GPS data: 81 individuals with useable GPS tracks (removed 2 that had few points) Penguins have been instrumented between 29th November 2018 – 14th January 2019 Captures were organised per round (8) at Bailey head + MAC point The GPS data have been preprocessed = Penguin_GPS_tidy Removed all Nas Removed locations with less than 4 satellittes Removed all locations that are positive (aka impossible) Match GPS data with penguins metadata Cut off by deployment date + 24h and by revovery date Removed D_R4_P10 that had a very high sampling rate (instrument failure?) Applied a speed filter (SDA filter) Build a CRAWL model for each track (no error model, couldn’t get it to work on GPS data with a constant error of 50m) Predicted regularly time spaced locations every 2 min based on the model (Penguin_GPS_tidyMOD) Penguin_GPS_tidy2 holds the clean dataset with the correct cut off dates at the start (right before the 1st trip and at the recovery date) TDR data: 3(4) instrument types: Axy-trek (“regular” and AGM), Mk9 WC and LOTEK 64 dive files associated with GPS data Each have a different decoding software and result in a different .csv file Each file is loaded and saved into a separate .Rds environment Each file is cleaned for Nas and only the TDR columns are kept The records are ordered by date Dives calibration For LOTEK instruments, I took the mean of the surface points (all the positive values after changing the sign of depth aka a point below the surface has a negative depth value). The mean of the surface points was between 2.3 and 1.7 m . That was the offset for each instrument used in the calibrDepth function. downsampled each record to 1 point every 15 sec. I tested that over a few chunks of records and between 1 sec resolution and 15 sec we are retaining over 92% of the dives and the ones that are lost are mainly very shallow dives (<5m) For MAC point we have 34 913 dives from 5 animals