Collisions with windows are an important human-related threat to birds in urban landscapes. However, the proximate drivers of collisions are not well understood, and no study has examined spatial variation in mortality in an urban setting. We hypothesized that the number of fatalities at buildings varies with window area and habitat features that influence avian community structure. In 2010 we documented bird-window collisions (BWCs) and characterized avian community structure at 20 buildings in an urban landscape in northwestern Illinois, USA. For each building and season, we conducted 21 daily surveys for carcasses and nine point count surveys to estimate relative abundance, richness, and diversity. Our sampling design was informed by experimentally estimated carcass persistence times and detection probabilities. We used linear and generalized linear mixed models to evaluate how habitat features influenced community structure and how mortality was affected by window area and factors that correlated with community structure. The most-supported model was consistent for all community indices and included effects of season, development, and distance to vegetated lots. BWCs were related positively to window area and negatively to development. We documented mortalities for 16/72 (22%) species (34 total carcasses) recorded at buildings, and BWCs were greater for juveniles than adults. Based on the most-supported model of BWCs, the median number of annual predicted fatalities at study buildings was 3 (range = 0–52). These results suggest that patchily distributed environmental resources and levels of window area in buildings create spatial variation in BWCs within and among urban areas. Current mortality estimates place little emphasis on spatial variation, which precludes a fundamental understanding of the issue. To focus conservation efforts, we illustrate how knowledge of the structural and environmental factors that influence bird-window collisions can be used to predict fatalities in the broader landscape.

Collisions with windows remain an important human‐related threat to bird survival in urban landscapes. Accurately estimating the magnitude of avian mortality at windows is difficult and may be influenced by many sources of error, such as scavenging of carcasses. Failure to account for removal of carcasses by scavengers can bias estimates of window mortality. We tested the hypothesis that carcass survival depends on local habitat factors known to influence scavenger behavior. Scavenger activity on bird carcasses was documented at 20 buildings in an urban landscape in northwestern Illinois for 1 week during each season of a year. Known‐fate models were used to relate carcass survival to local habitat composition and to evaluate temporal variation in survival. We also documented species of scavengers and the timing of scavenging using motion‐triggered cameras. Daily carcass survival was greater in winter than during spring, summer, and fall. Survival was related negatively to canopy cover (trees and shrubs within a 50‐m buffer) and window area, and positively to pavement cover. Using an exponential model of survival time, estimated mean time of survival of carcasses (t± SE) was 82.9 ± 11.7 d for winter and 11.8 ± 7.2 d for other seasons. Raccoons (Procyon lotor) scavenged more carcasses than other species. Our results suggest that (1) carcass survival times may be short at locations with preferred habitats of known scavengers and predictable sources of food, and (2) knowledge of scavenger distribution and activity can inform predictive models of persistence. In studies of bird‐window collisions, the influence of scavenger bias can be minimized by maintaining short time intervals between carcass searches. Search intervals can be inferred by estimating the number of days that a carcass should persist at a site, which can be calculated using predicted daily survival probabilities of carcasses at study buildings.

No abstract

No abstract