Final report for GS16-156
A diverse group of bee species, including both managed honeybees and wild native bees, are necessary to effectively pollinate commercially grown strawberries. Cultivated strawberries (Fragaria × ananassa Duch.) require pollination to produce marketable fruit, and strawberry seed set, fruit size, and fruit shape can be used to quantify pollination services. However, differences in pesticide application intensity may directly affect the pollination services provided by wild native bees.
Conventional strawberry growers routinely make over 20 synthetic pesticide applications annually to control pathogens, insects, and weeds during the season, most of which are preventative fungicide applications. Organic growers generally make fewer pesticides applications, typically less than 10 per growing season, and organically acceptable active ingredients may be less persistent in the enviroment. In order to measure the effects of pesticide exposure on native bees across a gradient of intensity, we will evaluate conventional and organic farms of both high and low pesticide intensity.
Higher pesticide intensity has been shown to reduce species abundance and richness, thereby reducing pollination services. Such reductions may reduce yield or increase the proportion of malformed berries, and may correlate with the gradient of pesticide intensity. Pesticides have been shown to reduce the pollination efficiency of bumble bees, but their effects on solitary bees are unclear. Further, greater pesticide intensity may reduce native bee immunity, increasing the likelihood of infection. We expected that farms with fewer pesticide applications, and organic farms which make fewer applications of less persistent insecticides, will support larger populations of native bees, which will in turn have greater immunity and result in measurably higher pollination services. This information will be used to both inform grower pest management recommendations to reduce impacts on pollinators and shared with policy makers to inform regulations aimed at improving pollinator health.
We identified a total of 22 strawberry farms in North and South Carolina, 7 of which used organic production practices. Sites were visited 3 to 4 times during the 2017 and 2018 growing seasons At each visit, three 50-meter long transects were established. Pan traps were placed along each transect and bees were sampled with a sweep net along the rows adjacent to transects. To relate pollinator abundance and diversity to fruit quality, up to thirty fully ripe berries were sampled at each visit and weight, seed set, symmetry were recorded. Pesticide records and honey bee stocking records were obtained from each site. Landscape composition was quantified using the National Land Cover Database and ArcGIS.
We defined bee health through estimates of census (bee counts) and effective population size (via molecular population genetics), immuno-competence, and pathogen intensity and are measuring these variables using molecular techniques described in detail in our Methods & Results.
Conclusions to date
We have completed field data collection. Molecular analyses will be completed in Fall/Winter 2018/2019 and statistical analyses will be completed in Spring 2019. Site characteristics varied widely between the selected farms. Pesticide applications over the two years ranged from 0 to 24, and six locations made no pesticide applications. Most foliar pesticide applications in conventional fields were fungicides, and fungicide, insecticide/miticide, and herbicide usage ranged from 0-15, 0-9, and 0-2 applications, respectively.
During 2017, Honey bee stocking rates at conventional farms ranged from 0 to 8 hives, and 6 of the 13 farms had no managed honey bees. Organic farms had between 0 to 5 hives, and 3 of the 7 had none. During 2018, honey bee stocking rates ranged from 0-4 hives per acre, including 9 sites which did not stock honey bees.
Landscape surrounding farms varied across our initial four habitat categories, but the dominant surrounding habitat varied by the size of the area assessed, either 500 or 1500 m surrounding the farm center.
