Scripps Researchers Launch GOFLOW AI to Map Ocean Surface Currents Using Existing Weather Satellites

New GOFLOW deep learning technique maps ocean surface currents in hourly detail using existing weather satellites. See how AI is solving the vertical mixing gap.

By: AXL Media

Published: Apr 13, 2026, 8:00 AM EDT

Source: Information for this report was sourced from University of California - San Diego

Revolutionizing Marine Observation Through Deep Learning

A collaborative team of oceanographers has introduced a transformative method for monitoring the movement of the world’s oceans by repurposing existing meteorological infrastructure. The technique, known as GOFLOW (Geostationary Ocean Flow), utilizes advanced neural networks to analyze thermal data from geostationary satellites that were originally designed for weather forecasting. Published in Nature Geoscience, the research led by Luc Lenain of Scripps Institution of Oceanography demonstrates that sophisticated AI can identify subtle deformations in surface temperature patterns to calculate water velocity. This approach bypasses the traditional requirement for expensive new hardware, instead turning the vast archives of GOES-East satellite imagery into a dynamic laboratory for physical oceanography.

Addressing the Critical Gap in Vertical Mixing Data

The movement of ocean currents is a fundamental driver of the planet’s climate, responsible for transporting heat, carbon, and life-sustaining nutrients across the globe. Despite its importance, a persistent observational gap has existed at scales smaller than 10 kilometers, where much of the ocean's vertical mixing occurs. Traditional satellite methods, which often rely on sea-surface height measurements, provide updates only every 10 days, a frequency insufficient for tracking currents that evolve over mere hours. GOFLOW addresses this limitation by producing hourly maps, enabling scientists to observe the rapid "up and down" movement of water that pumps carbon dioxide into the deep ocean and brings nutrients to the surface.

Training Neural Networks on Synthetic Ocean Dynamics

To bridge the gap between static temperature images and fluid motion, the researchers trained their AI model using high-resolution computer simulations of ocean circulation. These simulations provided the neural network with a comprehensive library of how various water velocities "bend and stretch" surface temperature gradients. By learning these complex visual signatures, the AI can now look at real-time thermal time-lapses and infer the underlying current dynamics with remarkable precision. This transition from simulation-based theory to observation-based reality allows for the tracking of the Gulf Stream's intricate structures in a way that was previously invisible to hum...