Evaluating Behavioural Modelling Predictions in the Blue Shark (Prionace glauca) Enables Greater Insight on Habitat Use from Location only Argos Data

The relationship between habitat and behaviour provides important information for species management. For large, free roaming, marine animals satellite tags provide high resolution information on movement, but such datasets are restricted due to cost. Extracting additional biologically important information from these data would increase utilisation and value. Several modelling approaches have been developed to identify behavioural states in tracking data. The objective of this study was to evaluate a behavioural state prediction model for blue shark (Prionace glauca) ARGOS surface location-only data. The novel nature of the six SPLASH satellite tags used enabled behavioural events to be identified in blue shark dive data and accurately mapped spatio-temporally along respective surface location-only tracks. Behavioural states modelled along the six surface location-only tracks were then tested against observed behavioural events to evaluate the model's accuracy. Results showed that the Behavioural Change Point Analysis (BCPA) model augmented with K means clustering analysis performed well for predicting foraging behaviour (correct 86% of the time). Prediction accuracy was lower for searching (52%) and travelling (63%) behaviour, likely related to the numerical dominance of foraging events in dive data. The model's validation for predicting foraging behaviour justified its application to nine additional surface location-only (SPOT tag) tracks, substantially increasing the utilisation of expensive and rare data. Results enabled the critical behavioural state of foraging, to be mapped throughout the entire home range of blue sharks, allowing drivers of critical habitat to be investigated. This validation strengthens the use of such modelling to interpret historic and future datasets, for blue sharks but also other species, contributing to conservational management.