the Efficient modelling of individual animal movement in continuous time

It is my belief (and sincere hope) that the FInCH approach opens new horizons in the world of statistical movement ecology, now that continuous time switching-diffusion models are more accessible to ecologists, which will hopefully increase their uptake. We also hope that this stokes new research efforts in the field. With regards to the FInCH approach, we see a plethora of new research directions, some of which I highlight in this section. 

So far, we have only looked in depth at FInCH in the spatially homogeneous case, where the probability of an animal changing behaviour is not affected by its environment. However, I’m extremely excited to look at FInCH with spatial covariates in the form of a case study on one-dimensional seal diving data obtained courtesy of a colleague at the University of St Andrews. We hope that the analytic work for the spatially inhomogeneous FInCH is already complete- which involves integrating the movement density of the animal within each local likelihood with respect to both the time and ‘location’ of each potential switch. 

Discrete-time models such as the HMM are typically formulated in terms of ‘steps’ and ‘turns’ (meaning the step length and turning angle of the animal between observations).  Allowing for directionality is advantageous as this may tell us a lot about an animal’s behaviour: for example, if an animal changes direction frequently it may be foraging, whereas if an animal moves in a more directed way, it may be migrating. Standard switching diffusion methods do not detect such directionality. However, to allow for the incorporation of direction, we believe that the FInCH approach may be amenable to the Continuous Time Correlated Random Walk (CTCRW) method of Johnson et al. (2008).  This approach assigns a diffusion process to the animal’s velocity, which may then be integrated over to produce a smooth movement process (which is also more realistic over a fine timescale). Work by Michelot & Blackwell (2019) introduces state switching CTCRWs, which I aim to extend in my work.

Finally, taking inspiration from popular R packages that exist for discrete-time modelling techniques, it is my hope to produce an accessible R package for ecologists to use to model individual animal movement in continuous time. 

 Please feel free to contact me at dpdgrainger1[at]sheffield.ac.uk