Submesoscale processes (length scales of 0.1–10 km) play a critical role in oceanic energy dissipation and mass transport, yet traditional satellite altimetry data—with coarse spatial (∼25 km) and temporal (daily) resolution—are unable to resolve these dynamics. Geostationary ocean-color imagers such as GOCI-II now revisit the sea surface hourly at 250 m resolution, but the nonlinear, filament-rich evolution of sub-mesoscale ocean processes obscures the underlying currents. Here, we develop the BAPDE-RAFT, a boundary-aware, Poisson-based, divergence-constrained extension of the RAFT optical-flow network, to retrieve pixel-scale surface velocities from consecutive images. Compared with the standard maximum cross-correlation (MCC) approach, BAPDE-RAFT lowers end-point and angular errors by 44 % and 38 %, respectively. Wavenumber analysis places the critical scale at which model error exceeds signal at λc ≈ 4.0 km—over an order of magnitude finer than the 60–80 km limits of traditional MCC algorithm, confirming that only BAPDE-RAFT retains spectral power throughout the 1–10 km sub-mesoscale processes. When applied to hourly GOCI-II chlorophyll-a images in the Japan Sea/East Sea, the model reproduces diurnal current variability and the expected dual cascade: an upscale kinetic-energy flux (~k⁻³) and a downscale tracer cascade (~k⁻¹). Despite occasional cloud gaps, these results demonstrate that high-cadence geostationary imagery can yield physically consistent maps of submesoscale flow, opening new avenues for satellite studies of ocean dynamics and mass transport.
 

Oceanic submesoscale process; Flow field retrieval; Geostationary satellite observation; Deep learning.