Scholarlink

<p>Understanding the spatiotemporal distributions of species interacting with fisheries is essential for their conservation and management, yet modeling wide-ranging, highly mobile and data-limited species presents significant challenges. This study applies three Species Distribution Modeling (SDM) approaches—Generalized Additive Mixed Models (GAMMs), Boosted Regression Trees (BRTs) and Bayesian Additive Regression Trees (BARTs)—to investigate the drivers influencing the occurrence of silky shark (<i>Carcharhinus falciformis</i>) and predict its distribution in the eastern tropical Atlantic Ocean. Models were built using European Union tropical tuna purse seine fishery observer data from 2010 to 2023. All three approaches showed good predictive performance (AUC &gt; 0.8) and identified consistent predictors. Fishery-related variables (set type, teleost catch) and environmental variables (sea surface temperature and sea surface height) were retained in all models, suggesting their role as key habitat drivers. Ensemble predictions highlighted four high-probability areas of silky shark occurrence: coastal waters off Guinea, Gabon, and Angola, and the offshore region of the southern and central tropical Atlantic (5–15° S, 30°–10° W). High occurrence probabilities coincided with regions of high primary productivity, such as the Guinea Thermal Dome, Gabon-Angola coastal upwelling, and equatorial upwelling zone, particularly from May to October. Integrating multiple modeling approaches strengthened confidence in spatial predictions by accounting for inter-model uncertainty. While these predictions can inform spatially-explicit ecological risk assessments and management strategies, further refinements—such as integrating longline and other fisheries data, incorporating fisheries-independent datasets (e.g., tagging), and extending to multispecies frameworks—will be essential to improve model validation and support integrated bycatch management.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Predicting silky shark distribution in the eastern tropical Atlantic Ocean using fisheries dependent data

  • 期刊论文

  • 发表日期：

    2026年3月2日

查看全文
  • Leire Lopetegui-Eguren,
  • Haritz Arrizabalaga,
  • Hilario Murua,
  • Nerea Lezama-Ochoa,
  • Shane Griffiths,
  • Jon Lopez,
  • Jon Ruiz Gondra,
  • Philippe S. Sabarros,
  • María Lourdes Ramos Alonso,
  • Maria José Juan-Jordá

摘要

Understanding the spatiotemporal distributions of species interacting with fisheries is essential for their conservation and management, yet modeling wide-ranging, highly mobile and data-limited species presents significant challenges. This study applies three Species Distribution Modeling (SDM) approaches—Generalized Additive Mixed Models (GAMMs), Boosted Regression Trees (BRTs) and Bayesian Additive Regression Trees (BARTs)—to investigate the drivers influencing the occurrence of silky shark (Carcharhinus falciformis) and predict its distribution in the eastern tropical Atlantic Ocean. Models were built using European Union tropical tuna purse seine fishery observer data from 2010 to 2023. All three approaches showed good predictive performance (AUC > 0.8) and identified consistent predictors. Fishery-related variables (set type, teleost catch) and environmental variables (sea surface temperature and sea surface height) were retained in all models, suggesting their role as key habitat drivers. Ensemble predictions highlighted four high-probability areas of silky shark occurrence: coastal waters off Guinea, Gabon, and Angola, and the offshore region of the southern and central tropical Atlantic (5–15° S, 30°–10° W). High occurrence probabilities coincided with regions of high primary productivity, such as the Guinea Thermal Dome, Gabon-Angola coastal upwelling, and equatorial upwelling zone, particularly from May to October. Integrating multiple modeling approaches strengthened confidence in spatial predictions by accounting for inter-model uncertainty. While these predictions can inform spatially-explicit ecological risk assessments and management strategies, further refinements—such as integrating longline and other fisheries data, incorporating fisheries-independent datasets (e.g., tagging), and extending to multispecies frameworks—will be essential to improve model validation and support integrated bycatch management.