Distance models reveal biases associated with passive trapping methods for measuring wild bee abundance

There is a need for increased monitoring to track wild bee population declines to inform conservation and restoration efforts. Bees can be sampled non-lethally through visual surveys (e.g., distance transects) or lethally through active (e.g., hand netting) or passive (e.g., trapping) methods. Lethal methods suffer from imperfect detection that is difficult to account for, and evidence suggests that lethal sampling can even invert habitat quality patterns such that high-quality sites yield fewer individuals and low-quality sites yield more individuals.To study potential biases associated with imperfect detection, we used hierarchical density estimation with visual surveys to estimate bee density within 40 young forest patches across Pennsylvania, USA. We surveyed bee communities non-lethally using visual surveys and lethally using blue-vane traps and bee bowls every two weeks between May and September 2019. We collected data on blooming flowers, vegetation structure, and weather. We found that bee densities estimated from distance transects had a positive relationship with floral resource availability. In contrast, abundance measured via bee bowls and blue-vane traps had no relationship, or negative trends with floral resource availability. Raw bee counts within 2-m of the transect always correlated with modeled densities, showing that some methods do not share the biases of attractive traps. Our study demonstrates that failing to account for imperfect detection can impact the interpretation of pollinator surveys. This study adds to a growing body of literature that acknowledges the value of distance sampling for insects like bees to better understand species’ habitat needs and monitor populations for conservation.