Kelsey DiPietro is accelerating research through her work in computer models.
In the world of multigigabit high-speed internet, pre-ordered fast food and Twitter pitch events, there is little patience for slow computer models — especially ones used to combat the Earth’s rapidly changing climate. To address this, Kelsey DiPietro, a Jill Hruby Fellow appointed at Sandia in 2019, has created algorithms that make computer models of complex systems more efficient by skipping over areas of datum where there is little change and honing in on those with observable shifts.
Kelsey, an applied mathematician with a doctorate from the University of Notre Dame, has made advances in creating fast 3D computational algorithms, with the aspiration of integrating them into DOE’s supercomputer-powered Energy Exascale Earth System Model, or E3SM.
“These advances will make them more tractable, or more easily solvable in a reasonable amount of time, and that is really key for my current project at E3SM where precise models can take an impossible amount of computation to solve,” she said.
She said in computational modeling, researchers adjust their models based on feedback they receive from the experts on her team, which consists of 100 researchers from eight DOE labs and several universities.
“We frequently start with very simplified models of a natural phenomenon, and then rely on other researchers to help validate how realistic or accurate our models may be,” Kelsey said. “We can create fast models all day, but if they don’t provide enough information or accurate enough information for our partners, then the model is not useful. It is a really dynamic relationship.”
The developments Kelsey has made during her time as a Hruby Fellow will allow her to take a previously intractable problem and apply it to adaptive digital meshes that can be manipulated into different shapes and forms and then refined based on user-specified criteria for existing Earth modeling applications.
“My advances will allow for easier integration without any prohibitively expensive computational costs,” she said. “This approach could be used for a wide range of applications, but there are certain problems where use of this model is more straightforward. For example, it’s much easier to target a problem that evolves on a slow time scale, such as moving ice sheets, than something on a short time scale, like a flash flood.”
The E3SM project, concluding in December, has been developing models and simulations that consider the interactions between the water cycle, the Earth itself and the atmosphere, known as Earth-system science drivers. The data for these three drivers is supplied by researchers from different disciplines, and this diversity is what fuels innovation, Kelsey said. “The diverse educational background of people involved in the project — physicists, engineers and mathematicians — helps bring a wide range of ideas.”