Dataset for 'Multiscale structure of mason bee pollen provisioning across heterogeneous agricultural landscapes'

Understanding how agricultural landscapes shape pollinator resource use is critical for sustaining ecosystem services in working agroecosystems. Land-cover datasets such as the USDA Cropland Data Layer (CDL) are widely used to infer pollinator habitat quality and guide agri-environmental management, yet it remains unclear whether coarse landscape metrics reflect the floral resources collected during maternal provisioning. We evaluated multiscale structure of pollen provisioning in mason bees (Osmia spp.) across heterogeneous agricultural landscapes in Ohio using pollen DNA metabarcoding (ITS1 and ITS2). Nesting shelters were deployed across 30 farms and rural properties through a coordinated citizen science network, enabling broad spatial sampling across diverse agroecosystem contexts; 19 sites produced completed nests for analysis. We quantified pollen diversity and composition from sequential brood cells and assessed the relative influence of within-nest provisioning dynamics, site-level environmental context, and surrounding 1-km land-cover composition on pollen community structure. Pollen composition was stable across sequential brood cells, indicating limited behavioral adjustment during nest construction. In contrast, site identity explained approximately 60 percent of variation in nest-level pollen composition, demonstrating strong environmental filtering at local scales. Across sites, provisions were dominated by woody, non-crop taxa, underscoring the importance of early-season trees, shrubs, and field-margin vegetation within agricultural systems for pollinator support. Despite pronounced among-site differentiation, proportional land cover within 1 km explained little additional variation in pollen community structure. Constrained ordination and mixed modeling revealed weak associations between pollen composition and CDL-derived crop, forest, pasture, or developed cover. These findings suggest a scale mismatch between commonly used land-cover datasets and the fine-scale floral resources that structure pollinator foraging in agroecosystems. Our results highlight the importance of localized habitat features for supporting pollinator nutrition and suggest that interspersing habitat enhancements within existing agricultural landscapes may more directly influence pollen provisioning than broader landscape composition. Integrating fine-scale floral assessments with landscape analyses will improve ecological inference and agri-environmental decision-making.