How to deal with birds that commute daily between separate areas

[Kristina Paxton]

Hi Chris,

I have a dataset of locations from GSM-GPS tags of I'o or Hawaiian Hawks that are residents on the Big Island of Hawai'i. Locations were taken ever hour during the day and at 3 time points at night with birds on average tracked for ~600 days and having 7000 - 9000 locations. When I used ctmm to get home range estimates, the variograms for half the birds had patterns indicating that birds were not range residents. From reading your publications and other conversations on the list serve I know that home range estimates using the ctmm package should not be calculated if an individual is not exhibit range residency. I used the segclust2D package to determine segments for the tracking period when a bird was a range resident with the goal of calculating a home range separately for each segment where a bird was a range resident and then averaging the home ranges to get an overall area of space use. However, in looking at the segmentation of the data birds are not shifting their home range to a new area and staying within that new area for a period of time, but they are commuting back and forth daily or every other day between their main/core area of use and the separate location for some set period of time (weeks to months). If I discard these commuting time periods from a home range analysis the areas they are commuting to will not be included in any of the home range analysis, greatly reducing the overall size of the spaced used by birds throughout the year. I have included a pdf with an example bird - showing all of the birds points on a satellite map, a variogram of all of the data, the results from the segmentation analysis, and variograms for the 7 segments identified. Given that the movements are not migratory and they are not shifting to a new location, but commuting between locations I'm not sure what the most valid approach is for calculating space use. Is it valid to calculate a home range for each of the 7 segments or should I use a different approach that doesn't require range residency? Any advice for this type of dataset would be greatly appreciated.

[Christen Fleming]

Hi Kristina,

That looks like behavioral switching between 3 behaviors. I would reduce the 7 states returned by segclust2D to 3 states, calculate 3 home-range areas, and then average them together with mean().

Best,

Chris