HOME IS WHERE THE HABITAT IS: MODELING SHORTFIN MAKO HABITAT SUITABILITY VIA MACHINE LEARNING METHODS

• Main Author: Garrison, Julian
• Format: Text
• Language: unknown
• Published: DigitalCommons@URI 2023
• Subjects: North Atlantic
• Online Access: https://digitalcommons.uri.edu/theses/2380; https://digitalcommons.uri.edu/context/theses/article/3324/viewcontent/Garrison_uri_0186M_13150.pdf

Given the mounting threats of species overexploitation, climate change, and other anthropogenic stressors to global biodiversity, there is a growing need for conservation and management efforts informed by the life history and ecology of target species. Apex marine predators such as the shortfin mako shark (Isurus oxyrinchus) are especially vulnerable owing to their life history traits, but accurately mapping habitat preferences remains challenging. Using a novel framework that combines multiple analytical techniques, I report on nearly a decade of habitat preferences of 106 shortfin makos in the Gulf of Mexico (GoM) and western North Atlantic Ocean (NAO). I leverage the predictive power of machine learning (ML) to generate region-specific habitat suitability models based on satellite telemetry and remote sensed environmental data. Ensemble-based models performed best in predicting shortfin mako habitat suitability, and variables indicating coastal proximity were consistently the most important for model predictions at broad scales. In the GoM, sharks concentrated their residency behaviors around the Yucatán Peninsula during the late winter and early spring but expanded home ranges to include much of the GoM during the summer. In contrast, NAO sharks concentrated their residency behaviors off the northeastern U.S. coast during the summer, whereas winter habitats were more diffuse and located further south along the U.S. East Coast and in the open western NAO. Predicted habitat suitability from ML models aligned well with these observed contrasting patterns in seasonal shortfin mako movements, while also demonstrating considerable interannual variability.