Abigail Bodner (MIT, USA); Dhruv Balwada (Columbia University, USA); Spencer Jones (Texas A&M University, USA); Leah Johnson (U Washington , USA); Tatsu Monkman (NYU, USA); Ryan Shijie Du (NYU, USA)

Learning submesoscale surface velocities from SWOT observations presents a challenge for at least two reasons: (1) submesoscale dynamics are characterized by O(1) Rossby number and exhibit a wide range of ageostrophic motions, including convergent fronts and strong vortical asymmetry; and (2) inertia-gravity waves (IGWs) strongly influence SSH at these scales. These wave motions are ageostrophic, and moreover, their O(1 day) timescales make it infeasible to remove them from the 21-day repeat cycle SWOT data using temporal filtering. Alternate methods, not relying on geostrophy or temporal filtering, need to be developed if we wish to quantify submesoscale transport from SWOT.

Our SWOT Science Team is focused on estimating the submesoscale, non-wave, ageostrophic velocity field from SWOT KaRin snapshots, in conjunction with other satellite and in situ data. In this presentation, I will discuss our team's approach to this problem, covering results from work over the past six years, emphasizing how they fit together and provide parts of the solution to this puzzle, and ending with a presentation of our current efforts and plans going forward. Highlights include the development of deterministic inversion methods to recover surface and subsurface balanced velocities using higher-order balanced models; use of joint probability distributions of vorticity, strain and divergence in the LLC4320 simulation to reveal statistical flow patterns as a function of region and season; using machine learning approaches to estimate wave-filtered SSH fields; and a newly-begun effort to use generative diffusion models to estimate surface velocities, along with their uncertainty quantification.