Parsimony in Integrated Age-Structured Assessment Models: Modeling of Time-dependent Parameters and Uncertainty in a Changing Environment
- Terrance Quinn II, Fisheries Division, School of Fisheries and Ocean Sciences, University of Alaska Fairbanks
- Peter-John Hulson
Without good information and analysis in a stock assessment, major errors in management can occur and not be observed until it is too late to take action. In this project, researchers will continue work that began with the Alaska Sea Grant project "Dealing with Uncertainties in Integrated Age-Structured Assessment Models" (R/31-16) to create and evaluate methods to incorporate uncertainty in integrated age-structured assessment models. Researchers will complete this undertaking with two refined objectives: (1) develop statistical theory that enables quantitative evaluation of optimal parameter complexity and model comparison, and (2) evaluate the risks to management using current and spatially explicit ASA models for walleye pollock in the eastern Bering Sea with time-varying parameters due to factors such as climate change.
A broad mandate for Ecosystem Based Fisheries Management (EBFM) has emerged from recent ocean commissions and advisory panels. To move toward EBFM, current stock assessment models need to be improved and extended, so that uncertainties in data sources, model parameters, and model selection are better understood. Errors in stock assessment then would be reduced and fisheries harvests would be more sustainable.
Why is this an Alaska Sea Grant project?
One of Alaska Sea Grant's six key goals outlined in the 2009–2013 Strategic Plan is sustained, well-managed, and healthy marine, coastal, and watershed ecosystems in Alaska. The program pursues this goal through support of research that provides decision-makers with science-based information that can be used to craft well-informed policies governing the use and conservation of Alaska's marine and coastal resources.
How will researchers conduct their study?
The overall goal of this project is to better understand uncertainties in age-structured assessment (ASA) models by constructing simulated populations with known population parameters. The population, fishery, and surveys will be simulated over the time series for the respective populations. A simulated stock assessment will be done and estimates will be compared to true values after replicating the simulations 1,000 times, so that bias and variance can be estimated. Given our knowledge of the true population parameters and structure, we will evaluate the risk in model misspecification and estimation of parameters that vary in time, such as from climate changes.
The main hypothesis to be tested is whether age-structured assessment models can be improved by making population parameters such as natural mortality and maturity time-dependent. Through computer simulation/estimation experiments, we will determine the factors that guide optimal model complexity using both Maximum Likelihood and Bayesian statistical procedures.
North Pacific Fishery Management Council Scientific and Statistical Committee
Alaska Department of Fish and Game
The main outputs of this project are mathematical models, computer programs, and research articles. The anticipated users of these results include other scientists at universities, government agencies, and nonprofit organizations. Researchers expect to provide insight into which model selection approaches lead to the most parsimonious and robust results in stock assessment outputs.
Accomplishment: September 2011
Alaska Sea Grant fisheries scientists boosted confidence and reduced uncertainty in fisheries models
Relevance: Alaska Sea Grant–funded scientists in 2008 undertook efforts to improve Age Structured Assessment (ASA) models, the primary type of model used to manage and allocate fisheries catches in Alaska’s $2 billion a year walleye pollock, Pacific Ocean perch, and other fisheries harvested in Alaska waters.
Response: A key goal of this study was to better understand and predict areas of uncertainty within the model. Researchers analyzed the fisheries models and found ways to improve confidence in the model’s sample size and reduce observational error in data collection. Using the improved simulation ASA models they developed, the researchers were able to estimate more accurately the mean effective sample size across several years of data collection. Finally, they evaluated the effect of model parameterization on stock assessment uncertainty. This work showed the bias and uncertainty in a spatially explicit age-structured assessment model for walleye pollock in the Bering Sea. As a result, researchers determined that fish stock movement can often be estimated without data on fish movement derived from tagging studies.
Results: Findings from this project will be made available to the North Pacific Fishery Management Council and the Alaska Department of Fish and Game. These findings should influence stock assessment scientists to improve their stock assessments by reducing bias and increasing precision. It should also lead statistical and fishery scientists to think about effective sample size and effective number of parameters in general modeling and in model selection.
Recap: Alaska Sea Grant research improved traditional Age Structured Assessment (ASA) models used to manage Alaska’s multi-billion dollar a year commercial fisheries. Federal and state fishery managers are expected to consider using these improved fishery models.