K. Shafer Smith (New York University, USA); Dhruv Balwada (Columbia University, USA); Spencer Jones (Texas A&M University, USA)

Since the start of data collection in 2023, the Surface Water and Ocean Topography (SWOT) mission has exceeded its original requirements and revealed exciting potential for studying previously unresolved submesoscale processes and the upper-ocean kinetic energy cycle. However, significant challenges remain in fully exploiting SWOT's sea surface height (SSH) data, particularly with respect to estimating and validating submesoscale surface and subsurface currents and vertical velocities. One key limitation is the temporal and spatial sparsity of SWOT observations, which complicates the use of traditional dynamical reconstruction methods such as those based on quasi-geostrophic (QG) theory that require SSH data on a uniform grid.
To address this challenge we aim to generate and validate a realistic, high-resolution 2D SSH field by extrapolating SWOT observations onto a regularly spaced grid using deep learning techniques. Our long-term objective is to create SSH fields suitable for enabling higher-order, QG-based submesoscale velocity reconstructions in the upper ocean. We evaluate several approaches for extrapolating SWOT SSH data beyond the SWOT swath, including modified versions of the video inpainting method behind NeurOST, as well as a score-based data assimilation method that leverages high-resolution model priors. As an initial validation step we assess each method’s performance in reconstructing SSH by evaluating physical metrics such as power spectral density and upper-ocean transport using high-resolution model data.