|500 m radius||Mostly Agricultural||Mostly Forested||Mostly Natural, Non Forested||Mostly Urban|
|1500 m radius||Mostly Agricultural||Mostly Forested||Mostly Natural, Non Forested||Mostly Urban|
In general, bee densities were low in southeastern strawberry fields. Nearly all bee samples were collected via targeted sweep netting during transect walks, and very few bees were collected in pan traps. Pan traps were more effective at capturing flies, which were more common than bees on strawberry flowers during the first half the fruiting period. In 2017, we collected a total of 692 honey bees (Apis mellifera), 33 punative Augochlorella spp. bees, 106 punative Andrena spp. bees, 420 punative Lassioglosssum spp. bees, and 39 bees belong to other genera via sweep net. Honey bees, Augochlorella, Andrena, and Lasioglossum bees were collected at 20, 11, 14, and 18 locations, respectively. Honey bees were also collected directly from 34 hives at 11 locations just after the bloom period and 28 hives at 8 locations in September 2017. In 2018, we collected 442 honey bees (Apis mellifera), 26 putative Augochlorella spp. bees, 107 putative Andrena spp. bees, 439 putative Lasioglossum spp. bees, and 27 bees belonging to other genera via sweep net. Honey bees, Augochlorella, Andrena, and Lasioglossum bees were collected at 18, 6, 15, and 17 locations, respectively. Honey bees were also collected directly from 22 hives at 8 locations just after the bloom period and will be sampled directly from the hive again in September 2018. Additionally, syrphid flies from all 18 locations were collected and stored on dry ice. Recent work internationally has determined that bee viruses can be detected in some species of syrphid flies, so we will use a subsample of our specimens to assess if bee viruses are present in common syrphid flies in our region. A total of 1380 ripe berry samples were collected from multiple visits at each location and weight, symmetry, and seed set have already been recorded for all berries.
While the overall goal of our project remained the same, our objectives shifted in response to the greater diversity and lower abundance of native bee species observed after the first season. Further, we developed alternative molecular approaches that are more suitable to our field collected specimens and are still assessing pollinator health in the context of farm management. Specifically, our objectives are:
Objective 1. Assess the impact of a gradient of pesticide use applications on wild and managed bee health.
Hypothesis 1. Sites with higher pesticide application frequency or more persistent active ingredients (conventional vs organic) will have smaller and less diverse native bee populations.
Objective 2. Evaluate the impact of honey bee stocking densities on managed and wild bee pollinator health.
Hypothesis 2. Sites with higher densities of honeybees will have smaller and less diverse native bee populations. These sites will also have greater prevalence of bee pathogens, Deformed wing virus and/or Black Queen Cell Virus.
Objective 3. Estimate the impact of landscape context on the overall health of bee populations.
Hypothesis 3. Sites with less natural land coverage or greater urbanization will host smaller and less diverse native bee populations.
Objective 4 (new): Identify potential strawberry pollinating species of flies and determine if these pollinators are impacted by above factors.
Hypothesis 4: Flies contribute a meaningful amount of pollination early in the strawberry fruiting period.
Data collection methods. We visited 20 farms in 2017 (4 from SC and 16 from NC; 13 conventional and 7 organic) at least three times during strawberry bloom, late-March through early June. In 2018, we included 16 sites from our 2017 sampling season and identified 2 new NC sites for a total of 18 locations (4 from SC and 14 from NC; 12 conventional and 6 organic) which were visited at least three times during the bloom period.
During each visit, we conducted timed transect walks wherein we collected and identified, to the extent feasible, all bees visiting strawberry flowers using sweep nets. We then conducted targeted sweep net sample collection of native bees and honey bees for use in additional molecular analyses. Bees collected via sweep net were immediately stored in dry ice to preserve RNA. During sweep net collection in 2018 we also sampled syrphid flies from each site, stored them on dry ice immediately, and will assess them for the presence of honey bee pathogens. We also placed florescent yellow and blue pan traps within 3 different rows (second row, 1/3 field distance, ½ field distance) of the strawberry fields to serve as an additional measure of pollinator diversity. Insects collected in pan traps were sieved and stored in ethanol for subsequent identification. Interestingly, many of the insects visiting strawberry flowers early during bloom (March and April) were not bees, but were instead flies. All flies collected in the pan traps have been counted and identified to family. Lastly, to get a metric of pollination efficiency, we collected 30 ripe berry samples at each site visit in order to measure seed set and fruit size. Comparing these berry metrics to our predictor variables and bee population sizes will help elucidate if particular farming practices may be influencing berry quality or yield.
Objective 1. Assess the impact of a gradient of pesticide use applications on wild and managed bee pollinator health.
