Observing the unwatchable through acceleration logging of animal behavior

This review synthesized how tri-axial accelerometers had become a workhorse for inferring behavior in animals that are difficult or impossible to watch directly, including cryptic and nocturnal carnivores. The authors explained how characteristic acceleration signatures can be classified into behavioral states such as resting, walking, running, and feeding, and how overall dynamic body acceleration serves as a proxy for energy expenditure. They reviewed analytical pipelines, from supervised machine learning trained on annotated data to validation against direct observation. For felid science, the framework enabled energetics and activity budgets to be estimated from collars on free-ranging cats. The paper is a foundational methods reference that helped standardize how acceleration logging is used to make the unwatchable observable.