Identifying the source of farmed escaped Atlantic salmon (Salmo salar): Bayesian clustering analysis increases accuracy of assignment

Farmed Atlantic salmon escapees represent a significant threat to the genetic integrity of natural populations. Not all escapement events are reported, and consequently, there is a need to develop an effective tool for the identification of escapees. In this study, > 2200 salmon were collected from 44 cages located on 26 farms in the Hardangerfjord, western Norway.

This fjord represents one of the major salmon farming areas in Norway, with a production of 57,000 t in 2007. Based upon genetic data from 17 microsatellite markers, significant but highly variable differentiation was observed among the 44 samples (cages), with pair-wise FST values ranging between 0.000 and 0.185.

Bayesian clustering of the samples revealed five major genetic groups, into which the 44 samples were re-organised. Bayesian clustering also identified two samples consisting of fish with mixed genetic background. Performing self-assignment simulations with the data divided into different sub-sets, overall accuracy of assignment was 44% within the entire material (44 samples), 44% for the 28 spring samples, 59% for the 16 autumn samples, and 70% for 8 autumn samples collected from a geographically restricted area.

Accuracy of assignment varied greatly among the individual samples. For the Bayesian clustered data set consisting of five genetic groups, overall accuracy of self-assignment was 99%, demonstrating the effectiveness of this strategy to significantly increase accuracy of assignment, albeit at the expense of precision.

This study demonstrates the potential to identify the farm of origin for escapees in a region with a large number of salmon farms. The approaches described here will be of relevance to a range of other species reared in culture where identification of escapees may be required.