Triaxial accelerometry allows to determine chick feeding events in a marine bird

DEL CAÑO, MONSERRAT; QUINTANA, FLAVIO; DELL'OLMO, G.; GOMEZ LAICH, AGUSTINA

Congreso; 3rd World Seabird Conference; 2021

Over recent years, accelerometers have proven to be a powerful tool for the identification and quantification of animal behavior. Even though in seabird species this technique has allowed to identify activities such as diving, floating and flying, until now it has never been used to identify and analyze parental food provisioning events. In the present study, we applied the Support Vector Machine (SVM) algorithm to identify food provisioning behaviors from triaxial accelerometer data from breeding Imperial shags (Leucocarbo atriceps). A total of 15 breeding females were equipped with an accelerometer on the head which was programmed to record data at 50 Hz and left on the bird for less than 24 h. In order to label acceleration data, females were filmed during chick feeding and non-feeding events by means of a video camera. Fed chicks were categorized according to age in young and old (less and more than 7 days, respectively). Acceleration features from feeding and non-feeding periods were calculated using a sliding window of 1 s with a 0.98 s overlap. Finally, one SVM was run for each chick category, using a cost value of 1. SVMs correctly classified the feeding and non-feeding events in the 84 and 93% of the cases, for young and old chicks, respectively. However, the specific classification of food provisioning events (recall score), was correct for 79% of young and 95% of old fed chicks. Of the calculated features, those derived from the surge and heave acceleration axes and the pitch angle were the most important for distinguishing feeding from non-feeding behaviors. Our results suggest that accelerometry can be successfully employed to recognize food provisioning events in the Imperial Shag and can be applied in a wider spectrum of seabird species.