To develop clear guidelines that farmers can use to design optimal pollinator habitat, we will select 33 sites in the greater Ithaca region, New York* across forest, agricultural, wetland, successional, and developed habitat typical of Northeast US agricultural landscapes. Within each of these sites, we will complete Objectives 1-3.
- Assess plant species abundance and richness (to be completed with existing data from collaborators)
- Document bee species abundance and richness.
- Measure soil moisture, fertility, and organic matter content.
Using the data from these objectives, we will complete the following analytical objectives:
- Determine relationships among local abiotic factors, including soil moisture, fertility, and proximity to water, and plant and wild bee abundance and richness. Expected Outcome: Plant and wild bee communities will be more diverse at sites close to water features, and sites with more spatially variable soil moisture, and lower soil fertility (as these sites are less likely to be dominated by a few very competitive plant species).
- Determine the relative contribution and interactions among local abiotic factors, plant diversity, landscape heterogeneity, and connectivity in explaining wild bee diversity and abundance. Expected Outcome: Local abiotic factors will interact with landscape heterogeneity and connectivity, so that there is a minimum amount of habitat necessary to support wild bees, but plant diversity and abiotic factors will significantly influence bee community diversity above this threshold.
*We originally proposed field data collection in Lancaster County, PA in arable, pasture, grassland, and forest habitats where Egan and Mortensen (2012) quantified plant diversity. In the process of preparing for the start of the funded project, Dr. Grozinger and I have developed a new collaboration with some researchers at Cornell University (see ‘Outcomes’ section below). Dr. Aaron Iverson has been leading an extensive plant community survey across the Finger Lakes region surrounding Ithaca, NY. We would like to measure bee diversity at a subset of the sites where Dr. Iverson and colleagues quantified the plant community. All of the project objectives and methods for bee and soil sampling are the same as the original proposal. Please see the ‘Materials and Methods’ below for a more detailed justification for the change in study location.
My project focuses on integrating multiple abiotic and biotic factors to predict spatial patterns of wild bee biodiversity in agricultural landscapes. The purpose of this research is to 1) address fundamental knowledge gaps about drivers of wild bee diversity, specifically how the quantity and diversity of floral resources, proximity to water, and soil conditions interact to dictate wild bee diversity at local and landscape scales, and 2) formulate practical guidelines for farmers in implementing pollinator conservation plantings.
Biodiversity conservation is a fundamental tenet of sustainable and organic agriculture, both for its intrinsic value and the ecosystem services provided by a diverse flora and fauna. More than 90% of flowering plants and nearly 75% of major crops benefit from animal pollination. For many important pollinator-dependent crops in the Northeast, such as apple, pumpkin, watermelon, and blueberry, wild bees are more efficient pollinators than managed honey bees and are responsible for a majority of pollination services for these crops. Globally, higher wild bee diversity translates to higher fruit set of many pollinator-dependent crops, regardless of honey bee abundance.
In agricultural landscapes, adjacent natural and semi-natural habitats host a majority of the available plant species and provide crucial food resources (nectar and pollen), water, and nesting sites, as well as shelter from pesticides for pollinators. Loss of semi-natural habitat reduces floral and nesting resources available for bees, and loss of high quality nutritional resources decreases bees’ resilience to other stressors like pathogens or pesticides. The potentially synergistic effects of habitat loss/degradation and pesticide use means that agricultural landscapes can be high-risk environments for bees.
In response to these challenges, many farmers, landowners, and conservation practitioners are increasingly interested in how farmland can be designed and managed for wild bees. Indeed, farmers can receive government conservation payments for managing surrounding wildlife habitat and planting field borders, hedgerows, shelterbelts or wildflower meadows for pollinators. While there is some evidence that the recommended plant mixes can attract diverse bee communities, supplementing with specific plant species or mixes does not reliably improve pollinator outcomes.
Landscape context appears to be critical for determining if these plantings are successful: for example, the amount and arrangement of habitat in the landscape (heterogeneity and connectivity) is crucial for pollinator movement and access to resources, and is positively correlated with higher bee abundance and diversity worldwide. Studies in Europe found that ignoring the context surrounding pollinator plantings can decrease the benefits of these plantings so that there is no detectable difference between conservation and control sites.
