Building the next generation of analysis tools for animal tracking data
Abstract
State-space models and hidden Markov models have revolutionized the analysis of animal tracking data, leading to a robust capability for the inference of unobserved movement behaviours from location data obtained via animal-attached telemetry devices. Of the two tools, state-space models are the preferred choice when dealing with animal tracking data that is highly prone to measurement error, eg. light-level geolocation data and Argos satellite data. Bayesian Markov Chain Monte Carlo methods are a widely-used, computer-intensive approach for fitting complex state-space models to data but they can be extremely slow to implement. This slow computation time prevents the development and application of more advanced models that can infer realistic and complex animal movement processes. Despite this hindrance, it has become clear that such models now are essential both for handling the high-volume, complex data collected by animal-attached devices and for understanding how animals may respond to current and future environmental change. This project aims to develop the next generation of animal movement analysis tools by building a powerful, lightning-fast and easily used toolbox of state-space models for multiple types of animal tracking data.We will use recently available software ??? Template Model Builder ??? that allows complex state space models to be fit, via maximum likelihood, up to 1000 times faster than the Bayesian approach. This reduces computation time from hours or days to seconds without loss of precision or accuracy. The Office of Naval Research (ONR) is a key partner of the Integrated Ocean Observing System and its Animal Telemetry Network (ATN). Our new, fast models contribute to the ATN???s commitment toward integration of animal-sensed ocean variables into operational ocean physics models. The models will be automated and integrated into operational ocean data workflows so that huge volumes of location error-prone animal tracking data can be quality controlled in near real-time. This fast quality control will allow more precise location of animal observed ocean structure data (Conductivity ??? Temperature ??? Depth) that can better inform operational ocean physics models, resulting in better understanding and prediction of the ocean environment.Our analysis tools contribute directly to the ONR Marine Mammals and Biology program???s Integrated Ecosystem Research thrust by improving on current tools for understanding marine mammal behaviour and distribution in an environmental context. Spatial habitat models are widely used to infer how the environment shapes animals??? distributions and habitat use but these tools are blind to animal behaviour. Our models will be augmented to estimate movement behaviour and how it may be influenced by environmental, physiological and/or human activity variables. By coupling our new models with habitat models, a more nuanced understanding of how animals actually use habitat is possible. This new approach can be used to refine dynamic management of threats, such as ship traffic, to ocean wildlife by accounting for how movementbehaviour modulates animals??? vulnerability to threats.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jul 10, 2018
- Source ID
- N000141812405
Entities
People
- Ian Jonsen
Organizations
- Macquarie University
- Office of Naval Research
- United States Navy