Network analysis of farmed Atlantic salmon movements in British Columbia, Canada

Abstract

An inherent issue to the Atlantic salmon aquaculture production is the possible transmission of infectious pathogens due to the transportation of live fish. This study employed network analysis to model the contribution of Atlantic salmon transfers to the spread of pathogens. We used a publicly available salmon transfer dataset covering the period 2015-2022. Official records showed that 812 transfers of Atlantic salmon occurred between various British Columbian (BC) salmon production units in that timeframe. For the purpose of evaluating changes in the network structure of farmed Atlantic salmon movements, the daily networks were aggregated into two-year periods to generate a time-ordered series of biennial movements. The freshwater hatchery and marine netpen sites comprised the two types of facilities that made up the Atlantic salmon transfer network, which consisted of 99 nodes (facilities) and 350 edges (links) overall. All the networks showed both scale-free and small-world topology, which would encourage the persistence and spread of pathogens in the Atlantic salmon facilities while simultaneously making it easier to develop risk-based surveillance techniques by focusing on high centrality nodes. Additionally, the rare occurrence of high betweenness and reach, presence of disassortative mixing, negative correlation between the in- and out-degree and between ingoing and outgoing infection chain of facilities, and the identification of freshwater hatcheries as potential superspreaders all suggest that Atlantic salmon transfers might not play a significant role in the spread of pathogens between facilities in the British Columbian Atlantic salmon farming industry. Community detection revealed two or three communities persistently in the aquaculture management unit (AMU) level network, and it would be more effective to make zoning based on AMU. In conclusion, targeted surveillance efforts on high-centrality facilities can be employed to combat any infectious outbreak in the BC Atlantic salmon industry caused by live Atlantic salmon movement.

Keywords: Atlantic salmon; disease control; fish diseases; network analysis; risk-based surveillance.

Figure 1

Visualization of the frequency of Atlantic salmon transfers between different facility types (FH-freshwater hatchery, MN-marine netpen site) from 2015 to 2022 in BC, Canada.

Figure 2

The proportion of intra-and inter-aquaculture management unit (AMU) Atlantic salmon transfer flows in BC, Canada, from 2015 to 2022 (biennial and overall network) is shown by edge bundling. The weight of the arrows reflects the frequency of transfer between AMU. While inter-AMU movement is represented by arrows pointing to various AMU, intra-AMU movement is represented by arrows starting and ending in the same AMU. Each color represents a specific AMU.

Figure 3

(A) Visualization of the Atlantic salmon transfer networks (biennial and overall network) in BC, Canada. Circles and squares represent marine netpen sites and freshwater hatcheries, respectively. (B) Distribution of superspreaders and supersinks in different AMUs.

Figure 4

(A) Distribution of centrality parameters (ID-in-degree, OD-out-degree, TD-total degree, IC-ingoing infection chain, OC-outgoing infection chain, and NB-node betweenness) in different facility types (FH-freshwater hatcheries, MN-marine netpen site); (B) Spearman rank correlation between centrality parameters for the biennial and overall Atlantic salmon transfer networks in BC, Canada.

Figure 5

Degree distributions of Atlantic salmon transfer networks (biennial and overall) in BC, Canada. The cumulative frequencies of the node’s degree distributions are shown on a log–log scale. The maximum likelihood approach (38) approximated each degree distribution as a power law (black dashed lines showing power law for degree).

Figure 6

Effectiveness of targeted removal of nodes over the size of LWCC in different biennial and overall networks. Over a thousand simulations, random node removal is represented by the median value (gray line) and its 95 percentile (light blue shaded area).

Figure 7

Trade communities in different AMU-level Atlantic salmon transfer networks. Each shaded color represents a specific community within the transfer network. Arrows indicate the direction of transfers between different AMUs (red arrow- transfers between AMUs of different communities, black arrow- transfers between AMUs of same community).