Thomas P. Hurst

Real-time reporting of baleen whale passive acoustic detections from ocean gliders

In the past decade, much progress has been made in real-time passive acoustic monitoring of marine mammal occurrence and distribution from autonomous platforms (e.g., gliders, floats, buoys), but current systems focus primarily on a single call type produced by a single species, often from a single location. A hardware and software system was developed to detect, classify, and report 14 call types produced by 4 species of baleen whales in real time from ocean gliders. During a 3-week deployment in the central Gulf of Maine in late November and early December 2012, two gliders reported over 25,000 acoustic detections attributed to fin, humpback, sei, and right whales. The overall false detection rate for individual calls was 14%, and for right, humpback, and fin whales, false predictions of occurrence during 15-min reporting periods were 5% or less. Transmitted pitch tracks--compact representations of sounds--allowed unambiguous identification of both humpback and fin whale song. Of the ten cases when whales were sighted during aerial or shipboard surveys and a glider was within 20 km of the sighting location, nine were accompanied by real-time acoustic detections of the same species by the glider within ±12 h of the sighting time.

Bottlenose dolphins modify behavior to reduce metabolic effect of tag attachment.

Attaching bio-telemetry or -logging devices (‘tags’) to marine animals for research and monitoring adds drag to streamlined bodies, thus affecting posture, swimming gaits and energy balance. These costs have never been measured in free-swimming cetaceans. To examine the effect of drag from a tag on metabolic rate, cost of transport and swimming behavior, four captive male dolphins (Tursiops truncatus) were trained to swim a set course, either non-tagged (n=7) or fitted with a tag (DTAG2; n=12), and surface exclusively in a flow-through respirometer in which oxygen consumption () and carbon dioxide production (; ml kg−1 min−1) rates were measured and respiratory exchange ratio (/) was calculated. Tags did not significantly affect individual mass-specific oxygen consumption, physical activity ratios (exercise /resting ), total or net cost of transport (COT; J m−1 kg−1) or locomotor costs during swimming or two-minute recovery phases. However, individuals swam significantly slower when tagged (by ~11%; mean ± s.d., 3.31±0.35 m s−1) than when non-tagged (3.73±0.41 m s−1). A combined theoretical and computational fluid dynamics model estimating drag forces and power exertion during swimming suggests that drag loading and energy consumption are reduced at lower swimming speeds. Bottlenose dolphins in the specific swimming task in this experiment slowed to the point where the tag yielded no increases in drag or power, while showing no difference in metabolic parameters when instrumented with a DTAG2. These results, and our observations, suggest that animals modify their behavior to maintain metabolic output and energy expenditure when faced with tag-induced drag.

Low complexity lossless compression of underwater sound recordings.

Autonomous listening devices are increasingly used to study vocal aquatic animals, and there is a constant need to record longer or with greater bandwidth, requiring efficient use of memory and battery power. Real-time compression of sound has the potential to extend recording durations and bandwidths at the expense of increased processing operations and therefore power consumption. Whereas lossy methods such as MP3 introduce undesirable artifacts, lossless compression algorithms (e.g., flac) guarantee exact data recovery. But these algorithms are relatively complex due to the wide variety of signals they are designed to compress. A simpler lossless algorithm is shown here to provide compression factors of three or more for underwater sound recordings over a range of noise environments. The compressor was evaluated using samples from drifting and animal-borne sound recorders with sampling rates of 16-240 kHz. It achieves >87% of the compression of more-complex methods but requires about 1/10 of the processing operations resulting in less than 1 mW power consumption at a sampling rate of 192 kHz on a low-power microprocessor. The potential to triple recording duration with a minor increase in power consumption and no loss in sound quality may be especially valuable for battery-limited tags and robotic vehicles.