Multifaceted framework for defining conservation units: An example from Atlantic salmon (Salmo salar) in Canada

Conservation units represent important components of intraspecific diversity that can aid in prioritizing and protecting at-risk populations, while also safeguarding unique diversity that can contribute to species resilience. In Canada, identification and assessments of conservation units are done by the Committee on the Status of Endangered Wildlife in Canada (COSEWIC). COSEWIC can recognize conservation units below the species level (termed “designatable units”; DUs) if the unit has attributes that make it both discrete and evolutionarily significant. There are various ways in which a DU can meet the criteria of discreteness and significance, and increasing access to “big data” is providing unprecedented information that can directly inform both criteria. Specifically, the incorporation of genomic data for an increasing number of non-model species is informing more COSEWIC assessments; thus, a repeatable, robust framework is needed for integrating these data into DU characterization. Here, we develop a framework that uses a multifaceted, weight-of-evidence approach to incorporate multiple data types, including genetic and genomic data, to inform COSEWIC DUs. We apply this framework to delineate DUs of Atlantic salmon (Salmo salar, L.), an economically, culturally, and ecologically significant species that is also characterized by complex hierarchical population structure. Specifically, we focus on an in-depth example of how our approach was applied to a previously data-limited region of northern Canada that was defined by a single large DU. Application of our framework with newly available genetic and genomic data led to subdividing this DU into three new DUs. Although our approach was developed to meet the criteria of COSEWIC, it is widely applicable given the similarities in the definitions of a conservation unit.

Methods
Genomic data here includes data used for Labrador, Canada, in the main analysis of the paper. Genomic data from a 220,000 SNP array (see Barson et al., 2015) developed using a targeted, biallelic SNP Affymetrix Axiom array by the Centre for Integrative Genetics (CIGENE, Ås, Norway) were compiled for populations of Atlantic salmon. Genotyping was performed using the Affymetrix Axiom array protocol (Thermo Fisher Scientific), with raw data processed using the Axiom Analysis Suite (AAS) software based on the manufacturer's best practices protocol. AAS assigns all SNP assays into different categories based on clustering patterns, and only SNPs that were classified as high quality were used for subsequent analyses. SNPs were filtered using PLINK v1.9 (Purcell et al., 2007) for minor allele frequency (MAF) of 0.05.