Bayesian hierarchical modeling of seven years of inter-stage survival rates of wild Atlantic salmon smolt and post-smolt from three rivers of eastern Canada

No abstracts are to be cited without prior reference to the author.

Mortality of Atlantic salmon post-smolt has been assumed to be highest in the first few months of migration at sea due to their small body size and the stress associated with acclimation to the marine environment. We report on research undertaken to estimate the location and timing of mortality of smolt during the first 2 months at sea. More than 1,700 wild Atlantic salmon smolt from three rivers of the Gulf of St. Lawrence (Canada) were acoustically tagged and released from 2003 to 2013. Acoustic arrays were first installed and monitored at the head of tide of each river, and at the exit of these rivers to the Gulf of St. Lawrence. In 2007 an array became fully operational in the Strait of Belle Isle (SoBI), the Gulf of St. Lawrence exit leading to the Labrador Sea, about 800 km from the point of smolt release. A Bayesian state-space model variant of the Cormac-Jolly-Seber model was used to disentangle the imperfect detection of tagged smolt on the acoustic arrays from apparent survival during their out migration. The model reduced uncertainty in expected values of the annual and river specific detection probabilities at the head of tide and bay exit arrays, however, it was not possible to independently resolve the detection probabilities at the SoBI array and the probability of survival through the Gulf of St. Lawrence. This telemetry research provides useful guidance in the design of such experiments and the treatment of data.