Abstract
Identifying at-sea foraging areas is a longstanding goal for seabird ecology and conservation. GPS tracks can reveal behaviour because slow, tortuous flight often indicates searching linked to feeding attempts, but two-dimensional (2D) paths may oversimplify three-dimensional (3D) flight. Here, we present the first 3D tracks for Ascension Frigatebirds Fregata aquila and assess whether incorporating flight altitude improves our ability to describe putative behavioural states. We compare results using altitude derived from GPS loggers and barometric altimeters deployed simultaneously. Tracked birds (three females) travelled at a mean altitude of 178 m and attained maximum heights of 1,658–1,871 m (measured by barometric altimeters). Hidden Markov models based on 2D tracks defined three states (interpreted as ‘search’, ‘slow travel’ and ‘fast travel’). However, with 59.5% of locations defined as ‘search’, identifying the most important foraging hotspots would be challenging. Including altitude was informative, allowing models to define two further states by introducing high-altitude ‘soaring/thermalling’ behaviour (3.4% of locations) and dividing ‘search’ into mid-altitude (44.7%) and low-altitude (12.2%), the latter being more likely to represent prey capture. Barometric altitude was less prone to large errors than GPS, but GPS altitude was highly correlated (r = 0.94) and state assignments overlapped by 88% overall. Using GPS altitude would reduce potential tag effects and allow us to model tracks in 3D for males and lighter females, which are too small to carry both loggers. Overall, incorporating flight height from either barometric altimeters or GPS loggers into behavioural models improved our ability to distinguish putative foraging events from high-altitude flight.