calculate area for individuals with return or commute behavior

[yu lei]

Hi Chris,

      Recently, I want to estimate the yearly home range and exploration behavior of a residency species. The population I am studying exhibits diverse spatial behaviors, such as philopatric (stay at home since birth), settlement (stay in a specific range all year), return (A→B→A), and commute (A ⇆ B). The first two types have a typical stable home range. For the latter two, I plan to calculate their stable home range as follows: First, segment the data using segclusted2.

1. For the return behavior, identify the three stable phases A, B, and A, and calculate AKDE for each. Since A is calculated twice and overlaps to a large extent, I cannot simply consider A+B+A as the annual home range. Therefore, I plan to import ABA into QGIS, merge them together, eliminate overlapping areas, and retain the outside boundary to serve as the annual home range.

2. For the commute behavior, I will try to identify relatively stable A and B. If distinguishable, I will calculate the area as in step 1. If it is difficult to distinguish (e.g., frequent and irregular commuting), I will not segment the data and will calculate it directly, as long as the variograms tend to asymptote.

3. Besides calculating the annual stable home range, I am also interested in understanding the frequency and duration of exploration behaviors. Is it feasible to overlay the home range shapefiles with the tracking positions in GIS to assess occurrences and durations outside the home range?

4. I am not sure if the above methods are optimal, and considering my study involves 41 individuals, each with 2-4 years of data, manually and visually segmenting or fitting models for each individual's annual spatial behavior would be very time-consuming. I hope there is a faster and more reliable method, or if there are any reference R codes available.

Thank you for any suggestions.

Best,

Yu Lei

[Christen Fleming]

Hi Yu Lei,

1. You can calculate the ctmm::mean() of A and B AKDEs, weighted by the time spent in each range.

3. You can assign p-values to locations by exporting and then extracting from the CDF.

4. I would start with some automated clustering, range estimation, and then overlap estimation; and then sort the results by overlap and by quality (as inferred from DOF[area] and tau[position]). Looking at the worst fits, you can then see how many individuals you need to tweak by hand or invest more effort into making the automation work better.

Best,

Chris