The contact structure of Great Britain's salmon and trout aquaculture industry

Abstract

We analyse the network structure of the British salmonid aquaculture industry from the perspective of infectious disease control. We combine for the first time live fish transport (or movement) data covering England and Wales with data covering Scotland and include network layers representing potential transmission by rivers, sea water and local transmission via human or animal vectors in the immediate vicinity of each farm or fishery site. We find that 7.2% of all live fish transports cross the England-Scotland border and network analysis shows that 87% of English and Welsh nodes and 72% of Scottish nodes are reachable from cross-border connections via live fish transports alone. Consequently, from a disease-control perspective, the contact structures of England and Wales and of Scotland should not be considered in isolation. We also show that large epidemics require the live fish movement network and so control strategies targeting movements can be very effective. While there is relatively low risk of widespread epidemics on the live fish transport network alone, the potential risk is substantially amplified by the combined interaction of multiple network layers.

Keywords: Aquaculture; Control policy; Fish diseases; Giant component; Graph; Network.

Fig. 1

Breakdown of the 3517 geographical sites by location, site type, species group and presence of live fish movement records in the database. Site count breakdown for species group (multi, salmon, trout) is as follows:. England and Wales: 2441 total with 196 farms (23,5,168), 2245 other (13,83,2149); Scotland: 1076 total with 595 farms (44, 498, 53), 481 other (1, 7, 473).

Fig. 2

Plot of the British salmonid aquaculture industry live fish movement networks for salmon (top), and trout (bottom) species groups. Nodes in the Marine Scotland network are blue and nodes in the Cefas network red. Node size is proportional to the log of the number of connections. Node color scales to white depending on proportion of connections for which node is the source (i.e. white = source, full color = sink). Links have the colour of the source node. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article).

Fig. 3

The multi-layer network of the GB salmonid aquaculture industry for salmon (top), and trout (bottom). Link layers are colored by type: blue = river, local = green, transport = black and marine = orange. Nodes are colored by their dominant link type. Node color scales to white depending on proportion of connections for which node is the source (i.e. white = source, full color = sink). Links have the colour of the source node. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article).

Fig. 4

Impact on expected out-component size, EOR, of removing an increasing number of highly connected nodes from the transport layer. Nodes are selected at each step by ordering according to different criteria as defined in the Methods section: by degree in (green), degree out (red), or betweenness (blue), calculated on either the transport network layer (dashed line) or the full network (solid line). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article).

Fig. 5

Maps of site density on a 10 km grid (a) and spatial patterns of mean in-component size (b, d) and out-component size or reach (c, e) for sites in each grid cell, for transport (b, c) and combined network (d, e).