Farmers' delivery of floral resources: to “bee” or not to “bee”
THE PROBLEM
Agriculture is the most widespread use of land on Earth and one of the greatest drivers of biodiversity loss (Green et al., 2019). In the contiguous United States, over half of the land is devoted to agricultural production (NASS, 2021). Globally, simplified and intensively managed agricultural landscapes have replaced diverse farming systems, reducing biodiversity and ecosystem services essential to agricultural production such as pollination, pest control, and soil health (Le Provost et al., 2020).
A PRODUCTIVE FRAMEWORK
Understanding how land use can optimize both biodiversity and food security is essential to halting the negative environmental impacts of increasingly intensive use of agricultural land. To address this challenge, we propose a research approach that integrates landscape-scale remotely sensed data with trait information of wild and crop plants. This integrative framework generates multiscalar (field to landscape) depictions of agroecological outcomes (yield and biodiversity) associated with a particular land-use context. We apply this approach (Box 1 and Figure 1) to pollinators, providers of vital ecosystem services to both crops and wild plants (Kremen et al., 2007), by linking land use to both pollinator health and crop yield.
Box 1. Protocol to generate a model to identify land-use scenarios that minimize biodiversity loss while supporting agricultural production
1. Define study area: Using open-source rasterized land use data, such as the USDA CropScape data set, identify the type of land-use in the region of interest (e.g., specific crops, semi-natural and natural habitat).
2(A). Quantify floral resources in the field: Collect field data from randomly sampled transects across the agricultural landscape by measuring traits on a plant such as nectar volume and sucrose concentration on a sample of flowers, quantify the number of flowers on a plant across the season to provide a temporal component, then determine biomass and density of plants.
2(B). Quantify indicators of environmental health: Sample response variables (abundance and diversity of pollinators) of each land-use patch using randomized standard sampling methods. These samples are taken simultaneously with sampling of floral resources (2A).
3. Geospatially anchor floral resource information to generate the floral resource landscape: Number of flowers are typically highly correlated with the size of the plant. Plant biomass combined with plant density can then be used to predict caloric resources on higher scales (e.g., plots in a field, fields, and landscapes within a given radius of the sample). Thus, for each land-use category, generate a distribution of plant density based on observed field observations. From plant density, project anticipated floral resources on each pixel based on relationships derived from field observations. The floral resources landscape is predicted through regression analyses: temporal and spatial distribution of floral resources = measurements of individual plants across the flowering phenology (number of flowers) and floral resources (nectar, pollen) weighted by crop or wild plant densities. Floral resource information (nectar volume and sugar concentration) for plant and crop species is also accessible from large-scale open-source data sets (e.g., Kattge et al., 2020).
4. Link agroecological outcomes to the floral resource landscape: Link response variables (crop yield, wild-plant seed production, pollinators) to the floral resource landscape. Estimate wild and crop yield based on fertile seed count from a random sample of focal individuals within each land-use type or this information can be accessed by yield estimations from county-wide U.S. Department of Agriculture yield data (if available for the region) or by remotely sensed vegetative proxies (Brogi et al., 2020). Thus, for every point in the landscape, one can predict (using regression approaches) the relationship between floral resources and agroecological outcomes based on survey information of floral resources provided by different plants and their densities and the agroecological parameters of interests.
5. Simulate optimized landscapes: How might land-use reconfiguration shift outcomes based on the estimated parameters? Link spatiotemporal composition and distribution of floral resources in an agricultural landscape with landscape indicators of biodiversity (Koh et al., 2016) and county-scale yield data. Thus, policy makers can identify scenarios of land-use composition and configuration at the landscape scale that minimize biodiversity loss while supporting agricultural and wildflower production (1B).
RESOURCES FOR POLLINATORS, AN EXEMPLAR APPROACH
Agriculture reconfigures the spatiotemporal distribution of wild and crop plants, that affects important ecological drivers of agricultural production: pollinator services and pollinator health (Hellerstein et al., 2017). Quantifying the distribution and composition of floral resources in agricultural landscapes allows us to understand the interactions between agricultural land use, pollinator health, and agricultural outcomes. Depending on the distribution and availability of floral resources across the landscape, pollinator diversity and abundance is promoted or depressed (Schulte et al., 2017). Thus, we focus on the resources (or interaction currencies that mediate pollinator–plant interactions, Kissling et al., 2012) directly related to the energetic requirements of pollinators (Heinrich and Raven, 1972), that is, the density of sugar resources or kilocalories per hectare. By classifying the landscape based on resources available to pollinators, we can reduce complexities such as plant species composition, competition, or facilitation among pollinators and or plants into a single parameter that has great explanatory power for community-level processes (Nottebrock et al., 2017). Reducing the landscape to a map of available resources may more easily contribute to land-use decisions that both improve pollinator nutrition and enhance pollination services to agricultural systems.
As depicted in Figure 1 and Box 1, we generate floral resource maps by linking field measurements of floral resources such as nectar and pollen with land-use categories derived yearly for the continental United States from remotely sensed data that is compiled into matrices (rasters) of pixels where each pixel value represents a land-use category (e.g., USDA CropScape). Land-use categories correspond to land planted with different crops (e.g., corn, soybean, fallow) or non-crops (e.g., herbaceous wetlands, grasslands) and are georeferenced satellite imagery with moderate resolution (30–50 m) and extensive agricultural ground truthing. Floral resource maps should reflect the composition of crops, field margins, and non-crop areas. CropScape does not provide species composition for non-crop areas that consist of native and non-native plants. Thus, the density of floral resources that specific non-crop plants provide need to be quantified within each locale.
