comparing hand to open pollinated branches because hand pollinated flowers should theoretically be maximally pollinated. Seed number per berry was also measured in 2022 (seeds were extracted from 27 ripe berries per site). Seed set is an indicator of pollination success with more seeds per berry indicating greater pollination, fertilization, and seed development.
Data were analyzed using linear mixed-effects models using R Studio software and transformed as needed to meet model assumptions. Data are presented in original units. Our threshold for statistical significance was set at P<0.05. P-values less than 0.05 were considered statistically significant.
Results and Discussion:
In 2021, average honey bee forager counts were the same between Beeflow treated and untreated sites (Fig. 1). Average return forager counts (an estimate of hive strength) were also not different between the Beeflow treated and untreated sites, with treated hives having 25.2 (±3.4) return foragers/minute relative to 31.6 (±7.9) for the untreated control. Observed return forager counts were low relative to the recommendation of a minimum of 100 returning bees/minute (Sagili and Burgett, 2011). Low return forager counts may be attributed to air temperatures frequently being below 64 °F, as the recommendation is based for temperatures of 64 °F and above. Similar return forager counts have been measured in other pollination studies in Washington blueberry pollinated by different beekeeping operations (DeVetter et al., 2016) and appears to be temperature sensitive. Fruit set was also unaffected by treatment, averaging 88 (±4.5)% and 94 (±0.85)% in treated and untreated sites, respectively. However, average berry weight was larger for Beeflow treated sites [averaging 1.8 (±0.03) g/berry] versus the untreated sites [1.2 ±0.05) g/berry] and could be due to elevated bee visitation not captured in our forager counts.
While average honey bee forager counts were the same between the treatments, there was a very observable increase in honey bee foraging on sub-optimal weather days for Beeflow treated sites indicating Beeflow stimulated foraging during suboptimal weather conditions, but the effect wasn’t strong enough when looking at averages across the season (Tables 1-2). Pollination conditions were overall very good for ‘Duke’ in 2021 and ‘Duke’ is a relatively easy cultivar to pollinate, which may have limited our ability to detect other potential effects from the Beeflow treatment.
Table 1. Honey bee forager counts on two sub-optimal weather days in ‘Duke’ highbush blueberry treated with Beeflow versus an untreated control. Data collected in northwest Washington in May 2021.
| Treatment | 2-May-21 | 4-May-21 | ||
| Honey bee countsz | SE | Honey bee counts | SE | |
| Control | 41.7 | 19.1 | 1.7 | 0.9 |
| Treated (Beeflow) | 106.0 | 17.6 | 27.7 | 2.3 |
zNumber of honey bees/half bush/328-ft (100 m) transect; SE denotes standard error.
Table 2. Environmental conditions in ‘Duke’ field sites immediately before and after honey bee foraging was measured on 2-May and 4 May 2021 in northwest Washington.
| Date | 2-May-21 | 4-May-21 | ||
| Variable | Before | After | Before | After |
| Temp (F) | 55 | 61 | 55 | 55 |
| Humidity (%) | 62 | 51 | 71 | 70 |
| Wind (mph) | 3 | 0 | 5 | 5 |
| Pressure (inches) | 30.24 | 30.25 | 30.24 | 30.24 |
| Cloud cover (%) | 80 | 50 | 70 | 80 |
| Bloom (%) | 60 | 80 | ||
‘Liberty’ was used in 2022 and was selected because it is known as a more challenging cultivar to pollinate due to its smaller flowers (Courcelles et al., 2013). Despite suboptimal bloom time conditions in 2022, honey bee activity measured as forager counts was greater in Beeflow–treated sites compared to untreated sites (P=0.03; Fig. 2). Farm site also had an effect (P= 0.02; Fig. 3), although there was no interaction between farm site and treatment (P=0.56). These data suggests that the Beeflow treatment can increase honey bee visitation to ‘Liberty’ blueberry flowers, but there are also landscape–level effects that can influence measures of honey bee activity. The effect of farm site is unsurprising given a recent publication that shows honey bee visitation is determined by the number of honey bee hives in the surrounding landscape within a 0.6 mile radius of a field and that field-level stocking densities can miss contributions from other bees within the landscape (Eeraerts et al., 2022). In addition, management could also influence honey bee visitation and is difficult to control for in field experiments. Despite the effect of field site, the Beeflow treatment effect was strong enough that it met our threshold of statistical significance set at P=0.05.
