On April 20, 2026, a research team published a study in Nature Geoscience detailing a new artificial intelligence framework called GOFLOW (Geostationary Ocean Flow). This technique leverages existing geostationary weather satellites to measure ocean surface currents with a level of detail previously unattainable through conventional satellite altimetry. By applying advanced computer vision and deep learning algorithms to high-frequency imagery, GOFLOW provides a continuous, high-resolution view of ocean dynamics across vast areas of the planet.
Traditional methods for measuring ocean currents primarily rely on nadir-pointing radar altimeters aboard polar-orbiting satellites, such as the Jason-3 or the Sentinel-6 Michael Freilich missions. While accurate, these systems typically provide data along narrow tracks with repeat cycles ranging from 10 to 35 days, leaving significant spatial and temporal gaps in the global data set. In contrast, GOFLOW utilizes the rapid-scan capabilities of geostationary satellites, such as the National Oceanic and Atmospheric Administration’s (NOAA) GOES-R series and the Japan Meteorological Agency’s Himawari-8 and Himawari-9. These satellites remain fixed over specific longitudes and capture images of the Earth’s surface every 10 to 15 minutes.
The GOFLOW system processes thermal infrared and visible spectrum data to track the movement of sea surface temperature patterns and chlorophyll concentrations. The AI architecture employs a specialized optical flow algorithm, specifically a deep residual neural network, that can distinguish between the actual horizontal advection of water and atmospheric interference, cloud shadows, or sensor noise. According to the study, GOFLOW can resolve ocean features at a spatial resolution of approximately 2 kilometers. This represents a substantial improvement over the 100-kilometer resolution typically associated with global gridded altimetry products currently in use by the oceanographic community.
Lead author Dr. Elena Vance noted that the integration of AI allows for the extraction of velocity vectors from sequences of satellite images that were previously used almost exclusively for meteorological forecasting. The researchers validated the GOFLOW data against a global network of thousands of drifter buoys and coastal high-frequency radar stations. The results showed a high correlation in both current speed and direction, particularly in the detection of sub-mesoscale eddies. The system is notably effective in coastal regions and areas with intense eddy activity, where traditional satellite data often fails to capture small-scale turbulence and rapid changes in flow.
The technical specifications of the GOFLOW model include a multi-scale convolutional neural network trained on over a decade of historical satellite and in-situ data. The researchers highlighted that because the system utilizes existing satellite infrastructure, it does not require the deployment of new hardware. This software-defined approach to remote sensing allows for the immediate scaling of ocean monitoring capabilities. The study concludes that the GOFLOW framework provides a critical tool for improving climate models, enhancing maritime navigation safety, and optimizing the response to environmental hazards such as oil spills and marine debris accumulation.