Crop diversity and composition are increasingly found to be important drivers of pest and natural enemy dynamics that can be obscured when using broad landcover categories (e.g., crop vs. semi-natural habitat). We leveraged four large-scale pest monitoring data sets, collectively spanning over 1000 site-years across Montana, North Dakota, Nebraska and Colorado, to examine the complex source-sink landscape dynamics of the wheat stem sawfly, Cephus cinctus and its associated parasitoids. Cephus cinctus is a habitat generalist that attacks a wide range of wild grasses and cultivated small grain crops, so both crop and semi-natural habitats have the potential to be important sources of insects. We modeled pest infestation and parasitism responses to landscape-scale semi-natural habitat cover (rangeland and set-aside Conservation Reserve Program (CRP) grassland), crop diversity and crop composition extracted from the USDA-NASS cropland data layer and USDA-FSA CRP data layers. Despite dramatic differences in cropping systems across regions (e.g., mean crop richness varying from 2 to 5 crops; complete turnover in the composition of rotational crops), some general patterns emerged. Semi-natural grassland habitats (rangeland or CRP) were not important predictors of pest infestation or parasitism levels in wheat. Instead, infestation was most consistently predicted by the area planted to wheat in the previous year. Additionally, in regions where multiple wheat types were planted one crop (winter wheat in MT and durum wheat in ND) drove the pattern, with the relationship disappearing when cover was aggregated across host crop types. The strength of the relationship between wheat cover and pest pressure also varied considerably year to year, with strongest effects observed in years of intermediate pest densities, illustrating the importance of multi-year data sets to detect landscape patterns. Overall, the results suggest that increasing the area rotated to non-host crops in the landscape, or even to less preferred hosts or resistant varieties, could dramatically reduce pest pressure in this system. Despite a frustrating lack of consistency in pest response to landscape structure that foils generalizations across systems, our analyses suggests that strong patterns, with direct management implications, can emerge from careful dissection of crop compositional influences for a particular pest or region. Building cross-regional, cross entity collaborations to bolster support for pest monitoring programs, standardizing methodologies, and increasing data accessibility will all be important to maximize the utility of such data in developing pest management solutions moving forward.
