Shark detection probability from aerial drone surveys within a temperate estuary

Abstract

Drones are easy to operate over metres-to-kilometre scales, making them potentially useful to monitor species distributions and habitat use in shallow estuaries with widely varying environmental conditions. To investigate the utility of drones for surveying bonnethead sharks (Sphyrna tiburo) across estuarine environmental gradients, we deployed decoys, fashioned to mimic sharks, in the field. Decoys were placed in two flight areas (0.8 km2 each) in shallow (<2 m) water near Beaufort, N.C., on five days during 2015–2016. Survey flights were conducted using a fixed-wing drone (senseFly eBee) equipped with a digital camera. Images were indexed for combinations of six environmental factors across flights. Images representative of all (N = 36) observed environmental combinations were sent to a group of 15 scientists who were asked to identify sharks in each image. Non-parametric rank-sum comparisons and regression tree analysis on resultant detection probabilities highlighted depth as having the largest, statistically reliable influence on detection probabilities, with decreasing detection probabilities at increased depth. Detection probabilities were higher during midday flights, with notable effects of wind speed and cloud presence also apparent. Our study highlights depth as a first-order factor constraining the temperate estuarine habitats over which drones may reliably quantify sharks (i.e., <0.75 m).