Our oceans are increasingly noisy due to increasing anthropogenic activities, particularly shipping. This noisiness impacts marine mammals and some fish species who rely on sound to communicate, find prey, and steer clear of hazards and predators. Anthropogenic noise sources obscure these biologically important sounds over large areas of the marine environment and can cause behavioural and physiological harm, including impacts on fish stocks. It is important to understand sound in the oceans as changes in sound levels can be used to monitor leaks from industrial infrastructure (pipelines and rigs), and from carbon capture and storage projects.
Marine autonomous vehicles are being increasingly used to sense and understand the oceans. Some of these vehicles are equipped with Passive Acoustic Monitoring (PAM) devices that can be configured to record a large bandwidth of acoustic frequency facilitating a high-fidelity and complete record of the marine soundscape. Interrogating the vast datasets that are recorded by fleets of autonomous data is a current challenge. This project will apply and further develop machine-learning techniques to discriminate between different sources of sound, and design and implement systems on board the autonomous platforms that can identify and communicate that particular sound sources are present in the environment.
This project will determine the relative intensities of natural and anthropogenic sounds in the marine environment using data from acoustic recorders mounted on autonomous systems and fixed buoys, with data available from both the Atlantic and Southern Ocean. A particular focus for this project will be data collected using Gliders - underwater autonomous systems that use an internal pump to change their buoyancy to move up and down in the water. These systems are intrinsically quiet, and so are platforms well suited for recording sounds in the oceans. The PhD student will go to sea to collect data, and process and interpret this data.
The student will apply and further develop existing software tools for analysing large acoustic datasets. After training, the student will apply machine learning techniques to enable discrimination of sounds from different sources (biological, seabed seeps, industrial infrastructure, shipping, hydrocarbon exploration). We shall exploit supervised learning techniques to manually label a subset of the data for training purposes. The student will also consider methods for determining the location of sounds sources using multiple receivers on a mobile platform. Methods for flagging with confidence particular sound sources within data sets will be developed, and implemented on an autonomous system.
The INSPIRE DTP programme provides comprehensive personal and professional development training alongside extensive opportunities for students to expand their multi-disciplinary outlook through interactions with a wide network of academic, research and The student will be registered at the University of Southampton and hosted at the School of Ocean and Earth Science at the National Oceanography Centre Southampton. Specific training will include attendance at courses in the Faculty of Engineering and Physical Sciences on marine acoustics, signal processing and/or machine learning. In addition, there are a wide range of masters level modules available in Oceanography at NOCS, and it would be expected that some of these modules would be taken, depending on the background of the candidate (e.g. Physical Oceanography, Marine Geology and Geophysics). The PhD student will benefit from Southampton’s membership of the Turing Institute which is a national focus of data science and artificial intelligence industrial/policy partners.
Please see https://inspire-dtp.ac.uk/how-apply for details.
- Solan, M., Hauton, C., Godbold , J.A., Wood, C.L., Leighton, T.G. and White, P.,Anthropogenic sources of underwater sound can modify how sediment-dwelling invertebrates mediate ecosystem properties, Scientific Reports, 6, 1-24, 2016, (doi:10.1038/srep20540).
- Gruden, P. and White, P. R., Automated extraction of dolphin whistles – a sequential Monte Carlo probability hypothesis density approach. The Journal of the Acoustical Society of America 148, 3014 (2020); https://doi.org/10.1121/10.0002257
- Zhang, Z, and White, P.R. A blind source separation approach for humpback whale song separation. The Journal of the Acoustical Society of America 141(4), 2017: 2705-2714.(doi:10.1121/1.4980856)