Ocean observation is essential for studying ocean dynamics, climate change, and carbon cycles. Due to the difficulty and high cost of in situ observations, existing ocean observations are inadequate, and satellite observations are mostly surface observations. Previous work has not adequately considered the spatio-temporal correlation within the ocean itself. This paper proposes a new method—convolutional long short-term memory network (ConvLSTM)—for the inversion of the ocean subsurface temperature and salinity fields with the sea surface satellite observations (sea surface temperature, sea surface salinity, sea surface height, and sea surface wind) and subsurface Argo reanalyze data. Given the time dependence and spatial correlation of the ocean dynamic parameters, the ConvLSTM model can improve inversion models’ robustness and generalizability by considering ocean variability’s significant spatial and temporal correlation characteristics. Taking the 2018 results as an example, our average inversion results in an overall normalized root mean square error (NRMSE) of 0.0568 °C/0.0027 PSS and a correlation coefficient (R) of 0.9819/0.9997 for subsurface temperature (ST)/subsurface salinity (SS). The results show that SSTA, SSSA SSHA, and SSWA together are valuable parameters for obtaining accurate ST/SS estimates, and the use of multiple channels in shallow seas is effective. This study demonstrates that ConvLSTM is superior in modeling the subsurface temperature and salinity fields, fully taking global ocean data’s spatial and temporal correlation into account, and outperforms the classic random forest and LSTM approaches in predicting subsurface temperature and salinity fields.

Physical oceanography models rely heavily on grid discretization. It is known that unstructured grids perform well in dealing with boundary fitting problems in complex nearshore regions. However, it is time-consuming to find a set of unstructured grids in specific ocean areas, particularly in the case of land areas that are frequently changed by human construction. In this work, an attempt was made to use machine learning for the optimization of the unstructured triangular meshes formed with Delaunay triangulation in the global ocean field, so that the triangles in the triangular mesh were closer to equilateral triangles, the long, narrow triangles in the triangular mesh were reduced, and the mesh quality was improved. Specifically, we used Delaunay triangulation to generate the unstructured grid, and then developed a K-means clustering-based algorithm to optimize the unstructured grid. With the proposed method, unstructured meshes were generated and optimized for global oceans, small sea areas, and the South China Sea estuary to carry out data experiments. The results suggested that the proportion of triangles with a triangle shape factor greater than 0.7 amounted to 77.80%, 79.78%, and 79.78%, respectively, in the unstructured mesh. Meanwhile, the proportion of long, narrow triangles in the unstructured mesh was decreased to 8.99%, 3.46%, and 4.12%, respectively.