Unsupervised Bayesian reconstruction of individual life histories from otolith signatures: case study of Sr : Ca transects of European eel (Anguilla anguilla) otoliths

The reconstruction of individual life histories from chemical otolith measures is stated as an unsupervised signal-processing issue embedded in a Bayesian framework. This computational methodology was applied to a set of 192 European eel (Anguilla anguilla) otoliths. It provided a robust and unsupervised analysis of the individual chronologies of habitat use (either river, estuary, or coastal) from Sr:Ca measures acquired along an otolith growth axis. Links between Sr:Ca values and habitat, age, and season and the likelihood of the transitions from one habitat type to another were modelled. Major movement characteristics such as age at transition between habitats and time spent in each habitat were estimated. As a straightforward output, an unsupervised classification of habitat use patterns showed great variability. Using a hidden Markov model, 37 patterns of habitat use were found, with 20 different patterns accounting for 90% of the sample. In accordance with literature, residence behaviour was observed (28% of the eels). However, about 72% changed habitat once or several times, mainly before age 4. The potential application of this method to any other measures taken along an otolith growth axis to reconstruct individual chronologies gives a new insight in life history tactics analysis.