Pesticide use (Predictor) – To assess the effect of pesticide use on pollinator health, we are using bee samples collected from 20 farms in 2017 and 18 farms in 2018 that vary across a continuum in their pesticide application regime. We retrieved pesticide records from each farm and are using two variables to quantitatively assess pesticide use: (1) Farming technique: there is large variation in the persistence of pesticide active ingredients of conventional vs organic farms; and (2) Number of pesticide applications: greatly varies depending on whether growers use scheduled applications or based their application regime on forecasting and scouting. Data analysis – We are determining the effect of pesticide use on pollinator health using Generalized Linear Mixed Models (GLMMs) incorporating farming technique as a categorical variable, number of pesticide applications as a continuous variable, and site as a random effect. Mixed effects models will allows us to determine the relative effect of different aspects of pesticide application practices on bee population size and stability, their levels of immunocompetence, and the intensity of common diseases that threaten pollinators. Outcomes – We will generate foundational data on how real-world levels of pesticide applications affect different aspects of the health of both managed (A. mellifera) and wild bee populations. These results are crucial for proper regulatory actions for pesticide use applications that achieve optimal pest control and pollination services.
Objective 2. Evaluate the impact of honey bee stocking densities on managed and wild bee pollinator health. Honey bee stocking densities (Predictor) – We collected the following information regarding honey bee colony density at each farm: (1) Number of colonies, (2) Date of colony establishment, (3) Date of colony removal; (4) Average distance between closest-neighbor colony; and (5) Size of the farm (acreage). Data analysis – We calculated honey bee stocking densities from the number of rented colonies corrected by size farm. We are in the process of checking for multi-collinearity among the honey bee density variables (density, establishment and removal date, distance between colonies) using the “vifstep” function in the R statistical package usdm (Naimi, 2013). GLMM models will be used to determine the effect of these predictor variables on (1) wild bee population sizes, and (2) pathogen intensity on both wild and honey bees. Pathogen data will be analyzed using a RDA analysis (Youngsteadt et al., 2015). We will choose the best-fit model using model comparison and AIC scores. We will correlate honey bee stocking densities and overall pathogen intensity to see if farms with closer and higher number of honey bee colonies enhance levels of disease in wild and managed pollinators. Outcomes – A common practice associated with renting honey bees for pollination services is to have high number of colonies per unit area to maximize pollination services (Goulson et al., 2015). This study will pioneer the assessment of the effects of honey bee stocking density on (1) pathogen abundance of both honey bees and wild bees, and (2) population sizes of wild pollinators. A negative effect of higher honey bee stocking densities on native bee population sizes could indicate competition between managed and wild pollinator species.
Objective 3. Estimate the impact of landscape context on the overall health of bee populations. Landscape context (Predictor) – The 22 separate study sites range in the proportion of agricultural, natural, and urban cover within a 2 km radius and are at least 5 km apart from each other (Figure 5). We have analyzed landscape surrounding each site at local (500 m radius) and regional (1500 m radius) scales using the USGS National Land Cover Database (30m resolution, NLCD 2014) in two ways. First, we grouped the land use types surrounding each site into four broad categories (Agricultural, Natural, Forest, and Human Development) to calculate the percentage of each type of habitat at local and regional scales. Second, we estimated landscape complexity surrounding each of our sites by calculating the total area comprising the different land use types used in the Land Cover Database. For all spatial analyses, we use the software ArcGIS v.10.2 (ESRI, 2013). Data analysis – To determine the relationship between landscape contexts and pollinator health, we will first screen the landscape context variables (Agricultural, Natural, Forest, Human Development, and Heterogeneity) for multi-collinearity using the “vifstep” function in the R statistical package usdm (Naimi, 2013). Because local and regional landscape variables are highly correlated, we will investigate them in separate GLMMs. Site will be included in the analysis as a random effect. Top models will be selected based on AICc using the “dredge” function in the R statistical package MuMIn (Barton, 2015). GLMMs will be performed using the ”glmer” function in the package lme4 (Bates et al., 2014). Outcomes – This project will elucidate for the first time how landscape context affects pollinator health. These results will also allow us to make landscape management recommendations for how to best integrate pollinators into more sustainable IPM tactics, as it will elucidate possible interactions between our three predictor variables. By quantifying the effects of these predictors on honey bees and native bee species, our goal is to provide broad and generalized recommendations that aim to provide improved environmental conditions for both managed and wild pollinators.
