Abstract:Satellite-based remote sensing missions have revolutionized our understanding of the Ocean state and dynamics. Among them, space-borne altimetry provides valuable Sea Surface Height (SSH) measurements, used to estimate surface geostrophic currents. Due to the sensor technology employed, important gaps occur in SSH observations. Complete SSH maps are produced using linear Optimal Interpolations (OI) such as the widely-used Data Unification and Altimeter Combination System (DUACS). On the other hand, Sea Surface Temperature (SST) products have much higher data coverage and SST is physically linked to geostrophic currents through advection. We propose a new multi-variate Observing System Simulation Experiment (OSSE) emulating 20 years of SSH and SST satellite observations. We train an Attention-Based Encoder-Decoder deep learning network (\textsc{abed}) on this data, comparing two settings: one with access to ground truth during training and one without. On our OSSE, we compare ABED reconstructions when trained using either supervised or unsupervised loss functions, with or without SST information. We evaluate the SSH interpolations in terms of eddy detection. We also introduce a new way to transfer the learning from simulation to observations: supervised pre-training on our OSSE followed by unsupervised fine-tuning on satellite data. Based on real SSH observations from the Ocean Data Challenge 2021, we find that this learning strategy, combined with the use of SST, decreases the root mean squared error by 24% compared to OI.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to learn sea surface height (SSH) interpolation through multivariate simulated satellite observations. Specifically, the paper focuses on the following aspects: 1. **Data Missing Problem**: Due to the limitations of sensor technology, there are large gaps in the sea - surface height data observed by satellites. These gaps lead to the need for interpolation when generating complete sea - surface height maps, and the existing linear optimal interpolation methods (such as the DUACS system) perform poorly when dealing with small - scale and rapidly evolving eddies. 2. **Utilization of Multivariate Observations**: There is a physical connection between sea - surface temperature (SST) and sea - surface height. SST can obtain a high spatial resolution through passive infrared technology. The paper proposes to use SST information to enhance the interpolation effect of SSH. 3. **Application of Deep - Learning Methods**: Traditional interpolation methods are mainly based on linear optimal interpolation, while deep - learning methods can better capture complex data relationships. The paper explores the applications of supervised and unsupervised learning in SSH interpolation and proposes a hybrid - learning strategy, that is, supervised pre - training on simulated data and unsupervised fine - tuning on real data. ### Main Contributions 1. **New OSSE Dataset**: The paper proposes a new Observing System Simulation Experiment (OSSE), which contains 20 years of SSH and SST data. These data have realistic simulated - observation characteristics and can be used to train neural networks. 2. **Comparison between Supervised and Unsupervised Learning**: The paper compares the training effects of a fixed neural - network architecture in supervised and unsupervised ways, evaluates the error of SSH interpolation and the accuracy of geostrophic - current reconstruction, and conducts quantitative and visual comparisons of eddy structures. 3. **Hybrid - Learning Strategy**: The paper proposes a hybrid - learning strategy, that is, supervised pre - training on simulated data and unsupervised fine - tuning on real data. This strategy combines the richness of simulated data and the practicality of real data and improves the interpolation accuracy. ### Specific Methods - **Physical Relationship**: The paper describes in detail the physical relationship between SSH and SST and explains why SST observations can provide important information for SSH reconstruction through the geostrophic - flow equation and the advection equation. - **OSSE Design**: The paper designs a high - resolution global ocean physical reanalysis product (GLORYS12) as a reference dataset to simulate satellite observations of SSH and SST. - **Neural - Network Architecture**: The paper uses an Attention - Based Encoder - Decoder (ABED) neural network, which can effectively fuse temporal and spatial information and improve the interpolation accuracy. - **Loss Function**: The paper proposes three loss functions, which are respectively used for supervised learning, unsupervised learning, and unsupervised learning with regularization, to evaluate the effects of different learning strategies. ### Conclusion Through multivariate simulated satellite observations and deep - learning methods, the paper significantly improves the accuracy of sea - surface - height interpolation, especially when dealing with small - scale eddies. The proposed hybrid - learning strategy also shows its superiority when applied to real data.

Learning of Sea Surface Height Interpolation from Multi-variate Simulated Satellite Observations

Learning Sea Surface Height Interpolation From Multi‐Variate Simulated Satellite Observations

Training neural mapping schemes for satellite altimetry with simulation data

A Deep Learning Approach to Spatiotemporal Sea Surface Height Interpolation and Estimation of Deep Currents in Geostrophic Ocean Turbulence

Multimodal 4DVarNets for the reconstruction of sea surface dynamics from SST-SSH synergies

Inversion of Sea Surface Currents From Satellite‐Derived SST‐SSH Synergies With 4DVarNets

OceanBench: The Sea Surface Height Edition

Data-Driven Interpolation of Sea Surface Suspended Concentrations Derived from Ocean Colour Remote Sensing Data

Multi-scale decomposition of sea surface height snapshots using machine learning

4DVarNet-SSH: end-to-end learning of variational interpolation schemes for nadir and wide-swath satellite altimetry

Satellite-Based Global Sea Surface Oxygen Mapping and Interpretation with Spatiotemporal Machine Learning.

An information flow-based sea surface height reconstruction through machine learning

A Deep Framework for Eddy Detection and Tracking From Satellite Sea Surface Height Data

Deep learning for the super resolution of Mediterranean sea surface temperature fields

Spatiotemporal Prediction of Monthly Coastal Upwelling Scenario in SST Fields Using Deep Learning Based Models

Application of deep learning technique to the sea surface height prediction in the South China Sea

Scale-aware neural calibration for wide swath altimetry observations

Deep learning for ocean remote sensing: an application of convolutional neural networks for super-resolution on satellite-derived SST data

High-Resolution Seamless Daily Sea Surface Temperature Based on Satellite Data Fusion and Machine Learning over Kuroshio Extension

An Observing System Simulation Experiment for the Calibration and Validation of the Surface Water Ocean Topography Sea Surface Height Measurement Using in Situ Platforms

An Ensemble-Based Machine Learning Model for Estimation of Subsurface Thermal Structure in the South China Sea