Understanding Variogram Output

[Alice Ball]

Hello,

I'm trying to understand how best to approximate the home range for a tagged individual, but I'm puzzled by the variogram output. The variogram (shown below) appears to show a steady increase for about 6 months before reaching something resembling an asymptote at around 6.5 months.

One task I was asked to achieve by the research team was to provide seasonal estimates for home range. The data were initially separated into summer and winter datasets. After playing with the data I divided the summer dataset further into an early summer and late summer dataset as these appeared to be two quite distinct clusters, and ran AKDEs on each of these separate datasets. The tagged individual does seem to have moved west to east during the course of the survey (see below). I thought it may be better, however, to pool all the data and look at the data as a whole, which is how I came up with the variogram above.

95% AKDEs - blue is "early summer", yellow is "late summer", and pink is "winter"

My question is how best to fit a KDE to the data as a whole and how best to interpret the results. Do the data show activity within a home range or does it show a migration over that time? Is it worth trying to fit an AKDE to all the data, or is it best to fit it to the separate clusters as I did initially and then give a measure of overlap? I have tried to fit an AKDE to all the data using both range=TRUE and range=FALSE, however am not able to get an acceptable movement model that fits the data well.

I'm new to AKDEs and the ctmm library so any advice would be greatly appreciated. The species we are looking at is very understudied with little information available about their ecology, so there is little prior knowledge about how they utilise their home ranges and any seasonal differences.

[Christen Fleming]

Hi Alice,

Yeah, it looks like there is a smaller daily/sub-weekly range that is slowly drifting within the larger seasonal range.

You can check the variograms of your segmented tracks for an asymptote, and even calculate a mean segment variogram.

You can plot the empirical variograms of the data atop the theoretical variograms of the stationary models, segmented or not, where you can check how large the discrepancy is.

Finally, you can mean() the AKDEs of the three segments and compare that to the AKDE of the entire dataset under an (incorrectly) assumed stationary model, to see how big of a difference that makes.

Best,

Chris