Using the particle filter to geolocate Atlantic cod (Gadus morhua) in the Baltic Sea, with special emphasis on determining uncertainty

The use of archival tags on fish gives information of individual behaviour with an unprecedented high resolution in time. A central problem in the analysis of data from retrieved tags is the geolocation, namely the infererence of movements of the fish by comparing the data from the tags with environmental observations like temperature, tide, day length, etc. The result is usually represented as a track; however, the spatial and temporal variability in the precision is often substantial. In this article, the particle filter is applied to geolocate Atlantic cod (Gadus morhua) in the Baltic Sea, leading to a representation of the results as probability distributions for each time step, thus giving an explicit representation of uncertainty. Furthermore, the method is used to estimate the magnitude of the error in the measurements by the tags and the swimming velocity of the fish. The average swimming velocity during a day was estimated to be around 0.20 m·s–1 for fish of ~60 cm length. The method is general and the presentation is formulated to facilitate implementation for different systems where other quantities are observed.