Using artificial neural networks to combine acoustic and trawl data in the Barents and North Seas
Groundfish have a wide and variable distribution making the use of trawling alone a highly inadequate sampling method. Trawl data provide species identification and numbers over a very small area and habitat type while acoustic data provide a wider coverage of the ecosystem, but fail to identify spe...
Main Authors: | , , , |
---|---|
Format: | Report |
Language: | English |
Published: |
ICES
2004
|
Subjects: | |
Online Access: | http://hdl.handle.net/11250/100719 |
id |
ftimr:oai:imr.brage.unit.no:11250/100719 |
---|---|
record_format |
openpolar |
spelling |
ftimr:oai:imr.brage.unit.no:11250/100719 2023-05-15T15:38:57+02:00 Using artificial neural networks to combine acoustic and trawl data in the Barents and North Seas Neville, Suzanna Hjellvik, Vidar Mackinson, Steven Kooij, Jeroen van der 2004 445932 bytes application/pdf http://hdl.handle.net/11250/100719 eng eng ICES ICES CM documents 2004/R:05 http://hdl.handle.net/11250/100719 20 s. acoustics trawl stock assessment VDP::Agriculture and fishery disciplines: 900::Fisheries science: 920::Fisheries technology: 924 Working paper 2004 ftimr 2021-09-23T20:15:55Z Groundfish have a wide and variable distribution making the use of trawling alone a highly inadequate sampling method. Trawl data provide species identification and numbers over a very small area and habitat type while acoustic data provide a wider coverage of the ecosystem, but fail to identify species. Both methods provide essential information required for assessing fish stock abundance and distribution, but no systematic method of combining these data has yet been identified. Acoustic and trawl data were collected for both the North and Barents Seas and their relationships investigated using artificial neural networks (ANNs). ANNs are information processing systems that were inspired by the structure of the brain. They can incorporate multiple variables and explore complex interactions between variables in far greater depth than many traditional statistical techniques. This modelling tool has had many successes in environmental modelling and is increasingly being applied in situations where the underlying relationships are poorly known. Network architectures, optimisation of connection weights (learning), model validation, and the reasons for the difference in performance between the North Sea and Barents Sea models are discussed. Report Barents Sea Institute for Marine Research: Brage IMR Barents Sea |
institution |
Open Polar |
collection |
Institute for Marine Research: Brage IMR |
op_collection_id |
ftimr |
language |
English |
topic |
acoustics trawl stock assessment VDP::Agriculture and fishery disciplines: 900::Fisheries science: 920::Fisheries technology: 924 |
spellingShingle |
acoustics trawl stock assessment VDP::Agriculture and fishery disciplines: 900::Fisheries science: 920::Fisheries technology: 924 Neville, Suzanna Hjellvik, Vidar Mackinson, Steven Kooij, Jeroen van der Using artificial neural networks to combine acoustic and trawl data in the Barents and North Seas |
topic_facet |
acoustics trawl stock assessment VDP::Agriculture and fishery disciplines: 900::Fisheries science: 920::Fisheries technology: 924 |
description |
Groundfish have a wide and variable distribution making the use of trawling alone a highly inadequate sampling method. Trawl data provide species identification and numbers over a very small area and habitat type while acoustic data provide a wider coverage of the ecosystem, but fail to identify species. Both methods provide essential information required for assessing fish stock abundance and distribution, but no systematic method of combining these data has yet been identified. Acoustic and trawl data were collected for both the North and Barents Seas and their relationships investigated using artificial neural networks (ANNs). ANNs are information processing systems that were inspired by the structure of the brain. They can incorporate multiple variables and explore complex interactions between variables in far greater depth than many traditional statistical techniques. This modelling tool has had many successes in environmental modelling and is increasingly being applied in situations where the underlying relationships are poorly known. Network architectures, optimisation of connection weights (learning), model validation, and the reasons for the difference in performance between the North Sea and Barents Sea models are discussed. |
format |
Report |
author |
Neville, Suzanna Hjellvik, Vidar Mackinson, Steven Kooij, Jeroen van der |
author_facet |
Neville, Suzanna Hjellvik, Vidar Mackinson, Steven Kooij, Jeroen van der |
author_sort |
Neville, Suzanna |
title |
Using artificial neural networks to combine acoustic and trawl data in the Barents and North Seas |
title_short |
Using artificial neural networks to combine acoustic and trawl data in the Barents and North Seas |
title_full |
Using artificial neural networks to combine acoustic and trawl data in the Barents and North Seas |
title_fullStr |
Using artificial neural networks to combine acoustic and trawl data in the Barents and North Seas |
title_full_unstemmed |
Using artificial neural networks to combine acoustic and trawl data in the Barents and North Seas |
title_sort |
using artificial neural networks to combine acoustic and trawl data in the barents and north seas |
publisher |
ICES |
publishDate |
2004 |
url |
http://hdl.handle.net/11250/100719 |
geographic |
Barents Sea |
geographic_facet |
Barents Sea |
genre |
Barents Sea |
genre_facet |
Barents Sea |
op_source |
20 s. |
op_relation |
ICES CM documents 2004/R:05 http://hdl.handle.net/11250/100719 |
_version_ |
1766370390456664064 |