CRIMAC: Centre for research based innovation in marine acoustic abundance estimation and backscatter classification
Echosounders are important tools for finding how much and what kind of fish are in the sea. Fishermen want to trawl on schools of fish of the right species and size. Marine scientists want to map which resources of fish and plankton are found in the sea. In both cases, echosounders provide useful information about what is in the water masses under the boat. This applies in particular to new types of broadband echosounders that can measure a continuous frequency spectrum. This type of broadband data gives opportunities for new methods that provide increased insight. This is a main focus for CRIMAC.
CRIMAC is an SFI center that aims to develop scientific methods for understanding and utilizing data from broadband echosounders. NORCE has relevant knowledge in acoustics, mathematical methods and data science and contributes with both theoretical method development and practical data processing. NORCE has also developed the software LSSS in collaboration with the Norwegian Institute of Marine Research, which is used to interpret acoustic data.
It is the swim bladder of fish that gives the greatest response when measuring with echosounders. For species without a swim bladder (such as sand eel and mackerel), the response is given by the body/tissue of the fish. Once the acoustic dispersion from the fish/swim bladder is understood, it is possible to say something about the characteristics of the fish based on acoustic measurements. The shape of a fish/swim bladder can, for example, be determined using ultrasound, and the corresponding scattering problem can then be solved numerically. Alternatively, a simpler model of fish/swim bladder can be defined which can be studied analytically. Both methods have advantages and disadvantages. NORCE contributes with the development of analytical methods where the fish's swim bladder is modelled as a prolate spheroid, i.e., a rotational ellipsoid which has been arrived at by rotating an ellipse around its long axis. The figures below show a sand eel, an ellipse and a prolate spheroid.
The data from the echosounder should be of the best possible quality so that the results of the analyzes carried out are as correct as possible. However, there are often different types of noise that affect the measurements. This can be noise from ships and waves or signals from other measuring instruments. NORCE contributes to the development of methods for the detection and removal of narrowband noise. This is a type of noise that occurs on a few individual frequencies, and which can affect the frequency spectrum of broadband data. In this case, one need methods to find which frequencies contain noise, so that one can afterwards remove the noise or simply ignore these frequencies.
NORCE also contributes with the development of automated processing and analysis of acoustic data. This is becoming increasingly relevant and also necessary due to the increased use of buoys and other autonomous vessels equipped with echo sounders and other measuring instruments. Then the amount of data increases so that it is not practically possible to analyze all the data manually. Therefore, an automatic processing is needed that starts with raw data from the, performs the relevant analyzes and finally ends up with reports that can more easily be use further.
For more information about CRIMAC, see https://crimac.no/.
Project facts
Name
CRIMAC: Centre for research based innovation in marine acoustic abundance estimation and backscatter classification
Status
Active
Duration
01.12.20 - 31.12.27
Total budget
214.726.000 NOK
Project website
Publications
Examples of AI and Machine Learning methods applied to data from marine ecosystems – Handegard, Nils Olav. 2024, . HAVFORSK.
Identifying size and behavior of Mueller’s Pearlside (Maurolicus muelleri) by combining measured broadband frequency responses with backscattering models and in-trawl observations – Tenningen, Maria; Pedersen, Geir; Khodabandeloo, Babak; Handegard, Nils Olav; Forland, Tonje Nesse; Klevjer, Thor A.; Johnsen, Espen. 2024, . HAVFORSK, UIB.
Analysing large amounts of echosounder data using cloud based data access combined with deep learning models – Handegard, Nils Olav; Bildøy, Leif; Brautaset, Olav; Furmanek, Tomasz; Holmin, Arne Johannes; Pala, Ahmet; Utseth, Ingrid; Malde, Ketil. 2024, . HAVFORSK, NR, UIB.
See all publications in Cristin
