Simulation driven Bayesian calibration of data weighting

E/PD-19

Fisheries managers use stock assessments—collections of demographic information about fisheries populations—to determine the health of the population and inform fisheries management. Modern stock assessments use models to compile data from different sources to inform management decisions about fish populations. The decision-making process, based on these models, can be simulated and evaluated via a process called Management Strategy Evaluation (MSE). Presently there is no clear, objective, and practical way of determining how competing sources of information should be weighted in stock assessment models.  

In this project, Grunloh will use Bayesian optimization methods together with MSE simulations to evaluate and optimize the effectiveness of stock assessment models with respect to how data sources are weighted. He aims to test these methods against existing assessment models of rockfish populations which live along the California coast.

This research builds upon the most current methods of data weighting, and is novel in its use of simulation together with efficient optimization algorithms to objectively calibrate stock assessment methods. Using these methods, he hopes to provide stock assessment scientists with tools to better serve population health and better inform management objectives.