V for (In)valid: This week almost all the media feeds carried the news that a consortium of researchers from Scotland, Norway and the Faroe Islands are to create a new system to validate the results of existing sea lice dispersal models. The project, known as SAVED – Sustainable Aquaculture Validating Ectoparasite Dispersal (models) has just received a funding boost from the Sustainable Aquaculture Innovation Centre. The news raises a number of fundamental issues.
A variety of dispersal models are already available; however each model works with a different set of underlying assumptions meaning that they tend to return different results. This can be illustrated using a slide taken from the presentation made by Dr Meadhbh Moriarity, the senior aquatic epidemiological modeller for the Scottish Government’s Marine Directorate at the last Sea lice meeting held in the Faroes. Three different models had r values of 0.22, 0.25 and 0.66 showing a significant difference between them. This is because the various models use different complex mathematical techniques therefore the intention of this project is to reduce variability and ensure that all models produce a similar valid result if the same data set is used. This will be achieved using a bespoke Python script that can be applied to each model.
From my perspective this is all very puzzling. Certainly, during my analysis of the Norwegian NALO sea lice data, I found that the Sea Lice Expert Group used three different models to assess the impacts of sea lice for the Traffic Light System. Surely, if sea lice have an impact on wild fish, then they impact them in just one way, not three different ways. As Dr Moriarty points out, different models are built on different underlying assumptions and if they give different results then clearly some of these assumptions must be wrong. In fact, I would argue that the basic assumption of sea lice dispersal is fundamentally incorrect but as the modellers are working in their own “Model Land” and not the real world, such an error does not seem to even merit consideration by the modelling fraternity. Perhaps rather than develop a bespoke Python script that can be applied to every model to iron out the differences, the modellers should sit round and agree on just one set of assumptions and develop one universal sea lice model.
However, even one universal model does not resolve the underlying problem with this approach to sea lice interactions. I would suggest that in pursuing this project, the relevance of the project name appears to have been ignored. Validating Ectoparasite Dispersal suggests to me, and to others with whom I have spoken, that validating the model means investigating whether the model accurately predicts what is actually happening in the open sea, not ensuring that all models speak the same language.
Model validation is something that is largely ignored simply because previous attempts to do so have failed. If the model predicts large numbers of sea lice larvae in the sea at certain locations and times, then it should be possible to sample the location at the right time and identify large numbers of sea lice larvae that are supposed to be there. The fact that no-one has yet managed to do this is very much an inconvenience that is simply ignored, presumably until someone does actually identify large numbers of sea lice in the open sea, that is if they ever bother to look for lice in the sea again. If the sea lice larvae are not present in the sea as predicted, then they cannot be considered a risk to wild fish. Yet, in the unreal world of “Model Land”, modelled sea lice appear to present a risk to wild fish populations that must be managed.
The SPILLS project attempted to validate the Scottish Government model by sampling the sea water for lice. However, they only found 20. The fact that such a low number was identified has not deterred the use of models for as the SPILLS report stated ‘the failure to find sea lice larvae does not mean that they are not there’. They seem unable to accept that the reason they cannot find these larvae is simply because they are not there as the research has shown. Unfortunately, previously published peer reviewed papers published in reputable journals are not an obstacle to the use of models. The relevant science is simply ignored.
As I have written before, a paper published in 1996 using data collected in 1994 – that is now thirty years ago – reported on attempts to identify sea lice larvae in an Irish fjord. The research team had been commissioned to look for lice larvae at various points throughout the fjord but failed to identify any in exactly in the same way as the SPILLS project over twenty-five years later failed to do. However rather than just give up on their quest, one of the Irish team suggested sampling for lice at the farm – the only one in the fjord – and then radiating out away from the farm, which is exactly what they did.
The researchers certainly found lice larvae in the sea around the farm at between 2-3 larvae /m3 but as the distance from the farm increased, the concentration of larvae decreased, so by 1,000 metres away, the concentration was virtually zero. I believe that the concentration at that point was 1 larva /4m3, which can hardly be described as a soup, cloud, or curtain of lice. I met with one of the researchers earlier this year just before he retired. He described that at the time, they received a great deal of abuse with claims that the findings were falsified. This was very much in the same vein as when the Jackson paper was published in 2013. Those who promote the idea that salmon farming is responsible for the decline in wild fish simply attempt to denigrate any science that does not support their established narrative. For the same reason, there is a great reluctance to discuss any science that does not fit in with the claims made against salmon farming.
A Canadian paper presented at Sea Lice 2016 took a similar approach to the earlier Irish one and found sea lice were also diluted to very low levels within a short distance from the farm that they could not represent any threat to wild salmon.
What is interesting about the SAVED project is that SEPA are included as an observer. However, SEPA are taking a different approach to validating their model. This does not involve checking to see if sea lice larvae are present in sufficient numbers at the locations predicted by the models. Instead, they intend to use sentinel cages as a way of validating the model. This is despite the fact that the use of sentinel cages has been downgraded by Norwegian researchers.
It is now over eleven years since sentinel cages were used in Scotland at sites in Loch Linnhe. The original study which ran from 2011 to 2013 was also used as part of the SPILLS project. It is unclear why, if sentinel cages are so useful, the SPILLS project did not deploy a new set of cages in areas linked to the Shuna dispersal model. Instead, they preferred to rely on old data.
