Reconstruction of Ocean Temperature Field Based on a Temperature Profile

Temperature is an important sea parameter that plays a significant role in the dynamic procedure, air-ocean interaction, and ecological environment of the ocean. It is closely related to ocean phenomena such as mesoscale eddies, ocean currents, sea surface winds, and the El Niño-Southern Oscillation. However, ocean temperature is difficult to collect and mainly relies on satellite remote sensing and floats. Both methods have limitations: remote sensing can only acquire the temperature at the sea surface, and floats cannot obtain large-scale ocean temperatures. Recent studies have tried to overcome these limitations by considering the correlation between the surface and subsurface parts of the ocean. They built models to represent the relationship between the subsurface temperature and the surface parameters, such as sea surface temperature, sea surface salinity, and sea surface height. Then, they inferred the subsurface temperatures from the sea surface parameters measured by the satellites. These studies provide a way to enrich the subsurface ocean temperature data to a certain extent. In addition to using sea surface information, the ocean temperature profile also contains the ocean temperature information of the adjacent area, which the artificial intelligence technique could mine. Thus, we trained an artificial neural network model using the HYbrid Coordinate Ocean Model data for the studying area. The model’s input is the ocean temperature profile and acquisition time; its output is the ocean temperature field around the temperature profile. The results show that the model reconstructed the temperature field better than the climatological data.