Source: NOAA//Unsplash
Computer scientists at MIT worked with oceanographers to develop a machine-learning model that uses fluid dynamics to create more accurate predictions about ocean currents. This will allow scientists to better forecast weather, predict the spread of oil spills and measure energy transfer from the ocean, as MIT News reports.
As you probably already know from looking at surf forecasts, ocean currents are tracked using GPS-tagged buoys. As the buoys travel through the water, scientists record the GPS data to reconstruct the movement of ocean currents.
According to a new study, a team of MIT scientists and oceanographers discovered that the standard model used to analyze this buoy data often inaccurately reconstructed currents, because of assumptions it made about the behavior of water. By incorporating fluid dynamics data into their model, they were able to create more accurate predictions, while only using a small amount of additional computational processing power.
To estimate currents, oceanographers previously used a machine-learning technique known as a Gaussian process, which makes predictions using relatively little data by making assumptions about the data. However, it turned out that some of the assumptions being made were known to be inaccurate, from a physics perspective.
The authors of the study built a new model that incorporates fluid dynamics using a process called Helmholtz decomposition. This method models an ocean current by breaking it down into two factors: vorticity (whirling motion) and divergence (rising or sinking). They then tested this new model using synthetic and real ocean data and discovered they were able to more accurately predict currents and identify divergence.
“Our method captures the physical assumptions more appropriately and more accurately,” said senior author Tamara Broderick, “In this case, we know a lot of the physics already. We are giving the model a little bit of that information so it can focus on learning the things that are important to us.”
Now that they have this new model, oceanographers hope to be able to make more accurate estimates from buoy data, which will allow them to monitor the movements of aquatic plants, carbon, plastics and oil. It will also be important for understanding and tracking climate change. Hopefully, the improvement in data will make its way down to surf forecasting as well.
