Accelerometer Data from Collared Female Polar Bears in the Beaufort Sea, 2009-2016

Data were collected from accelerometers built into Telonics Gen 4 satellite radio-collars deployed on adult female polar bears in the southern Beaufort Sea between 2009 and 2016. Some bears wore their collars for multiple years and some were collared more than once in non-continuous years. Collars provided GPS locations every hour with an accuracy of 30 m. Accelerometers recorded data from 3 axes and internally processed those values before outputting the data as a single activity value representing the 'active' seconds per minute. Internal processing of the data to determine 'active' was setup by the manufacturer of the collars - Telonics. Prior to satellite transmission or storage of data within the collar, the one-minute accelerometer activity counts were summed into one of three pre-deployment defined intervals (15 min, 30 min, 2 h and 3 h). Following internal processing, data were either transmitted via satellite or downloaded from recovered collars. Accelerometer data was quality controlled as previously described by Ware et al. (2017). Polar bear location data were not always collected at intervals that matched collection of the activity sensor data. Therefore, using the observed locations and associated accuracies, we used the R statistical computing package ‘CRAWL’ to predict polar bear locations for each activity observation as described in Rode et al. (2015). The CRAWL model accounts for variable location quality and sampling intervals and allows for location estimates to be obtained at user-defined intervals. We predicted locations using the CRAWL method only for data collected < 7 days apart though CRAWL has been validated to provide accurate locations when data gaps were up to 14 days (Rode et al. 2015). With the CRAWL-based location and the associated activity sensor value, we determined if a bear was on land or on sea ice at the time the activity data were recorded. Denning events were identified via temperature data using a control chart based algorithm to identify extended periods of warm temperatures that are indicative of denning events (Olson et al. 2017; Rode et al. 2018). Denning events have been removed from this data set.