Dissertation Defense: Kayla Mitchell
Title: Oxygen to Carbon: Constraining the Climatological Biological Carbon Pump with an Oxygen-based Inverse Model
Abstract: The long-term storage of carbon in the ocean interior is largely mediated by the biological carbon pump (BCP), which transfers photosynthetically fixed organic matter from the surface ocean to depth. Quantifying the global BCP, even its climatological mean state, remains an outstanding problem in ocean biogeochemistry. Estimates of annual global carbon export span a factor of two or more. Estimates of transfer efficiency, the fraction of exported carbon that reaches the deep ocean and remains sequestered for hundreds to thousands of years, range more widely still. This spread in the mean state BCP obscures estimates of its temporal variability, including the climate-driven long-term trends we most need to detect. Despite numerous studies, there is consequently no consensus on whether the BCP has changed since the preindustrial era, or whether current methods could detect such a change if it had.
This dissertation presents OXCAR (Oxygen to Carbon), an inverse biogeochemical model that couples a steady-state oxygen mass balance to the OCIM2 circulation model to estimate the mean state BCP from the WOA23 oxygen climatology. In Chapter 1, we compare a hierarchy of carbon remineralization parameterizations to identify which environmental controls are the dominant drivers of carbon export, transfer efficiency, and their spatial patterns. The result is a 3-D, oxygen-constrained climatology of the BCP that offers an independent check on existing estimates and enables a cleaner separation of structural from parametric uncertainty than is tractable in parameter-rich, mechanistically explicit models. In Chapter 2, we develop the Bayesian-geostatistical framework used for OXCAR’s uncertainty estimates. We quantify the spatial correlation of model-data residuals to estimate the true effective degrees of freedom of independent observations and show how accounting for this correlation is necessary for credible uncertainty estimates and model selection. In Chapter 3, we present a synthesis of BCP estimates from the field’s current ensemble of numerical ocean models, comparing them to observation-based estimates, to each other, and to OXCAR. The synthesis includes primary results from the RECCAP2 (Regional Carbon Cycle Assessment and Processes, phase 2) BCP chapter. We extend the RECCAP2 synthesis with a diagnosis of model-model and model-observation discrepancies for one of the most consequential and uncertain BCP variables: transfer efficiency. This analysis surfaces an outstanding large-scale disagreement among CMIP6-class models in the latitudinal pattern of transfer efficiency, whose dominant physical or biogeochemical driver remains unresolved. OXCAR provides an independent, subsurface constraint on this disagreement. Because its alternative remineralization parameterizations are evaluated within a unified framework, OXCAR shows how different structural choices result in the distinct latitudinal patterns seen across the ensemble. Combined with the uncertainty framework developed in Chapter 2, it also identifies which of these patterns the oxygen data most strongly support.