Behavioural state inference from movement and environmental data using Markovian step selection functions

Animal movement data often contain multiple behavioural states that vary across space and time, but identifying these states in a biologically interpretable way remains challenging. We propose an ecologically guided framework for inferring individual behavioural states from movement trajectories using hidden Markov models with step selection functions (HMM-SSFs). The method combines a small set of a priori, biologically motivated state hypotheses with explicit ordering and distinctness criteria on movement signatures, together with model-adequacy diagnostics and a stable fitting strategy. Unlike fully unsupervised approaches, our framework is designed to select among a limited number of ecologically meaningful behavioural states rather than exhaustively search a large latent state space. Simulations show that the method can recover the intended state structure under different behavioural scenarios, and that inference is sensitive to the specification of control locations in the SSF component. An illustrative application to a plains zebra trajectory demonstrates how the framework can be used to infer and label behavioural states from movement data. Overall, the proposed workflow provides a practical and reproducible way to infer individual-level behavioural states while maintaining ecological interpretability.