Estimating diadromous fish abundance during their migration by mark–recapture: remodeling combined with sequential Bayesian inference

In diadromous species, entire cohorts migrate through narrow river corridors. Capture–mark–recapture (CMR) protocols with double trapping allow to quantify abundance and to monitor migration phenology: individuals are marked at a first trap, some of which being recaptured at a second facility, that also collects unmarked individuals. We propose a novel model that includes a day of recapture effect on the probability of passage of marked individuals at the recapture trap. A Bayesian hierarchical approach allows joint analysis across a series of years using the same protocol. Annual migrants numbers are estimated using mild assumptions about the distribution of daily numbers, and an original Bayesian sequential procedure by which we first estimate the variations of the daily capture probabilities and then transfer this information into the estimation of migrants numbers. We demonstrate our approach with a case study of salmon smolts migration in the Scorff River (France) over 25 years. We discuss the merits of the Bayesian sequential procedure which prevents from any undue influence of the numbers of unmarked fish caught on capture probabilities estimation.