Background
The question of species survival and dispersal in the face of increasing habitat fragmentation is a key issue for the 21st century. How habitat fragmentation affects species survival, and how much habitat needs to be kept intact, or reintroduced, to ensure ecosystem stability, are questions difficult to answer with field work alone. These questions have thus been an active area of inquiry for mathematical modelers. While this work has led to general insights on the effect of habitat fragmentation on species survival, there is a paucity of research on the effect of landscape manipulations on crop yields and pollination deficits. Existing studies give us insight into the features that characterize bee-friendly landscapes, but do not make it possible to determine optimal enhancements for specific landscapes.
It is evident that landscape configuration is strongly related to bee survival and pollination services, but the relationship is complex. Field work shows that wildflower plantings and natural habitat do increase pollinator density within crops, but the response varies from one location to another along with crop type, presence of resources in the surrounding landscape, and composition of the native pollinator community. Two aspects related to pollination service in this kind of intensive agroecosystem remain relatively neglected: The influence of density and phenology of flowers other than the focal berry crop, and The reproductive success of crop-visiting insects. Bees choose flowers to visit based on flower profitability (nectar/pollen). Profitability is both density-dependent, and distance-sensitive.
Given the complexity of the relationship between pollination services and floral resources on the landscape, it is difficult for Fraser Valley blueberry growers to justify investing in expensive or time-consuming landscape manipulations, when is it unclear exactly what modifications would most benefit their particular farms. There is thus a need for modelling work, informed by parallel field work, to provide a quantitative means of comparing different wildflower planting arrangements in different agricultural landscapes. This project will combine current knowledge and a new field study of wild bee behaviour in the Fraser Valley with modelling expertise on insect movement and optimization, to develop an estimation tool for wild pollination services in agricultural landscapes
Objectives
Our primary objective is to combine current knowledge and a new field study of bumble bee foraging and reproductive success in the Fraser Valley with our expertise on mathematical descriptions of insect movement and model optimisation, to develop a novel modelling tool, with a grower-friendly interface, that will make it possible to determine optimal arrangements of wildflower patches in Fraser Valley blueberry farms.
1. Field Study
The field study component will gather data relating the spatial and temporal landscape context to the reproductive success of native bees surrounding cultivated blueberry crops. Using ground-installed domiciles for bumble bees, we will measure the reproductive success of nests. Multiple blueberry fields, representing growing regions across the Fraser Valley, will be used. Nests will be placed across a targeted gradient of distance from sizeable native habitat, and each location will be characterised by the density and phenology of alternative (non-blueberry) floral resources. The independent variables are thus distance from native habitat, distance from blueberry fields, and density and phenology of alternative floral resources. We will manipulate the latter by planting wildflowers in field margins or other unused spaces, and observe how these additional resources affect the reproductive success of bumble bee colonies and the pollination services provided to the blueberry crop. We expect the results to depend on the size of the wildflower patch, as well as the independent variables identified above. Pollen recovered from domiciles will provide a record of colony visitation of cultivated blueberry, integrated over the period of nest provisioning.
2. Insect Movement Models
The modelling study will include the creation of three separate models, which we describe in general terms.
The first model will be used to make preliminary predictions regarding the effect of wildflower patches. Instead of modelling space explicitly, the landscape will be represented simply by the proportions that are covered by floral and non-floral resources: blueberry flowers, wildflowers, native habitat, and other habitat. Since the model is spatially-implicit, it can be created quickly enough to provide input to the selection of study sites for the first field season. Using a stochastic renewal process, this model will be used to predict the probability of colony survival and the expected colony size depending on the coverage of wildflowers and crop bloom. These relationships can then be translated into overall pollination effectiveness of a colony. Blackbox optimisation will be used to identify optimal sites for the placement of wildflower patches.
While the spatially-implicit model can generate useful predictions, we expect that landscape configuration has a strong effect on pollination services. That is, if 25% of the landscape contains alternative floral resources, these could be scattered throughout the landscape or concentrated in one far flung corner (for example). These two landscape scenarios yield very different availability of floral resources to foraging bees. Consequently, the second and third models will be spatially explicit and more complex. The second model will be an individual-based model (IBM), while the third will be a partial differential equation (PDE) model. The IBM, a stochastic process model, has the advantage that individual bee behaviour can be described in greater detail, but the approach can rapidly become computationally burdensome as new agents (bees) are added. There will thus be an upper limit to the number of bumble bee nests whose pollination services can be investigated via the IBM. This model will be ideal for studying and predicting the pollination services of a few bumble bee nests on a landscape corresponding to the size of the nests’ foraging range and using a detailed description of forager movement
The PDE model will describe bee behaviour in a more averaged way, but will be able to handle much larger numbers of bees for the same computational effort as the IBM. The PDE model will be ideal for exploring the effect of indirect competition for resource between bumble bee nests (several hundred bees) and honey bee hives (tens of thousands of bees). The second advantage of the PDE model is that it can be run for longer periods of time (multiple bumble bee generations) at reasonable computational cost. This model will thus be helpful for examining the long-term sustainability of enhanced bumble bee pollination services as a function of the spatio-temporal distribution of floral resources.
3. User Friendly Tool
Of the three models discussed above, the IBM is the most useful for explaining our work to non-specialists, as the construction and behaviour of IBMs is generally highly intuitive - indeed, it is one of the main advantages of this approach. Our IBM will show individual bees moving across a spatial landscape of blueberry and non-blueberry flowers, providing pollination services to some locations and not others. Bee movement can be seen when watching the simulation, as with a video of real bee movement, and so the modelled bee behaviour is easy to observe and understand, and also easy to mentally compare with real bee behaviour. Furthermore, IBM software packages such as NetLogo are freely available, can use spatial GIS landscape data, and are specifically designed in such a way that non-technical users can still run the the IBMs and experiment with them in a straightforward way. The user-friendly modelling tool for the BCBC and growers will thus be based on the IBM.