Careers in underwater acoustics

For an island nation, we don’t have enough experts working in underwater acoustics. According to UKAN’s Professor Kirill Horoshenkov, this country desperately needs people to work in this area so he got in touch with Professor Traci Neilsen, Brigham Young University, Physics and Astronomy and asked her to share the story of her work in the field. Hers is a timely and very welcome article.

By Professor Traci Neilsen, Brigham Young University, Physics and Astronomy

Acoustics is fundamental to many different areas of study, technology and innovation and although careers in acoustics cover a variety of fields, few young people know of these options. Most acousticians describe discovering acoustics at some point along their career path so the IOA and the UK Acoustics Network (UKAN) have both developed fabulous resources describing different careers. As a complementary effort, I recently led the Committee for Outreach and Education of the Acoustical Society of America to create infographics for select careers in acoustics https://exploresound.org/acoustics-careers/ that are intended to attract the attention of secondary school (high school) students and university students who are searching for a career path. Links to the UK Acoustic Network Resources Archive – The UK Acoustics Network and other career information are provided in the ‘Learn More’ link on each career page. (If you have ideas for additional resources that could be included, please email me at tbn@byu.edu with your suggestions.) As current Chair of ASA’s Committee on Outreach and Education, I can say that we hope to continue working with the IOA and UKAN to increase understanding about the rewarding careers in acoustics.

Two of these 16 careers focus on underwater acoustics:
1. Acoustical oceanography (https://exploresound.org/acoustics-careers/acoustical-oceanography/) and
2. Underwater communications (https://exploresound.org/acoustics-careers/underwater-communication/).
The spread of careers in underwater acoustics, however, covers a much broader range and touches on many of the 10 challenges identified by the UN Decade of Ocean Acoustics (https://oceandecade.org/challenges/).

Sound in the ocean is used to conduct environmental impact studies, monitor ambient ocean noise levels, as well as seismic activity (https://acousticstoday.org/2021-summer-twenty-thousand-leagues-under-the-sea-recording-earthquakes-with- autonomous-floats-frederik-j-simons-joel-d-simon-and-sirawich-pipatprathanporn/) and tsunamis. Marine bioacoustics focuses on monitoring species populations and migrations to assist conservation efforts. Shipping noise is monitored to quantify the amount of anthropogenic noise in the ocean (https://acousticstoday.org/the-bureau-of-ocean-energy-management-and-ocean-noise-shane-guan-jill-lewandowski-and-erica-staaterman/) and can be used to monitor fishing activities.

Clever noise mitigation technologies have been developed to limit the impact of marine construction noise and the noise produced by offshore wind (https://acousticstoday.org/the-underwater-sound-from-offshore-wind-farms-jennifer-amaral/) and marine energy (https://acousticstoday.org/listening-to-the-beat-of-new-ocean-technologies-for-harvesting-marine-energy-joseph-haxel-christopher-bassett-brian-polagye-kaustubha-raghukumar-and-cailene-gunn/) is closely monitored. Acoustic tomography involves measuring ocean currents and temperature and provides estimates of how the ocean environment is changing.

The rapidly changing Arctic is being closely monitored with acoustics (https://acousticstoday.org/ocean-acoustics-in-the-rapidly-changing-arctic-peter-f-worcester-matthew-a-dzieciuch-and-hanne-sagen/), as well as the sounds produced by glaciers (https://acousticstoday.org/the-underwater-sounds-of-glaciers-grant-b-deane/). Similarly, acoustic recordings are used as a rain gauge for the open ocean to provide information about the changing climate (https://acousticstoday.org/rainfall-at-sea-using-the-underwater-sounds-of-raindrops-as-a-rain-gauge-for-weather-and-climate-barry-b-ma-brian-d-dushaw-and-bruce-m-howe/). Ocean technologies also rely on acoustics; active sonars are used to survey the seafloor and for sub-bottom profiling. Underwater communication is enabled via underwater acoustic modems leading to the need for developing improved underwater acoustic transmitters and sensors.