Environmental context for pollinator conservation practices has typically been defined in terms of local plant diversity, landscape heterogeneity, and connectivity, but abiotic site conditions (such as soil quality) can also influence plant nutritional quality and pollinator visitation, and local water features provide important resources for bees. Thus, it is critical that conservation practices for pollinators take into account both biotic and abiotic context. However, accessible guidelines are very sparse for growers on how and where to implement pollinator practices in terms of abiotic and biotic factors at multiple scales. In my proposed research, I will identify the relationships among local abiotic factors, landscape context, and plant and pollinator communities to help create predictive models that can be used by land managers to optimize placement of new pollinator habitat.
Objective 1: Plant species richness, abundance, and floral area was measured by Iverson et al (in preparation) in the greater Ithaca region, New York in 2015 and 2016. Iverson et al. surveyed 144 sites across 20 habitat types (including forest edge, dry oak forest, conifer and mixed forest, row crop, hedgerow, shrub wetland, emergent wetland, roadside ditch, etc.) that span 5 macro-habitat classes (forest, agriculture, wetland, successional, and developed). The number of sites for each habitat type ranged from 3-20 based on the plant community variability, with a median of 5 sites/habitat type. Plant communities were sampled with a half scale Modified Whittaker plot sampling design. This sampling method quantifies plant abundance as percent ground cover in 10 0.5m2 quadrats and species richness in a 500m2 plots. Iverson et al. also measured floral area for all species flowering at the time of each site visit. With multiple observations of each habitat type at different times of the year, Iverson et al generated a floral area database for all major flowering plant species within each habitat. By combining the empirical flora area data with plant flowering time drawn from regional flora references, they calculated mean floral area for each site and habitat type over the growing season.
Objective 2: We measured bee species richness and abundance using 12oz. Solo polystyrene plastic cups as bee bowls according to the protocol of a long-term bee monitoring program in our region. We filled fluorescent blue, fluorescent yellow, and white bee bowls with 50:50 mix of propylene glycol and water and placed them at the height of dominant vegetation for 7-14 days of sampling. We arranged bee bowls in 100m transects in visible areas, alternating bowl color with 10 meters between each bowl, for a total of 9 bowls per site. We sampled wild bees at 33 field sites of 7 habitat types (forest, forest edge, floodplain forest, old field, roadside ditch, mixed vegetable farm, and apple orchard) that span a range of semi-natural to agricultural land use. From Iverson et al. plant sampling locations, we selected bee sampling sites based on landowner willingness to participate in our study and distance between sites. To ensure we were sampling independent bee communities at each site, We chose sampling locations that were at least 1km from all other sites, a distance greater than the mean foraging range of a typical mid-Atlantic wild bee community (Kammerer et al 2016). We sampled bees in late April and again in mid-July, guided by peak floral abundance in forest, wetland, and successional habitat. After collection, bee specimens were be stored in 70% ethyl alcohol solution until pinning and sorting. After the field season ended, we washed and pinned all specimens, and began identifying bee specimens to genus, or species, when possible, as in my previous publication , with taxonomic assistance from collaborators within the PSU Center for Pollinator Research, including Dr. David Biddinger, and Sam Droege at the USGS Bee Inventory and Monitoring Lab.
Figure 1: Left, a yellow bee bowl sampling bees at an organic vegetable farm. Center, wild bees captured in bee bowl surrounded with spring ephemeral flowers at a floodplain forest site. Right, spring beauty (a spring ephemeral flower visited by a specialist wild bee, Andrena erigeniae) flowering at a forested patch within a crop farm.
Objective 3: In May 2018, we collected soils at each of the bee sampling sites. Along the bee sampling transect, we collected five soil samples with a bucket auger to a depth of 9-18 cm, depending on rock and moisture content of subsoil. Shallower soil cores (9-12cm) were taken at sites with very rocky or wet (floodplain habitat) subsoil due to sampling constraints. Also, wild bee nesting would likely be inhibited by very high rock content or completely saturated subsoils, so we considered the shallower sampling depth representative of the most favorable zone for soil nesting wild bees. At two locations along the bee transect, we collected three undisturbed soil cores (0-3 cm, 4-6 cm, and 7-9cm deep) with a slide hammer soil core sampler (Soilmoisture Equipment Corp, Goleta, CA) to quantify bulk density. Due to above described sampling constraints, we were only able to sample 2 bulk density soil cores deep at some locations, but the number of cores was recorded for each sample. Bulk density cores at all depths were combined for processing and analysis.
Figure 2: Kammerer Allen collecting a soil core to measure bulk density in an apple orchard near Geneva, NY.
