Researchers develop automated method to identify fish calls underwater
Editor’s Note: The following article was written by Sean Nealon of Oregon State University and adapted for NOAA Research. The Oregon State University press release is available online.
CORVALLIS, Ore. – A research team led by the Cooperative Institute for Marine Ecosystem and Resource Studies of NOAA at Oregon State University has developed an automated method that can accurately identify calls from a family of fish.
The method takes advantage of data collected by underwater microphones called hydrophones and provides an efficient and inexpensive way to understand changes in the marine environment due to climate change and other anthropogenic influences, researchers from the Cooperative said. Institute for Marine Ecosystem and Resource Studies. .
The results were published in the journal Marine Ecology Progress Series.
Hydrophones are increasingly deployed in the world’s oceans. They offer advantages over other types of surveillance as they operate at night, in low visibility conditions and for long periods of time. But the techniques for effectively analyzing hydrophone data are not well developed.
Hydrophones record the sounds of fish
Hydrophone deployed in a tropical reef region in American Samoa National Park. Credit: Tim Clark / National Park Service.
This new research, led by Jill Munger when she was an undergraduate student, is starting to change that. Munger came to the state of Oregon after having worked for more than 20 years in the corporate world.
Passionate about scuba diving, she wanted to study the ocean. She received a fellowship from CIMERS to conduct research in underwater acoustics with Joe Haxel, who was at the time at the Hatfield Marine Science Center in Newport and working with the acoustics program at the Pacific Marine Environmental Lab at the National Oceanic. and Atmospheric Administration.
Haxel gave him a hard drive containing 18,000 hours of acoustic data collected over 39 months in a tropical reef region within American Samoa National Park. American Samoa is an American territory located in the western Pacific Ocean.
The data was collected via a 12-station hydrophone area maintained by NOAA and the National Park Service which is distributed worldwide in waters controlled by the United States. The hydrophones were designed and built by NOAA and CIMERS researchers at the Hatfield Marine Science Center.
Munger decided to focus on the ladies’ calls, in part because they are distinctive. They cringe to create pops, clicks and chirps associated with aggressive behavior and nest defense. She compared the sound to the purring of kittens. Soon it became clear to him that manual listening to the recordings was not going to work.
“It’s such a slow and tedious process,” she recalls thinking. “I have all of this data, and I only look at a tiny part of it. What happens in all the other parts that I haven’t had the chance to listen to? “
A conversation with his brother, Daniel Herrera, a machine learning engineer, sparked an idea. Could they use machine learning to automate data analysis?
Machine learning algorithms build a model based on data samples, called training data, to make predictions or decisions without being explicitly programmed to do so.
Machine learning techniques have been used to automate the processing of large amounts of data from passive acoustic monitoring devices that have collected sound data from birds, bats, and marine mammals. Techniques have been used for fish calls, but this is an underdeveloped field of science, Munger said.
The damsels gather on the reef
Damsels bring a bright blue color to a tropical reef region in American Samoa National Park. Credit: National parks network.
In this case, the machine learning sample or training data was 400-500 Miss Munger calls identified by manually listening to the hydrophone recordings. With that start, Herrera, co-author of the article, built a machine learning model that accurately identified 94% of damsel calls.
“We built a machine learning model on a relatively small set of training data and then applied it to a huge data set,” Munger said. “The implications for environmental monitoring are enormous.”
Munger, who now works in the lab of Scott Heppell, an associate professor in the Oregon Department of Fisheries, Wildlife, and Conservation Sciences at the College of Agricultural Sciences, believes machine learning will increasingly be used by scientists to monitor many species of fish in the ocean because it requires relatively little effort.
“The advantage of observing fish calls over a long period of time is that we can begin to understand how this relates to changing ocean conditions, which influence our country’s living marine resources,” said Munger. “For example, the abundance of damselfish calls may be an indicator of the health of coral reefs.”
Munger received comments from National Park Service staff on the biology of damselflies and reef habitats near the hydrophone.
The other co-authors of the article are Haxel, Heppell, and Samara Haver, all from the state of Oregon; Lynn Waterhouse, formerly of the John G. Shedd Aquarium in Chicago; Megan McKenna, formerly of Natural Sounds and Night Skies Division, National Park Service, Fort Collins, Colorado; and Jason Gedamke and Robert Dziak of the NOAA Office of Science and Technology, Silver Spring, Md., and Pacific Marine Environmental Laboratory, Newport, respectively.
Photos of damsels and hydrophone: https://flic.kr/s/aHsmXbbend