Objective 4. Identify potential strawberry pollinating species of flies and determine if these pollinators are impacted by above factors. We have identified flies collected during both years of the project in order to determine the diversity of species visiting strawberries and the abundance of each species. During 2018, we also preserved flies following the same methods as bees in order to assess virus load as an initial measure of pollinating fly health.
In 2017, we collected a total of 692 honey bees (Apis mellifera), 33 punative Augochlorella spp. bees, 106 punative Andrena spp. bees, 420 punative Lassioglosssum spp. bees, and 39 bees belong to other genera via sweep net. Honey bees, Augochlorella, Andrena, and Lasioglossum bees were collected at 20, 11, 14, and 18 locations, respectively. Honey bees were also collected directly from 34 hives at 11 locations just after the bloom period and 28 hives at 8 locations in September 2017.
In 2017, we collected between 0-14 native bees and 0-8.3 syrphid flies per visit in pan traps among all sites. Pesticide records were obtained for 19 of 20 locations and application number ranged from 0-24 applications at a given location, including 6 sites with no pesticides applied. Specifically, fungicide, insecticide/miticide, and herbicide usage ranged from 0-17, 0-10, and 0-2 applications, respectively. Honey bee stocking rates ranged from 0-4 hives per acre, including 9 sites which did not stock honey bees. Landscape composition was determined at 500m and 1500m for all locations. For both 2017 and 1028, and at 1500m among all sites, natural land coverage varied between 3-75% of the landscape while urban land coverage was between 2-46%. For both 2017 and 1028, and at 500m among all sites, natural land coverage varied between 1-78% while urban land coverage was between <1%-46%. A total of 1710 ripe berry samples were collected from multiple visits at each location and weight, symmetry, and seed set have already been recorded for all berries.
In 2018, we collected 442 honey bees (Apis mellifera), 26 putative Augochlorella spp. bees, 107 putative Andrena spp. bees, 439 putative Lasioglossum spp. bees, and 27 bees belonging to other genera via sweep net. Honey bees, Augochlorella, Andrena, and Lasioglossum bees were collected at 18, 6, 15, and 17 locations, respectively. Honey bees were also collected directly from 22 hives at 8 locations just after the bloom period and will be sampled directly from the hive again in September 2018. Additionally, syrphid flies from all 18 locations were collected and stored on dry ice. Recent work internationally has determined that bee viruses can be detected in some species of syrphid flies, so we will use a subsample of our specimens to assess if bee viruses are present in common syrphid flies in our region.
In 2018, we collected between 0-29 native bees and 0-17.3 syrphid flies per visit in pan traps among all sites. Pesticide records were obtained for 17 of 18 locations and application number ranged from 0-21 applications at a given location, including 6 sites with no pesticides applied. Specifically, fungicide, insecticide/miticide, and herbicide usage ranged from 0-15, 0-9, and 0-2 applications, respectively. Honey bee stocking rates ranged from 0-4 hives per acre, including 9 sites which did not stock honey bees. Landscape composition was determined at 500m and 1500m for the 2 new locations in 2018. A total of 1380 ripe berry samples were collected from multiple visits at each location and weight, symmetry, and seed set have already been recorded for all berries.