Many marine operations are being conducted by autonomous underwater vehicles (AUVs) that require advanced navigation and robotic operations. To support all these activities, underwater sound propagation must be modelled (https://acousticstoday.org/computational-acoustics-in-oceanography-the-research-roles-of-sound-field-simulations-timothy-f-duda-julien-bonnel-emanuel-coelho-and-kevin-d-heaney/), and the interactions between the sound and the seafloor need to be understood (https://acousticstoday.org/the-acoustics-of-marine-sediments-by-megan-s-ballard-and-kevin-m-lee/). Advanced digital signal processing techniques (https://acousticstoday.org/model-based-ocean-acoustic-signal-processing-edmund-j-sullivan/) have been developed to filter and enhance data from arrays of hydrophone to acoustically speaking; ’find a needle in a haystack’ as they identify quiet noise sources amid the cacophony of noise in the ocean (https://acousticstoday.org/physics-based-signal-processing-approaches-underwater-acoustic-sensing-lisa-m-zurk/). The detecting, localising and tracking of sound sources actively seeks to ensure the safety of the seas and harbours.

Entry level information
Although there is great need for experienced researchers in all of these areas, limited opportunities exist for students to get formal training specifically in underwater acoustics. There are several resources which can be of great benefit to those who are learning about underwater acoustics though – entry level information on many topics is nicely organised on the Discovery of Sound in the Sea (DOSITS) website (dosits.org). The DOSITS project continues to provide reliable information on underwater acoustics for a wide variety of audiences, from students to policy and decision-makers at environmental regulatory agencies. (Morin, 2016) DOSITS hosts webinar that hundreds of people watch live or view later on the webpage (https://dosits.org/decision-makers/webinar-series/). DOSITS also have an extensive audio gallery with examples of hundreds of sounds recorded in the sea (https://dosits.org/galleries/audio-gallery/). Additional marine sounds are available from the Ocean Conservation Research group at https://ocr.org/sound-library/ and a list of labeled datasets of ocean sounds is hosted by the UKAN at https://acoustics.ac.uk/open-access-underwater-acoustics-data/

Online tutorials
For those wishing for more advanced examples of how to model and analyse ocean sounds, several open-source resources are available such as the Ocean Acoustics Library (OALIB), which is supported by the US Office of Naval Research (https://oalib-acoustics.org). OALIB provides links to sound propagation models based on rays, normal modes, the parabolic equation, wavenumber integration and benchmarks for the latest 3D sound propagation models. Recent ONR-funded efforts to increase exposure and training in ocean acoustics have led to an incredible collection of online tutorials, two examples are Ocean Hack Week (https://oceanhackweek.org/about/pasthackweeks.html) and the Bridge to Ocean Acoustics and Technology (BOAT) workshops (https://boat-fundamentals.readthedocs.io/en/latest/landing.html).

Many GitHub repositories are available that can be used for modeling and processing ocean acoustics data, one of these was created for a special session I co-chaired for the Acoustical Society of America’s virtual meeting in November 2024. our experts in key fields of ocean acoustics were asked to contribute video descriptions and a Jupyter notebook to be hosted in an open GitHub repository (https://github.com/tbneilsen/underwater-acoustics-data-processing). John Ragland, University of Washington, provided a Jupyter notebook (https://github.com/tbneilsen/underwater-acoustics-data-processing/tree/main/noise_statistics_with_OOI_hydrophones) and a video at (https://www.youtube.com/watch?v=RVE_0-kFffE) showing how to access OOI data and calculate the noise statistics. As an example of active sonar, Wu-Jung Lee, Applied Physics Laboratory (UW-APL), University of Washington, shared an example of Echopype (https://github.com/OSOceanAcoustics/echopype-examples), which is an open-source Python library for processing echosounder data in a scalable and interoperable approach.