Average return forager counts (an estimate of hive strength) were numerically greater for the Beeflow treated sites when averaged across sampling dates in 2022 (Table 3). Although the treatment effect was not statistically significant given the traditional threshold value for statistical significance is P=0.05, the P-value for this measurement was 0.07, which is 0.02 units away from statistical significance and could be biologically meaningful. In contrast, the effect of farm site and the interaction between Beeflow treatment and farm site was not significant (P=0.9 and 0.47, respectively). Average return foragers were also compared between the treatments at two temperature ranges, 50-59 °F and 60-70 °F, which represented the air temperatures during pollination data collection. Overall, number of return foragers did not appear affected by Beeflow treatment at these temperature ranges (Table 4).
Table 3. Honey bee return forager counts in ‘Liberty’ highbush blueberry treated with Beeflow versus an untreated control (n=4 sites; 2 sites per treatment). Data collected in northwest Washington in May 2022.
| Treatment | Average honey bees/minz | SE |
| Control | 26 | 3.9 |
| Treated | 35 | 3.3 |
zAveraged across four hives/site.
Table 4. Honey bee return forager counts in ‘Liberty’ highbush blueberry treated with Beeflow versus an untreated control (n=4 sites; 2 sites per treatment) by ambient air temperature. Data are means ± standard error. Data collected in northwest Washington in May 2022.
| Average honey bees/minz | ||
| Treatment | 50-59 °F | 60-70 °F |
| Control | 24±5 | 37±6 |
| Treated | 19±5 | 41±4 |
zAveraged across four hives/site.
Similar to previous years, fruit set was not affected by our Beeflow treatment (P=0.93; Fig. 4). In contrast and as shown in Table 5, Beeflow treated sites had greater average berry weight than the untreated control (P=0.006), although there were also farm site effects (P=0.0001) with no interaction between Beeflow treatment and farm site (P=0.08). Seed set was lower for Beeflow treated sites (P=0.01) with significant farm and farm by treatment interactions (P=3.5e-10 and 0.0009, respectively). The interaction shows the effect of Beeflow on seed development is contingent upon the site and likely site management. This is consistent with the results presented by farm location (Table 6). Regarding comparisons between bagged–, open–, and hand-pollinated branches, bagging effectively reduced berry weight and seed set, but no significant differences were observed between open- and hand- pollinated treatments (Table 7). This may be due to accidental damage caused to flowers that are hand pollinated.
Table 5. Average berry weight and seed number per berry of ‘Liberty’ treated with Beeflow versus an untreated control (paired design; n=4 sites). Berry weight was greater for Beeflow treated sites relative to the untreated control (P=0.006), while seed set was lower for Beeflow treated sites (P=0.01). Data were collected Aug. 2022.
| Treatment | Average berry weight (g) | Seed #/berryz |
| Control | 1.3 | 11.9 |
| Treated | 1.6 | 9.2 |
zDetermined from 27 berries per treatment per site.
Table 6. Average fruit set, berry weight, and seed number of ‘Liberty’ blueberry by farm location and treatment. Data were collected in 2022.
| Farm ID | Treatment | Fruit set (%) | Berry weight (g) | Seed no/berryz |
| 1 | Beeflow | 76.9 | 1.3 | 11.1 |
| 2 | Beeflow | 34.9 | 1.8 | 7.4 |
| 3 | Control | 35.9 | 1.2 | 17.3 |
| 4 | Control | 70.6 | 1.4 | 6.4 |
zDetermined from 27 berries per treatment per site.
Table 7. Berry weight and seed number of ‘Liberty’ blueberries treated with or without Beeflow in northwest Washington. Branches were subject to three sub-treatments within Beeflow and untreated sites: 1) Bagged = pollinators excluded/negative control; 2) Hand pollinated = full pollination/positive control; and 3) Open pollinated = natural field-level pollination.
| Treatment | Average berry weight (g) | Seed #/berryz |
| Beeflow | | |
| Open | 1.6z | 9.2 |
| Bagged | 0.0 | 0.0 |
| Hand | 1.3 | 6.9 |
| Control | | |
| Open | 1.3 | 11.9 |
| Bagged | 0.3 | 0.7 |
| Hand | 1.4 | 10.6 |
zDetermined from 27 berries per treatment per site.
