Glue-on accelerometers decode Eurasian otter behaviour in zoo trial

A new methodological study now shows that a small tri-axial accelerometer glued temporarily to an otter’s fur can be used to recognise a range of behaviours from acceleration patterns alone, offering a potential route into the private lives of this vulnerable species.

Researchers tagged three adult male Eurasian otters (Lutra lutra) living under human care in zoos. Each tagged otter was observed for four to five days, with behaviours recorded simultaneously by video cameras and direct visual observation, creating a labelled dataset linking specific activities to their acceleration signatures.

The team combined these visual observations with the acceleration data to train an extreme gradient boosting model, a machine-learning approach, to classify 13 different behaviours during the tracking period based solely on the accelerometer output. When the trained model was applied to behavioural events that were new to it, it predicted the correct activity with an overall accuracy of 53.6%, with an F1 score of 0.506, a recall rate of 0.536 and a precision of 0.538.

Performance, however, varied considerably between behaviours. In particular, stationary behaviours were often not clearly distinguished from one another by the model. When the researchers grouped the stationary behaviours together — with the exception of latrine behaviour — accuracy rose to 68.2%, with an F1 score of 0.66.

Accelerometers of this kind measure body posture and movement-based proxies for energy expenditure, so refining the classification could eventually allow researchers to estimate not only what a free-ranging otter is doing but also how much energy different activities cost. The authors note that improving the attachment method with more glue to keep the tag on for longer would strengthen the approach, and they argue the method has great potential to reveal the ecological significance of behaviours and energetics in wild Eurasian otters.

Combined with spatial location data, the researchers expect such accelerometry to answer what, where, when and how wild otters allocate their activities. Understanding wild Eurasian otter behaviour, spatial movements and habitat use is expected to have great significance for the conservation and management of this vulnerable species. Because the current results come from three adult males in zoos, the findings represent a validated methodological step rather than a direct measurement of wild otter behaviour, and further work will be needed to confirm how the classifiers perform on free-ranging animals.

Subject of Research: Using glue-on tri-axial accelerometers and machine learning to classify Eurasian otter behaviours

Article Title: Glue-on tri-axial accelerometers code for different behaviours in Eurasian otters (Lutra lutra)

Article References: Rosell, F., Lodberg-Holm, H. K., Robstad, C. A., Kambestad, M., Wilson, R. P., & Mortensen, R. M. (2026). Glue-on tri-axial accelerometers code for different behaviours in Eurasian otters (Lutra lutra). Animal Biotelemetry. https://doi.org/10.1186/s40317-026-00504-x

Image Credits: AI Generated

DOI: 10.1186/s40317-026-00504-x

Keywords: Eurasian otter, Lutra lutra, biotelemetry, tri-axial accelerometer, animal-borne sensors, behaviour classification, machine learning, extreme gradient boosting, zoo study, wildlife conservation, Glue-on, tri-axial