Quantitative method to estimate species habitat use from light-based geolocation data

Abstract

The development of biologging techniques has been instrumental in studying the behaviour of wild animals and interpreting it with respect to the bio-physical features of their habitat. Light-based geolocation currently appears to be the only technique suitable for the study of farranging small species, particularly marine species, over long periods, but it provides locations with low precision. In this study, we sought to improve the exploitation of these data. Specifically, the goals were to (1) correct rather than reject estimates, especially during equinox periods, (2) perform reproducible time-saving routine analyses, and (3) determine the spatial likelihood of the estimations to adapt inferences on habitat use by the population. We therefore applied an existing data-processing method based on spatial template fitting, using Markov Chain Monte Carlo and state-space modelling (Kalman filter) improved by a facultative sea surface temperature-matching procedure and a land mask. The main functions used for geolocation are grouped under the R package 'TripEstimation', freely available online. We focused on a typical example of animal movement that at present can only be ethically obtained from light-based geolocation. The method made it possible to estimate realistic positions during equinox periods and to routinely process the 12 available datasets. We thus obtained the most probable location for each sunrise/sunset as well as the posterior distribution around each estimated location, allowing an intuitive habitat use investigation at a scale of 100 km. This paper describes the method used and provides the complete and comprehensively annotated commands required for its use.

Stage-dependent distribution of the Critically Endangered Amsterdam albatross in relation to Economic Exclusive Zones

“…We used average travelling speed calculated from the GPS surveys (22.6 km h −1 ) to set up the specific movement model used to constrain location estimates. We summarized early model runs for this analysis following Thiebot & Pinaud (2010); however, proper implementation of these scripts should now involve summarizing from the posterior after running the chain for a large number of iterations (see Sumner et al 2009). The current implementation nevertheless produced valid location estimates from the GLS loggers compared with the satellite tracks of the individuals surveyed during a similar stage (Fig.…”

“…We used the maximum travelling speed value measured with the more precise GPSs (99 km h −1 ) to set up a specific speed threshold in this case. From GLS data sets, the most probable location estimates were generated following Thiebot & Pinaud (2010) with the package 'tripEstimation' in R 2.9.0 (R Development Core Team 2009, http:// cran. r-project.…”

Inter-population segregation in the wintering areas of macaroni penguins

“…This was a trade-off between having a sufficiently large period enabling us to describe a significant habitat use by the birds, but which was not too large to result in temporal smoothing and homogenisation of the habitat parameters. We therefore used a 1 mo window according to the spatial likelihood of the locations obtained by the geolocation technique (Wilson et al 1992, Hill 1994, Thiebot & Pinaud 2010, and the level of environmental change usually found between seasons (Clarke 1988). To determine which month to consider, we relied on changes in mean swimming speed.…”

“…Geolocation data were analysed following Thiebot & Pinaud (2010), assuming a mean travel speed of 2 km h -1 in order to estimate the most probable track, using the package 'tripEstimation' in R 2.9.0 (R Development Core Team 2009, http://cran.r-project.org/web/ packages/tripEstimation/index.html). Two fixes per day (1 every 12 h) were produced along the tracks, with a mean spatial accuracy of 180 km calculated on albatrosses (Phillips et al 2004) that may be better for the slower penguins (114 km, estimated by Thiebot & Pinaud 2010). Tracks were described and compared using a set of common parameters, based on the start and end dates of the trip; these dates were derived from the time of the first and last temperature records from each logger.…”

Biologging technologies: new tools for conservation. Introduction

“…Bailey et al (2009) applied a switching state-space model to satellite tracks of blue whales to identify seasonal and interannual variability in the location of migratory pathways and foraging hot spots. Thiebot & Pinaud (2010) extend the use of a spatial template fitting method, using Markov chain Monte Carlo and state-space modeling, to include a sea surface temperature matching procedure and land mask, thus improving the quality and potential use of light-based geolocation data.…”