A comparison of two techniques for modeling native bee habitat in alternative future landscapes that incorporate perennial bioenergy crops

Background/Question/Methods

Native bees provide vital ecosystem services, but their populations are at-risk under current land management techniques. Efforts are underway to improve conditions for native pollinators, as reflected in President Obama’s 2014 Pollinator Health Task Force.  Simultaneously, energy policy supports investigation of alternative future landscape patterns (FLPs) that include perennial bioenergy crops (PBCs) within a conventional corn/soy matrix.  Minimal changes to land management may support larger, more resilient pollinator populations.  We hypothesize that different FLPs, all with implications for PBC in a conventional agricultural matrix, will have positive, but different, impacts on pollinator populations, depending on the PBC varieties chosen and the specific layout of the FLP.  In order to test this hypothesis, we model native bee populations using two fundamentally different modeling techniques for predicting the effects of landscape change on native bees.  The first, the InVEST model, uses species metrics and land cover attributes to predict the probability of bees nesting or foraging at an individual landscape pixel.  The second, a recently published regression model, predicts bee abundance, diversity, and community composition in a pixel from various land cover proportions within 1500 m of the pixel. Both modeling techniques provide spatially explicit predictions of the suitability for native bees. 

Results/Conclusions

Since the two models do not necessarily predict the same effect for a specific landscape configuration, we employ both techniques on a series of PBC-incorporating FLPs in a 20,000 ha agricultural watershed in Illinois. Within each model, we compare results across FLPs to test the hypothesis that different FLPs will have different effects on bee populations.  Next, we compare the results of both models within each FLP to compare the different modeling techniques with each other.  Finally, we develop and model neutral landscape models that duplicate the patch metrics of each FLP, in order to explore the hypothesis that PBC layout influences native bee populations.  We find that InVEST and regression modeling techniques predict different impacts of landscape configuration and composition on native bee populations, with InVEST providing more detail.  Using both models together helps bound the bee population response and provides insight into uncertainty about the effects on native bee populations, particularly when examining a suite of different landscape patterns that incorporate PBC.  Minimal changes to landscape management, for instance including small areas of perennial cover to provide floral resources or nesting habitat and reducing the use of agrochemicals, may indeed support larger, more resilient native bee populations.