A Bird's-Eye View: Novel Use of Drone Images to Quantify Differences in Altitudinal Reflections in Bird-window Collision Studies

Abstract. Bird-window collisions (BWCs) constitute a significant source of mortality for both resident and migratory birds. Because windows reflect surrounding landscape components, such as vegetation or sky, birds do not always perceive glass as a barrier. Here we demonstrate a novel technique to classify and quantify reflections in windows on the Radford University campus in southwest Virginia, U.S.A. We deployed a consumer-grade Unmanned Aerial Vehicle, i.e., drone, to photograph 14 contiguous or near-contiguous window columns across five campus buildings in August 2020. For each study aspect, the drone (equipped with an RGB camera) captured images at ca. 5 m altitudinal increments from the ground floor to the roof of each building (three to six images/vertical column). We then manually classified each image in ImageJ to calculate approximate proportions of reflected: (1) vegetation, (2) sky, (3) and artificial structures or impervious surfaces, plus (4) nonreflective glass. We used a generalized linear model to determine how proportional reflections of vegetation, sky, buildings, and nonreflective glass varied across vertical increments. We found the proportion of sky significantly decreased with increasing photo heights, whereas proportion of nonreflective glass significantly increased with increasing heights. This supports previous findings that because birds are drawn to sky reflections, they may collide at relatively lower positions on buildings. Inconsistency in landscape design and building positioning on the campus precluded trends in vegetative or building reflections by height. Our pilot study demonstrates the applicability of a consumer-grade drone for investigating visual characteristics of reflections that influence BWCs from variable observation angles. We suggest the expanded use of drone images as a straightforward technique to measure changes in reflection characteristics from varying degrees of observation. They are a novel method in developing a BWC risk assessment as well as potential mitigation strategies in a suburban or campus environment with buildings of intermediate heights.