The point of SEPA’s risk framework is to limit sea lice numbers at the time of the smolt migration in spring, yet despite running the sentinel cages for three years in Loch Linnhe, not one sentinel cage showed any infestation during the spring migration window. Instead, any infestation linked to these cages occurred during the autumn period when adult fish carrying adult breeding sea lice could have infected the sentinel cages as they swam by on the return to their home river. Unfortunately, because the sentinel cages were deployed in relation to a model of sea lice dispersal, no consideration was given to other ways that the fish in the sentinel cages could be infested. It seems that any planned sentinel cage work demanded by SEPA will adopt a similar approach.
SEPA have identified 19 sites that they consider to be high risk in terms of sea lice infestation. One of these is located in Loch Harport on the south-western side of the Isle of Skye. Unfortunately, there is little information as to why this site is considered high risk. The farm is relatively isolated from other farming sites in the area and the local salmon rivers are not key rivers for salmon fishing. Because the farm is so isolated, there is a question mark as to how it might become infested when there are no other farms in the vicinity. It is also protected from any lice that might be carried from more distant farms by wind and current (although as discussed, this is extremely unlikely). The more likely route of infestation would be from passing wild fish and if this is how farming sites can become infested then this could also be true for sentinel cages. The accepted science is that once larvae latch onto a fish, they remain there until they become one of the mobile stages. However, there are examples of fish that have lost some of their non -mobile lice between two sampling occasions. As sampling the same fish in the wild is relatively rare, the issues this raises in terms of fish-to-fish transfer have never been explored.
Finally, the reason the story about SAVED was in the news was because of a press release from SAIC concerning the funding of this project. The fact that SAIC funded this project comes as something of a surprise, given their limited budget. Heather Jones of SAIC said that the project will help improve fish health by ensuring the prediction tool give reliable results. It is unclear how this tool will improve farmed fish health at all. The dispersal modelling is about impacts on wild fish not fish health. She suggests that the benchmark determined by this ‘fantastic example of international collaboration’ could have significant benefits in terms of helping bring about proportionate regulation and enabling the future growth and development of farming. Unfortunately, Heather is sadly misinformed. What it will do is yet further disadvantage the salmon farming industry by limiting development through use of a totally ineffective model.
Perhaps if SAIC wanted to advance work on sea lice, then it should have used the funding to honour its commitment to host Sea Lice 2024.
Of course, if I am wrong in my assessment of this project, I would be more than happy to discuss the issues with the Scottish Government’s Marine Directorate. This is extremely unlikely as they have made it clear in writing that the science of sea lice has been settled and is not open for further discussion. Perhaps, if they were so confident about their science then they would be more than willing to engage in such discussion, but they are not.
Lune too: The Lancaster Guardian reports that the Lancaster & District Angling Association have warned that wild salmon in the river Lune are heading towards extinction. According to the newspaper, avian predation, extremes in sea temperature and sea lice from commercial salmon farms are contributing to a dramatic drop in stocks.
Sadly, because the wild salmon sector refuse to discuss the issues involved in the interactions between wild and farmed salmon, they ensure that salmon farms continue to be blamed for wild salmon’s path towards extinction by an ill-informed angling community.
The angling club say that the fish counter at Forge Weir indicated that there has been a significant decline from around 8,000 salmon in the early 2000s to 3,500 in 2015. Since then (eight years) the counter has been out of action but information suggests that in 2018 only 1,908 salmon were counted. However, the same source told the paper in 2022 that the count was 1,384 in 2021 so it is strange that the 2018 figure was used this week.
I have uncovered a graph of counts prior to 2015 when the counter was damaged by Storm Desmond. The graph comes from the Environment Agency in a document dated 2020.
The decline in catches in the English river Lune appears to have more in common with the collapse of salmon catches from Scotland’s east coast rivers since 2010 than anything to do with salmon farming.
The angling association says that experience elsewhere has shown that improvements in habitat alone are not sufficient to repopulate failing rivers. The association is therefore calling for an expansion of hatchery activity as they believe that hatcheries will play an important part in securing the future of river Lune salmon.
Spey Hatchery: By coincidence, the Spey Fishery Board has posted that they have seen an incredible 900% increase in salmon fry at one of their hatchery stocked sites. The Green Burn, a tributary of the river Spey recorded less than 10 salmon fry last year but after changing their stocking strategy from planting out eyed ova to unfed fry, they have identified nearly 100 fry this year. These include some exceptionally large and healthy individuals. Currently, the Spey Fishery Board stocks out around 235,000 ova or unfed fry into burns which salmon cannot reach due to man-made barriers such as distillery weirs. However, if salmon cannot naturally reach these burns, how will any of these hatchery raised fish be able to reach the same burn if they manage to return as adults?
Last year, despite being the worst year for salmon catches on record, anglers fishing the Spey still managed to catch and kill 109 ready-to-breed adult salmon. Perhaps, if these fish had been allowed to breed, there would be more juveniles in the river without the need for a hatchery programme. Because the Scottish Government have yet to demand a real time catch recording system be imposed, there is no information available as to how many salmon have already been killed this year.
I find it rather strange that whilst the Scottish Government only permits conservation stocking, this is occurring in a long-established Grade One conservation river system. Either the river is Graded One and of premier conservation status or it’s not. Yet again, the angling sector effectively are allowed to do what they want, both in terms of recording data and stocking, whilst the salmon farming sector is strictly regulated.