Subsurface Ocean Structure Reconstruction Using Remote Sensing Data

Reconstructing the three-dimensional structure of the ocean interior relies on combining satellite observations of the surface with sparse in situ measurements to infer subsurface temperature, salinity and velocity fields. Satellite sensors provide uninterrupted, global coverage of sea surface temperature, sea surface height, ocean colour and surface salinity, but cannot directly penetrate below the upper few centimetres. In situ programmes such as the Argo array furnish vertical profiles at discrete locations, yet lack the spatial density to resolve mesoscale and basin-wide patterns. To bridge this gap, researchers have developed statistical and dynamical models, data assimilation frameworks and a range of machine-learning algorithms. Techniques range from linear regression and optimal interpolation to convolutional neural networks, long short-term memory networks and ensemble methods that exploit correlations between surface signatures and subsurface properties. These approaches yield high-resolution fields of temperature, salinity and currents throughout the upper few thousand metres, enabling improved monitoring of ocean heat content, circulation changes and marine ecosystem dynamics. The resulting reconstructions are of global significance, informing climate change assessments, seasonal forecasting and resource management by providing continuous, data-driven views of the ocean’s hidden interior.

Research from Nature Portfolio

No recent Nature Portfolio content available.

Subsurface Ocean Structure Reconstruction Using Remote Sensing Data publication trend

The graph below shows the total number of articles in subsurface ocean structure reconstruction using remote sensing data across all publications each year (not limited to Nature Index journals).

Technical terms

Remote sensing: Acquisition of data about the ocean surface using satellite-borne instruments.

Argo float: Autonomous drifting device that records vertical profiles of subsurface temperature and salinity.

Thermohaline structure: Three-dimensional distribution of temperature and salinity that determines seawater density and circulation.

ConvLSTM network: Deep learning model combining convolutional layers with long short-term memory units to capture spatial and temporal patterns simultaneously.

Ensemble learning: Machine-learning approach that integrates multiple predictive models to improve overall accuracy and robustness.

References

1. High resolution 3-D temperature and salinity fields derived from in situ and satellite observations. Ocean Science (2012).
2. Estimating Subsurface Thermohaline Structure of the Global Ocean Using Surface Remote Sensing Observations. Remote Sensing (2019).
3. A Deep Learning Network to Retrieve Ocean Hydrographic Profiles from Combined Satellite and In Situ Measurements. Remote Sensing (2020).
4. Subsurface Temperature Reconstruction for the Global Ocean from 1993 to 2020 Using Satellite Observations and Deep Learning. Remote Sensing (2022).