NEWPORT, R.I. –
Dr. T. Aran Mooney, Dr. Seth McCammon and Dr. Yogesh Girdhar from the Woods Hole Oceanographic Institution (WHOI) in Massachusetts shared their studies on coral reefs, mapping sounds and the use of autonomous underwater vehicles (AUVs) with the Naval Undersea Warfare Center Division Newport workforce on June 5.
WHOI’s Reef Solutions program, formed in 2012, was designed to monitor and explore soundscapes of a range of coral reefs in the study site of the U.S. Virgin Islands, Mooney said. Coral reefs are hubs of biodiversity as they support over one billion people by contributing to coastal protection, fisheries, livelihood, tourism, food resources and pharmaceuticals. Mooney said 25% of all marine species live on coral reefs at some point in their lifetime.
Coral reefs also play an important role in breaking up and reducing storm surges by dissipating 97% of wave energy. The value of coral reefs is estimated to be up to $2.6 trillion.
To date, 25% of coral reefs have been lost within the past 30 years, Mooney said, calling them the most imperiled ocean ecosystem. He noted that the United Nations predicts 90% of coral reefs could be lost by 2050. Some of the impacts to the loss of coral reefs include changes in pH range, eutrophication, diseases, and increases in water temperature and ocean noise.
Healthy coral reefs are rich with marine life, which produces a lot of sound, piquing WHOI’s interest in measuring the health of the reefs through sound. However, a rich acoustic environment creates barriers to obtaining accurate data.
Mooney began observing the snap rates on snapping shrimp on the Yawzi and Tektite reefs and what environmental factors drive them. After one week of studying the snap rates he was able to determine the correlation between dawn and dusk, water temperature, and different seasons.
Mooney then created a snapping shrimp monitoring system where he adjusted the water temperature and matched different pairs of snapping shrimp to further observe their social patterns and snapping rate.
“From the observational studies, we wanted to see if we could predict or drive these patterns to happen,” Mooney said. “We collected these animals where we could focus on particular variables in the laboratory. We looked at behavioral context and how that may influence snap rate, as well as that might affect snap rate.”
McCammon discussed how WHOI seeks and maps hotspots to develop ways to protect and preserve the reefs. He highlighted three key steps: identifying biological sound signals from overall reef soundscapes; localizing source locations relative to the AUV in real time; and using next-best viewpoint algorithms to determine optimal sensing locations.
By using an acoustic recorder, such as the Curious Underwater Robot for Ecosystem Exploration (CUREE) AUV, McCammon and his team built maps of the reefs and created a generative model for bearing observations.
“We are very interested in machine learning to identify biological sound signals from our soundscape,” McCammon said. “We encode the knowledge from the experts that annotate the spectrograms into a machine learning tool.”
One downside to using an AUV is the thruster noise that is generated as it moves through the water, hindering hearing the biological sound. A solution to this is to pause the thrusters, have the AUV stop mid-water column, and record a little bit of sound before moving on.
“This turns the sound source into a discrete sensor placement problem,” McCammon said. “I have to figure out where and when to stop to make those observations.”
During the sensing windows, McCammon and his team receive a lot of information. Referring to Mooney’s example of the snapping shrimp, McCammon said many different fish calls also can be identified. In their real-time beamforming for snapping shrimp, McCammon uses a statistical model for the distribution of angles of the different shrimp snaps they observe within a three- or 10-second drifting window. The data is then utilized as the bearing observation the AUV uses as it moves through the water.
The goal for Girdhar and his team is to define an exploratory path through the environment that maximizes the information gained by the AUV. He curates these paths by running an unsupervised learning system where the AUV is put into an environment without any pre-training. Wherever the AUV is placed, it learns about the environment in real time.
The information gained from the robot’s observations is then used to rapidly analyze the missions. Girdhar demonstrated how an AUV would operate while exploring the ocean floor. It will naturally stay around patchy regions, then drift away and backtrack. This type of behavior is useful for setting patching phenomena, he said.
CUREE, developed by Girdhar, is the main AUV used in WHOI’s laboratory. It has hydrophones as well as a stereo camera system pointed forward for pelagic observations and another camera pointed downward for benthic analysis. Vision is the main sensing platform on the CUREE, which allows the scientists to follow the sea floor contours more tightly than using just sonar. CUREE has a huge amount of computing power, so a lot of the processing can be done onboard.
Aside from CUREE, Girdhar also utilizes a highly maneuverable, engine-based underwater AUV called ReefGlider. By developing an AUV with buoyancy control engines rather than using thrusters, Girdhar can maintain the low-noise level and increase the acrobatics.
“Coral reefs have so much information in them, structural, acoustic and visual,” Girdhar said. “As a computer scientist, my job is to find interesting problems and how we monitor all the complex processes happening there and how we use robots and artificial intelligence to make that happen.”
Moving forward, Girdhar would like to develop an AUV that is quieter to allow for making passive acoustic measurements, as well as more maneuverable and acrobatic so they can follow fish. The idea is to build something that is flipper-based that can roam the ocean and find visual and acoustic anomalies. Although not yet developed, Girdhar and his team have been experimenting with direct drive brushless motor flipper actuation systems combined with cameras.
“It will be hard to control them, but the use of reinforcement learning will be a way to control them,” Girdhar said. “Once we have them, it should be a game-changer. It will be a way to monitor complex environments such as coral reefs.”
To learn more about the Reef Solutions team at WHOI and the work it is doing, visit https://reefsolutions.whoi.edu/
NUWC Newport is the oldest warfare center in the country, tracing its heritage to the Naval Torpedo Station established on Goat Island in Newport Harbor in 1869. Commanded by Capt. Chad Hennings, NUWC Newport maintains major detachments in West Palm Beach, Florida, and Andros Island in the Bahamas, as well as test facilities at Seneca Lake and Fisher's Island, New York, Leesburg, Florida, and Dodge Pond, Connecticut.