Discussion
This research compared area, shape, and locational accuracy of time- and non-time-based home range estimators to assess the impacts of sample size and GPS-point pattern shape. In general, GPS-point pattern shape had more impact than sample size on results. BRB was the method that maintained accuracy most consistently. The reasons for that are not completely clear, however the localized variation of time and reorientation with the advective coefficient might explain its overall higher accuracy. BRB uses successive locations, rather than generating global variables from an entire trajectory, which is the case in the other time-based home range estimators, so BRB likely takes more advantage of the spatiotemporal variability of telemetry data.
That sampling size greater than 125 points (the 05% sample size) had less impact on accuracy than point pattern shape was consistent with previous research comparing KDE and MCP (Downs and Horner 2008). The comparison here also agrees with the findings of Downs and Horner (2008), which found that KDE does not estimate home range area as accurately as earlier studies suggest.
The findings for this research also correspond with Walter et al. (2015), who found that time-based home range methods output home ranges that more accurately fit GPS-point pattern shapes than point-based methods. The research here extends the research by Walter et al. (2015) and also found that time-based estimators are better at accurately estimating area and location. However, KDE maintained location accuracy reasonably well for different GPS-point pattern shapes, except for perforated patterns.
It is worth noting that T-LoCoH offers the most input from a user standpoint. Given the large number of tests ran for this research, T-LoCoH was automated. During each simulation, the time was weighted at a 60% level using the TSD metric, which is half way between the 40-80% value that is recommended in method-1 of the T-LoCoH user guide (Lyons 2014). T-LoCoH estimates can vary depending on these inputs and further tests of T-LoCoH would be warranted for future research. Yet, even with automation, T-LoCoH still had a relatively high accuracy in estimating home range area.