After collection, we measured the wet mass of all soil samples, then dried them at 60 degrees C for five days, or until the mass did not decrease. We sent the bucket auger samples to the Penn State Agricultural Analytics lab and they measured pH, P, K, Mg, Ca, Zn, Cu, S, total nitrogen by combustion, percent organic matter, and percentage sand, silt, and clay via standard laboratory methods (https://agsci.psu.edu/aasl/soil-testing/soil-methods). We calculated soil bulk density of each sample as the total dry mass of bulk density cores divided by the volume of the sampling cylinder multiplied by the number of cores. To identify patterns across the 14 soil characteristics we measured, we used a principal components analysis. We conducted the analysis using the prcomp function in R on centered and scaled soil variables.
Objective 4: Following the Kammerer et al  analytical approach, I plan to use statistical models to quantify the relationship between plant richness and local soil moisture, fertility, organic matter content, proximity to water features, and landscape heterogeneity and connectivity at each site. I will calculate proximity to water features using ArcMap GIS software version 10.5 and a dataset of water features from the Tompkins County GIS office. Landscape heterogeneity will be represented with two metrics, the percent of perennial habitat land cover, and diversity of land cover types in the landscape. Landscape heterogeneity and connectivity metrics will be calculated from the 2016 USDA National Crop Data Layer using FRAGSTATS landscape analysis software and geospatial tools in the R statistical computing language. All landscape metrics will be computed at 3 scales (500m, 1000m, and 1500m) centered around each plant sampling site.
For statistical analyses, I will use either general linear mixed models, generalized linear mixed models, or generalized additive models depending on the distribution of richness data, and specify a habitat random effect to account for variance due to habitat specific variables that were not measurable. Statistical model fits will be compared using Akaike information criterion and variance explained values. All analyses will be conducted in the R statistical computing language.
Objective 5: Analyses for this objective will use the same predictor variables and statistical methods as Objective 4, except plant species richness, evenness, and floral area at each site will be included as a predictor of bee richness.
Study location: Conducting the study in the greater Ithaca region, NY will allows us to build on an incredibly detailed, rich dataset of plant diversity and floral area across the region to study how bee diversity responds to plant community characteristics, soil, and local abiotic factors. There are a couple advantages of moving the study to New York instead of Lancaster County, PA as we originally proposed.
- Iverson et al. measured flowering plants’ floral area, which is not available in the plant dataset we have for Lancaster County, PA.
- The NY plant survey work was conducted more recently (2015/2016 compared with 2008), which decreases the likelihood for management or land use changes in the period between the plant and bee surveys. We expect this will decrease the number of sites that are not suitable for the bee survey due to land use change since the plant survey, and increase the strength of the relationship between plant diversity and bee diversity in our dataset.
- There are many more sites and habitat types represented in the Iverson et al. plant survey (144 sites over 20 habitats in NY vs. 80 sites over 5 habitats in Lancaster, PA).
- The Finger Lakes region has a larger area of pollinator dependent crops (primarily apples near the Iverson et al sites), which increases the audience for our farm design website and other outreach activities.
In year one of the project, we captured 1677 wild bees and 241 Apis mellifera. Surprisingly, we collected more wild bees in the April sample than the July sample. In April, we left the traps out for 14 days instead of 7 because it was very cold during the first week of sampling and there were very few bees in the traps that we inspected after one week. It is possible that this increased sampling time led to higher number of specimens, but we also observed greater diversity of wild bees in April than July. After taxonomic identification is complete, we will verify our field observations by analyzing species richness at each site. We will also adjust the the April and July samples to a common sampling effort using species rarefaction curves (iNEXT package in R statistical computing language). In the field, we observed that, in April, the old fields and floodplain forests had the most bees, while the roadside ditches yielded the greatest number in July. As we expected, the floodplain forest, forest, and forest edge sites had many more bees in late April than July, as the latter sample occurred after the forest canopy closed and traps were shaded. Surprisingly, the old field sites had more bees in late-April than July, even though the diversity of flowering plants at these sites was much higher later in the season.
We found significant variation in soil characteristics between the site and habitat types we sampled. There were two main gradients in our soil dataset revealed by the principal components analysis (Figure 1). Explaining 27.5% of the variation in our soil data, the first principal component was correlated with soil texture (percent sand, silt, and clay). Roadside ditches had sandier soil than any of the other habitat types, except some floodplain forest samples. All of the other habitat types had loam to silt-loam soil. Interestingly, soil texture was highly variable between floodplain forests, with soil texture in this habitat encompassing the full range of texture classes represented at all other sites. The second principal component explained 19.5% of the variation in soil characteristics, and was associated with several soil fertility variables. Specifically, some vegetable farm and orchard samples had much higher potassium, phosphorus, copper, and zinc content than the other habitats, likely due to fertilizer or manure application to support crop growth. Forested sites were most highly correlated with higher total nitrogen in the soil, probably due to high organic matter content from leaf litter accumulation. In future analyses, we will analyze if and how these patterns in soil characteristics explain patterns in plant and bee abundance and species richness.
