Assessing performance of Bayesian state-space models fit to Argos satellite telemetry locations processed with Kalman filtering.

Mónica A Silva,Deborah J F Russell,Mark F Baumgartner,Ian Jonsen,Rui Prieto,Dave Thompson

doi:10.1371/journal.pone.0092277

Abstract

Argos recently implemented a new algorithm to calculate locations of satellite-tracked animals that uses a Kalman filter (KF). The KF algorithm is reported to increase the number and accuracy of estimated positions over the traditional Least Squares (LS) algorithm, with potential advantages to the application of state-space methods to model animal movement data. We tested the performance of two Bayesian state-space models (SSMs) fitted to satellite tracking data processed with KF algorithm. Tracks from 7 harbour seals (Phoca vitulina) tagged with ARGOS satellite transmitters equipped with Fastloc GPS loggers were used to calculate the error of locations estimated from SSMs fitted to KF and LS data, by comparing those to “true” GPS locations. Data on 6 fin whales (Balaenoptera physalus) were used to investigate consistency in movement parameters, location and behavioural states estimated by switching state-space models (SSSM) fitted to data derived from KF and LS methods. The model fit to KF locations improved the accuracy of seal trips by 27% over the LS model. 82% of locations predicted from the KF model and 73% of locations from the LS model were <5 km from the corresponding interpolated GPS position. Uncertainty in KF model estimates (5.6±5.6 km) was nearly half that of LS estimates (11.6±8.4 km). Accuracy of KF and LS modelled locations was sensitive to precision but not to observation frequency or temporal resolution of raw Argos data. On average, 88% of whale locations estimated by KF models fell within the 95% probability ellipse of paired locations from LS models. Precision of KF locations for whales was generally higher. Whales’ behavioural mode inferred by KF models matched the classification from LS models in 94% of the cases. State-space models fit to KF data can improve spatial accuracy of location estimates over LS models and produce equally reliable behavioural estimates.

Highlights

The collection of individual animal movement data has become widely utilized by ecologists in the last decade due to the improvement of the underlying technologies and reduction of operational costs involved in animal telemetry
Data Collection and Processing In the interest of clarity we’ll use the following terminology throughout the paper: i) Least Squares (LS) locations/data and Kalman filter (KF) locations/data refer to the locations/data provided by Argos that were derived from the application of the LS and KF algorithms, respectively; and ii) LS or KF model refer to the state-space models fit to data derived from the application of either the LS or KF algorithm
A total of 368 LS and 375 KF model locations were within 30 min of a GPS position and were used to compare spatial accuracy of locations derived from each hierarchical SSM (hSSM) (Table 1)

Summary

Introduction

The collection of individual animal movement data has become widely utilized by ecologists in the last decade due to the improvement of the underlying technologies and reduction of operational costs involved in animal telemetry. Of the several technologies available, one of the most popular is that based on satellite tags (platform transmitter terminals, PTTs) using the Argos system [1]. The Argos service provider assigns a quality index, or location class (LC), to each position based on its estimated precision. Attempts to measure spatial error of Argos locations using either stationary tests or double-tagging experiments with free-ranging animals consistently reported larger errors than those indicated by Argos. Many of these studies provided error estimates for location classes A and B, showing these could be in the range of tens to hundreds of kilometres (reviewed in [3])

Objectives

Methods

Results

Conclusion