To check ocean currents, scientists release GPS-tagged buoys within the ocean and record their velocities to reconstruct the currents that transport them. These buoy data are also used to discover “divergences,” that are areas where water rises up from below the surface or sinks beneath it.

By accurately predicting currents and pinpointing divergences, scientists can more precisely forecast the weather, approximate how oil will spread after a spill, or measure energy transfer within the ocean. A brand new model that comes with machine learning makes more accurate predictions than conventional models do, a brand new study reports.

A multidisciplinary research team including computer scientists at MIT and oceanographers has found that a normal statistical model typically used on buoy data can struggle to accurately reconstruct currents or discover divergences since it makes unrealistic assumptions concerning the behavior of water.

The researchers developed a brand new model that comes with knowledge from fluid dynamics to higher reflect the physics at work in ocean currents. They show that their method, which only requires a small amount of additional computational expense, is more accurate at predicting currents and identifying divergences than the normal model.

This latest model could help oceanographers make more accurate estimates from buoy data, which might enable them to more effectively monitor the transportation of biomass (equivalent to Sargassum seaweed), carbon, plastics, oil, and nutrients within the ocean. This information can also be necessary for understanding and tracking climate change.

“Our method captures the physical assumptions more appropriately and more accurately. On this case, we all know numerous the physics already. We’re giving the model a bit little bit of that information so it may well concentrate on learning the things which are necessary to us, like what are the currents away from the buoys, or what is that this divergence and where is it happening?” says senior creator Tamara Broderick, an associate professor in MIT’s Department of Electrical Engineering and Computer Science (EECS) and a member of the Laboratory for Information and Decision Systems and the Institute for Data, Systems, and Society.

Broderick’s co-authors include lead creator Renato Berlinghieri, an electrical engineering and computer science graduate student; Brian L. Trippe, a postdoc at Columbia University; David R. Burt and Ryan Giordano, MIT postdocs; Kaushik Srinivasan, an assistant researcher in atmospheric and ocean sciences on the University of California at Los Angeles; Tamay Özgökmen, professor within the Department of Ocean Sciences on the University of Miami; and Junfei Xia, a graduate student on the University of Miami. The research might be presented on the International Conference on Machine Learning.

Diving into the information

Oceanographers use data on buoy velocity to predict ocean currents and discover “divergences” where water rises to the surface or sinks deeper.

To estimate currents and find divergences, oceanographers have used a machine-learning technique often called a Gaussian process, which might make predictions even when data are sparse. To work well on this case, the Gaussian process must make assumptions concerning the data to generate a prediction.

A regular way of applying a Gaussian process to oceans data assumes the latitude and longitude components of the present are unrelated. But this assumption isn’t physically accurate. As an illustration, this existing model implies that a current’s divergence and its vorticity (a whirling motion of fluid) operate on the identical magnitude and length scales. Ocean scientists know this just isn’t true, Broderick says. The previous model also assumes the frame of reference matters, which implies fluid would behave in another way within the latitude versus the longitude direction.

“We were considering we could address these problems with a model that comes with the physics,” she says.

They built a brand new model that uses what’s often called a Helmholtz decomposition to accurately represent the principles of fluid dynamics. This method models an ocean current by breaking it down right into a vorticity component (which captures the whirling motion) and a divergence component (which captures water rising or sinking).

In this manner, they provide the model some basic physics knowledge that it uses to make more accurate predictions.

This latest model utilizes the identical data because the old model. And while their method will be more computationally intensive, the researchers show that the extra cost is comparatively small.

Buoyant performance

They evaluated the brand new model using synthetic and real ocean buoy data. Since the synthetic data were fabricated by the researchers, they might compare the model’s predictions to ground-truth currents and divergences. But simulation involves assumptions that won’t reflect real life, so the researchers also tested their model using data captured by real buoys released within the Gulf of Mexico.

In each case, their method demonstrated superior performance for each tasks, predicting currents and identifying divergences, compared to the usual Gaussian process and one other machine-learning approach that used a neural network. For instance, in a single simulation that included a vortex adjoining to an ocean current, the brand new method accurately predicted no divergence while the previous Gaussian process method and the neural network method each predicted a divergence with very high confidence.

The technique can also be good at identifying vortices from a small set of buoys, Broderick adds.

Now that they’ve demonstrated the effectiveness of using a Helmholtz decomposition, the researchers want to include a time element into their model, since currents can vary over time in addition to space. As well as, they need to higher capture how noise impacts the information, equivalent to winds that sometimes affect buoy velocity. Separating that noise from the information could make their approach more accurate.

“Our hope is to take this noisily observed field of velocities from the buoys, after which say what’s the actual divergence and actual vorticity, and predict away from those buoys, and we predict that our latest technique might be helpful for this,” she says.

“The authors cleverly integrate known behaviors from fluid dynamics to model ocean currents in a versatile model,” says Massimiliano Russo, an associate biostatistician at Brigham and Women’s Hospital and instructor at Harvard Medical School, who was not involved with this work. “The resulting approach retains the pliability to model the nonlinearity within the currents but also can characterize phenomena equivalent to vortices and connected currents that will only be noticed if the fluid dynamic structure is integrated into the model. This is a wonderful example of where a versatile model will be substantially improved with a well thought and scientifically sound specification.”

This research is supported, partially, by the Office of Naval Research, a National Science Foundation (NSF) CAREER Award, and the Rosenstiel School of Marine, Atmospheric, and Earth Science on the University of Miami.