Integrated retrieval of ocean surface wind and current fields from single-geometry SAR observations through a physically constrained Bayesian framework

Ocean surface winds and currents are coupled through air–sea interaction but are commonly retrieved independently from spaceborne synthetic aperture radar (SAR). For operational single-geometry SAR, simultaneous wind–current retrieval is underdetermined and remains reliant on background information. We propose a physically constrained Bayesian framework that combines radar backscatter and Doppler centroid anomaly measurements to retrieve 10-m wind and radial surface current. An empirical wind-drift relation softly constrains the radial-current increment relative to a product-derived background current, and the solution is obtained by weighted maximum a posteriori estimation. The framework is evaluated using multi-year Sentinel-1 observations collocated with scatterometer, buoy, and high-frequency radar measurements across different sea states. Parameters calibrated in a Gulf Stream-influenced U.S. East Coast region are fixed for temporally held-out events and an independent Southern California test. Across the East Coast data, wind-speed root-mean-square difference (RMSD) decreases by 0.59−0.77ms−1 relative to ERA5 atmospheric reanalysis. Against high-frequency radar, radial-current RMSD is 0.21ms−1, compared with 0.59ms−1 for the operational Sentinel-1 radial-velocity product, 0.48ms−1 after a conventional wind-wave Doppler correction, and 0.38ms−1 for an ocean-model projection. Event-level ablations show that the physical constraint provides further RMSD reductions of 0.171ms−1 and 0.324ms−1 in the two independent tests. These results isolate radial-current skill attributable to the physical term beyond joint data fusion alone and demonstrate the value of jointly exploiting complementary SAR observables. The framework provides a practical pathway toward joint kilometer-scale mapping of coastal winds and radial currents from routine single-geometry acquisitions.