SHREVEPORT -- Six LSUS computer science students collaborated on an artificial intelligence tool to help the Air Force Global Strike Command better predict the weather.
As part of a senior capstone class, the students assembled an advanced weather visualization tool that aids AFGSC in making medium-range weather forecasts to enhance mission planning.
The app is powered by GraphCast, an artificial intelligence model by Google DeepMind that specializes in medium-range weather forecasts (3-10 days).
Students developed a user-friendly app that integrates GraphCast overlayed with global maps, predictions for temperature, precipitation and wind speed, comparative analysis between the AI model and traditional weather models, and traditional weather alerts among other functions.
U.S. Air Force Tech. Sgt. Brian Tubbs said the LSUS team’s weather app delivered an immediate solution for Kirtland Air Force Base in New Mexico, where the app is being beta-tested as part of “Project Airstorm.”
“Collaborating with LSUS students on ‘Project Airstorm’ through STRIKEWERX and the LSUS Collaboratory has been exceptional,” said Tubbs, 377th Test Support Squadron flight chief of mission weather operations who serves as the Air Force’s primary point of contact for the project. “The students, including those in the National Security Fellowship Program, brought technical expertise in computer science, mathematics and AI along with enthusiasm for addressing real-world Air Force challenges.
“Their work – from coding and integrating GraphCast to designing a user-friendly interface – has delivered a practical tool for Kirtland’s immediate needs. The LSUS team’s creativity and adaptability shone in design sprints and collaborative efforts, making them invaluable partners.”
The 58th Special Operations Wing will use the weather app to plan training missions and develop maintenance and installation operations.
Meet Heimdall, the weather visualization app
The Air Force has an affinity for Norse mythology. So Heimdall, who can see into the future as the guardian of the bridge connecting the world of gods and the world of humans, is a fitting moniker.
Heimdall the weather app integrates GraphCast and its prediction model into global maps, allowing a user to pick any point in the world and see current and predicted weather conditions as far out as 16 days.
Measured conditions include actual temperature, “feels like” temperature, wind speeds and gusts, cloud cover, humidity and air pressure to name a few.
“The whole purpose is not only to implement artificial intelligence for the app, but to improve the existing models the Air Force uses for prediction,” said Eliana Gafford, a computer science senior from Shreveport who is a front-end developer. “Right now, the Air Force relies mainly on numerical weather prediction models, and our goal was to improve those forecasts with AI implementation.
“I’m very grateful for this experience because we finally get to implement these things we’ve been practicing and studying into something that’s bigger than all of us.”
The user can customize favorite locations, such as Kirtland or Barksdale Air Force bases in addition to gathering data about any point on a flight path for a specific mission.
Traditional weather practices are typically better at observing current conditions, so the team included severe weather alerts from the National Weather Service.
But AI models are having success in terms of medium- and long-range forecasting.
The Heimdall tool has an analysis function in which the user can compare the accuracy of the AI forecast to the traditional forecast in terms of temperature, precipitation and wind speed.
For example, the AI model more accurately predicted wind speeds at Kirtland Air Force Base in the past seven days, whereas the traditional model better forecasted temperature and precipitation levels.
In contrast, the AI model more accurately predicted temperature and precipitation levels in Shreveport for the past seven days, while the traditional method guessed wind speed more accurately.
“GraphCast has been able to provide more detailed forecasts from a more global standpoint and not necessarily in precise locations,” said Joshua Francis, an LSUS graduate student from Haughton who is the student team project lead and full-stack developer. “The AI model relies on historical weather data to predict future weather conditions, so the weather forecasts should only get better as AI gets more data.
“We’re fortunate to have the opportunity to work with the [AFGSC], and this is something great to put on our resume. We worked with Global Strike Command in implementing additional features after meeting our initial project requirements.”
The app is complete with log in, chat, and feedback features in addition to resource links that provide important information to further develop the app.
The app development was part of the computer science software engineering capstone class, which spans two semesters.
Students conducted research this fall while app building and testing mostly occurred this spring.
Other students involved in this project include Antonio Mata (user experience/user interface designer), Connor Zittrauer (solutions architect and full-stack developer), Vincent Hartline (full-stack developer) and Duaa Khawaldeh (data analyst).
Real-world experience
The LSUS team got the full professional experience complete with project deadlines, budgets, and Air Force requests.
While GraphCast serves as the backbone of the app, the team pivoted to other services for different parts of the app because of budget and time restrictions.
“We needed a different AI model to do the map imaging, and we found Media Source map data to generate our map tiles,” said Mata, a senior from Shreveport. “We needed more time and money to use GraphCast for this, but we were able to integrate Media Source into the app.
“We definitely had to learn how to communicate with each other, with our points of contact, and solve problems.”
One specific addition to the app came in the form of map files in which lightning strike points are viewable.
“That’s useful to the Air Force because they can see if it’s safe for a plane to go into this area or tell them to avoid this area,” Mata explained.
LSUS computer science faculty member Keyvan Shahrdar said experiential learning is a key part of the curriculum.
“This kind of project gives students practical experience working in a collaborative software development environment and helps bridge the gap between classroom theory and real-world application,” Shahrdar said. “They’re not only writing code but also managing timelines, understanding budgets, selecting tools, and adapting to evolving requirements – skills that are critical in any professional tech role.
“The real strength of AI lies in its ability to analyze vast datasets and uncover trends that would be difficult or impossible for humans to detect manually.”
AI in weather forecasting and LSUS’s involvement
Tubbs said the Air Force is leaning into the future of AI in weather forecasting, and LSUS student participation constructed a usable weather app that allowed the Air Force to implement AI as they continue to study its best use.
LSUS became involved in this effort through partnerships with the AFGSC, STRIKEWERX, and the Cyber Innovation Center (CIC), the latter two of which are located on and around Barksdale Air Force Base.
The CIC, a nonprofit organization that aims to orchestrate partnerships to satisfy the needs of government, industry and academic entities, approached the LSUS computer science department about the project.
STRIKEWERX is a technology transfer and transition partner for the AFGSC.
Tubbs added that Heimdall is a significant improvement to traditional weather models.
“GraphCast has shown superior accuracy in medium-range forecasting, particularly in tracking complex weather patterns and extreme events,” Tubbs said. “It processes data faster than legacy systems, and this speed and precision allows Kirtland AFB to make better-informed decisions.
“As an interim solution driven by the initiative through STRIKEWERX, Heimdall provides immediate value while the 557th Weather Wing at Offutt AFB develops additional AI-driven tools for future Air Force-wide deployment.”