Figure 3: Principle components ordination plot of soil characteristics at 33 sites in the Finger Lakes region, NY. Sample colors correspond to the following habitat types: ‘apple’= apple orchard, ‘ditch’ = roadside ditch, ‘edge’ = forest edge, ‘field’ = old field, ‘flood’ = floodplain forest, ‘forest’ = mesic upland remnant forest, and ‘veg’ = mixed vegetable farm.
Education & Outreach Activities and Participation Summary
We are still analyzing preliminary results for this project, so have not yet included any specific results from this project in outreach presentations. However, we plan to target farmers and landowners, NRCS and Xerces conservation staff, and pollinator researchers through several outreach activities and products, including an interactive pollinator landscape assessment website, presentations at farmer and scientific society meetings, and peer reviewed publications.
One of the key outreach materials from this project will be an interactive ‘Landscapes for Bees’ website. We are partnering with Azavea, a geospatial technology company based in Philadelphia, PA to create an interactive web map that allows users to analyze the quality of their landscape for bees. During this first phase (beta version set to go live in Feb 2019), the site is primarily designed for beekeepers seeking to compare the quality of the landscapes surrounding their apiaries, but in the second phase, we plan to expand the target audience to include farmers, conservation professionals, and landowners. We will also expand the geographic coverage to include New York state. This interactive portal characterizes landscape quality for bees in terms of floral resources, nesting resources, and a novel pesticide index and makes these indices publicly available to beekeepers, farmers, researchers, or any other interested parties.
I have been part of the discussions with Azavea regarding the audience, function, and design of the ‘Landscapes for Bees’ interactive website, and have generated all of the landscape quality data provided to end users through the portal. In the next phase of the website development, I will be actively involved in developing the verbage targeting farmers and landowners interested in supporting wild bees. I also plan to include a brief description of the ‘Landscapes for Bees’ portal in my presentations at farmer and scientific society conferences to increase user engagement. Lastly, I will reach out to Kelly Gill, Xerces Society for Invertebrate Conservation Pollinator Conservation Specialist in the Mid-Atlantic and Northeast to review the website and facilitate its inclusion with Xerces’ library of pollinator conservation resources.
Figure 4: Screenshot of beta version of ‘Landscapes for Bees’ interactive portal, live version to be released in February 2019.
During year one of the project, I presented my work at the Penn State Center for Pollinator Research annual symposium and Pennsylvania Beekeepers Association Meeting in State College, PA. In year two, I plan to present at the US landscape ecology association meeting in Fort Collins, CO, International Pollinator Conference in Davis, CA and national Entomology meeting in St. Louis, MO hopefully planning a symposium on pollinator landscape models (pending symposium acceptance by conference organizers).
The U.S. National Strategy to Promote the Health of Honey Bees and Other Pollinators set a goal to restore or enhance 7 million acres by 2020, and federal agencies, farmers, and land owners need practical guidelines in deciding where and how to achieve this goal. The results of my research will be used to formulate best practices for farmers, conservation professionals, and researchers to design effective, site-specific conservation practices to support and restore diverse wild bee communities and pollination services in agricultural landscapes. For example, if my results support the expectation that sites close to water features have higher plant and wild bee diversity, choosing to preserve or enhance forest fragments surrounding streams would likely influence pollinator diversity more than conservation of similar forest patches without water. Effective guidelines for how to implement pollinator conservation will improve the return on investment for government conservation programs, supporting diverse wild bee communities essential for crop and wild plant pollination across the Northeast.
Through this project, Dr. Grozinger and I have developed a new collaboration with some researchers at Cornell University. Dr. Aaron Iverson, a post-doctoral fellow working with Dr. Scott McArt has been conducting an extensive plant community survey across the Finger Lakes region surrounding Ithaca, NY. We have started collaborating with Drs. Iverson and McArt and are surveying bee diversity at a subset of the sites where Dr. Iverson and colleagues quantified the plant community. This allows us to leverage an incredibly detailed, rich dataset of plant diversity and floral area across the greater Ithaca region to study how bee diversity responds to plant community characteristics as well as soil and local abiotic factors.