DNA/RNA extraction protocols were validated on both small native bees and honey bees. RNA/DNA extraction has begun for Lasioglossum and honey bee samples in order to measure titers of Deformed Wing Virus and Black Queen Cell Virus against a standard curve. We will additionally assess the relative gene expression of honey bee immune genes via qPCR. Standard curve development and verification of qPCR protocols are nearly completed, so Lasioglossum and honey bee samples from 2017 are being assessed currently. Synthetic gene block fragments containing the two primer-amplified viral fragments together are being used in serial dilution to generate a standard curve for viral titers. The relative gene expression of antiviral immune genes, Argonaut-2 and Vago, will be assessed for honey bee samples and compared among sites with variable pesticide usage intensity or landscape complexity.
Species identification for Lasioglossum samples collected in 2017 is nearly complete after modification of forward primers to correctly amplify native bee rather than bacterial endosymbiont genes. Lasioglossum bees collected in 2018 from sites with greater than 20 specimens will be identified to species as well. Fly and bee specimens collected in pan traps from both years have been processed and have been identified to family.
Educational & Outreach Activities
The target audiences for this project are, in the near term, strawberry growers, extension personnel, beekeepers, and other agricultural professionals. Efforts to reach these audiences during the past two project years include:
- Establishment of on farm research sites. We have established a total of 22 on farm research sites, 18 in North Carolina and 4 in South Carolina. Sites were visited at least three times per year from April through June or until harvest ceased. Growers were directly engaged in project data collection and rationale.
- Development of online tools. We have developed two background online publications on strawberry pollination and will expand online information to a new pollinator-focused extension website, planned at pollinators.ces.ncsu.edu, during winter 2018. https://entomology.ces.ncsu.edu/small-fruit-insect-biology-management/strawberry-pollination-basics/ & https://entomology.ces.ncsu.edu/strawberry-pollinating-insects/
- Grower extension meetings. We have presented an overview of our project and expected results at four strawberry extension meetings to a total of 169 attendees in three states. Nash County Strawberry Preplant Meeting (8/14/18 for 50 attendees), Strawberry preplant insect update. Georgia Strawberry School (8/9/17, presented remotely for 14 attendees), Designing biology-based management programs for southeastern strawberries. Virginia Strawberry School in Virginia Beach (2/28/17, for 85 attendees), Designing biology-based management programs for Mid-Atlantic strawberries. Nash County Strawberry Preplant Meeting (8/16/16, presented remotely for 20 attendees), Strawberry preplant insect update. Results were also presented to researchers and farmers at the Entomological Society of America (ESA): South-eastern branch meeting (3/6/2018, for about 20 attendees), the SARE: Our Farms Our Future conference in St. Louis (4/4/2018, with a poster), and the upcoming international ESA conference in Vancouver Canada, in November 2018.
The mid and long term target audiences for this project are growers of pollination dependent crops, regulatory agencies, and the public. Our public facing online resources are available to this audience currently. Future efforts will further reach out to these audiences.
This project will lead to a better understanding of how current pesticide use practices, landscape management, and honey bee stocking rates may be affecting berry quality and yield as well as the diversity, abundance, and pathogen prevalence of early spring native bee communities in our region. Further, this work will increase our understanding of pathogen titers in native bees relative to honey bees and whether these detected pathogen loads may be influencing native bee community composition or health. Our hope is that this work will provide evidence for developing pollinator recommendations in strawberries. Specifically, to reduce reliance on managed honey bees to reach yield goals by enhancing populations of native bees and other beneficial insects around the farm. Future work should continue to assess native bee populations throughout the year and how pathogen dynamics and farming practices are shaping community assemblage and size. Already our cooperating growers are now better informed about the value of native pollinators and in the future may be more likely to reduce pesticide application frequency or include pollinator friendly habitat in landscapes where food resources may be scarce.
On farm cooperators were informed throughout the study of relevant insect scouting observations and were updated on summaries of results. Growers were shown different beneficial insects, including native bee species and syrphid flies, which are responsible for much of the necessary pollination of strawberries in our region. Although our growers are unlikely to sample pollinators in future seasons, they are now aware of the diversity of native species present on their farms and various recommended techniques to improve overall pollinator health. An interesting and unintended outcome was also education concerning the value and benefits of pollinating fly species, which are present in larger densities and earlier than most early season native bees.