Estimating time-variation in confounded processes in population dynamics modeling: A case study for snow crab in the eastern Bering Sea

Population dynamics models used to provide management advice for harvested natural resources often estimate population processes that can be mutually confounded. These processes (e.g., natural mortality (M) and catch ability (q)) are often assumed to be time-invariant, but this assumption can be violated in real populations. Not allowing for time-variation in these processes can result in retrospective patterns in estimated biomass, which can result in errors in management advice. Allowing for time-variation in these processes can also result in errors in management advice if the estimated time-variation does not reflect the true underlying variation. Here I present a population dynamics model for snow crab in the eastern Bering Sea that tests different combinations of implementation of time-variation in M and q. The North Pacific Fisheries Management Council did not adopt any of the models with time-variation in M or q for management because the estimated variability in M and q was difficult to explain with any known mechanism and management advice differed by nearly an order of magnitude among models. Large differences in management advice underscore the importance of evidence-based approaches to incorporating time-variation in population processes into stock assessment models.

Automated summary

Population dynamics models used for management advice often estimate confounded processes, and not accounting for time-variation in these processes can lead to errors in management advice. This study emphasizes the importance of evidence-based approaches when considering time-variation in population processes to ensure accurate management advice.