Publication Abstract

Geolocation of North Sea cod using Hidden Markov Models and behavioural switching

M. W. Pedersen, D. Righton*, U. H. Thygesen, K. H. Andersen and H. Madsen

The assessment of the uncertainty of a tag based geolocation is essential to properly report the results of an analysis of archival tag data. Here, we describe an application of a direct Fokker-Planck based method to depth data from Atlantic cod (Gadus morhua) in the North Sea area providing the result in the form of a probability  distribution of the position and the most probable migration trajectory. In this study the geolocation relies mainly on the tidal patterns observed when a demersal fish spends a prolonged period of time at the sea bed. {Classification of differences in the vertical movement behaviour of cod provides a way to partition the time series into intervals of high and a low activity and allows for the application of separate movement parameters that increase the realism and the precision of the geolocation method}. A likelihood ratio test will reveal the statistical significance of this separation and tell if the fish shifted its behaviour during the time at liberty.  Furthermore will a two parameter model improve the uncertainty estimate compared to a time  constant activity level which tends to under- and overestimate the uncertainty in cases of high and low activity respectively.  Access to the probability distribution of the position of the fish provides a range of descriptive statistics and is a step towards making detailed population inference from archival data.

Reference

M. W. Pedersen, D. Righton*, U. H. Thygesen, K. H. Andersen and H. Madsen (2008) Geolocation of North Sea cod using Hidden Markov Models and behavioural switching. Canadian Journal of Fisheries and Aquaculture Science, 65, 2367-2377