Linear Inverse Model Approach to Forecasting Regional Dissolved Oxygen

To the first order approximation, dissolved oxygen is determined by the competition between ocean ventilation and biological productivity. The oxygen levels vary significantly over space and time, and its pattern is crucial for the marine habitats and cycling of redox-sensitive elements. There is a growing interest in predicting the oxygen levels in the oceans, and this study examines the Linear Inverse Model (LIM) as a tool to predict dissolved oxygen in the oceans. The variability of oxygen concentration at a point may depend on many factors, but the LIM assumes the evolution of oxygen separated into deterministic (e.g. large-scale transport) and random components (e.g. eddies). The LIM extracts the linear dynamics from the statistics of observed and/or simulated ocean tracers. In this study, simulated oxygen field in the Community Earth System Model (CESM) is used as a perfect observation to develop and validate the LIM for the dissolved oxygen in the upper thermocline near the Ocean Station Papa (OSP). A simple example including the effects of upstream concentrations can significantly improve the forecast relative to the model based on the local concentration only. Constructing LIM with the leading principle components (PCs) is a feasible approach to provide temporal and spatial information. Furthermore, the LIM can be optimized to a specific forecast goals by selecting the set of PCs based on their contributions. This approach can also include other variables such as sea surface temperature or dynamic height in order to further improve the forecast skill. As the observational platform advances and data coverage improves, this approach may become useful to gain predictive skills for biogeochemical tracers.