Pacific juvenile loggerhead sea turtle trajectories : improved understanding through machine learning

In recent years, the critical situation of sea turtles has been brought to light. There is an urge to protect them. To do so, greater knowledge about the turtles' behaviour is needed. Only with adequate knowledge can appropriate conservation measures be implemented. This study aims at filling a knowledge gap regarding the seasonal and longitudinal migrations of the Pacific juvenile loggerhead sea turtle through a novel approach: machine learning. This offers an objective approach without underlying human assumptions. The master thesis compares different machine learning models in order to find the one that explains the migrations undertaken by the loggerhead sea turtles with the highest degree of certainty. Environmental factors are analysed to understand the reasons. The study shows that a support vector regressor is the most suitable model reaching R² scores above 0.9 when predicting a turtle's geographical position after a day and R² scores above 0.3 when predicting the change of positions after ten days. Through an analysis of variables, it can be deducted that the seasonal migration is dictated by food sources, water temperature and day length, while longitudinal migrations are mainly driven by the velocity of currents. The work aims at laying the ground for future studies and better conservation measures.