12.07.2015 Views

Fraser River sockeye salmon: data synthesis and cumulative impacts

Fraser River sockeye salmon: data synthesis and cumulative impacts

Fraser River sockeye salmon: data synthesis and cumulative impacts

SHOW MORE
SHOW LESS
  • No tags were found...

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

Limitationso We only considered the most basic c<strong>and</strong>idate models. We did not incorporate any nonlinearor interaction terms. Even if we had incorporated these, it is doubtful that wewould be able to correctly identify the functional form of the relationships given thecomplexity of the underlying system. Regression would not be expected to detectrelationships that are not explicitly incorporated into the c<strong>and</strong>idate models <strong>and</strong> thereforeregression is not the correct approach for identifying complex relationships (e.g,involving interactions among more than 2 factors, time lags etc.) that have not previouslybeen hypothesized.o Many of the life stages or stressor categories had very sparse <strong>data</strong> <strong>and</strong> so conclusionsregarding the importance of a given life stage or project are limited to the covariates thatwere available.o There were no <strong>data</strong> available for the disease/pathogen stressor category <strong>and</strong> so nostatements can be made about the relative likelihood of disease/pathogens based on thequantitative analysis, although there is belief among many experts that this may be animportant factor.o There were no <strong>data</strong> available for the incubation-rearing life stage <strong>and</strong> so no statementscan be made about the relative likelihood of the incubation-rearing life stage beinglimiting based on the quantitative analysis.o Due to the large number of covariates we were asked to consider simultaneously we hadto assume the same relationship between covariates <strong>and</strong> all stocks. We couldn’t estimatea separate parameter for every stock, because it would increase the number of parameters18 fold. For example, we use a single parameter to represent the relationship betweenproductivity <strong>and</strong> sea surface temperature for all stocks. If we believed the relationshipwas different for every stock <strong>and</strong> wanted to estimate this relationship separately wewould need to include a unique parameter for all 18 stocks. The only stock specificparameter we estimated was the density dependence relationship (Ricker b term) as it wasshown to vary substantially among stocks (Peterman <strong>and</strong> Dorner, 2011).o Some of the aggregate indices we generated to reduce the number of parameters (e.g.,April-Aug mean sea surface salinity) may actually mask true underlying relationships.Determining appropriate ways to generate annual estimates of the physical covariates torelate to the annual biological response (<strong>sockeye</strong> productivity) was a major difficultyencountered during this project. Functional <strong>data</strong> analysis may be a better approach foraddressing this challenge (Ainsworth <strong>and</strong> Routledge, 2011), but will also be limited bythe sheer volume of <strong>data</strong> sets, potential hypotheses, <strong>and</strong> limited expert knowledge aboutthe underlying system.223

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!