Geolocation of North Sea cod (Gadus morhua) using hidden Markov models and behavioural switching

Abstract

When geolocating fish based on archival tag data, a realistic assessment of uncertainty is essential. Here, we describe an application of a novel Fokker–Planck-based method to geolocate Atlantic cod ( Gadus morhua ) in the North Sea area. In this study, the geolocation relies mainly on matching tidal patterns in depth measurements when a fish spends a prolonged period of time at the seabed with a tidal database. Each day, the method provides a nonparametric probability distribution of the position of a tagged fish and therefore avoids enforcing a particular distribution, such as a Gaussian distribution. In addition to the tidal component of the geolocation, the model incoporates two behavioural states, either high or low activity, estimated directly from the depth data, that affect the diffusivity parameter of the model and improves the precision and realism of the geolocation significantly. The new method provides access to the probability distribution of the position of the fish that in turn provides a range of useful descriptive statistics, such as the path of the most probable movement. We compare the method with existing alternatives and discuss its potential in making population inference from archival tag data.

Link between vertical and horizontal movement patterns of cod in the North Sea

“…This is likely to give a conservative estimate of location accuracy since in practice the fish may have been moving, to some extent, during these periods, so the scatter of locations we generated would be produced by a combination of geolocation errors and fish movement. The mean error estimated from our method was approximately 8 km, which accords with estimates provided from data collected by stationary tags tested in the hidden Markov model geolocation method (Pedersen et al 2008). To put this error into context, it is comparable to the worst Argos locations (Hays et al 2001b) and is an order of magnitude better than light-based geolocation (Teo et al 2004).…”

“…This method is referred to as the tidal location method (TLM, as described in Metcalfe & Arnold 1997;Hunter et al 2004b). We used a novel Fokker-Planck-based method that combines the TLM with a hidden Markov model to estimate, for each day at liberty, the non-parametric probability distribution of geographic position (for details see Pedersen et al 2008).…”

“…However, for cod and many other demersal species, the linkage between vertical and horizontal movements is poorly defined (Arnold et al 1994). Given this interest in both vertical and horizontal movements and the phenology of migration, combined with the commercial value of cod, here we set out to use a recently developed tide-based geolocation method (Pedersen et al 2008) to assess the horizontal movement of Atlantic cod G. morhua. In the present study, we reconstructed the tracks of cod in the North Sea and English Channel over periods of many months and assessed how their horizontal movements were linked to their vertical movement behaviour, and how these varied over periods of up to 14 mo.…”

Over 10 million seawater temperature records for the United Kingdom Continental Shelf between 1880 and 2014 from 17 Cefas (United Kingdom government) marine data systems

“…Data from tags that were returned from recaptured cod were downloaded and the depth time series was used to estimate daily geographic location. This was done by matching the tidal and maximum depth data to known dates and locations as per the method described in Pedersen et al (2008). Temperature data from each tag were binned into 10 m depth intervals and then averaged.…”

Enhanced polymeric encapsulation for MEMS based multi sensors for fisheries research

“…The DST is aimed to be placed on fish and when the fish are re-caught the data is read out either by physical contacts in the DST or by powering the system through a wireless interface. The data can then be used to estimate the fish migration route by geo-location [8][9][10].…”