Hi Colin,

It sounds like you’re starting some interesting work. It seems that you’re already familiar with the work of Ran Nathan and Iain Couzin (among others). Their work provides a solid infrastructure on which to layer may other levels of ecologically relevant complexity. This will, of course, allow you to add your own natural history knowledge or empirical data to the mix. Really, any extension of those baseline models for your specific system will make for a significant contribution, especially given that your movement ecology models would be constructed using empirical data.

Are you looking for specific R tools that you can use in the construction of movement ecology models? If so, one option is the R wrapper for the NetLogo program (https://ccl.northwestern.edu/netlogo/). You’ve no doubt already seen the R wrapper, but I've attached some links to this response that may be useful.

NetLogo also has an associated edited volume by Westervelt that you might find useful. As it’s a Springer book, you should be able to obtain chapters via the Berkeley library. Reference is attached.

If, somehow, I have completely botched this response and you are still no more informed than you were when you posted your question, please ask again. This is an important topic, and I wouldn’t want my response to derail the discussion of it.

I hope the Berkeley weather is treating you well.

Best,

Patrick

http://www.jstatsoft.org/v58/i02/paper

http://rnetlogo.r-forge.r-project.org/

http://link.springer.com/book/10.1007%2F978-1-4614-1257-1

Hi Colin,

If you have not yet come across the following reference - then it may be of use:

Langrock, R., King, R., Matthiopoulos, J., Thomas, L., Fortin, D. and Morales, J. M. (2012) Flexible and Practical Modeling of Animal Telemetry Data: Hidden Markov Models and Extensions. Ecology 93 2336-2342

This paper is available at:

http://www.esajournals.org/doi/abs/10.1890/11-2241.1

Associated R code for fitting HMMs is available within the online appendix. I hope that this may be of some use. Again no covariate information is included in this example but it may be a start for coding such models in R.

Ruth King

One of the challenges of including covariates in HMMs for animal movements lies in what aspect of the model the covariates should influence. At least that is the case for the model types I’ve worked with, which is similar to the Morales et al approach (Ecology, 85(9), 2004). If the covariates are weather data, it could influence the probability of being in a certain state (e.g. more likely to be dispersive when it’s warm), or it could influence the movement parameters of each state (e.g. dispersive movement distance is longer when it’s warm), or it could be both. I know very little about your system, but it may require quite a bit of data to distinguish between models. Then there is another complexity if you want to extend the model to a hierarchical Bayesian model (for some reason I am assuming you are doing this Bayesian). I.e. should the effect of the covariate be a common parameter for the population, or is it an individual parameter? Depending on your data, it can be quite challenging getting the MCMC sampler to mix well.

As a first step before constructing a potentially complex model, is it perhaps worth fitting a model without covariates and then do a secondary analysis to look for correlation between posterior estimates of individual parameters and the covariates of interest? I did something along those lines in this paper: http://www.pnas.org/content/early/2013/07/24/1303157110.abstract

We didn’t find any clear trends there. (That’s not saying that movement isn’t influenced by weather conditions. This was most likely because we used tracking data from wet season, where pretty much every day is a god day for movement.) Using a similar approach might at least give you an idea about which covariates you should include and propose some relevant candidate models.

Also, sorry if the answer is too biased by my own experience. There are likely other approaches to HMM for animal movement, some of which may be more relevant to your system.

Hi Colin,

the code and papers of Roland would be a very good start from my point of view.

Just in case you look for something more basic explaining HMMs and ML-estimation, and don't have direct access to Zucchini and MacDonald (2009): attached an introductory text to HMMs. However, please pay attention: these are lecture notes, and may contain some errors.

Also this one may help!  http://www.crcpress.com/product/isbn/9781439817087

Have a look inside.

Hi Colin, I work in the ocean so the variables are a bit different but I think Pedersen's thesis http://geolocation.umassd.wikispaces.net/file/view/Pedersen_2007.pdf and code http://mwpedersen.dk/tracking.html   are great, as well as http://geolocation.umassd.wikispaces.net/HMM+Source+Code and an excellent book is http://www.cin.ufpe.br/~tfl2/artificial-intelligence-modern-approach.9780131038059.25368.pdf especially the chapter on probabilistic reasoning. Good luck!