Species traits and local abundance affect bird-window collision frequency

Abstract

Studies on bird-window collisions have generally drawn inferences about species' differential vulnerability from collision tallies. However, this common methodology is potentially biased because the number of collisions may simply reflect prevalence of species at the study site rather than species-specific vulnerability. Building on recent studies of abundance and collision rates, we offered a complementary methodology based on point count data that could be widely applied alongside carcass surveys. Additionally, we broadened our analysis beyond previously applied taxonomic and migratory classifications to include functional classifications of feeding guild, breeding status, and synanthropy. Our null hypothesis was that collision frequencies reflect a species' or classification group's prevalence at study sites. To test this possibility, we used collision data collected at three sites in the Research Triangle Area of North Carolina, United States. At one of these sites, Duke University's Main Campus, we also gathered relative abundances from the local bird community to develop a case study assessment of how background prevalence compared to number of collisions. Using the larger, three-site dataset, we developed an initial picture of collision susceptibility based solely on frequency, the standard practice. Then, by bootstrapping our Duke abundance data, we generated confidence intervals that simulated collision based on chance versus prevalence. We identified several instances where collision tallies produced misleading perception of species-specific vulnerability. In the most extreme case, frequencies from our Triangle Area dataset indicated locally breeding species were highly vulnerable to collisions while our abundance-based case study suggested this same group was actually adept at avoiding collisions. Through our case study, we also found that foliage gleaning was linked to increased risk, and omnivory and ground foraging were associated with decreased risk. Although our results are based on a limited sample, we argue that abundance needs to be incorporated into future studies and recommend point counts as a noninvasive and adaptable alternative to area-searches and mist netting.

Keywords

• carcass survey
• classifications
• collision vulnerability
• local abundance
• point count
• window strikes