In her PhD thesis, Clea Parcerisas (VLIZ, UGent) focused on developing automated methods for analysing and characterising the totality of underwater sounds in shallow and heavily exploited coastal areas. In order to use this so-called soundscapes – caused by both nature and human activities – as an early warning signal or 'canary' when monitoring noise pollution and other human impacts in the sea.
Sound travels further underwater than in the air, and many marine animals use it to obtain information about their surroundings and to communicate. Characterising the totality of underwater sounds (the 'soundscape') can teach us a lot about the composition and state of marine ecosystems. This can be done through the collection of sound data with 'passive acoustic monitoring' (PAM), a non-invasive technique. These data can then be used to monitor acoustic pollution and other human impacts over long periods, and to understand changes in ecosystem dynamics. To this aim, long-term recordings are necessary. However, manually analysing months and years of data is unfeasible, and scientists cannot keep up with the speed at which large volumes of data are collected. For this reason, automatic methods to process these data are necessary.
The thesis of Clea Parcerisas (VLIZ, UGent), presented on 28 June 2024 in Ostend, focuses on the development of automated methods to analyse and characterise soundscapes in shallow, coastal, and heavily exploited areas. In these marine areas – such as the Belgian part of the North Sea – there are several factors which make the analysis challenging, such as biofouling (undesirable fouling of organisms on equipment), flow-noise, masking and other factors that interfere with the propagation or reception of the sounds. The developed methods are tailored for the analysis of these areas, where little is known about their acoustic characteristics. These methods include supervised and unsupervised machine learning techniques, including the categorization of soundscapes, where different soundscape ‘types’ are found
and described, and deep learning techniques to disentangle specific sounds of interest from the long-term recordings.
The methods are tested in data collected under the LifeWatch framework, a new passive acoustic monitoring network. Finally, the researcher zooms in on how marine invertebrates use information from these soundscapes to decide where to settle. In a lab experiment with larvae of the Pacific oyster, Clea and her colleague Sarah Schmidlin demonstrate how these larvae take the soundscape into account in their choices. The results show that oyster reefs’ sounds trigger a higher settlement than sounds from sandy areas or vessel sounds.
Clea Parcerisas verdedigde haar doctoraat “Marine Soundscapes in Shallow Water: Automated Tools for Characterization and Analysis” op 28 juni op de InnovOcean Campus in Oostende. Promotoren van dit doctoraat zijn Prof. Dr. Dick Botteldooren (UGent), Dr. Elisabeth Debusschere (VLIZ) en Prof. Dr. Paul Devos (UGent).
Download the PhD-thesis of Clea Parcerisas via the VLIZ-bib >