Development and Validation of an Hmm-Based Geolocation Method Using Archival Tagging Data for Atlantic Cod

Geolocation methods using data from archival tags (including pop-up satellite tags) have been commonly applied to estimate daily positions of pelagic species.  However, the development and validation of alternative methods is required for geolocation of demersal species, because of considerable error in estimated positions that are based on light-based geolocation. We developed geolocation methods for Atlantic cod off New England using hidden Markov models (HMMs).  The approach is based on a modification of the observation likelihood and behavior models of an existing HMM framework and addresses both region- and species-specific challenges. The HMM emission probabilities are described by the likelihood model which compares environmental data recorded by the tag with those derived from an oceanographic model. Likelihood distributions are based on depth, temperature and tidal features. The transition probabilities are described by the behavior model which constrains the horizontal movement of the fish. Validation experiments were performed on stationary tags moored on the seafloor, double-tagged fish (archival tag and acoustic transmitter), and simulated tracks. Known data, including fish locations and activity level metrics, showed good agreement with those estimated by the HMM geolocation model. The developed methods  are expected to benefit geolocation applications for other species, regions, and tags types.