Conclusions:
Our interpretation to date is that Beeflow is a promising technology. We observed increases in honey bee visitation to ‘Liberty’ flowers and numerical increases in hive activity as measured by average return forager counts. Ultimate effects on fruit set, berry size, and seed set were inconsistent and likely depends on other management factors outside of pollination such as pruning, disease/pest pressure, and nutrient and irrigation management. However, we did observe consistent increases in berry weight in Beeflow treated sites in both ‘Duke’ and ‘Liberty’. Furthermore, a recent publication shows landscape density of hives within a 1000 m (0.6 mile) radius of a field has a greater impact on honey bee visitation to blueberry flowers than field level stocking density (Eeraerts et al., 2022). Therefore, Beeflow treatment effects could potentially become diluted across landscapes depending on the proportion of hives treated with this supplement within an area of 1000 m. These management and landscape factors are difficult to parse out from data collected on multiple cooperator sites and is a topic of future investigation being undertaken by DeVetter and collaborators. However, the numerical increases in honey bee foraging in 2021 and significant increase in 2022 combined with increased hive activity and berry size is promising and shows the potential of Beeflow to be used as an aid to promote honey bee pollination.
As noted, we did have a beekeeper drop out of our study in 2021 because they did not want to disperse hives along the field perimeter as required for the Beeflow treatment. It was also reported that the Beeflow feed made honey bees more “aggressive” and prone to stinging, although no actual declines in bee health were observed. Introducing Beeflow technology to blueberry growers may be challenging and require coordinating with the beekeepers growers are currently contracting with. These beekeepers may be reluctant to adopt Beeflow technology without additional engagement, information, and/or incentives. Additional economic analyses are also recommended to deduce if the investment in Beeflow technology leads to improved profitability at the farm level.
Acknowledgements: This project was funded by the Washington Blueberry Commission. We are grateful for the support provided by Eric Thompson of Belleville Bees and the multiple grower cooperators we worked with. Technical and material support were provided by Angelita De La Luz and Matias Viel of Beeflow for which we are appreciative of. Lastly, we are grateful for the assistance of Kayla Brouwer, Emma Rogers, Brian Maupin, Adriana Barsan, Alexandre Dias Da Silva, and Spencer Fiser who helped with data collection and entry.
References:
Arenas, A., V.M. Fernández, and W.M. Farina. 2007. Floral odor learning within the hive affects honeybees’ foraging decisions. Naturwissenschaften 94(3):218-222.
Balbuena, M.S., A. Arenas, and W.M Farina. 2012. Floral scents learned inside the honeybee hive have a long-lasting effect on recruitment. Animal Behaviour 84(1):77-83.
Courcelles, D.M.M., L. Button, and E. Elle. 2013. Bee visit rates vary with floral morphology among highbush blueberry cultivars (Vaccinium corymbosum). J. Appl. Entomol. 137: 693-701.
DeVetter, L.W., S. Watkinson, R. Sagili, and T. Lawrence. 2016. Honey bee activity in northern highbush blueberry differs across growing regions in Washington State. HortScience. 51(10):1228-1232.
Eeraerts, M., E. Rogers, B. Gillespie, L. Best, O.M. Smith, and L.W. DeVetter. 2022. Landscape-level honey bee hive density, instead of field-level hive density, enhances honey bee visitation in blueberry. Landscape Ecology. 1-13. https://doi.org/10.1007/s10980-022-01562-1.
Farina, W.M., A. Arenas, P.C. Díaz, C.S. Martin, and M.J. Corriale. 2022. In-hive learning of specific mimic odours as a tool to enhance honey bee foraging and pollination activities in pear and apple crops. Scientific Reports 12: https://doi.org/10.1038/s41598-022-22985-5.
Negri, P., L. Ramirez, S. Quintana, N. Szawarski, M. Maggi, Y. Le Conte, and M. Eguaras. 2017. Dietary supplementation of honey bee larvae with arginine and abscisic acid enhances nitric oxide and granulocyte immune responses after trauma. Insects 8(3):85.
Sagili, R.R. and D.M. Burgett, D. M. 2011. Evaluating honey bee colonies for pollination: A guide for commercial growers and beekeepers. PNW 623.
Stine, L., 2019. Beeflow raises $3m seed round for pollination-as-a-service. AgFunder News. <https://agfundernews.com/Beeflow-raises-3m-for-pollination-as-a-service.html>.
Winston, M. L. 1987. The biology of the honey bee. Harvard University Press, Cambridge, MA.