Assessing the utility and limitations of accelerometers and machine learning approaches in classifying behaviour during lactation in a phocid seal

Table 1 Ethogram of female grey seal behaviour during lactation

Behaviour	Description	Prop.
Rest	Seal is motionless, head on ground	0.581 ± 0.21
Alert	Seal is stationary, head is up and moving, may look at pup	0.257 ± 0.15
Presenting/Nursing	Female is rolled laterally so that nipples are presented to pup; pup may be nursing	0.075 ± 0.06
Flippering pup	Female using fore-flipper to lightly stroke/scratch pup	0.021 ± 0.02
Misc. mother–pup interaction	Any other active interaction between female and pup	0.009 ± 0.01
Locomotion	Directed movement > 0.5 body lengths; seal is moving on ventral surface using fore- and hind-flippers	0.014 ± 0.03
Comfort Movement	Female performing low intensity movements that do not change geographic location (< 0.5 body lengths)	0.053 ± 0.02
Social/aggression	Any non-aggressive behaviour such as nosing other females (social), or aggressive interaction such as open-mouth threats or biting/lunging (aggression)	0.017 ± 0.02

Behaviours observed in lactating grey seals (n = 53) over two breeding seasons sampled over 137 h of video data. Behaviours in italics were excluded from the final analysis due to their extremely rare occurrence in the video footage (< 2 min per individual in most cases) and inconsistent feature characteristics in the accelerometer signal. Behavioural categories were largely adapted from [38, 51, 52]. The mean proportion of time in each behaviour across individuals (prop ± standard deviation) is included to give an indication as to the approximate time and variability for each behavioural state

ISSN: 2050-3385