{"title":"Using fishery-related data, scientific expertise, and machine learning to improve marine habitat mapping in northeastern Mediterranean waters","authors":"Loukas Katikas , Sofia Reizopoulou , Paraskevi Drakopoulou , Vassiliki Vassilopoulou","doi":"10.1016/j.ecoinf.2025.103154","DOIUrl":null,"url":null,"abstract":"<div><div>Marine habitat mapping is an essential tool for planning conservation efforts and sustainable management of marine activities. High spatial resolution in marine habitat maps is of utmost importance, as it may encompass more detail in imagery and reveal important biotopes. This level of detail supports directing monitoring and analysis efforts for effective implementation of European Union (EU) environmental policies and provides more relevant advice for robust decision-making under sectorial policies (e.g., the Common Fisheries Policy) and more integrated policies (e.g., marine spatial planning). In this study, sea bottom type data recorded during national monitoring of commercial fishing vessel operations and fishery surveys in the Greek Seas were used. These data were then assigned to the EU EMODnet seabed habitats using local ecological knowledge. Two machine-learning algorithms, i.e., random forest classifier (RFC) and gradient boosting classifier, were trained on the entire national-scale dataset and subsequently applied to assess their performance in predicting habitat types in the Saronikos Gulf (regional scale) using various environmental factors as predictors. The borderline synthetic minority oversampling technique was applied to manage inherent data class imbalances. A validation dataset and georeferenced data from previous studies were used to compare the accuracy and predictive performance of the models. Using this approach, the Saronikos Gulf was enriched with five more habitat types than visualised in the EMODnet portal, which also filled habitat gaps in areas where no data existed. Results from application of the RFC-Borderline Smote (BS) model (62 % accuracy, 0.51 kappa score) were then used to address conservation planning commitments recently made by the Greek government. The vast majority of marine seabed priority habitats in the study area appeared to fall outside the borders of the current Natura 2000 sites, which served as the baseline for the declared trawl bans in Greek waters, following the provisions of the EU Marine Action Plan.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"88 ","pages":"Article 103154"},"PeriodicalIF":5.8000,"publicationDate":"2025-04-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Ecological Informatics","FirstCategoryId":"93","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S1574954125001633","RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"ECOLOGY","Score":null,"Total":0}
引用次数: 0
Abstract
Marine habitat mapping is an essential tool for planning conservation efforts and sustainable management of marine activities. High spatial resolution in marine habitat maps is of utmost importance, as it may encompass more detail in imagery and reveal important biotopes. This level of detail supports directing monitoring and analysis efforts for effective implementation of European Union (EU) environmental policies and provides more relevant advice for robust decision-making under sectorial policies (e.g., the Common Fisheries Policy) and more integrated policies (e.g., marine spatial planning). In this study, sea bottom type data recorded during national monitoring of commercial fishing vessel operations and fishery surveys in the Greek Seas were used. These data were then assigned to the EU EMODnet seabed habitats using local ecological knowledge. Two machine-learning algorithms, i.e., random forest classifier (RFC) and gradient boosting classifier, were trained on the entire national-scale dataset and subsequently applied to assess their performance in predicting habitat types in the Saronikos Gulf (regional scale) using various environmental factors as predictors. The borderline synthetic minority oversampling technique was applied to manage inherent data class imbalances. A validation dataset and georeferenced data from previous studies were used to compare the accuracy and predictive performance of the models. Using this approach, the Saronikos Gulf was enriched with five more habitat types than visualised in the EMODnet portal, which also filled habitat gaps in areas where no data existed. Results from application of the RFC-Borderline Smote (BS) model (62 % accuracy, 0.51 kappa score) were then used to address conservation planning commitments recently made by the Greek government. The vast majority of marine seabed priority habitats in the study area appeared to fall outside the borders of the current Natura 2000 sites, which served as the baseline for the declared trawl bans in Greek waters, following the provisions of the EU Marine Action Plan.
期刊介绍:
The journal Ecological Informatics is devoted to the publication of high quality, peer-reviewed articles on all aspects of computational ecology, data science and biogeography. The scope of the journal takes into account the data-intensive nature of ecology, the growing capacity of information technology to access, harness and leverage complex data as well as the critical need for informing sustainable management in view of global environmental and climate change.
The nature of the journal is interdisciplinary at the crossover between ecology and informatics. It focuses on novel concepts and techniques for image- and genome-based monitoring and interpretation, sensor- and multimedia-based data acquisition, internet-based data archiving and sharing, data assimilation, modelling and prediction of ecological data.