Identifying salmon lice transmission characteristics between Faroese salmon farms

9 Sea lice infestations are an increasing challenge in the ever-growing salmon aquaculture sector 10 and cause large economic losses. The high salmon production in a small area creates a perfect 11 habitat for parasites. Knowledge of how salmon lice planktonic larvae disperse and spread the 12 infection between farms is of vital importance in developing treatment management plans to 13 combat salmon lice infestations. Using a particle tracking model forced by tidal currents, we 14 show that Faroese aquaculture farms form a complex network. In some cases as high as 10% 15 of infectious salmon lice released at one farm site enter a neighboring fjord containing another 16 farm site. Farms were characterize as emitters, receivers or isolated, and we could identify 17 two clusters of farms that were largely isolated from each other. The farm characteristics are 18 a valuable input for the development of management plans for the entire Faroese salmon industry. 19 20 Running page head: Salmon Lice Transmission Characteristics between Faroese Salmon Farms 21


INTRODUCTION 23
In salmonid aquaculture, the infestation of parasitic organisms is a major challenge and causing 24 significantly economic losses in the aquaculture industry. The treatment cost are estimated to be 25 around e 0.25/kg (Costello 2009)   As in other temperate coastal areas, the production of Atlantic salmon in the Faroe Islands has 33 expanded to become a major activity. With an annual production now exceeding 80,000 ton the 34 Faroe Islands is currently the fifth largest salmon producing country, and the aquaculture industry found on salmonids, often referred to as salmon lice, while C. elongatus is a more opportunistic 44 parasite, and has been found on 80 different fish species (Kabata 1979). Of the two species L. 45 salmonis has by far the largest economic impact on the salmonid aquaculture industry due to its 46 damaging effect on its host (Boxaspen 2006). 47 The increased salmon production has elevated the density of the naturally occurring salmon lice 48 in the water column primarily due to the large growth in the number of hosts. The high density 49 further increases the chances of transmitting salmon lice between hydrodynamically connected farms. 50 Additionally, experience shows that sporadic uncontrolled treatment of salmon lice in an area yields 51 a resistant salmon lice population over time (Murray 2011, Aaen et al. 2015). Farms with intense 52 1 treatment might quickly develop resistant salmon lice strains. In inter-connected farm networks the 53 resistant strains spread out to the whole network after a given period. Therefore, in connected farm 54 networks, there may be a point when the external infection pressure reaches a stage when, from 55 a farmers perspective, treatment becomes virtually pointless. Coordinated treatment management 56 plans are thus essential to achieve long term sustainable control of salmon lice. Achieving control 57 requires a thorough understanding of the sea lice dispersion patterns, a factor highly dependent on 58 regional and local hydrography (Adams et al. 2015).

59
Dispersion of salmon lice larvae has been studied using numerical models in Scotland (Amundrud  (2008) showes that the tides through rectification are the main driver of a steady current clockwise 91 around the islands (Fig. 2).

92
In some of the fjords, and in the strait between the two biggest islands (white arrowhead, Fig. 1 where ϕ i is the Greenwich phase lag, θ i the inclination and a i and b i are the major and minor semi- lated with an Euler scheme where the position (x, y) of particle q at the next time step, t + ∆t, is 126 calculated from the velocity, u and v: where R is a uniformly distributed random number between [−1, 1]. D h is the diffusion coefficient  to the fjord opening is included (Fig. 3). The model is barotropic, and thus no attempt is made 175 to simulate the estuarine circulation which may appear in fjords with little tidal influence (Fig. 1).

176
The neglected estuarine circulation is, however, of importance for the dispersion within these fjords, 177 and consequently the receiving farm area (or "hit area") is subjectively defined as the entire fjord if 178 the farm is located inside a fjord or the long narrow strait between the two largest islands (Fig. 3).

179
A copepodid larvae entering these areas is considered an infection risk to the whole area. Farms 10°C water temperature. A particle was only allowed to infect a given farm site once, meaning that 194 if they re-enter the same farm site the particles were not recorded. However the particle was allowed in tidally relatively exposed areas allowing larvae to escape their initial release site, but at different 209 rates.

210
Particles disperse rapidly from farm 5 (Fig. 4b), which is located in the energetic strait on the 211 west side of the largest island in the center. In contrast, the particle dispersion from the other two 212 farm sites (Fig. 4a and c)  week lifespan, provided that they manage to escape their release area. Further, the clockwise residual 218 current also implies that only few particles are lost from the system.

