University of Houston

Despite advances in tools for tracking movement of animals, and in the ability of computers to process massive amounts of data, our ability to predict the migratory movements of animals is quite limited. A mastery of animal movement patterns will improve our understanding of the interactions between the animal world and the ecosystem and therefore help ensure the sustainability of our ecosystem. To address this need, this thesis develops computational frameworks and algorithms to analyze motion pattern of animals, which are modeled as trajectories. In particular, distance functions for spatial and spatio-temporal trajectories are investigated and trajectory clustering and classification algorithms are designed and implemented. This framework is evaluated on the motion patterns of cattle, mule deer, and elk from the Starkey Experimental Forest and Range in northeastern Oregon. Our clustering results indicate that elk, mule deer, and cattle tend to stick with their own group. The occurrence of mix-and-match type of movement pattern is minimal. The overall purity of trajectory clusters, excluding about 20% outliers, is almost 90%. The average accuracy accomplished by a k-nearest neighbor trajectory classification approach is a modest value of about 66%. We also compared day and night movement patterns of animals, but found no significant differences between them: animals seem to move with the same companions during day and night.