Retrieving the Ocean Interior Structure from Surface Data:FOAGRNN and SQG-mEOF-R

As many oceanic observations can only reflect the sea surface, such as satellite data, the retrieval of the ocean interior structure from sea surface information is of great importance for us to understand the three-dimensional ocean. In order to retrieve the three-dimensional salinity and density structure from surface data, two new method are proposed as follows. One proposed method is called generalized regression neural network with the fruit fly optimization algorithm (FOAGRNN), which is a nonlinear method and used to estimate subsurface salinity profiles from sea surface parameters. Compared with linear methodology, the estimated salinity profiles from the FOAGRNN method are in better agreement with the measured profiles at the halocline.The results suggest a potential new approach for salinity profile estimation derived from sea surface data. The other proposed method is called SQG-mEOF-R, which estimate the interior density from the sea surface density (SSD) and sea surface height (SSH) by combining the dynamical surface trapped mode derived from the Surface Quasi-Geostrophic (SQG) function with the statistical mode calculated from multivariate EOF reconstruction (mEOF-R) method and. This method is applied to the eddy-resolving OFES (Ocean General Circulation Model For the Earth simulator) simulation and compared with the conventional SQG or isQG (interior plus SQG) and mEOF-R methods. The results manifest that, no matter in the NorthWest Pacific (NWP) region dominated by surface-intensified eddies or the SouthEast Pacific (SEP) region characterized by subsurface-intensified eddies, SQG-mEOF-R perform a robust work in mesoscale density reconstruction.