75 procedure was implemented following the Wildlife Computers’ protocol, designed to minimize physical and behavioral impacts on the tagged individuals (Wilson and McMahon 2006; Hamelin and James 2018). Raw Argos locations are less precise than GPS positions, with errors generally exceeding several hundred meters from the actual position (Costa et al. 2010; Patterson et al. 2010; Jonsen et al. 2020). In order to improve localization accuracy, Argos data were filtered and smoothed using the R package ‘aniMotum’ (Jonsen et al. 2023) to provide quality-controlled locations (Jonsen and Patterson 2019; Jonsen et al. 2020, 2023). It is based on a continuoustime state-space model, with assigned error classes 3, 2, 1, 0, A and B provided in the Argos Least Squares format. Parameters and corrected positions were assessed through a ‘Move Persistence’ model. It accounts for irregular timing of observations due to data gaps as it supplies an index of how movement behavior varies in space and time, based on the autocorrelation of successive movements (Jonsen et al. 2023). For this study, we only focused on movements during the breeding period, in opposition to the migration period. Migration initiation was identified as a steep and persistent increase in movement speed, without a return movement. On the other hand, movements associated with the breeding period are characterized by low daily speeds and spatially limited movement patterns (Bourjea et al. 2013; Beal et al. 2022). The beginning of the migration was set as the first day of a minimum threshold of 3-consecutive days with a speed > 20 km.day-1 (Corbel 2020). In order to assess movement patterns and to identify potential shifts between the RP and the GLS areas, we calculated for each tagged individual: The duration of the telemetric survey; The estimated size of the distribution range, defined as the area bounded by the outermostrecorded positions of the tagged individual; Nm = the number of times individuals tagged on the RP nesting area traveled to the GLS nesting area, or the other way around (i.e. individuals tagged on the GLS nesting area traveling to the RP nesting area). Time duration and date of these nesting area shifts were also reported, as to take into account the inter-nesting interval temporality. The latter is estimated to be two weeks on the RP beach (Barbier et al. 2023). This information would allow us to distinguish putative nesting related behaviors from random roaming; Δ = the percentage of the distribution range overlapping that of the individuals tagged on the other nesting area (i.e. the combined distribution range of all individuals tagged on the GLS nesting area for individuals initially tagged on the RP nesting area, or the other way around).
RkJQdWJsaXNoZXIy MjE1NDI=