Data processing was carried out in the R statistical environment (version 4.1.0). The diveMove package (Luque, 2019) was used to zero-offset correct and account for drift in the pressure readings in the data downloaded from the dive behaviour loggers. Moreover, through the same package, tables were created summarising dive metrics including time of dive, dive duration, bottom time, post-dive duration, and maximum depth. To exclude nonforaging activities, a 5-meter threshold was applied. Furthermore, to omit dives that were followed by the animal resting, ones with post-dive duration above 2 minutes were not included in the analyses (Arnould & Hindell, 2001). Moreover, the descent and ascent rate (m/s) were calculated.

Dives were then categorised, firstly dividing them between normal dives and pause dives, with pause dives being characterised by more than 50% of the bottom time being at 0 m/s velocity. Normal dives were further divided into normal dives (ND) and APC dives (APC), with APC being characterised by a descent rate greater than 1.6 m/s as described by Volpov et al. (2015). On the other hand, pause dives were further separated into pause dives with a spike in velocity (SD) and pause dives without a spike in velocity (PD).

Statistical analyses were carried out in the R statistical environment. To understand if it is possible to recognise different patterns in descent rate, ascent rate, bottom time, and post-dive duration, Linear Mixed Effects (LME) models were fitted with an ANOVA through the lme4 package (Bates et al., 2015), giving the typology of the dive and the time as factors for the analysis. Moreover, “individual” was used as a random effect to understand the variance associated with the seal.

Linear and non-linear relationships were found between variables and descent rate, ascent rate, bottom time, and post-dive duration, therefore Generalised Additive Mixed Effects Modelling (GAMM) was used. mgcv (Wood, 2017) version 1.8-33 was used fitting GAMMs with a Gaussian distribution, testing for intercept effect by dive typology, time of the day, whether it was day or night, and a random intercept effect was included using individuals. GAMMs were used even to test the distribution of dives per typology during the day. Lastly, a correlation between the number of PD done by one individual, and the number of SD done by the same individual were tested with Spearman correlation.