Analysis of movements and behavior of caribou (Rangifer tarandus) using hidden Markov models

We explore how doubly stochastic, multiple-observation hidden Markov models (HMMs) may infer meaningful descriptions of woodland caribou (Rangifer tarandus) movement and behavior. Parameterized models allowed us to predict behavioral states (bedding, feeding and relocating), relative bout length and transitions, as well as most likely behavioral state sequences. Identification of state transitions and bout lengths appear specific to individuals and may identify dissimilar strategies of resource selection, behavior-specific habitats that are more important than is simply suggested by time spent there (pattern) and transitions between the same or different states that may be evidence for decision-making (process). Using only estimated model parameters, multiple-observation HMMs permitted us to successfully simulate movement and behavior representative of individual caribou through space and time.