- Researchers have developed a web application to enable citizen scientists to listen for the sounds of a population of killer whales off North America’s northeast Pacific coast in real time.
- A network of underwater microphones will stream sounds from under the sea to citizen scientists, who can then report any unusual noises and help decode orca language.
- The researchers have found that human listeners can readily detect unusual sounds amid a stream of underwater noise, and their participation can complement machine-learning algorithms being developed.
Researchers have developed a web-based application to enable citizen scientists to listen to the sounds of killer whales in the northeast Pacific in real time.
Publicly launched on Nov. 1, after several months of testing, the app livestreams audio from underwater microphones, called hydrophones, located in the Salish Sea off the U.S. state of Washington, through the internet to a user’s laptop, phone or tablet.
The Orcasound project produced the app to help raise awareness and understanding about ocean noise, and where and when humans and whales overlap, by broadening the audience listening for the signals of known pods of killer whales (Orcinus orca) and other novel sounds, such as other marine animals or human activity.
Scott Veirs, an oceanographer and bioacoustics specialist, and his father, Val Veirs, a physicist, lead the project, which started with funding from the U.S. National Oceanic and Atmospheric Administration (NOAA). The project’s partners now include United States and Canadian fisheries agencies and several nonprofit organizations, including scientists who have studied ocean sounds in the region for more than 15 years to call attention to the damage caused by noise pollution in the ocean.
The younger Veirs, Orcasound’s lead researcher, presented the new web app and spoke about the value of citizen scientists in acoustic research at the Acoustical Society of America’s recent meeting in Victoria, Canada, earlier this month.
“We want to make it really easy for citizen scientists to listen to signals,” Veirs said in a statement about the new app.
Orcas: easier to hear than to see
Killer whales, or orcas, live in groups called pods and communicate using sound. To human ears, their calls sound like whistles, screams and squeaks, which you can hear in this recording, courtesy of Orcasound:
Orcas, along with dolphins, porpoises and other toothed whales, also echolocate, emitting clicking sounds and assessing their relative position by the “echo” that bounces off nearby objects, such as prey. Hydrophones can pick up the clicks, whistles, calls and gurgles made by fish and marine mammals day and night.
Hydrophones also capture human noises, including from passing ships, construction or drilling. Sound travels easily through water, and the very loud noises produced by human sources such as sonar or pile drivers appear to harm whales and other cetaceans. Ships can also disrupt whale communication because they produce noise at either the lower frequencies used by baleen whales or the higher frequencies similar to those of killer whale echolocation clicks.
Veirs and his colleagues have intensively studied the Salish Sea’s southern resident killer whales (SRKW), a population of salmon-eating orcas that lives close to both shipping lanes and human communities that benefit from ecotourism.
“Orcas are heard every other day in July through September; humpbacks increasingly in October and November, otherwise once per month,” Veirs told Mongabay. “There are about 20 ships per day on average, each lasting 20-40 minutes.”
A network of hydrophones positioned throughout these waters helps the researchers locate the whales, even at night or in poor visibility, so they can send boats out to collect prey fragments or fecal samples to better understand what the whales are eating. The researchers also use hydrophones to study the whales’ communication and the types of human noises in the study area.
Promoting human listeners
The project benefits from people’s unique ability to recognize the sounds around them. “From decades of listening, we’ve learned that humans are amazing signal detectors,” Veirs said in his ASA presentation. “Humans are really good at hearing unusual sounds, and they’re really good at ignoring boring sounds.”
The Orcasound web app enables citizen scientists to listen to livestreaming audio from hydrophones at a depth of about 8 meters (26 feet), 30 meters (98 feet) off the coast of Washington state and British Columbia, Canada.
Anyone interested can visit the app’s citizen science listening page to learn what to listen for, how to listen live to the hydrophones (by clicking a pair of buttons), and how to inform the researchers if you hear an interesting sound. The page also lets you listen to recorded calls of specific whale pods, so that you can learn which group you are likely hearing in the live feed.
Here are the favorite calls of each of three pods, nicknamed J, K and L:
The Orcasound team developed the app to more reliably stream whale calls and other live ocean audio to a broad audience of people using a range of browsers and devices. They built it with new, affordable tools and funding from a 2017 Kickstarter campaign.
Each of the five underwater microphones is connected to a low-cost Raspberry Pi computer armed with audio hardware to analyze the incoming signal and stream it to app users via the internet. Veirs said he hoped to add six hydrophones to the network in 2019, including restoring hydrophones at the Port Townsend Marine Science Center and at the Seattle Aquarium.
The computers run on the open-source Linux operating system and open-source software to encode and send the sound files using data formats compatible with those used by online video streaming services. The project stores the audio data to online cloud storage servers for later analysis by both humans and algorithms.
“We’re setting the stage for AI and machine learning, but it’s not happening yet,” Veirs said. “This winter we’ll roll out an ‘I hear something!’ button. This will start the process of citizens tagging sounds for which development of an automated classifier may be worth the effort.”
The project team has found that humans listening to the sounds can complement machine-learning algorithms that are increasingly able to analyze sound data from the hydrophones.
“We actually have some listeners who will sleep with the hydrophones on in their bedroom,” Veirs said, “and their brains can wake up when they hear a killer whale sound, even if it’s just a faint killer whale sound in some ship noise. That sort of signal detection is right at the cusp of what machines are struggling to do.”
The aim of Orcasound is to harness this capacity by providing people interested in whales and marine conservation with an inexpensive and user-friendly way to participate in research.
“[Volunteer listeners] are really good at not only hearing the sonar event but in real time calling the relevant coast guard,” Veirs said. “They’ll learn which country’s sonar they’re hearing and call the right coast guard. It’s amazing.”
Risks and challenges
Although testing has confirmed the app’s ability to connect listeners to the range of sounds underwater, the project continues to face several challenges.
Among these is equipment failure underwater. Orcasound partners have agreed to monitor and help maintain specific hydrophones. Veirs said one of the hydrophones is under repair, although the associated computer is still working fine and streaming audio data. In this case, the host organization (Orca Network) will troubleshoot, together with the hydrophone manufacturer, and send its volunteers (boat drivers, SCUBA divers) underwater to help out if needed.
Keeping listeners engaged over time is also a concern. “We provide learning materials and eventually will expand to other areas around the globe so you can switch to an active node if yours gets too boring,” Veirs said. “Examples are at http://orcasound.net/learn, and we also hope to embed some of our citizen science within platforms like zooniverse.org which motivate participation through tracking effort (gamification).”
In addition, to train people listening to the streaming audio to be better detectors of whales while ensuring data quality, Veirs says, the project will compare the acoustic data submitted by volunteers with expert classifications.
Seamlessly livestreaming audio data from under the sea to various devices also poses a challenge. Maintaining real-time communications for different types of internet browsers, mobile devices and audio formats will require ongoing software tweaking to avoid limiting the app’s functionality for some users.
Veirs said his team learned from beta-testing that the technology they used to stream the audio wasn’t working well on iOS devices (iPhones and iPads). This resulted in about a 35 percent failure rate for the app. He added that the project’s tech team had since addressed the issue, though they were still facing a 20 percent failure rate after the app’s first week of live test data, and so were still eager for feedback.
Veirs, S., Veirs, V., Williams, R., Jasny, M., & Wood, J. (2018). A key to quieter seas: half of ship noise comes from 15% of the fleet. PeerJ Preprints, 6, e26525v1.
FEEDBACK: Use this form to send a message to the editor of this post. If you want to post a public comment, you can do that at the bottom of the page.