Schrödinger's fish: Examining the robotic observer effect on pelagic animals

Abstract

Robust sampling of animals is necessary for understanding ocean ecology, but evaluating the effectiveness of our samplers is a challenge. Scientific echosounders were added to two robotic platforms carrying video imaging systems: a remotely operated vehicle (ROV) and an autonomous underwater vehicle (AUV). The vehicles were used to quantitatively sample midwater life in Monterey Bay along horizontal transects at incremental depths ranging from 25 to 1000 m. The echosounders allowed the bulk behavioral responses of animals to be observed up to 200 m forward of each platform. These responses observed included no response, continual avoidance, avoidance to a fixed range resulting in a patch, and attraction. There were strong and interacting effects of depth and platform type on behavioral responses. Measurements of acoustic backscatter showed that animals responded more strongly to the AUV than the ROV. During AUV surveys, there were effects of day/night and the use of artificial illumination on animal responses. Behavioral responses to our sampling were both species- and context-dependent. These data inspired the expansion of an existing mathematical framework that formalized the processes affecting the sampling of motile ocean organisms. Originally developed for net sampling, we generalized the equations to be platform- and sensor-agnostic and incorporated animal decision-making processes to allow for behaviors consistent with the full range of responses we observed. These results and the framework can help move toward more effective sampling of motile animals in the ocean.