WASHINGTON – A U.S. Naval Research Laboratory  (NRL) college intern has used machine learning to improve Navy wave forecasting.

Braedon Kimball, a senior software engineering major at Mississippi State University’s Bagley College of Engineering, joined NRL through the Naval Research Enterprise Internship Program (NREIP). He learned about neural networks and data framing by using Python software, which helps programmers write clear, logical code for small- and large-scale projects at NRL’s Ocean Sciences Division.

This research could help the Navy provide more accurate wave forecasting, as well as a foundation of machine learning that could be extensible to other forecasting tasks.

“We aim to improve on numerical wave-forecast models by using machine learning to make predictions,” Kimball said. “The methods used to accomplish this were to train an artificial neural network using historical observations and model data with the effect of improving existing model predictions in areas the model does not normally do well.”

Kimball’s NREIP mentor, James Dykes, a physical scientist with NRL’s Ocean Sciences Division at Stennis Space Center, Mississippi, said neural networks show promise to improve weather and wave forecasting with numerical models.

“This project was about improving performance of ocean surface wave forecasts using the WAVEWATCH III model and applying machine learning ideas,” Dykes said. “Braedon was familiarized with the model output, how to process statistics and plots regarding the WAVEWATCH III model's performance, and extending that to applying some machine learning concepts to the problem.”

Dykes said the research is groundbreaking. “It has been very rewarding mentoring Braedon, imparting to him ideas about wave modelling and possibilities in machine learning.

“In the future, this work will be a launching point for other weather prediction neural networks being expanded into other aspects of climate data,” Kimball said.

With access to the Navy Department of Defense Supercomputing Resource Center, Kimball built a rudimentary neural network to demonstrate the possibility of improving wave forecasts.

After Kimball collected the necessary model and observation data using high performance computing, he trained the network to build a suite of classes to handle the data collection, as well as a neural network to make wave height predictions.

Although the topic presented to Kimball was new to him, his mentor was pleased Kimball could quickly learn about the tools to apply machine learning to the problem that in the end increased his overall understanding of the subject.

NREIP typically provides an opportunity for about 800 college students to participate in ten weeks of hands on research at 45 Navy laboratories during the summer, encouraging participants to pursue science and engineering careers. However, this year’s NREIP program was done virtually.

“I think we had a constructive experience and accomplished a lot,” Dykes said. “The remote connection environment provided a lot of flexibility, for instance Braedon could work from home nearby, as well as from his school hundreds of miles away at varying times without the need for an office building at a set 8-5 schedule. If Braedon comes back again soon, we can further build on what he has already learned.”

NRL mentors virtually hosted 67 college students and 11 high school students this summer and they hope students further their education via mentoring by laboratory personnel.

“At the end of the day, this method of mentoring can truly work,” Dykes said.

Learn more about NREIP at: https://nreip.asee.org/

About the U.S. Naval Research Laboratory

NRL is a scientific and engineering command dedicated to research that drives innovative advances for the U.S. Navy and Marine Corps from the seafloor to space and in the information domain. NRL is located in Washington, D.C. with major field sites in Stennis Space Center, Mississippi; Key West, Florida; Monterey, California, and employs approximately 2,500 civilian scientists, engineers and support personnel.