Using accelerometers to remotely and automatically characterize behavior in small animals.

Area of Science:

• Animal behavior
• Biologging technology
• Wildlife research

Background:

• Direct observation of animal activity budgets is difficult due to time constraints and observer effects.
• Tri-axial accelerometers are useful for animal behavior studies but are often too heavy for small animals.
• Mapping accelerometer data to specific behaviors requires sophisticated analytical methods.

Purpose of the Study:

• To develop and validate lightweight accelerometers for small-bodied animals.
• To create a machine learning system for translating accelerometer data into behavioral categories.
• To apply this technology to study the behavior of free-living chipmunks.

Main Methods:

• Development of novel, miniature acceleration-logging devices (1.5-2.5 g).
• Collection of paired accelerometer and direct behavioral observation data from captive chipmunks.
• Application of machine learning algorithms to classify accelerometer data into distinct behaviors.
• Deployment and retrieval of accelerometers from wild chipmunks.

Main Results:

• Successful development of miniaturized accelerometers suitable for small mammals.
• Creation of an automated system for behavior classification using machine learning.
• Generation of behavioral data from free-living chipmunks using accelerometers.
• Demonstration of the first application of accelerometers for behavioral studies in free-living mammals under 100 g.

Conclusions:

• Miniaturized accelerometers offer a viable method for studying the behavior of small, wild animals.
• Machine learning effectively translates accelerometer data into meaningful behavioral insights.
• This technology opens new avenues for non-invasive wildlife behavior research.