Effects of environmental factors on the detection of subsurface green turtles in aerial drone surveys

Abstract

Abstract Context Aerial video surveys from unpiloted aerial systems (UAS) have become popular in wildlife research because of increased accessibility to remote areas, reduction of anthropogenic disruption to habitats and wildlife, low operating costs, and improved researcher safety. In shallow marine systems, they can provide opportunities to rapidly survey species that cannot easily be surveyed using boat- or land-based techniques. However, detectability of subsurface animals in marine habitats may be affected by environmental factors. Aims We investigated the effects of water depth, seagrass cover, surface glare, and observer numbers and expertise on the probability of detecting subsurface green turtles in UAS video surveys. Methods We deployed inanimate green turtle decoys at randomised intervals along 24 pre-determined transects across a depth gradient in a seagrass-dominated bay off Great Abaco, The Bahamas. We collected aerial videos of the transects by flying a DJI Phantom 3 Advanced quadcopter drone at an altitude of 10 m over each transect. Three independent observers watched each video and recorded decoy sightings to compare detection probabilities across observer experience levels. We used a generalised linear model to test for the effects of glare, water depth, wind speed, and seagrass cover on the detectability of turtle decoys. We also recorded glare conditions with aerial videos taken at 2-h intervals over a still body of water on cloudless days off North Miami, FL. Key results Individual observers performed similarly, but adding one additional observer increased detection by 11–12% and adding a third observer increased detections by up to 15%. Depth, seagrass cover, and glare significantly affected decoy detections. In both summer and fall, the optimal times and directions to minimise glare in aerial video surveys were 0800 hours, facing any direction other than north, and 1800 hours, facing any direction other than south. Conclusions The number of human observers and environmental variables, especially depth, seagrass cover, and glare, are important to explicitly consider when designing and analysing data from UAS surveys of subsurface animal abundances and distribution. Implications Our study draws attention to potential limitations of UAS-acquired data for subsurface observations if environmental conditions are not explicitly accounted for. Quantifying the effects of environmental factors, designing surveys to minimise variance in these factors, and having multiple observers are crucial for optimising UAS use in research and conservation of sea turtles and other marine fauna.