Application de méthodes d'apprentissage statistique à la reconnaissance des opérations de pêche des fileyeurs

Since 2016, strandings of small cetaceans showing signs of capture have reached significant levels, which could question the viability of the North Atlantic common dolphin population (ICES 2022). It is against this backdrop that CNRS and Ifremer, in conjunction with the OFB, have co-constructed the...

Full description

Bibliographic Details
Main Author: Sans, Mathurin
Format: Report
Language:French
Published: 2023
Subjects:
Online Access:https://archimer.ifremer.fr/doc/00851/96275/104520.pdf
https://archimer.ifremer.fr/doc/00851/96275/
id ftarchimer:oai:archimer.ifremer.fr:96275
record_format openpolar
spelling ftarchimer:oai:archimer.ifremer.fr:96275 2023-10-01T03:58:04+02:00 Application de méthodes d'apprentissage statistique à la reconnaissance des opérations de pêche des fileyeurs Sans, Mathurin 2023-07-14 application/pdf https://archimer.ifremer.fr/doc/00851/96275/104520.pdf https://archimer.ifremer.fr/doc/00851/96275/ fre fre https://archimer.ifremer.fr/doc/00851/96275/104520.pdf https://archimer.ifremer.fr/doc/00851/96275/ info:eu-repo/semantics/openAccess restricted use text Report info:eu-repo/semantics/report 2023 ftarchimer 2023-09-05T22:51:06Z Since 2016, strandings of small cetaceans showing signs of capture have reached significant levels, which could question the viability of the North Atlantic common dolphin population (ICES 2022). It is against this backdrop that CNRS and Ifremer, in conjunction with the OFB, have co-constructed the Delmoges program (Delphinus Mouvement Gestion), which is developing a multidisciplinary scientific approach aimed at gaining a better understanding of the mechanisms involved in the accidental capture of dolphins. In particular, developments are being proposed to provide a more detailed description of the activities of fishing fleets deemed to be most at risk of bycatching cetaceans. The qualification of fishing routes is currently based on simple decision rules using speed thresholds, so this study focuses on other qualification methods that have proved highly effective in other work: machine learning. Using the qualified positions and fishing operations of the 20 gillnetters in the OBSCAME database as a training set, we trained different models (CART, Random Forest, XGBoost) to predict a gillnetter's fishing operations based on its positions. The XGBoost models give the best results, with almost 90% accuracy using fishing trips based cross-validation. The final step, which is still incomplete at the moment, is to model the nets of the vessels in the dataset based on the predictions, with the aim of assessing the fishing effort of Bay of Biscay gillnetters using new gear metrics. Depuis 2016, les échouages de petits cétacés présentant des traces de capture ont atteint des niveaux importants, qui pourraient remettre en question la viabilité de la population de dauphins communs de l'Atlantique Nord (ICES 2022). C'est dans ce contexte que le CNRS et Ifremer en concertation avec l'OFB ont co-construit le programme Delmoges (Delphinus Mouvement Gestion) qui développe une approche scientifique multidisciplinaire visant à une meilleure compréhension des mécanismes de captures accidentelles de dauphins. Des développements ... Report North Atlantic Archimer (Archive Institutionnelle de l'Ifremer - Institut français de recherche pour l'exploitation de la mer) Dauphins ENVELOPE(141.589,141.589,-66.771,-66.771)
institution Open Polar
collection Archimer (Archive Institutionnelle de l'Ifremer - Institut français de recherche pour l'exploitation de la mer)
op_collection_id ftarchimer
language French
description Since 2016, strandings of small cetaceans showing signs of capture have reached significant levels, which could question the viability of the North Atlantic common dolphin population (ICES 2022). It is against this backdrop that CNRS and Ifremer, in conjunction with the OFB, have co-constructed the Delmoges program (Delphinus Mouvement Gestion), which is developing a multidisciplinary scientific approach aimed at gaining a better understanding of the mechanisms involved in the accidental capture of dolphins. In particular, developments are being proposed to provide a more detailed description of the activities of fishing fleets deemed to be most at risk of bycatching cetaceans. The qualification of fishing routes is currently based on simple decision rules using speed thresholds, so this study focuses on other qualification methods that have proved highly effective in other work: machine learning. Using the qualified positions and fishing operations of the 20 gillnetters in the OBSCAME database as a training set, we trained different models (CART, Random Forest, XGBoost) to predict a gillnetter's fishing operations based on its positions. The XGBoost models give the best results, with almost 90% accuracy using fishing trips based cross-validation. The final step, which is still incomplete at the moment, is to model the nets of the vessels in the dataset based on the predictions, with the aim of assessing the fishing effort of Bay of Biscay gillnetters using new gear metrics. Depuis 2016, les échouages de petits cétacés présentant des traces de capture ont atteint des niveaux importants, qui pourraient remettre en question la viabilité de la population de dauphins communs de l'Atlantique Nord (ICES 2022). C'est dans ce contexte que le CNRS et Ifremer en concertation avec l'OFB ont co-construit le programme Delmoges (Delphinus Mouvement Gestion) qui développe une approche scientifique multidisciplinaire visant à une meilleure compréhension des mécanismes de captures accidentelles de dauphins. Des développements ...
format Report
author Sans, Mathurin
spellingShingle Sans, Mathurin
Application de méthodes d'apprentissage statistique à la reconnaissance des opérations de pêche des fileyeurs
author_facet Sans, Mathurin
author_sort Sans, Mathurin
title Application de méthodes d'apprentissage statistique à la reconnaissance des opérations de pêche des fileyeurs
title_short Application de méthodes d'apprentissage statistique à la reconnaissance des opérations de pêche des fileyeurs
title_full Application de méthodes d'apprentissage statistique à la reconnaissance des opérations de pêche des fileyeurs
title_fullStr Application de méthodes d'apprentissage statistique à la reconnaissance des opérations de pêche des fileyeurs
title_full_unstemmed Application de méthodes d'apprentissage statistique à la reconnaissance des opérations de pêche des fileyeurs
title_sort application de méthodes d'apprentissage statistique à la reconnaissance des opérations de pêche des fileyeurs
publishDate 2023
url https://archimer.ifremer.fr/doc/00851/96275/104520.pdf
https://archimer.ifremer.fr/doc/00851/96275/
long_lat ENVELOPE(141.589,141.589,-66.771,-66.771)
geographic Dauphins
geographic_facet Dauphins
genre North Atlantic
genre_facet North Atlantic
op_relation https://archimer.ifremer.fr/doc/00851/96275/104520.pdf
https://archimer.ifremer.fr/doc/00851/96275/
op_rights info:eu-repo/semantics/openAccess
restricted use
_version_ 1778530472845901824