To study ocean currents, scientists launch GPS-tagged buoys into the ocean and record their speeds to reconstruct the currents they transport. This buoy data is also used to identify “differences,” which are areas where water rises from below the surface or sinks below it.
By accurately predicting currents and identifying variations, scientists can more accurately predict weather, approximate how oil will spread after a spill, or measure energy transfer in the ocean. A new model incorporating machine learning provides more accurate predictions than traditional models, a new study reports.
A multidisciplinary research team including MIT computer scientists and oceanographers found that the standard statistical model typically used for buoy data can struggle to accurately reconstruct currents or quantify variations because it makes unrealistic assumptions about water behavior.
The researchers developed a new model that incorporates knowledge from fluid dynamics to better reflect the physics at work in ocean currents. They show that their method, which requires only a small amount of additional computational overhead, is more accurate at predicting currents and identifying variations than the conventional model.
This new model could help oceanographers make more accurate estimates from buoy data, which would enable them to more effectively monitor the transfer of biomass (such as sargassum seaweed), carbon, plastics, oils and nutrients in the ocean. This information is also important for understanding and tracking climate change.
“Our method captures the physical assumptions more appropriately and accurately. In this case, we already know a lot of the physics. We give the model a little bit of that information so it can focus on learning things that are important to us like what the currents are away from the buoys or what the difference is and where it is.” happening?” says senior author Tamara Broderick, associate professor in the Department of Electrical Engineering and Computer Science (EECS) at MIT and a member of the Information Systems and Decision Making Lab and the Institute for Data, Systems, and Society.
Broderick’s co-authors include lead author Renato Berlingheri, a graduate student in electrical engineering and computer science. Tripp, a postdoctoral researcher at Columbia University; David R. Bert and Ryan Giordano, postdoctoral researchers at the Massachusetts Institute of Technology; Kaushik Srinivasan, Research Assistant in Atmospheric and Oceanic Sciences at UCLA; Tamai Ozgukman, Professor in the University of Miami’s Department of Oceanography; and Junfei Chia, a graduate student at the University of Miami. The research will be presented at the International Machine Learning Conference.
Dive into the data
Oceanographers use data on buoy speed to predict ocean currents and to identify “differences” as water rises to the surface or sinks deeper.
To estimate currents and find differences, oceanographers have used a machine learning technique known as the Gaussian process, which can make predictions even when the data is sparse. To work well in this case, the Gaussian process must make assumptions about the data to create a prediction.
The standard method for applying a Gaussian process to ocean data assumes that the latitude and longitude components of the stream are unrelated. But this assumption is not physically accurate. For example, this current model implies that the divergence of the stream and its cyclicity (rotational motion of the fluid) operate on the same scales of volume and length. Oceanographers know that’s not true, Broderick says. The previous model also assumes that the frame of reference matters, meaning that the fluid will behave differently in the latitude versus longitude direction.
“We thought we could tackle these problems with a model that incorporated physics,” she says.
They built a new model that uses what is known as Helmholtz decomposition to accurately represent the principles of fluid dynamics. This method models the ocean current by dividing it into a eddy component (which captures the rotational motion) and a divergence component (which captures the rise or sink of the water).
In this way, they give the model some basic physics knowledge that it uses to make more accurate predictions.
This new model uses the same data as the old model. And while their method could be more computationally intensive, the researchers showed that the additional cost is relatively small.
They evaluated the new model using synthetic and real ocean buoy data. Because the synthetic data was synthesized by the researchers, they were able to compare model predictions to ground truth streams and variances. But the simulation includes assumptions that may not reflect real life, so the researchers also tested their model using data captured by real buoys launched in the Gulf of Mexico.
In each case, their method showed superior performance for both tasks, predicting currents and identifying variations, when compared to the standard Gaussian process and other machine learning approaches that use a neural network. For example, in one simulation involving a vortex adjacent to an ocean current, the new method correctly predicted no divergence while the previous Gaussian process method and the neural network method both predicted divergence with very high confidence.
Broderick adds that this technique is also good at identifying eddies from a small group of floaters.
Now that they have demonstrated the effectiveness of using Helmholtz decomposition, the researchers want to incorporate the time component into their model, as currents can vary over time as well as space. In addition, they want to better capture how noise affects the data, such as wind sometimes affecting a buoy’s speed. Separating this noise from the data may make their approach more accurate.
“Our hope is to take the noisily observed field of velocities from the floaters, then say what the actual divergence and eddy is, and predict farther from those floaters, and we think our new method will be useful for this,” he says.
This research is supported in part by the Office of Naval Research, a National Science Foundation (NSF) CAREER Award, and the University of Miami’s Rosenstiel School of Marine, Atmospheric, and Earth Sciences.