Scott A Martin , Georgy Manucharyan, Patrice Klein

Ocean eddies affect large-scale circulation and transfer energy between scales through non-linear eddy interactions. This eddy-induced kinetic cascade depends on the strain rate, which is strongly sensitive to the precise geometry and configuration of eddies. However, surface currents estimated globally from altimetry smooth and distort eddies, severely underestimating the strength of non-linear eddy interactions and the resulting cascade. Here, we present the first global deep learning estimate of surface currents from multi-modal satellite observations of sea surface height and temperature. We achieve a 30% improvement in spatial resolution over the community-standard sea surface height product and demonstrate that it significantly improves our observations of eddy dynamics. In many regions, this improved resolution leads to nearly an order-of-magnitude increase in the upscale kinetic energy cascade, emphasizing its crucial role in the seasonality of large mesoscale eddies. Our study suggests that deep learning can be a powerful paradigm for satellite oceanography.