AYK Chinook Salmon: Bayesian Lifecycle Models for Evaluating Marine and Freshwater Drivers of Survival
Concern over the sustainability of subsistence and commercial fisheries for Chinook salmon stocks in the Arctic-Yukon-Kuskokwim region of Alaska has driven increased interest in understanding the abiotic and biotic factors influencing survival in both marine and freshwater environments, and quantifying the impacts of bycatch interactions with the BSAI Pollock trawl fishery. I am using stage-structured Bayesian population dynamics models to estimate the influence that a range of potential environmental covariates have on Chinook salmon survival. Bayesian variable selection methods are implemented to account for model structural uncertainty in predictor variables by estimating the inclusion probability for each covariate. This estimation structure is used to identify key drivers of AYK Chinook salmon survival and generate predictions for trends in abundance and recruitment under alternative climate change scenarios. Funding provided by: Pollock Conservation Cooperative Research Center.
Management Strategy Evaluation of the Bristol Bay Commercial Salmon Fishery
Management of the commercial fishery for sockeye salmon in Bristol Bay, Alaska has traditionally operated under a policy of fixed escapement goals and active inseason regulation of fishing effort to achieve these targets. Previous escapement goal analyses have failed to account for both the biological uncertainty in future production regimes and implementation uncertainty in the management process. To address these concerns, we conducted a management strategy evaluation (MSE) of the Bristol Bay fishery to explore: (1) alternative escapement goals, (2) the value of information provided by existing survey and enumeration projects, (3) the potential for implementation of fixed fishing schedules to avoid costs associated with active inseason management. The MSE framework I built for this project simulates both the future production dynamics of the salmon populations based on time-varying stock-recruitment relationships, and the behavior of district managers during the season to account for the imprecision associated with mixed-stock management. MSE results lead to changes in management targets for Bristol Bay beginning in 2015.
Sacramento River Chinook Salmon: A Statistical Lifecycle Model for Evaluating the Environmental Drivers of Survival
In order to investigate the influence of anthropogenic and natural environmental factors on the survival of threatened and endangered Chinook salmon stocks in the Sacramento River, CA watershed, I utilized a stage-structured population dynamics model to estimate the impact of a range of environmental factors hypothesized to effect maximum survival and habitat capacity, across freshwater and marine life-stages. Results of this analysis are then used to predict population viability and abundance trends under a range of future climate and water use scenarios. In order to evaluate the efficacy of the estimation model, I also implemented a simulation-testing framework with an operating model that generates abundance trends with known responses to environmental covariates and randomly drawn observation and process errors, to which the estimation model is fit and parameter estimated compared with the “true” values specified to the operating model. Funded by: Delta Science Program, Delta Stewardship Council.
Preseason and Inseason Salmon Forecasting Methods for Bristol Bay, Alaska
Accurate forecasts for salmon returns assist management, and inform effort allocation and preseason planning by fishermen and processors. A large component of my current research is focused on improving statistical methods for predicting future returns of sockeye salmon, both prior to and during the commercial fishery. In additional to traditional cohort survival models for preseason forecasting, I’m evaluating the efficacy of environmental predictors of early marine survival, the use of dynamic linear models to evaluate temporal variation in survival relationships and maturation schedules, and benefits of Bayesian model selection methods. To generate predictions for sockeye abundance during the commercial fishery, I’ve redesigned an integrated inseason assessment model that generates forecasts based upon preseason information, Port Moller test fishery cpue, inshore catch and escapement, as well as age and genetic composition information. Currently, I’m working to integrate the inseason assessment model with predictions for salmon inshore arrival timing based upon spatially explicit winter sea surface temperature trends and spring SST acceleration rates.
See UW Alaska Salmon Program Bristol Bay Preseason Forecast
See Bristol Bay Inseason Updates
A Bayesian Analysis of Alternative Biological Escapement Goals for Bristol Bay, Alaska:
Paleolimnological Priors and Shifting Productivity Regimes
Traditional methods for modeling spawner-recruit relationships with Ricker or Beverton-Holt models in order to define MSY-based management goals have traditionally suffered from: 1) an inability to accurately define stock-specific capacity estimates given the relatively small range of spawning abundances observed, and 2) the assumption of temporal stability in relationships across marine and freshwater productivity regimes. Therefore, I have created a Bayesian hidden Markov version of the Ricker model to evaluate changes in river system productivity over time that incorporates prior information on system capacity from paleolimnological data collected by Dr. Daniel Schindler that reconstructs historical salmon abundance based on analysis of Nitrogen isotopes in lake-bottom sediment cores. Using this model we were are able to estimate difference in potential yield and optimal biological escapement goals across productivity regimes as well was the probability of occupying those regimes in the future, and develop management recommendations that are robust to future environmental fluctuations. Funded by: Bristol Bay Economic Development Corporation.
Curry Cunningham, PhD 