Synthetic Aperture Radar (SAR) and Surface Waves Investigation and Monitoring instrument (SWIM) both provide the global measurement of the ocean surface wave spectrum. The performance of these two sensors varies with sea states, and additionally, is affected by their inherent disadvantages. For SAR, the retrieved wave spectrum would be biased differently from the truth according to the goodness of the first guess, and the velocity bunching effect blocks the measurable range in the frequency domain. For SWIM, the wave propagation direction couldn't be resolved solely on itself and the spectrum amplitude is always slightly underestimated due to the distance bunching effect. Over the years, researchers have had to select one particular sensor based on their objectives and the sea state. Fortunately, we found that they possess clues to address each other’s drawbacks. SAR and SWIM generally measure consistent ocean surfaces due to the short revisit period and small distance of sampling spots between them, allowing us to merge the two datasets and produce a better wave spectrum estimate. In this work, we propose the specific wave spectrum fusion method (WSF) based on Image Fusion Transformer (IFT) along with customized data augmentation methods to tackle the observation dilemma. The wave spectra output from WAVEWATCH III are regarded as the ground truth and are subsequently transferred to SAR and SWIM simulators at different levels to train and evaluate the performance of the proposed approach. The evaluation considers both individual and combined effects of the pure/general velocity bunching, quasi-linear and approximated/fully nonlinear mapping relation from the SAR measurement, and the distance bunching effect and spectral ambiguity from the SWIM. Furthermore, the wave spectrum of buoys would be employed to validate the effectiveness of the WSF in practical usage, because the higher-order momentum of the wave spectrum is extremely sensitive to the noise, making it an excellent verification metric.

ocean remote sensing, wave spectrum, SAR, SWIM, data fusion, transformer