Discipline: Ecology Environmental and Earth Sciences
Bethany Johnson - Sonoma State University
Co-Author(s): Patricia Puerta, Oregon State University, Corvallis, OR
Stock assessment of Pacific cod in the Eastern Bering Sea relies heavily on accurate estimates of size-at-age relationships of the species. Presently, there is speculation that traditional assessments of Pacific cod size-at-age are inconsistent with a true representation of the population, due to an inadequate age data sampling strategy. It is debated whether the sampling strategy in practice appropriately accounts for environmental variables, which are known to affect fish growth and influence the size-at-age function. By evaluating discrepancies between the traditional length stratified sampling and a novel random stratified sampling, we aim improve stock assessment of one of the most ecologically and economically important Alaskan groundfish species. Data obtained by these two methods in 2015 were used to characterize difference in determining size-at-age models, interpreting effects of environmental variability in size-at-age, and predicting the age distribution of the population. Additionally, a complete sampling strategy evaluation in which both sampling strategies were simulated on a virtual population of Pacific cod was used to test the efficacy and accuracy of each protocol. Results indicated that the length stratified sample is biased in that it does not adequately account for spatial variability in Pacific cod size-at-age. Alternatively, it was found that the random stratified sampling strategy is able to represent spatial heterogeneity, making the design a more precise predictor of age data observed in the population. However, the sampling strategy evaluation suggests that although random stratified sampling is more precise, both sampling strategies lack in accuracy, thus implying that there is some bias in biomass estimates from both sampling strategies. Additional analyses on the effects of environmental conditions (e.g., warm vs cold years), sampling, and age measurement errors are currently underway to characterize the conditions under which one sampling strategy may prevail over the other.
Funder Acknowledgement(s): We would like to thank the National Science Foundation for generously providing funding that made this research possible.
Faculty Advisor: Lorenzo Ciannelli, email@example.com
Role: My contribution to the project involved statistical comparisons of two sampling strategies: random stratified sampling and length stratified sampling. Using R software, I developed size-at-age models with nonlinear regression techniques, and I evaluated possible models for goodness of fit. I helped determine which sampling strategy accounts for spatial heterogeneity and precisely predicts the age structure of the population. I was given the additional task of developing a simulation of the sampling process on a virtual population. This process involved an extensive amount of coding in R.