.png){kind=icon}
.png){kind=icon}
Return
Experimentally derived likelihoods for light-based geolocation
.png){kind=icon}
OA
AI
DOI：10.1111/2041-210X.12555.png)
.png){kind=icon}
.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Electronic tags have revolutionised animal movement studies, but the reliability of subsequent ecological inference hinges on being able to quantify uncertainty in location estimates. Light-based geolocation, which uses the time series of light intensity during twilight events, remains the only viable technology for many species (e.g. fish and small birds) despite its limited accuracy. Modern approaches to movement modelling, such as Kalman filters and gridded hidden Markov models, require a valid likelihood for each observed twilight. It is difficult to directly construct such a likelihood (i.e. the probability density of the light data during a twilight period) given any location on the globe, because of complicated autocorrelation structures and non-standard statistical distributions. We therefore use data from moored tags at known locations to construct a transformation that turns a simple one-dimensional statistic into a quantity with the properties of a log-likelihood. The result is a set of calibration splines that can be used with light data from a similar tag deployed on a real animal: for each twilight, the one-dimensional statistic is calculated for any location (e.g. on a grid of, or all possible, locations) and then transformed into a likelihood using the calibration splines. The likelihoods can then be input to any state-space model to estimate a track. We show an example track from a grid-based hidden Markov model applied to light data from a tag deployed on a southern bluefin tuna. This approach to light-based geolocation provides the flexibility to integrate movement and behaviour modelling in a novel way. The likelihood surfaces from our approach can be used in any state-space model of animal movement and behaviour, irrespective of whether estimation is by maximum-likelihood or Bayesian methods. Our approach is primarily aimed at users interested in developing and fitting their own state-space models to explore biological hypotheses about animal behaviour.
Keywords:
animal movement
location uncertainty
movement modelling
state-space models
AI Summary
Key information extracted from the uploaded paper, including a brief overview, abstract, background, key highlights, visual analysis, and future outlook.
Journal
IF:
6.2
Papers:
2.9K
Citations:
2.9W