219
The maximum Euclidean distance traveled by a particle varies greatly between farms. Particles

227
The relative distribution of nauplii and copepodid particles in each 100x100m grid cell over the 228 whole simulation period, including mortality, is shown with a heat map (Fig. 5). Here, particles stage. On the other hand particles released from farms in more exposed areas as farms 5 and 9 have 232 9 relativity low densities at their initial release sites, indicating the quick dispersion away from these 233 areas after being released. No clear "cold" spots, i.e. areas with relatively low density of copepodids, 234 are in the coastal regions. The exception is some of the fjords with little tidal influence where 235 estuarine circulation, which is neglected in the present simulation, must be expected to dominate 236 the dispersion (Fig. 5b). The highest density is found around the northern group of islands, which 237 is also where the majority of the farms are located.

238
Connectivity 239 The proportion and mean age of infectious salmon lice larvae that disperse between Faroese salmon 240 farms is summarized in three connectivity matrices (Fig. 7). The connectivity matrices reveal gen- the area. Farm 5, 9, and 13, which are located in tidally exposed areas (Fig. 3), are clearly not 244 self-infectious ( Fig. 7a and c). The farms 10-12, 20, 22, and 23 are close to 100% self-infectious and 245 emit very few larvae to other farms, while also having a relatively (to the amount emitted) high 246 infection rate from other farms.

247
The mean age in the highly self-infectious connections is close to 3.7 days, the time when larvae 248 become infectious, while the mean age for non-self-infectious farms is much higher, between 8-14 249 days (Fig. 7b). Larvae from farm 5 infecting farm 9 are very old (14-16 days) and vice versa, even 250 though they are very close geographically implying that the particles have traveled a long distance 251 before they enter the neighbour site. Interestingly, however, larvae from these farms infect nearby 252 farms (4, 6, 7 and 8) with younger larvae. km for a few sites (Fig. 6a). Including the life span of the copepodid stage there is a fairly high 276 probability for the majority of the sites that the maximum distance is beyond 50 km and some few 277 sites there is up 10% probability that they even reach beyond 80 km (Fig. 6b), which are distances

286
The self-infection in a number of farms is quite high, as seen by the low mean age and small 287 dispersion range in these connections (Fig.'s 7 and 5). The high self-infection is partly caused by the 288 low water fluxes in and out of these fjords, which may be underestimated due to the omission of the 289 estuarine and wind driven circulation, and that the particles stay within the initial receiving area 290 when becoming infections and therefore recorded the moment they become infectious (3.7 days).

291
Identifying critical nodes in the farm network is highly valuable information when developing 292 a management plan. We were able to identify farms either as emitters, receivers, or isolated. The to the close neighbour, farm 15, which is the highest receiving farm (Fig. 3). Farm 16 emits over 296 10% of its infectious salmon lice to farm 15, which must be considered a very strong connection.

297
One reason farm 15 is the highest receiver is that the defined receiving area is quite large as this 298 fjord is relatively wide. In addition, the connection to the strait outside this fjord includes a tidally

305
The connectivity matrices suggest that the three farms at the southernmost islands (farms 1-3) 306 are largely separated from the other islands, but are internally well connected. The rest of the farm 307 network seems to be one cluster with negligible contribution from the three isolated farms (10, 12 308 and 20) and the three relatively isolated farms (11 and 22-23) (Fig. 8). These isolated farms are 309 in the narrow strait between the two main islands and in fjords in the northeast group of islands, 310 which all are characterized by weak tidal currents (Fig. 1). Likely the dispersion within these areas these farms must be considered as isolated and highly self-infectious. in our simulations did not leave their own receiving area before becoming infectious because we 329 defined the whole fjord to be a potential infectious risk. This is evident when looking at the mean 330 age in most of the self-infection connections (Fig 7b). It would be unpractical in our setup to 331 take most particles out of the simulation after one connection as we would then in most cases 332 only observe self-infection. Therefore particles were allowed to continue in the simulation but they 333 were only allowed to infect once in any given connection. One downside with this method is that in the wind forcing it is likely that the overall distribution will be even more smeared out than 360 obtained here from the underlying tidal forcing only, but this still remains to be investigated.

361
The results presented have many implications which can benefit the Faroese aquaculture industry, responsible for the connectivity between farms sites as well as acting as a retention mechanism for 393 the resident sea lice population. 394 We acknowledge that wind and freshwater forcing, which are not included in the present study, 395 will influence the dispersion dynamics, especially on shorter time scales in the more sheltered fjords.

396
However on longer timescales the highly dominating Faroese tidal forcing will reflect the mean 397 dispersion pattern, enabling valuable insight on the background connection between farms in Faroe 398 Islands.

399
In summary, the basis is developed to create a robust biophysical model which can help find an 400 optimal treatment and management plan for the Faroese aquaculture industry.   There is water connection between farm areas 11 and 12, but not between 19 and 22 and 23 and 24.