FY 2026

A biogeochemical model intercomparison for the eastern Bering Sea shelf

Kearney, K.A., W. Cheng, and A.J. Hermann

J. Geophys. Res., 130(12), e2025JC022634, doi: 10.1029/2025JC022634, View open access article at AGU/Wiley (external link) (2025)

Uncertainty related to biogeochemical model structure—the equations, parameters, and variables used to simulate lower trophic level dynamics—can contribute significantly to overall uncertainty of regional model predictions of living marine resources metrics such as primary production. This may be particularly true in shallow coastal regions, where there is growing interest in using these types of regional models to inform ecosystem management. Here, we use a biogeochemical model intercomparison to analyze the divergence of ecosystem metrics across models for the eastern Bering Sea shelf region, and identify the biogeochemical processes that may lead to this spread. We run three biogeochemical models with varying complexity coupled to the same regional ocean model and run 30-year hindcast simulations spanning 1990–2020. We find that the models differ widely in their spatial and temporal patterns of simulated primary production, and that these differences propagate to most of the higher trophic level metrics examined. We highlight structural elements that lead to these differences, including (a) representation of benthic processes and their role in retaining nitrogen on the shelf, (b) the role of grazing control on spring bloom timing, and (c) the role of zooplankton groups in supporting regenerated production through the summer months. Overall, we conclude that the potential uncertainty associated with even well-established biogeochemical models may be high, particularly when these models are pushed beyond the original contexts under which they were developed. End users should strive to acknowledge and communicate this, particularly when using biogeochemical model output in management contexts.

Plain Language Summary. Ocean models are increasingly used to analyze and predict changes in marine resources (e.g., managed fisheries). Although many parts of these models have been rigorously tested for skill, other processes are more difficult to validate due to the sparsity of real-world measurements available. Biogeochemical models, which simulate the growth of marine algae, production, and decomposition of organic material, fall into this latter category. In this study, we compare three biogeochemical models applied to the Bering Sea region to see if they agree in their simulation of metrics that are important in marine resource management. We found many differences in the seasonal and multiyear primary production simulated across the three models, even though all three models experienced the same physical conditions, and these differences propagated throughout the food web. We highlight a few processes that contribute to these differences including (a) assumptions each model makes regarding material that sinks to the ocean floor, (b) the role of zooplankton in limiting the accumulation of their phytoplankton prey in spring, and (c) the role of zooplankton in keeping nutrients in the surface waters in the late summer months.