Page 41 - Summer2022

“During the integration of an ultra-short baseline receiver (USBL) on an autonomous underwater vehi- cle, I spent a few frustrating weeks trying to figure out exactly why I was not getting the accuracy I was expect- ing. Eventually, I realized that the position at which the USBL receiver was located on the vehicle changed the accuracies that I obtained — ultimately, I discovered that local acoustic effects between the received signal and the body of the vehicle created extremely repro- ducible biases in the resulting angle estimate” (personal email, 2022, used with permission). Biological factors are notorious for obscuring or chang- ing expected returns in all types of active sonar systems; fish and plankton scatter sound so they often show up when they are not the object of measurement. This was observed by James Ian Vaughn of WHOI when mapping bottom topography: “We were surveying the Kick’em Jenny volcano off the coast of Grenada with an EM302 multibeam some time ago. We saw a bunch of large discrete scatterers sitting in/over the caldera. We quickly followed up the multi- beam survey with an ROV dive. Turns out the scatters were a school of large tuna. Too deep to fish for, unfor- tunately” (personal email, 2022, used with permission). Hunting down the reasons for initially unexplained scattering and reflections has led to many revelations and discoveries across the ocean disciplines. A perfect example of this is the deep scattering layer. Early SONAR systems observed a so-called “false bottom” at around 500 meters deep. This scattering layer in the twilight zone of the ocean consists of enormous numbers of animals that migrate daily in depth, including fishes, squid, and siphonophores. Understanding this deep scattering layer has been a major objective in marine science and acous- tics this decade because disruption due to fishing and deep-sea mining could have profound implications for both biodiversity and the global carbon cycle (Boscolo- Galazzo et al., 2021). The Value of Weird Data “Blink our eyes, and the world you see next did not exist when you closed them. Therefore, he said, the only appropriate state of mind is surprise. The only state of the heart is joy. The sky you see now, you have never seen before” (from the Thief of Time by Terry Pratchett). The attitude of underwater acoustics users in relating tales of weird acoustic data is, indeed, joy mixed with some amount of chagrin. These incidents are clearly not isolated. And when organized into the categories illus- trated in Categories of Interference, they begin to paint a picture of the types of interference any given system may experience in the ocean. Instead of simply filtering out these surprises, they can be treated as interesting and worth preserving, sharing, and publishing. Over the last five years, there has been an increasing effort to centralize and process underwater acoustic data (e.g., Wall et al., 2021) and to share code and processing tips among the acoustics community. A great resource list of underwater acoustic datasets is maintained by the United Kingdom Acoustics Network and includes no fewer than 33 separate databases as of March 2022 (see acoustics.ac.uk/open-access-underwater-acoustics-data). Each database has different sensor characteristics in a mix of active and passive acoustics. In terms of processing code, the development of MATLAB toolboxes and open-source packages in Python for underwater signal processing and array pro- cessing make sophisticated analysis of acoustic data far more accessible. There are also domain-specific efforts to connect community members for information shar- ing, such as a new Bioacoustics Stack Exchange (see acousticstoday.org/wPQmt), built to provide a central- ized discussion space for conversations about processing and understanding bioacoustics data in particular. Including specific identification and labeling of weird data within broader datasets and processing tools as a part of this community-wide effort feels both natural and necessary, for three main reasons. First, there is an incredible potential of “found data” for other researchers, where weird noise in one discipline can be identified as a signal by another. Second, cross-sensor, labeled exam- ples of interference would pave the way for the use of machine-learning tools in the development of ubiquitous underwater acoustic classification tools for autonomous identification of interference sources. Finally, these types of tools would make underwater acoustic data interpre- tation significantly easier and would also provide more information about the oceans, feeding back into providing found data across disciplines. Summer 2022 • Acoustics Today 41