Calibrating tri-axial accelerometers for remote behavioural observations in Bewick's swans

Accelerometers in tracking devices are increasingly used to identify behaviour leading to detailed insights into the lives of free-ranging animals. To make proper use of an accelerometer, their settings and signals need to be tested and calibrated. Calibration of an accelerometer can be done by directly observing an individual animal of the species of interest, while an accelerometer is simultaneously measuring the movements of this individual. In case direct observations are difficult to obtain, this procedure can be performed with captive individuals. This study sought to calibrate the accelerometer sensor in GPS/GSM neck-collars in Bewick's swans Cygnus columbianus bewickii with observations of captive individuals in three zoos in the Netherlands. Using a random forest classification model, five behavioural classes were classified with an overall accuracy of 91%. An additional behavioural class (aquatic foraging) was identified based on a water sensor that was also included in the GPS/GSM collars. This classification was subsequently applied to accelerometer data from 12 free-ranging Bewick's swans equipped with these neck-collars to identify their behaviour during two spring migrations (2017 and 2018). The resulting time-activity budgets were in general agreement with current knowledge based on fragmentary field observations of Bewick's swan flocks along the flyway. The study shows how observations of zoo individuals can be instrumental to derive time-activity budgets of free-ranging individuals that can contribute to further research into the ecology of the species.