Fig. 2
From: Incorporating periodic variability in hidden Markov models for animal movement
Comparison of BIC-optimal state predictions for panther data for homogeneous transition HMMs (left panel) and heterogeneous transition HMMs (right panel), with movement step distribution models of varying complexity. For the time-heterogeneous model, all model variants estimate the same BIC-optimal number of behavioural modes (5); the homogeneous model is more variable, with the number of modes ranging from 5 to 7 or higher. Solid line: univariate response HMMs (turning angles ignored); dotted line: multivariate response HMMs (turning angles included, assuming a von Mises distribution). Red lines: log-normal step-length distribution; blue lines: Weibull step-length distribution