Sander M. von Benda-Beckmann
Dr. Sander von Benda-Beckmann, an astrophysicist, works in the field of underwater sound and is interested in applying crowdsourcing to the field of bioacoustics.
Peter Lloyd Tyack
Prof. Peter Lloyd Tyack studies the biology of marine mammals, especially the ways they communicate.
Prof. Lech Mankiewicz is director of the Center of Theoretical Physics at the Polish Academy of Sciences and a designer of small telescopic robots. His efforts to propagate open science and popularize scientific knowledge have earned him several prestigious awards.
Whales have very sophisticated hearing organs and can produce loud sounds for communication, orientation, and searching for food. Many scientists see those calls as an important element in communication between whales and in their social development. Special underwater microphones (called hydrophones) enable researchers to listen in. Our team, composed of scientists from the Woods Hole Oceanographic Institution and the University of St. Andrews, has assembled many hours of such recordings by listening to animals in their natural environment as part of long-term scientific projects. We have observed how animals respond to various underwater noises, chiefly human-generated ones. Most recordings therefore contain not only sounds made by killer whales and pilot whales but also noise generated by sonar, echo sounders used on ships, or other sources.
In order to understand how these animals communicate under water, we have categorized the recordings based on the similarity of calls. Killer whales and pilot whales make repetitive and rather easily distinguishable calls. But the multitude of sounds that comprise them is so great that it is difficult to understand the meaning behind specific calls. Most mammals have a sound repertoire specific to their species. Killer whales, however, appear to learn calls from their groups, with every group having its own “dialect.” Calls made by related groups are similar. But much less is known about the sounds made by pilot whales. Our goal is to determine the size of the sound repertoire of pilot whales and answer the question whether different groups of pilot whales have different sound repertoires, just like killer whales. Another purpose of our studies is to assess the impact of growing noise exposure on whales in their natural environment.
We have concluded that creating an acoustic map would go a long way towards helping us study whale communication. If analyzed and put into categories, the recordings we have assembled could help us unravel the mystery of vocalization and communication in marine mammals. By categorizing specific sounds, we could determine the call repertoires of killer whales and pilot whales and discover when they make specific calls and why. Unfortunately, a small team of scientists would need several years to go through such a huge amount of data collected for several years in various locations (in the North Atlantic and in the Bahamas). Some scientists believed that we should set aside most of the recordings and just focus on a small dataset. A different approach arose out of a discussion between Peter Tyack, Sander von Benda-Beckmann, and Robert Simpson of Zooniverse, when Sander referred to GalaxyZoo, a crowdsourcing-based project that he had used when working towards his doctorate in astrophysics. That was how the idea of Whale FM was born.
Crowdsourcing is a method of acquiring knowledge and resolving problems based on communication between a huge number of people, not necessarily experts. Although the practice of asking many people about their opinions is nothing new, crowdsourcing as a method has grown in popularity thanks to the Internet. Initially, it was used chiefly by the world of business. Over time, however, academics also started to turn to crowdsourcing as a way to resolve scientific problems.
Everyone who visits the website of the Whale FM project can help scientists determine the similarity of recorded sounds. The Citizen Science Alliance has partnered with Scientific American to set up a networked science project using the Zooniverse web platform: http://whale.fm. Its Polish partners are the PAS Center of the Theoretical Physics, the web portal Astronomia.pl, and the Hel Marine Station of the Institute of Oceanography at the University of Gdansk. Thanks to their involvement, the project is also available in Polish: http://spiewwielorybow.pl. The purpose of the project is to categorize the acoustic database of sounds made by marine mammals. It is addressed to Internet users of any age or education level who are interested in marine mammals.
The project allows scientists carry out a relevant analysis of the data and mitigate the risk of the subjective interpretation of sounds by individual scientists. A high number of potential volunteers reduces the risk putting sounds into wrong categories. Likewise, the project offers participants an excellent opportunity to gain further insight into whale biology. Internet users can listen to recordings of the sounds made by killer whales and pilot whales, find out more about their habits and behavior, and share their fascination with the subject along with ideas for improving the project. All this helps broaden public awareness of marine mammals, endangered by human activities.
More than 220,000 calls
The project was launched in 2011. Since then, we have been flooded with letters from volunteers. To date, over 13,000 users have listened to and categorized over 220,000 calls.
Volunteers listen to recordings and choose a sound from a random pool of sounds that they find most similar to a given pattern. Humans can generally hear sounds up to 20 kHz, in many cases even less (many adults have an upper limit of around 12 kHz), whereas the sounds made by killer whales and pilot whales can reach 30 kHz or even further out of the human range of hearing. On top of that, some recordings are quite short, which makes it significantly more difficult to find possible matches. This is why calls have been slowed down, so that it is easier for volunteers to hear them and compare them with other sounds.
If we analyze matches made by a single person (usually without relevant education), the probability of success is around 50%, not much better than a coin toss. But if a larger number volunteers all believe a given pair of sounds are similar, the statistical reliability of the outcome rises to 80%, even 90%. Right now, we are working on algorithms that allow us to extract call categories.
What are they saying?
Online participants often ask how the project can help us understand communication between whales. After all, matching sounds together will not actually make us understand what they mean. That is true. But we want to create an acoustic database to later understand their meaning and the situations in which individual whales make specific calls. In addition to collecting sounds, we are also observing the marine mammals. Consequently, we can say what a given animal was doing when making a certain sound. We work together with experienced biologists who watch the behavior of animals in groups. They know whether they are socializing, resting, or looking for food. Likewise, we can gain more insight into the behavior of mammals by tagging them – we know their movements, for example we know that they can dive down to depths of over 1,000 meters. Once whales are tagged using special non-invasive D-tags (they are attached to animals using suction-cups and fall off over time), we can not only listen to the sounds they make but also record what they hear.
Why is it so important to categorize matching sounds? If a similar sound is made in two different situations (for example, when an animal stops fighting and during eating), we want to know whether the sounds are truly identical or maybe different. Online participants are verifying our findings. He hope that their involvement in the experiment will help bring answers to many questions related to the communication of large marine mammals.
Marques T. A., Thomas L., Martin S., Mellinger D., Ward J., Moretti D., Harris D. V., Tyack P.L. (2013) Estimating animal population density using passive acoustics. Biological Reviews of the Cambridge Philosophical Society. 88, 2: 287–309.
Academia nr 2 (38) 2013