Large-bodied, kleptoparasitic, and social bee species are overrepresented on the iNaturalist platform in California

Participatory science platforms such as iNaturalist have created large databases of insect observations, including of wild bees. However, these data are rarely used in formal research contexts to inform bee monitoring or conservation. This is in part because iNaturalist data is prone to spatial, temporal and taxonomic biases, and little research has characterized the differences between iNaturalist data on wild bees or other insects with data generated by standard lethal collections. Here, we identify biases in bee species data on the iNaturalist platform, with the goal of informing which bee species and functional groups are good candidates for study using iNaturalist data as they are typically collected. We compared four previously published bee community datasets that were obtained using standardized lethal sampling in California, USA, with bee observations on iNaturalist from the same ecoregions and habitat types in which sampling occurred. We found compositional differences between iNaturalist and collections data, though both datatypes picked up on regional differences in bee community composition. We identified 101 bee species that were overrepresented on iNaturalist, including 33 species that were overrepresented in two or more regions. Preliminary analyses suggest that large-bodied, social, and brood-parasitic species are more likely to be overrepresented. Our results are congruent with findings of the only other study to compare iNaturalist with lethal collections for bees, despite the fact that these two studies occurred on opposite sides of North America. Careful consideration of biases will therefore be crucial for successful leveraging of iNaturalist data for bee research and conservation.