Subscribe to Holyrood updates

Newsletter sign-up


Follow us

Scotland’s fortnightly political & current affairs magazine


Subscribe to Holyrood
by Sofia Villegas
25 January 2024
AI-powered tech could tackle underwater noise pollution

An offshore wind farm in Aberdeen | Alamy

AI-powered tech could tackle underwater noise pollution

A Scottish university has developed a new AI-powered system with the potential to tackle the impact of noise pollution on marine life.  

The system can accurately model how sound waves travel underwater, which has far-reaching implications for sea life.  

Researchers from the University of Glasgow partnered with the University of British Columbia in Canada to make this “significant breakthrough”, which could dictate new regulations.  

Loud sounds created by cargo ship propellers and offshore wind farms can disrupt the migration patterns of marine mammals and affect their ability to navigate by echolocation.  

But predicting precisely how this noise pollution travels, and therefore being able to limit its impact, is complicated by how sound waves reflect off the surface of the sea and other underwater surfaces. 

To date, users require large amounts of computer processing power to create accurate models, with larger-scale projects taking days to complete.  

The new system, called the CRAN, simplifies high-dimensional modelling data for the AI model to later analyse, based on its prior knowledge of underwater physics, and predict the propagation of underwater sound waves over time. It uses deep neural networks to provide live feedback on the propagation of sound waves. 

By working from a simplified model and expanding it using machine learning, the system provides results much faster than conventional modelling processes. 

This innovative technology could allow for a better understanding of how sound waves move underwater, helping policymakers to mitigate the impact of shipping and offshore turbine construction on sea life.  

Wrik Mallik, author of the research paper, said: “Waiting seconds instead of days to produce models of underwater acoustic scattering would be a significant breakthrough for this field of research, and this paper shows how we’ve taken one step closer to making that happen. 

“Having real-time feedback on devices which could be used out on the ocean would allow much more effective planning to help mitigate the effects of noise pollution on marine animals.”   

To train the system, researchers created 30 different two-dimensional simulations of underwater environments to help it learn the physics of underwater sound waves. 

After the training stage, they asked the CRAN to predict how the sound waves would behave in 15 new underwater scenarios, and the model accurately predicted wave propagation with less than 10 per cent error for a duration more than five times longer than the data it was trained on. 

Mallik added: “Although this early-stage study demonstrated the effectiveness of the CRAN on two-dimensional data, we’re confident that the technology can be scaled up to meet the challenge of dealing with fully 3D acoustic simulations. We’ve already begun work to further develop and refine the system, and we plan to test it in real-world situations in the months ahead.”  

The team’s paper, titled ‘Deep neural network for learning wave scattering and interference of underwater acoustics’, has now been published in the scientific journal Physics of Fluids. 

The research was primarily conducted using Cirrus UK National Tier-2 High Performance Computing Service at EPCC and was partly funded by the Natural Science and Engineering Research Council of Canada.

Holyrood Newsletters

Holyrood provides comprehensive coverage of Scottish politics, offering award-winning reporting and analysis: Subscribe

Read the most recent article written by Sofia Villegas - TikTok fined £1.8m after providing Ofcom with inaccurate information.

Get award-winning journalism delivered straight to your inbox

Get award-winning journalism delivered straight to your inbox


Popular reads
Back to top