Experimental Validation of Variational Methods in the Three-Dimensional Reconstruction of Temperature and Salinity Fields within Mesoscale Eddies in the Ocean

Mesoscale eddies are ubiquitous in the world's oceans, and understanding their three-dimensional subsurface temperature and salinity (T-S) structures is crucial for deciphering their complex dynamical processes. This study applies variational methods to this domain, leveraging satellite observations and historical Argo data to successfully reconstruct the three-dimensional T-S fields of mesoscale eddies. Subsequently, by introducing a three-dimensional variational multiscale assimilation model, in situ underway observations of eddies were effectively integrated, significantly enhancing the accuracy of the reconstruction results. Comparisons with extensive Conductivity-Temperature-Depth (CTD) profile data revealed that while the preliminary variational reconstruction captured the basic structure of cold eddies, it underestimated the eddy strength and did not clearly depict the low-salinity center between 400-700 meters. After assimilating underway observation data of cold eddies, the eddy strength was markedly strengthened, and the low-salinity center became distinctly visible, consistent with observational data. Furthermore, the assimilation process notably increased the correlation coefficient between the reconstructed results and observational data while reducing the root mean square error. Compared to the MODAS method, the variational approach demonstrated superior reconstruction performance. This study not only validates the effectiveness of variational reconstruction methods for near-real-time, rapid reconstruction of subsurface T-S fields in oceanic mesoscale eddies but also highlights the pivotal role of assimilation techniques in improving reconstruction accuracy, providing a novel avenue for the quasi-real-time three-dimensional T-S reconstruction of mesoscale eddies in the ocean.

Keywords: mesoscale eddies in the ocean, three-dimensional reconstruction, multiscale three-dimensional variational assimilation