Spatial and Temporal Overlap

[John Stuart]

Hi there,

I have just come back to using the CTMM package after a lengthy break. I am just wondering if temporal overlap overlap is factored into the overlap analysis in the CTMM package or if there is a way to factor this in? 

For example if I have tracked two individuals that display a high degree of spatial overlap, but they used the same spaces at different times (e.g. they avoided each other or went to the same places at different times). Is there a way to represent this statistically through the overlap analysis? AI would this is tot too simple a question. Thanks you.

Kind regards,

John

[Christen Fleming]

Hi John,

overlap() compares the stationary distributions and does not factor in time. Ultimately, I'd like to code in support for correlated movement models, encounter rates, etc..

I've seen people test whether or not the distances between individuals (at the same time) are greater or lesser than what you would expect by random chance (at different times), but I don't think they adjusted for the non-independence in the samples, so the effective sample sizes should be reduced for that test.

Another thing you could try is to fit an autocorrelation model to the location difference vector and see if the variance comes out smaller or larger than what you would expect from two independent movement models.

Best,

Chris

[laura labarge]

Hi Chris and John, sorry to intrude on this conversation - was wondering if you might have any examples from the literature on these approaches to understanding if the movements of two individuals are correlated? (particularly the latter approach sounds interesting but I haven't seen it done before!)

Best wishes,

Laura

[Christen Fleming]

Hi Laura,

Unfortunately I don't know of any published examples offhand, but I could potentially code up the last example when I get some free time.

Best,

Chris

[Jenny Hansen]

To Laura and John,

   I would suggest taking a look at the 'wildlifeDI' package in R, which has several measures for investigating animal attraction and avoidance. There are a couple of vignettes floating around, but I would also check out these papers, which describe the indices and methodologies for quantifying animal interactions:

A critical examination of indices of dynamic interaction for wildlife telemetry studies - https://besjournals.onlinelibrary.wiley.com/doi/full/10.1111/1365-2656.12198

Measuring Dynamic Interaction in Movement Data- https://onlinelibrary.wiley.com/doi/abs/10.1111/j.1467-9671.2012.01353.x

Mapping areas of spatial-temporal overlap from wildlife tracking data- https://link.springer.com/article/10.1186/s40462-015-0064-3

Andrew Whetten has also developed some methods for looking at interactions. You can find his work on social interactions and pairwise associations on Researchgate- https://www.researchgate.net/profile/Andrew-Whetten/research

Hope this helps!

Cheers,

Jenny

[Christen Fleming]

Thanks Jenny,

It looks like the "IAB" index from that package accounts for serial correlations, though I'm not familiar with it.

The first paper you cited details the issue of autocorrelation, which in this context makes any finding seem more significant that it is, because if individuals are near (or far) from each other at one time, then they will tend to be near (or far) from each other at the next time.

Whatever method you use, I would suggest putting together some correction for the autocorrelation. In the simplest case, you can take any metric that assumes independence, thin down your data to independence by taking one point every time-to-independence, and then calculate the metric with the thinned data. With multiple data points per time-to-independence, you can thin the data multiple ways and average the results for a more robust final estimate.

Best,

Chris