Public Critiques and Responses

On this page I am posting public critiques of my work on sea lice and salmon as well as responses from the coauthors and me.
Please contact me to contribute critiques.

Declining wild salmon populations in relation to parasites from farm salmon
Krkosek, M., J. S. Ford, A. Morton, S. Lele, R.A. Myers, & M.A. Lewis, 2007. Science. 318, 1772-1775

For more information on this paper see the paper itself, the public summary of the research results, and the press release.

There is a published critique and rebuttal between us and Ken Brooks (aquaculture consultant) and Simon Jones (DFO) in the journal Reviews in Fisheries Science. Read the full exchange here.

Criticism Wild salmon populations naturally fluctuate because of many factors. The study did not consider all the factors and so it is not possible to isolate the effect of sea lice.

Response We carefully controlled for the other factors that affect pink salmon populations and were able to isolate the effect of sea lice by using a comparative analysis. We studied pink salmon populations in a large region of British Columbia containing groups of populations that are exposed and unexposed to salmon farms. It is already known that all the populations fluctuate in synchrony (Pyper et al. 2001 Can J Fish Aquat Sci, 1501-1515). In the analysis we first compared the unexposed and exposed populations before the sea lice outbreaks and the productivity of these groups were nearly identical. Then when sea lice infestations occurred for the exposed group, the productivity of the exposed populations declined sharply whereas it remained unchanged in the unexposed group. Since all other factors except sea lice infestations are common to both exposed and unexposed populations, the natural conclusion is that sea lice cause the difference between the populations. This is based on standard statistical principles, dating back to R A Fisher, and is how causal inference is often conducted in observational studies.

Criticism By excluding the Glendale river from the analysis, the data were selectively analyzed to find an effect of sea lice on wild pink salmon populations.

Response  It is statistically necessary to exclude rivers that recently had spawning channels constructed. This was done systematically for rivers in the Broughton as well as the unexposed area to the north. Increased productivity due to spawning channels make it impossible to evaluate recent trends relative to historical abundance - which is the basis of analysis in the paper.

However, in response to this public criticism we have reanalyzed the salmon data with the Glendale data included. The re-analysis shows that the conclusions of the paper are upheld even when the Glendale data are included. See the plots below showing the increased productivity in the Glendale after the spawning channel was built (left plot) and the re-analysis with Glendale data included as black dots (right plot).

With and without the Glendale data included, the population growth rate for Broughton pink salmon is significantly negative. The negative growth rate means the populations are declining to extinction.

Population growth rate, r, with 95% confidence intervals:

With Glendale data r= - 1.002, 95% CI: -1.53 to -0.52
Without Glendale data r = -1.17, 95% CI: -1.71 to -0.59

The mean time to extinction (99% loss) is about 4-5 generations from the estimate including Glendale data. The mean time to extinction calculated in the original paper is 4 generations.

In summary, it is statistically appropriate to exclude rivers, such as the Glendale, where spawning channel enhancements were added during the study time. The results as reported in the paper are correct. These results, and the related conclusions regarding population decline, apply to the rivers included in the analysis. Although inclusion of the Glendale data is not statistically appropriate, this does not change the results of the paper.

Time series of relative abundances of Glendale pink salmon showing when the spawning channel was built, the subsequent increased productivity, and then declines in productivity during the sea lice infestations. The fallow cohort is also shown, indicating the strong return in that year. Relative abundance is calculated by first dividing the annual escapement estimates by the time series mean escapement and then taking the natural logarithm. The calculations were done separately for the odd and even year lineages of pink salmon.

Fit of the Ricker model to Broughton pink salmon escapement data during infestation years with Glendale data included. The y-axis is the log survival and the x axis is the lagged abundance. The y-intercept identifies r, the population growth rate.

Epizootics of wild fish induced by farm fish published in the Proceedings of the National Academy of Sciences of the USA. For more information on this paper see the paper itself, our public summary of the research results, and the press release. 

1. 'Critical Review' by Ken Brooks posted on the BC Salmon Farmer's Association Website - review and response
    

2. Press release from  Positive Aquaculture Awareness - press release and response
    

3. Press release from the British Columbia Salmon Farmers Association - press release and response
    

4. Press release from the Canadian Aquaculture Industry Alliance - press release and response