Hi Andrew,

I agree with you the use of SECR method is quite probably assuming "way too much" in repect to an animals territory/home range being random.  What I have done in the past is to create the boundries of the "territory" then perform a Dirichlet tesselation on the data.  The Dirchlet tesselation then allows you to fit the internal configuration of the territory/home range.  This method does not make any assumptions of convexity and describes the spatial pattern of telemetric locations purely in terms of their relative position.  Once you have this "range" then it is safe to assume that positions within are random, driven by potential predator encounters, food, shelter, water etc.  Hope this helps

Tracy

Thanks Tracy, I've seen tesselations used to delineate home ranges before, but I've never really taken the time to examine their broader merits. Looks like I should.

Hi, there is some work being done to address this assumption, i.e. to deal with non independence of activity centers/home ranges. Also, I believe there has been some simulation studies to assess biases associated with violating that assumption (on the road and cant remember them off the top of my head).

This comes down to your objectives I guess.

With a large number of telemetered individuals, I imagine you could investigate characteristics of home ranges with respect to a bunch of covariates, including proximity to other individuals, using methods that can deal with territoriality (e.g., tesselations). This of course requires some assumptions, specifically about how representative your telemeterd individuals are of the population, but does sound very interesting.

SECR on the other hand has the added value of providing estimates of detectability and movement parameters along with estimates of overall density. So, if that is what you are after then it would be worth doing some simulations to mimic the territoriality of you specific species and see if it would bias things.

Chris

Thanks Chris, I'm glad to hear that someone has been looking into this already.

I suppose there's not going to be a simple answer to how important the assumption of independent home range distribution is; it'll vary among species and populations.

It seems to me then, that if you're going to be using SECR to estimate densities of territorial species (or other species with non-random spatial organisation), then you really should present some sound information to explain how you expect this to effect your estimates. Simulations using data relevant to your subject population would be a good place to start.

Agreed, i think we should all be aware of the limitations and assumptions of the tools we use. By the way,feel free to get in touch if you have any more questions/queries about these models - I think they have great potential and encourage their use - within reason of course.

Chris

Hi. I'm sure Chris is thinking about this article: A spatial capture-recapture model for territorial species (Brian J. Reich, Beth Gardner) http://arxiv.org/abs/1405.1976.

Jose

Thank you Jose, that's great - much more than I'd hoped for.

I was a bit worried that authors and reviewers were ignoring or unaware of this issue, so it's encouraging to see work like this.

Andrew

Hi again Andrew. After your question, I've been simulating different data scenario, generating a Strauss point process. Actually, there not seem a big problem to deal with using a SCR model for a territorial species.

Maybe you can replicate the simulations. I've sent a piece of code in R I've used (modificated from Royle et al. scrbook package)

traplocs

Thanks again Jose,

I'll have to learn a bit more about point processes, Reich and Gardner's methods and scrbook before I can do a lot with this. I have some GPS tracking data on the movements of individually recognisable feral cats on a camera trap grid that I really should use as well. Unfortunately, I don't have a lot of time to spend on this at the moment, but I'll start working through it all slowly.

Just noticed this nearly a year on... I simulated both dispersed and random Poisson home range centres in my 2004 SECR paper and found the method robust - this is not something that has been naively ignored. I suggest you look closely at what Reich and Gardner actually found before getting too excited about the claim that dispersion is important.

Thanks Murray, yes I've had a good look at your paper since I posed this question. I'm happy that I've got a better understanding now. Cheers, Andrew