Sarah Gille (UCSD, United States); Matthew Mazloff (UCSD, United States); Bruce Cornuelle (UCSD, United States)

SWOT's high-resolution measurements detect ocean features that are small enough to include ageostrophic or unbalanced motions, which necessitates refinements to data assimilation strategies. Using 4D-Var, observations from SWOT and the Global Ocean Observing System are assimilated into a regional model of the California Current System. This is achieved through a regional 2-km configuration of the MIT general circulation model (MITgcm) and its adjoint, with a 31-day assimilation window. The assimilation is run during the SWOT Cal-Val period (May 1-31, 2023). The first-guess model run is initialized with GLORYS ocean state and forced with ERA5 hourly atmospheric state; those fields are adjusted through iterative adjoint model runs to fit the observational constraints. Provided that the initial state is generally consistent with the SWOT observations at large scale, the 4D-Var method can successfully reduce model-data misfits over the month-long assimilation window, despite intrinsic nonlinear variability and instability in the high-resolution model. Assimilating SWOT improves the fit to independent (not assimilated) observations from Cal-Val mooring and gliders, as well as high-frequency radar; withholding the data allows us to quantify its impact. Because of the dynamical propagation of adjoint sensitivities over 31 days, surface observations help constrain the interior ocean state. Oxygen, carbon, and other biogeochemical fields are simulated using the optimized ocean state.