Open Access Repository

Analysis of marine animal behaviour from electronic tagging and telemetry data using state-space models


Downloads per month over past year

Patterson, TA (2009) Analysis of marine animal behaviour from electronic tagging and telemetry data using state-space models. PhD thesis, University of Tasmania.

[img] PDF (Whole thesis)
whole_Patterson...pdf | Request a copy
Full text restricted
Available under University of Tasmania Standard License.


The last few decades have seen a technological revolution in the collection
of movement and other data from free roaming animals. Electronic tagging
and telemetry (ETT) methods have been advanced by miniaturization of
components and electronic storage capacity. Biologists now routinely collect
data from animals in some of the least accessible habitats. However,
until recently this data avalanche has not been matched by accompanying
developments in analytical methods. In this thesis I present a series of
methods and case-studies for analysis of ETT data using State-space models
(SSMs). The SSM combines a stochastic model of movement or behaviour,
and potentially other biological or physiological processes, with a model of the
observation process.
Chapter 1 provides a review of individual movement analysis and recent SSM
developments. Common threads running through the multitude of approaches
of movement data proposed in the literature are contrasted with the SSM.
In chapter 2 the machinery of Hidden Markov Models (HMMs) is outlined.
The HMM is the main SSM employed in this thesis for categorisation of
discrete hidden behavioural states from movement data. To demonstrate HMMs incorporating an in situ covariate, I analyse data from juvenile southern
bluefin tuna ( Thunnus maccoyii) on their migration route across the Indian
Ocean. Diagnostics are used to compare a two-state switching model against
a two-state memoryless model and a one-state model.
While considerable effort has been applied to analysis of horizontal movements,
analysis methods for vertical movement (VM) data are underdeveloped.
Commonly used data compression systems make VM data analysis challenging.
In chapter 3, I present a HMM approach for detecting behavioural shifts
from compressed VM data. The HMM is made from two nested-HMMs;
a continuous-time Markov model to handle the effects of data compression
routines is nested within a discrete time HMM used to detect shifts in
behaviour through time. The nested HMM is applied to data from summaries
of southern bluefin tuna vertical movement data. The predictions of the HMM
based only on summary data are compared to the actual behaviour in the raw
Chapter 4 quantifies error from state-space models fitted to data from service
Argos. Service Argos is the ubiquitous provider of satellite telemetry for
wildlife tracking. Therefore, correcting errors in Argos positions is a critical
part of making movement inferences for a host of species. We evaluate a
hybrid approach which removes aberrant positions using a heuristic speed
filter and then use Kalman filtering to construct a movement path between
corrected positions. The method is applied to Argos data from grey seals
( H alichoerus grypus) fitted with tags that also provide Fastloc G PS estimates
of location. Importantly I present an approach for quantifying the accuracy of
error estimation which can be applied to any state-space technique.
Expanding upon the methodological developments given in chapter 2 and 4,
chapter 5 presents a broader analysis of data from a highly-mobile, ocean
top-predator; the southern elephant seal (Mirounga leonina). Movement data
from four subpopulations are analysed using three different types of HMM. The chapter demonstrates the utility of the approach for comparison of behaviours
between groups of animals; in this case from four regions throughout the
Southern Ocean. Seals from each region experience different oceanographic
regimes resulting in the adoption of particular strategies and habitat selection.
The links to demographic differences between each population are explored.
I conclude the thesis with a discussion of further developments required in
the analysis of movement data. These include the spatial prediction of
core habitats from Markovian behavioural models. The interpretation of
the stationary distribution of estimated Markov chains and how this may
be affected by an animals choice of covariates is considered. Finally, the
methods developed in this thesis are situated within the context of a synthesis
of movement ecology.

Item Type: Thesis (PhD)
Copyright Holders: The Author
Copyright Information:

Copyright 2009 the author

Additional Information:

Chapter 1 appears to be the equivalent of an article published as Patterson, T., L. Thomas, C. Wilcox, 0. Ovaskainen, and J. Matthiopoulos. 2008. State-space models of individual animal movement. Trends in Ecology and Evolution 23:87- 94.

Chapter 2 appears to be the equivalent of a peer reviewed version of the following article: Patterson, T., Basson, M., Bravington, M.V. and Gunn, J.S (2009) Classifying movement behaviour in relation to environmental conditions using hidden Markov models. Journal of Animal Ecology:78, 1113-1123, which has been published in final form at [10.1111/j.1365-2656.2009.01583.x. This article may be used for non-commercial purposes in accordance with Wiley Terms and Conditions for Self-Archiving."

Chapter 4 was later published as Ecology as Patterson, T., McConnell, B.J., Fedak, M.A., Bravington, M.V. and Hindell, M.A. 2010 Using GPS data to evaluate the accuracy of state-space methods for correction of Argos satellite telemetry error, Ecology, 91(1), 273-285 and that article is copyright by the Ecological Society of America

Date Deposited: 03 Feb 2015 03:10
Last Modified: 04 Apr 2016 22:25
Related URLs:
Item Statistics: View statistics for this item

Actions (login required)

Item Control Page Item Control Page