Thursday, December 23, 2010

Sea Lice PR Science Runs Wild

For about two weeks now the media has been flooded with stories about a new publication examining the impacts of salmon farming on populations of wild pink salmon. The study was done by a group of three researchers using data on lice levels in fish farms in the Broughton Archipelago. Data which until now has been proprietary and the industry has refused to release to independent scientists. Previous work by Marty Krkosek and collaborators in 2007 have documented the impact of the salmon farming industry on wild salmon and a 2010 paper by Brendan Connors found that sea lice infestations were also depressing the productivity of Broughton coho salmon. Now as the Canadian public and government appear poised to push the industry towards closed containment that would finally eliminate the transfer of parasites and disease from salmon feedlots to wild juveniles, the aquaculture industry releases this piece of PR science, hiring three virtually unknown scientists, including a veterinarian who has worked for the salmon farming industry for over 15 years. This conflict of interest is clearly stated on the first page of the paper.

With pressure mounting on the salmon farming industry they sought to infuse uncertainty into the debate, and rather than testing a set of hypotheses in an objective fashion the scientists appear to have set out to show that salmon farming has no appreciable impact on the productivity of wild salmon. This is glaringly obvious in a few parts of the paper. Among the most astonishing quotes are,

"all published lab and field data support the conclusion that something other than fish farms caused the population decline in 2002."

This is simply not true. Previous population modeling has absolutely attributed declines to high parasite loading coming from salmon farms. At the height of the sealice outbreaks upwards of 90% of wild juveniles were infected with sealice. Pink salmon juveniles lack scales and therefore are highly vulnerable to parasites.

They then go on to discuss laboratory studies of sea lice infection on juvenile pink salmon. Saying,

"Because most or all of the initially infested fish shed their lice before they died, the cause of death remains unknown, and the evidence points to something other than sea lice killing most of the fish."

This is absolutely ludicrous. Ecotoparasites like sea lice are more than capable of leaving hosts at any time and have even been shown to move from juvenile pinks to larger predatory coho salmon during the act of predation. The fact that they aren't on the juvenile pink salmon at the time of death by no means vindicates the sea lice. That would be the equivalent of saying that because someone died of a bullet wound after the murderer has fled the scene that they did not die because of the individual who shot them. The authors also cite relatively low mortality of infected pink salmon in the lab environment, again this is just not biologically relevant. In the wild, juvenile pink salmon don't just die. They fall victim to predators, they are lost to disease, they starve for lack of food. In a laboratory environment the effect of sea lice cannot be fully accounted for because the actual sources of mortality are largely absent and the fish are fed on a daily basis. In the wild, infected pink salmon fry grow more slowly and are much more likely to fall victim to predators and disease.

that's not a healthy salmon

The analysis done by Marty et al. found what it was looking for. Part of the problem is the sea lice records only go back as far as 2000. Prior to that the farms did not document sea lice levels, however they had become a problem before that. Pink salmon are naturally extremely variable in their productivity and with a short time series the effect of a perturbation like high sea lice loading is likely to be masked by natural variability. Furthermore for the entire second half of the sea lice timeseries 2000-2009 the farms were aggressively treating fish with slice, an anti-sealice medication that can dramatically reduce sea lice densities, at least until they develop resistance.

I wondered however if I did a similar analysis, what I would find. Just so happened that I had the DFO escapement/runsize data for the Broughton archipelago on hand from a class project. By comparing the number of spawners in a given year to the number of recruits two years later (all pink salmo return to spawn at age two) we can make inferences about the relative productivity in a given year and model the effects of various factors ranging from environmental conditions, and spawner density to the number of farm salmon in the broughton archipelago. My time series went from 1976 to 2004 so I was unable to capture the latter portion of the period in question, but given the fact that the salmon farming industry started in the Broughton in the early 1990s that gave me 14 years of salmon farm data to use. Keep in mind this is by no means peer reviewed, but I think it provides some valuable insights.

I used what is called a stepwise model selection, basically starting by just modeling the relationship between spawners and spawners/recruits log transformed so that the data fits the assumptions of a linear model.

relationship between spawner abundance and log(r/s) in the Broughton Archipelago. x-axis is spawner abundance, y is log (r/s)

Not surprisingly I found a highly significant negative relationship between spawner abundance and the per spawner productivity. This is commonly seen in productive populations of salmon where densitry dependence can limit population growth. Basically the more spawners, the less available resources for each individual fish and the population growth rate declines in the subsequent generation.

The next step was to add in farm salmon abundance into the model. Data is in gigatons. Using a full factorial linear model, there is a nearly significant negative relationship between productivity and spawners*farmed fish (p=0.085) (see below for an explanation of p-values) adding farm salmon to the model also improved our model fit (R^2) from 0.56 to 0.62. While interactions between factors in a model can be difficult to interpret, this one is actually fairly straightforward. Basically in our linear equation spawners*farmsalmon has the effect of -0.00000008391, the number may seem small, but in multiply that coefficient by the spawners*farmsalmon and we see it can actually have a pretty profound impact. For 2000 spawners*farm salmon is 25276888 meaning that on average log (recruits/spawner) is reduced by -2.12. That's a considerable impact.

This is because, as Marty et al. demonstrated convincingly in their paper, the density of sealice on salmon farms during the spring outmigration is actually closely associated with the abundance of adult pink salmon the previous fall. Adult pink salmon naturally carry sea lice, however normally they die prior to the emergence of their offspring meaning that juveniles rarely encounter sealice until they are large enough to fend off the parasites. Instead, salmon farms serve as refuges for sealice, which become extremely abundant in the high density feedlots. So despite the lack of sea lice data before 2000, the spawners*farmsalmon interaction probably does the best job of characterizing sea lice abundance in the Broughton prior to 2000.

Figures from Marty et al. in PNAS

A is the abundance of sea lice relative to the abundance of spawning pink salmon. Note that in years of high spawner abundance there are more lice in farms.

B is the prevalence of sealice on juvenile pink salmon relative to the abundance of lice on farms. No surprise there, lice from farms are ending up on wild juveniles.

While this is basically a back of the napkin analysis, it demonstrates just how robust the conclusions drawn by Marty et al. PR science has no place in a reputable journal like the Proceedings of the National Academy of Science and the salmon farming industry quickly revved its. PR machine to ensure maximum media coverage. Lets hope the reading public are smarter than the gullible journalists because the wild salmon are the ones paying the price for the industry's unwillingness to take farms out of the ocean and move to closed containment.

*a p value simply describes probability of observing the given result randomly. In ecology a p-value below 0.05 is considered significant, however in other fields a p of 0.10 is significant, and given the high degree of variability this sort of result should give readers of the Marty et al, sea lice PR paper more than a little pause in accepting their results at face value.

1 comment:

Media Mentions said...

You know, I have to say that I’m actually incredibly relieved that some salmon farms are finally taking an initiative into a more sustainable direction and away from past practices, such as the ones recently outlined online in the press ( Still, with the increasing population and thus demand, it will take more than lukewarm measures to ensure a good, stable source of salmon and other seafood.

Well anyway, some food for thought :]