Accounting for animal movement during aerial imaging surveys

Abstract The response of wildlife to the movement of aircraft during low‐altitude image collection can lead to double‐ or undercounting animals, thereby biasing abundance estimates. The preservation of spatial arrangements of animals, particularly in spatially overlapped images, provides new opportunities for measuring such bias compared with observer surveys. We present a novel method to assess count error and platform bias in aerial imaging surveys by comparing counts and relative positions of animals in spatially overlapped image pairs. We analyzed drone imagery of overwintering migratory waterbirds at state and federal wildlife conservation areas in New Mexico, USA. To assess count error, we compared counts of animals in spatially overlapping images along transects and between adjacent transects. To detect the movement of animals in response to the movement of fixed‐wing and rotary‐wing drones in grid sampling surveys, we calculated vectors between the geographic median of the animal locations in spatially overlapping image areas, normalized these vectors across surveys relative to the heading of the drone, and used circular statistics to test for directional bias. We found that ducks, geese, and cranes showed low but differing degrees of count bias throughout surveys, and that the rotary‐wing drone generated more directionally biased movement than the fixed‐wing drone. Cranes and geese showed decreases in aggregate counts of −9.9% and −9.5% between adjacent transects, indicating movement out of the imaging area, whereas ducks showed a small net increase in the aggregate count of 2.5% between adjacent transects, indicating that they may be “pushed” slightly by the movement of the drone. We present a broadly applicable method for quantitatively assessing animal responses to aircraft, to subsequently improve the credibility of abundance estimates.