Multispecies integrated distribution models (iSDMs) to evaluate the role and efficacy of citizen science in monitoring bumblebees across the upper Midwest

Bumble bee species across North America have undergone substantial range contractions driven by climate change, and face further compounding threats from land use change and agricultural intensification. An emerging concern is that traditional monitoring schemes are not expansive enough to effectively monitor these changing range dynamics. Citizen scientists are highly effective for increasing sampling density and extent, but pose data challenges associated with misclassification and preferential sampling (i.e., bias). Here, we integrate two volunteer-based data collection programs (Bumble Bee Watch [BBW] and Wisconsin’s Bumble Bee Brigade [BBB]) that leverage expert verification of species and identities and involve with both presence-background (BBW, BBB) and presence-absence (BBB) sampling protocols to investigate multi-species patterns in bee occurrence across the upper Midwest and explore sensitivities to these data frailties. We fit combinations of multi-species species distribution models that varied with respect to the data inputs considered, whether reported or expert-verified species identities were used, and whether strategies designed to account for preferential sampling were employed or not. Results indicate great sensitivity both to data volume, sampling assumptions, and species prevalence. Reliable invertebrate monitoring over broad scales via community science programs will likely require design or model-based strategies to improve data quality and impose additional structure on sampling designs across a reasonable data subset, but encouragingly, multi-species iSDMs appear to provide a useful framework for improved inference.