How many collisions do surveys miss?
Bird–window collision monitoring usually depends on researchers finding evidence of collisions near buildings. But even well-designed surveys provide only a partial view. Wildlife may remove collision evidence, grounds or custodial activity may disturb it, and environmental conditions may make it harder to detect over time. If evidence disappears before researchers arrive, the collision is never recorded.
The Duke Bird–Window Collision Detectability Study examines this hidden source of uncertainty. By learning how long collision evidence remains detectable in different settings, we can better understand what our monitoring data show—and what they may miss.
Watch the project video
Duke Kunshan University student Zeheng Li introduces the detectability study and explains why accounting for undetected collisions is important to bird conservation.
How the study works
Researchers place standardized, commercially sourced study items near selected buildings at Duke University and Duke Forest. A small infrared, motion-activated camera monitors each item for approximately one week, recording whether it remains visible and, when it disappears, what may have removed or disturbed it.
The cameras are directed only toward the immediate study area—not toward building entrances, walkways, or general human activity. Any incidental images of people are excluded from the research.
By repeating these trials at seven campus buildings and seven Duke Forest sites across multiple seasons, the research team can compare detectability in highly developed and more forested environments. These comparisons will help us determine whether collision evidence persists differently across settings and seasons.
Why detectability matters
Duke researchers have monitored bird–window collisions for more than a decade. These long-term observations help identify buildings, architectural features, and landscape conditions associated with greater collision risk. Yet raw survey counts cannot tell us about collisions whose evidence disappeared before a monitoring visit.
The detectability study will allow researchers to estimate the proportion of collision evidence that surveys are likely to miss. These estimates can then be used to develop adjustment factors for Duke’s long-term dataset, producing a more complete picture of collision frequency and helping researchers communicate both the value and the limitations of observational data.
More accurate estimates can strengthen recommendations for bird-friendly building design, guide decisions about where collision-prevention treatments are most needed, and improve the methods used by other monitoring programs.
A Duke–Duke Kunshan collaboration
The project is led by Dr. Nicolette Cagle in collaboration with Dr. Binbin Li and students from Duke University and Duke Kunshan University. Duke Kunshan students Hanyong Cai and Zeheng Li have contributed to field deployment and project communication, including the video featured above.
This collaboration builds upon a longer shared history. Duke’s monitoring methods helped inform bird–window collision research in China beginning in 2017, and the present study reconnects the Duke and Duke Kunshan teams through comparative research, student field training, and applied conservation.
The study is conducted with the support and approval of the Duke Forest, Duke’s Office of Climate and Sustainability, and the appropriate university research oversight offices.