The quantity of floral resources provided by each land-use category for pollinators can then be assigned simplified energetic resource values (kcal/ha). Each land-use type at a geospatial location is also linked with field measurements of agroecological outcomes, e.g., crop yields and pollinator diversity and abundance. By linking the different categories of land use to the amount of floral rewards for each land use collected from field surveys, we can estimate the spatiotemporal distribution of floral resources on the landscape. We can then link this distribution of land-use types with key agroecological outcomes (i.e., crop yield and pollinator health) across landscapes covered by different combinations of crop, natural and semi natural environments. Because we are integrating across individuals to generate predictions of pollinator resources (ecosystem function) at a given scale, we call this individual trait surface a floral resource landscape.
Landscape-level analyses matter not only for ecological outcomes but also for the formulation of best farming practices. For example, Stiles et al. (2021) demonstrated that increased landscape diversity (measured as Shannon's diversity index with land use quantified in CropScape) in the South Dakota Prairie Coteau (a roughly 50,000-km2 plateau that rises from the flat prairies of western Minnesota and eastern South Dakota) resulted in higher pollinator diversity and higher yield in experimental plots of the new oil crop, Brassica carinata. Thus, landscape-level management that increases pollinator diversity could increase farmers' net profitability by replacing costs associated with inputs such as tillage and neonicotinoid seed treatments. Similarly, analyses at the national level find that more diverse agricultural landscapes are associated with significantly higher yields of the major crops (Nelson and Burchfield, 2021).
While there are an increasing number of studies demonstrating that both ecological outcomes and the economic interests of farmers can be simultaneously optimized (e.g., Dainese et al., 2019), the resource landscape approach may enhance our mechanistic understanding of how shifts in land-use composition and configuration iteratively alter key coupled agricultural and ecological outcomes. The floral resource landscape approach can generate the necessary information about where and how we need to improve the availability and distribution of floral resources in the agricultural landscape to improve pollinator services and environmental health. Since pollinator health is correlated to landscape attributes that generally enhance biodiversity (Kremen et al., 2007), our landscape resource approach may reveal the relationship between landscape use and other biodiversity driven ecosystem services.
We acknowledge that diverse resources are also important to sustain the biodiversity of pollinators, in addition to our proposal of the simple measure of kilocalories per hectare. Different pollinators do not necessarily forage on the same plants, and different plant species produce pollen and nectar of different nutritive values that can differentially affect pollinator health. We suggest that plant diversity, including crop and non-crop species, may be used as a proxy for resource diversity. Since pollinator services to crops depend on pollinator access to natural habitat for foraging and nesting (Kremen et al., 2007), models to assess pollinator services to crops should also assess aspects of landscape diversity and plant diversity, including the contribution of natural areas. However, the novelty and advantage of quantifying floral resources is that this approach reduces a complex landscape to a relatively easily measurable parameter that can be compared across spatial and temporal scales and plays an intuitively important role in pollinator health. For example, in our studies of pollinators communities found in remnant grasslands of the Prairie Coteau, landscape composition surrounding the sampling sites did predict pollinator diversity, but the sugar resources at the site were a better predictor of pollinator diversity and plant–pollinator interactions (Vilella-Arnizaut et al., 2021 [Preprint]). Since landscape use and their policy drivers are so important, we suggest that the floral resource landscape is, at the very least, a valuable informative parameter in addition to other measures of landscape diversity that have been previously studied.
INCENTIVIZING LANDSCAPE TRANSFORMATION
Because humans are directly responsible for the composition of agricultural and surrounding semi-natural landscapes (i.e., humans as keystone species; Moll et al., 2021), increasing the sustainability of these landscapes requires further understanding of the underlying political and economic factors that shape land-use decisions. The research approach described here can be used by policy makers to optimize biodiversity and food security. For example, in the United States, land use is strongly influenced by U.S. agricultural policy, particularly the U.S. Farm Bill, which is the major policy mechanism driving how and where food is produced and how land is used in the United States (Spangler et al., 2020). Federally subsidized crop insurance reduces the risk of bringing new, predominantly marginal land into production (Lark et al., 2015), which has contributed to declines in agrobiodiversity (Aguilar et al., 2015) and ecological health (West and Six, 2007), causing concern for the sustainability and viability of crop production.
Sustainable landscapes can then become additional targets for policy changes to enhance regenerative farming practices and ecosystem services that include maintaining pollinator diversity and abundances. The individual-trait-based landscape approach that we advocate can be used to optimize the construction of sustainable landscapes by quantifying how land use affects the configuration and complexity of floral resources. Changes as simple as increasing the number of crops eligible for crop insurance, including native grasses for biofuels in the crop insurance program, or increasing funding per acre of land transferred to the Conservation Reserve Program would make for a more sustainable agricultural landscape. Furthermore, the social factors, such as adherence to different measures of the U.S. Farm Bill at a county or finer scale, can be overlaid on the previous geospatial data to directly relate social behaviors with land use and ecosystem function and services. To summarize, the novelty and advantage of quantifying floral resources is that they can be integrated with other landscape descriptors, including landscape composition, diversity, and configuration, and related to agricultural policy driving human choices to provide a roadmap to environmental sustainability.
ACKNOWLEDGMENTS
We acknowledge the helpful comments from two anonymous reviewers. This research was funded by the North Central Sun Grant Initiative (USDA/DOE) SA1500640 to C.F. and in part by U.S. Department of Agriculture (USDA) National Institute of Food and Agriculture Grant No. 2020-67019-31157 to E.B. Any opinions, findings, conclusions, or recommendations expressed in this publication are those of the authors and do not necessarily reflect the view of the U.S. Department of Agriculture.
AUTHOR CONTRIBUTIONS
H.N. led the conceptual development and writing of the manuscript. E.B. and C.F. contributed to the conceptual development and writing of the manuscript.
