Polar Bear Continuous Time-Correlated Random Walk (CTCRW) Location Data Derived from Satellite Location Data, Southern Beaufort Sea, 1986–2016

Data were obtained through the Argos System (www.argos-system.org) or the Iridium satellite system (Iridium Satellite Communications, McLean, VA, USA).

All polar bear satellite locations collected between 1986–2003 were Doppler-shift derived and estimated by the Argos system (CLS; http://www.argos-system.org/manual/ ). We performed an initial filter to remove implausible Doppler locations using the Douglas Argos-filter (DAF) algorithm (Douglas et al. 2012). DAF retained all standard-quality locations (Argos location classes 3, 2, and 1), and auxiliary location classes (0, A, B and Z) when corroborated by a consecutive location within a 10 km radius, or when movement rates were <10 km hr-1 and the internal angles (α, in degrees) formed by preceding and subsequent vectors (of lengths d1 and d2 km) were not suspiciously acute (α > -25 + β × ln[minimum (d1,d2)], where β = 15). We assigned β = 15 because it performed well for our specific tracking data across seasons and regions. Most polar bear locations collected after 2003 were derived by on-board geographic positioning system (GPS) receivers whose horizontal accuracy has been estimated as 30 m (Adams et al 2013). Doppler PTTs transmitted data during duty cycles that averaged 4–8 hr at intervals separated by 1–7 days. GPS PTTs recorded locations every 1–4 hr. As a result, the GPS data, which were collected after 2003 consisted of more frequent locations than the duty-cycled Doppler PTTs. To account for this discrepancy in sampling frequency, we subsampled the GPS location data to emulate the temporal gaps introduced by duty cycles that were used for the Doppler PTTs (Durner et al. 2017). Specifically, we quantified the frequency distribution of temporal gaps within the Doppler PTT location data, then for each GPS track, we randomly introduced temporal gaps ranging from 1–7 days in a manner that emulated the frequency distribution of 1–7 day gaps in the Doppler PTT data. To standardize location intervals across all annual tracking data subsets, we used continuous time-correlated random walk (CTCRW) model predictions to estimate a single location every 24 hr at 12:00 Coordinated Universal Time (UTC) based on the Argos and subsampled-GPS locations. The CTCRW model further provides standard error estimates for the average predicted locations based on the estimated error for Argos Location Classes and GPS locations, and the elapsed time differences with the empirical tracking locations.