Motion-triggered video cameras to record the composition and spatial structure of orchid pollinators

The pollinators for most of the 230 native orchid species in North America are either unknown or known from only a few locations. We cannot effectively conserve orchids without knowing the pollinators they depend on and how those pollinators change across a species’ range. Much of the challenge to documenting pollinators is the need to sit and observe flowers for long hours, and the limited number of people who are able to make observations at a time. Decline of pollinators is thought to be an important driver of orchid decline worldwide. Because of this, documenting the pollinators of all North American orchids and how they differ among locations is a focus of the North American Orchid Conservation Center (NAOCC). We constructed inexpensive Raspberry Pi-driven diurnal and nocturnal (infrared) motion-triggered video cameras to identify pollinators of orchid species in Maryland, Pennsylvania, Virginia, North Carolina, and South Carolina. Over 3 years, with the help of multiple volunteers, interns, and technicians, we captured videos over 8,000 hours of observation, and used a custom machine learning algorithm to expedite reviewing videos by excluding those that were triggered by movement other than pollinators.
We recorded 855 pollination events on 18 orchid species by 41 pollinator species. Notably, with orchids it is possible to distinguish casual floral visitors from pollinators, because orchid pollinia are clearly visible. Pollinators differed among species and among locations within species. The study included three species of fringed orchids (genus Platanthera) and two naturally occurring hybrids. We documented that the species shared pollinators at sites where they co-occurred and related the shared pollinators to genetic and morphological evidence of hybridization at those sites. We also found significant variation in pollinator visitation rates among sites for the Platanthera species.
This project combines pollinator documentation using computer-controlled video cameras, machine learning to sort through videos, and, for some species, genomic analysis, to understand how pollinators differ among species and sites and how they may contribute to genetic intermixing. This project is ongoing and we will continue adding species and sites in the future.