Estimating Three‐Dimensional Structures of Eddy in the South Indian Ocean From the Satellite Observations Based on the isQG Method

Abstract

By incorporating the high‐resolution satellite remote sensing sea surface temperature (SST) with low‐resolution sea level anomaly and sea surface salinity (SSS), this study explores the reconstructability of the three‐dimensional (3D) eddy structures via the “interior + surface quasigeostrophic” (isQG) method in the South Indian Ocean. We apply the incorporation of high‐resolution SST to improve the spatial resolution of the reconstruction. We also propose a correction scheme for density reconstruction within the mixed layer to offset the absence of mixed layer dynamics in the SQG framework. Comparison against the in situ observations demonstrates a satisfactory reconstructability for subsurface velocity and density anomalies. Statistically, the zonal velocity reconstructability outperforms its meridional counterpart and the corresponding velocity phase. The reconstructed shallow‐layer velocity exhibits a superior skill in eddy‐active regions, when compared with drogued drifter observations. Reconstructed subsurface velocities reproduce the spatial structures of eddy‐induced velocity anomaly along the GO‐SHIP observation transect, although present smaller magnitudes. Results demonstrate the potential applicability of the isQG method for reconstructing mesoscale eddies, particularly in the ocean at mid‐to‐high latitudes, where subsurface dynamics are strongly influenced by barotropic and the first baroclinic modes. With the upcoming high‐advanced satellite observations, the isQG framework is expected to achieve better subsurface estimations.

Subsurface temperature estimation of mesoscale eddies in the Northwest Pacific Ocean from satellite observations using a residual muti-channel attention convolution network

“…Currently, the combination of various satellite ocean surface data and Argo profiles constitutes an effective method for accurately estimating the subsurface three-dimensional temperature structure (Chen et al, 2023d;Chen et al, 2024). Firstly, relatively straightforward univariate or multivariate linear regression (MLR) is one of the most common approaches.…”