Kathleen Wage, George Washington University, presented a tutorial at (https://github.com/tbneilsen/underwater-acoustics-data-processing/tree/main/freq_wavenumber_tutorial) on frequency-wavenumber array processing. The final contribution was a detailed explanation (https://www.youtube.com/watch?v=- M_1YkVI5F8) and description (https://github.com/tbneilsen/underwater-acoustics-data-processing/tree/main/das_tutorial) of how to retrieve and process fibre optic data for distributed acoustic sensing provided by TJ Flynn, Johns Hopkins University Applied Physics Laboratory Lab.

Below: Figure 1. Schematic of many noise sources in the ocean (Credit: NOAA Fisheries)

Sonar challenges
A large percentage of work in ocean acoustics falls into two categories:
• active sonar; and
• passive sonar.
Active sonar refers to scenarios in which a known sound is purposefully generated and the reflected sound is recorded, similar to echolocation. One example of active sonar is the multibeam echosounder illustrated in Figure 1. Passive sonar, on the other hand, deals with sounds that are being made by unknown or uncontrolled sound sources and recorded on underwater microphones (hydrophones). My student-centred research group is studying a passive sonar application in which characteristics of the seabed are inferred from the sounds of transiting cargo ships. Specifically, we are training deep learning (DL) algorithms to identify the type of seabed in the area where a cargo ship is traveling and one or more hydrophones are listening. The input data samples for the deep learning algorithm are spectrograms of the shipping noise.

Above: Figure 2. Synthetic spectrogram of shipping noise for an area with 75m water and seafloor covered with sand or clay, generated using the Wales-Heitmeyer source spectrum (Wales, 2002) and the range-independent normal mode model ORCA (Westwood, 1996)

Several challenges exist when applying DL to passive sonar applications. First, supervised training of deep learning neural networks requires labeled data. In my application, this requires a label for the seabed class associated with each ship noise, because, in general, the seabed class is not known. My group is using synthetic ship noise spectrograms over the cavitation noise from ships as the training and validation data (Van Komen, 2021; Escobar, 2021). An example of the synthetic ship noise spectrograms generated for the same ship but with different seabed types are shown in Figure 2 to illustrate the impact of the seabed on the broadband sound propagation. After the DL models are trained and validated on the synthetic spectrograms, the generalisability of the trained models is demonstrated when they are applied to measured ship noise spectrograms to obtain an effective seabed type.

A second challenge is how to evaluate the generalisation results when the correct answer is unknown. Accuracy cannot be used, so a statistical representation of the results is needed to obtain a measure of the precision – to obtain the statistical distribution of results, an ensemble learning approach is used in which the results from multiple trained models are combined. The precision of the aggregated results can be obtained using the information entropy of the distribution and thus, a measure of uncertainty is obtained (Lau, 2025).

Sliding in to underwater acoustics
Additional challenges in ocean acoustics applications of DL include the temporal and spatial variability of the sound speed in the water and interference from additional noises, both natural and anthropogenic, some of which are illustrated in Figure 3. Ongoing work seeks to tackle these and many additional challenges in finding robust passive sonar deep learning approaches to seabed characterisation. Now that I have shared a bit about my research, I return to where this column began: How did I find a career in underwater acoustics? In short, I slid into it, departed for a while and then dived in again! As a graduate student pursuing a PhD in physics at the University of Texas (UT) at Austin, I decided to study acoustics because I enjoyed the fundamental physical principles and felt that acoustics would provide me with a broad range of job possibilities. I then contacted faculty members with research positions in acoustics, one involved highway noise barriers and the other underwater acoustics. It was an easy decision for me and I began learning about underwater acoustics, signal processing, sound propagation modeling and optimisations with Dr Evan Westwood at the Applied Research Laboratory, UT at Austin.

