Dissertation
The Development and Calibration of Techniques to Measure Energy Expenditure and Activity in Grizzly Bears
Washington State University
Doctor of Philosophy (PhD), Washington State University
2023
DOI:
https://doi.org/10.7273/000005379
Abstract
The manuscripts within this dissertation relate to using sensor data to identify grizzly bear (Ursus arctos horribilis) behaviors and quantify their energy expenditure. In chapter one we trained grizzly bears to walk on a treadmill on the horizontal, uphill, and downhill and measured their energy expenditure. We found that the costs of moving upslope increased linearly with slope angle and speed. Moving downhill was more costly than traveling horizontally at slower speeds but became less costly at higher speeds. We applied our estimates for the cost of travel to GPS data of grizzly bears in the Greater Yellowstone Ecosystem. Although their most efficient walking speed was 4.3±0.4 km h-1, they instead moved at an average velocity of 2.2±1.0 km h-1 and preferred traveling on near-horizontal slopes. In chapter two, we used three types of activity sensors; two reported a single variable and one reported raw triaxial accelerometer data. Sensor data was paired with video recordings of captive bears and used to develop random forest models. The single variable sensors were able to identify lying, stationary, walking, and foraging behaviors although accuracy was poor. However, accuracy improved by incorporating behavior derived from the previous minute in a new random forest model. Our triaxial sensor random forest model with seven behavior classes had an out-of-bag (OOB) error rate of 7.7%. Random forest models with 7 or 15 behavior classes and 1 s of previous behavior had OOB error rates of 0.13% and 0.04%, respectively. Our final model using 27 behavior classes and 4 s of previous behaviors had an OOB error rate of 3.1%.
In chapter three, we used doubly-labeled water to determine the daily energy expenditure (DEE) of captive grizzly bears. Indices such as growth rate and heart rate were significantly correlated with grizzly bear DEE. However, none were as precise as needed to understand the determinants of grizzly bear energetics in the wild. Grizzly bears, polar bears, and giant pandas had similar DEE when corrected for variation in distance traveled.
Metrics
8 File views/ downloads
78 Record Views
Details
- Title
- The Development and Calibration of Techniques to Measure Energy Expenditure and Activity in Grizzly Bears
- Creators
- Anthony Michael Carnahan
- Contributors
- Charles T Robbins (Advisor)Lisa A Shipley (Committee Member)Patrick A Carter (Committee Member)R. David Evans (Committee Member)
- Awarding Institution
- Washington State University
- Academic Unit
- Biological Sciences, School of
- Theses and Dissertations
- Doctor of Philosophy (PhD), Washington State University
- Publisher
- Washington State University
- Number of pages
- 76
- Identifiers
- 99901031441001842
- Language
- English
- Resource Type
- Dissertation