Yes, that is correct. You need to specify that the longer distances are the exception due to transit over water :)

Yes, you should exclude GPS fixes that are over water (because the owl is not actually using these areas) and also exclude random points located in water (because these areas do not provide available habitat for the owls). Also, you might want to think about just using 90% or 95% kernel or MCP to exclude occasional long-distance exploratory forays from your estimation of home range.

How will you ensure that the location points are random? KDE assumes independence and we know movements are not random. Also, make sure your sample sizes are appropriate for KDE. It is sensitive to sample size.

In that case I would not use MCP and KHR only with strict settings.

You could of course use concave polygones.

However, I think it would be better to use LoCoH.

And do not forget: also corridors for crossing are belonging to the home range ;-)

Best

Oliver

Regarding Katherine's remark you could use aKDE, an approach that accounts for autocorrelation present in movement data. MCP are not a good choice anyway.

Regarding your question I also agree that you can exclude water bodies from the analysis since you can found that step by your question about habitat selection. However, as Oliver mentioned, when talking about space use those points should be included.

If you use Kernel home range estimates, or a similar frequency-based spatial distribution estimator, your home range can be estimated along a continuum of percent use, such as 100%, 90%, 50%. The 100% estimate will include the surrounding water and may not accurately reflect the total area used by the individual, but 90% estimates, for example, will more accurately reflect the total are used by the individual. An interesting hypothesis is whether the 90% home range (or whatever percent reflects appropriate use) increases as the the landscape becomes more fragmented. Good luck!

It is a bit tricky situation which we have also faced with our Dalmatian pelican study (see relevant question I have made a few months ago).

Regardless of the method used in estimating the home range, these sea crossings between the mainland and the islands are certainly parts of the home range of these particular owls and their area should be included in it. However, when working on analysis of habitat use they should be excluded because they are entirely unsuitable habitat for these animals and the probability to occur at these habitats has nothing to do with covering any of their life needs, so it would be a mistake to take the into the account in a way equivalent to all other types of habitats.

Thank you for all the good advice. Seems to be agreement of the best procedure for the habitat analysis is to exclude the large water bodies. And including water bodies (crossings) in the home range estimations. Although, I was not aware of that KDE has assumption of independence between fixes, how could this assumption be fulfilled when studying animal movement by GPS tracking?

Note: I have updated the original post to include the temporal resolution of my fixes.

As mentioned, guess aKDE is a good choice for such cases: Article Rigorous home-range estimation with movement data: A new aut...

I agree with you Ronny. We have to accept some dependence (literal meaning) between animal location at the time of a previous fix, where it is now, and where it can potentially go before the next fix. But I understand mathematical people focus on something different, statistical independence of points as a basic assumption of kernel density estimation. Nevertheless, in many previous animal tracking studies it was acceptable to use kernel methods with long intervals between fixes (reducing the issue of auto-correlation), and could okay for your fixes 3 times per night with 2 h intervals. Newer methods like aKDE from Fleming et al and Brownian Bridge movement models I think are more important for people who get high frequency fixes. Might be interesting to try both with your owl data. It sounds a very interesting study, best wishes for your analysis.

The question you rise is a challenge for many statistical analysis using data from individual animal movements, however also a chance. The recent development in GPS tracking technology has even increased the pressure on finding solutions for this issue, as the memory and battery capabilities allow very fine time-steps between the GPS fixes.

The first question to ask your data, is, if the time steps you've chosen really result in autocorrelated data. Did you assess this already statistically?

As discussed in several prior studies in movement ecology, a few approaches have typically been taken, including calculating and correcting the confidence intervals based on rarefied data where locations are no longer correlated (e.g. Fortin et al. 2005 Wolves influence elk movements: behavior shapes a trophic cascade in Yellowstone National Park, Ecology) or by including an autocorrelative structure (e.g. van Beest et al. 2012 Temperature-mediated habitat use and selection by a heat-sensitive northern ungulate, Anim Behav ). We typically included temporal variables as predictor covariates in the models.

You might also want to have a look at publications such as :

Perotto-Baldivieso et al (2012) Detecting autocorrelation problems from GPS collar data in livestock studies, Applied Animal Behavior Science

or

Anderson, D.M. (2010) Geospatial methods and data analysis for assessing distribution of grazing livestock, in Proc. 4th Grazing Livestock Nutrition Conference. B. W. Hess, T. DelCurto, J.G.P. Bowman, and R.C. Waterman eds. West. Sect. Am. Soc. Anim. Sci., Champaign, IL.

As already mentioned, the temporal autocorrelation can also be in interesting asset in the analysis of animal movements and general animal behavior. Finally, you should bear in mind that even the landscape has to a certain extent an autocorrelated nature.

For the analysis of your movement data, you might furthermore consider the application of step-selection functions (SSFs), which bear the "...potential to depict resource selection at multiple spatial and temporal scales..." (Thurfjell et al. 2014 Application of step-selection functions in ecology and conservation, Movement Ecology).

I recommend Andrew's answer

Best Regards