Modelling harbour seal habitat by combining data from multiple tracking systems

Abstract

Technological developments over the last 20 years have meant that telemetry studies have used a variety of techniques, each with different levels of accuracy and temporal resolution. This presents a challenge when combining data from these different tracking systems to obtain larger sample sizes or to compare habitat use over time. In this study, we used a Bayesian state-space modelling approach to integrate tracking data from multiple tag types and standardise position estimates while accounting for location error. Harbour seal (Phoca vitulina) telemetry data for the Moray Firth, Scotland, were collated from three tag types: VHF, Argos satellite and GPS–GSM. Tags were deployed on 37 seals during 1989 to 2009 resulting in 37 tracks with a total of 2886 tracking days and a mean duration of 87 days per track. A state-space model was applied to all of the raw tracks to provide daily position estimates and a measure of the uncertainty for each position. We used this standardised tracking dataset to model their habitat use and preference, which was then scaled by the population size estimated from haul-out counts to give an estimate of the absolute number of harbour seals using different parts of the Moray Firth. As expected for a central place forager, harbour seals most frequently occurred in areas close to their inshore haul-out sites. However, our analyses also demonstrated consistent use of offshore foraging grounds, typically within 30 km of haul-out sites in waters < 50 m deep. The use of these statistical models to integrate and compare different datasets is especially important for assessing longer-term responses to environmental variation and anthropogenic activities, allowing management advice to be based upon datasets that integrate information from all available tracking technologies.