After receiving my PhD and completing a part-time postdoc, I took a break from research while raising my three children. My husband became a faculty member in the Department of Physics and Astronomy at Brigham Young University and I taught a few classes each year at the university as an adjunct instructor. Once our children were all in school, I did part-time research with other acoustics faculty members and had the opportunity to apply sound propagation modeling and optimisation work to jet aircraft noise. When our youngest child was entering high school, I decided to apply for a full-time faculty position and in May 2018, I began that position and dived back into underwater acoustics. I observed the progress that had been made in my absence and understood that I did not have time to catch up but needed to begin my research group with the current hot topic: machine learning. My post-doctoral advisor, Dr David Knobles, invited me to participate on a grant and helped me get re-established, the underwater acoustics community was welcoming and the transition has been better than I ever expected.

I have particularly enjoyed the opportunity to connect with scientists and engineers throughout the world as we apply acoustics to study our amazing oceans. One such venue is the International Conference on Underwater Acoustics (ICUA), sponsored by IOA. I hope everyone takes full advantage of networking opportunities at such conferences (https://acousticstoday.org/home/networking-up-tracianne-b-neilsen/). My attendance at ICUA in 2022 and 2024, paved the way for my six-month sabbatical at the University of Southampton and the National Oceanographic Center. I will have returned to Utah by the time this article is published, but I am truly grateful for the hospitality shown by so many as I visited different UK universities – I am pursuing extended collaborations with several groups and looking forward to visiting again. In sharing this personal narrative, I hope it may make room for squiggly careers (Tupper, 2020). I also hope that it motivates people to pursue or to share information about careers in underwater acoustics. By working together, we can help achieve the goal of the UN Decade of Ocean Science to find “the science we need for the ocean we want.” (https://oceandecade.org/)

Below: Figure 3. Visualization of how multibeam sonar is used to map the seafloor. (By NOAA’s National Ocean Service – Flickr: Collecting Multibeam Sonar Data (Original source: National Ocean Service Image Gallery), Public Domain, https://commons.wikimedia.org/w/index.php?curid=30679792)

References

DOSiTs: Morin, Holly, Kathleen J. Vigness-Raposa, Christopher Knowlton, Gail Scowcroft, James H. Miller, Darlene R. Ketten, and Arthur N. Popper. “Implementing multiple digital platforms to effectively communicate research on underwater acoustics.” In Proceedings of Meetings on Acoustics, vol. 27, no. 1. AIP Publishing, 2016. https://doi.org/10.1121/2.0000255

Escobar-Amado, C. D., Neilsen, T. B., Castro-Correa, J. A., Van Komen, D. F., Badiey, M., Knobles, D. P., & Hodgkiss, W. S. (2021). Seabed classification from merchant ship-radiated noise using a physics-based ensemble of deep learning algorithms. The Journal of the Acoustical Society of America, 150(2), 1434-1447.

Lau, G. E., Mortenson, M. C., Neilsen, T. B., Van Komen, D. F., Hodgkiss, W. S., & Knobles, D. P. (2025). Ensemble approach to deep learning seabed classification using multichannel ship noise. The Journal of the Acoustical Society of America, 157(3), 2127-2149.

Tupper, H., & Ellis, S. (2020). The squiggly career: Ditch the ladder, discover opportunity, design your career. Penguin Business.

Van Komen, D. F., Neilsen, T. B., Mortenson, D. B., Acree, M. C., Knobles, D. P., Badiey, M., & Hodgkiss, W. S. (2021). Seabed type and source parameters predictions using ship spectrograms in convolutional neural networks. The Journal of the Acoustical Society of America, 149(2), 1198-1210.

Wales, Stephen C., and Richard M. Heitmeyer. “An ensemble source spectra model for merchant ship-radiated noise.” The Journal of the Acoustical Society of America 111, no. 3 (2002): 1211-1231.

Westwood, Evan K., Chris T. Tindle, and N. Ross Chapman. “A normal mode model for acousto‐elastic ocean environments.” The Journal of the Acoustical Society of America 100, no. 6 (1996): 3631-3645.