Date: 17 December 2019
Venue: School VI
Using time series models to make biological inferences from animal tracking data
Animal location data can be thought of as time series of individual animal behaviour. This type of animal movement data can be effectively analysed using statistical models for time series to help understand what the animal is “doing”. One class of time series models that has become popular for analysing animal tracking data are hidden Markov models (HMMs). These models fall under the umbrella of state-space models, where we assume that the time series of observations is generated by an underlying “hidden” time series of system states, and that the observations and underlying states are linked in some way. The dependence structure between the observations and underlying, unobserved states allows us to make inferences about the unobserved time series, from the observed one. I will show examples of what we can learn about animal movement data using HMMs, based on two case studies from marine systems: acoustic detections of great white sharks and satellite locations of Weddell seals.
Theoni Photopoulou is a Newton International Fellow in the School of Biology. She has a background in marine ecology and statistics and splits her time between the Scottish Oceans Institute and the Centre for Research into Ecological and Environmental Modelling.