Identifying leatherback turtle foraging behaviour from satellite telemetry using a switching state-space model

Abstract

Identifying the foraging habitat of marine predators is vital to understanding the ecology of these species and for their management and conservation. Foraging habitat for many marine predators is dynamic, and this poses a serious challenge for understanding how oceanographic features may shape the ecology of these animals. To help resolve this issue, we present a switching state-space model (SSSM) for discerning different movement behaviours hidden within error-prone satellite telemetry data. Along with modelling the movement dynamics, the SSSM estimates the probability that an animal is in a particular discrete behavioural mode, such as transiting or foraging. Using Argos satellite telemetry for leatherback sea turtles, we show that the SSSM readily identifies distinct classes of movement behaviour from the noisy data. Moreover, patterns in simultaneously collected diving data, to which the model is blind, match well with behavioural mode estimates. By combining behavioural mode estimates from the model with the diving data, we show that while transiting, leatherbacks make longer, deeper dives; and while foraging, they encounter cooler waters that range from 13 to 22°C. These differences are consistent among the turtles studied and within the same turtle in different years. This modelling approach can enhance standard kernel density estimators for identifying habitat use by incorporating behavioural information into the estimation procedure. Ultimately, we can build predictive models of habitat use by incorporating environmental data and diving behaviour directly into the SSSM framework.

Bahamas connection: residence areas selected by breeding female loggerheads tagged in Dry Tortugas National Park, USA

“…The model was described in 2005 [41] and has previously been applied to the movement of marine animals including turtles [1,3,4,42,43,[46][47][48][49][50]56,80]. Location data obtained through satellite transmitters are often received at irregular time intervals and sometimes involve large gaps and positional errors.…”

“…SSM has been used to identify locations when turtles show directed movements versus area-restricted search (ARS) patterns-deemed previously as migration and inter-nesting or foraging 'modes' [1,3,4,[40][41][42][43][44][45][46][47][48][49][50]. When combined with MCP (simple polygon created with home-range locations [51,52]) or KDE (a nonparametric method used to identify one or more areas of disproportionately heavy use [i.e.…”

Development of foraging strategies with age in a long-lived marine predator

“…Eight hours is the smallest time step we could use given the temporal resolution of the Argos data being fit (see Breed et al 2009Breed et al , 2011. As in other implementations of this model, the proportion of MCMC samples in each respective behavioural state was used to classify behaviour, with proportions above 0.7 assigned to travel, proportions below 0.3 assigned to foraging and intermediate proportions assigned as uncertain (Jonsen et al 2007, Breed et al 2009). In practice, the uncertain behavioural state tended to show some autocorrelation between consecutive steps, so for the trip analyses undetermined locations were included in travel segments.…”

Active dispersal in loggerhead sea turtles ( Caretta caretta ) during the ‘lost years’

“…Locations were collected by ARGOS-CLS. All raw surface locations were filtered using a Bayesian state space switching model [42], which regularized track positions at daily 24 h intervals. Such methodologies have been used in sea turtle movement studies to estimate animal location while accounting for satellite positioning errors [18].…”