Bailey Jenkins

Abstract:

The shortfin mako shark (Isurus oxyrinchus) is an ecologically important apex predator but is unfortunately susceptible to overfishing due to slow growth and reproduction rates. To gain a more thorough understanding of how mako sharks use their habitat in the Northwestern Atlantic, satellite tag data from 50 individuals was analyzed using a switching state space model and subsequently incorporated into a machine learning model to predict behavior. This model used corresponding location based environmental data as predictors of behavior state and implemented the use of five different machine learning algorithms to make these predictions. The logistic regression model yielded the highest accuracy score of 87.9%, and demonstrated a strong positive correlation between depth, chlorophyll-a concentration and behavior state. Using machine learning to understand mako behavior shows promise as a tool to validate fisheries regulations, and can be widely applied to other marine megafauna, given there is sufficient tagging and environmental data. Future studies should consider using SPLASH tags to gain more location specific temperature data and use depth as another predictor of behavior.