Scale-tamer

Applied mathematician and computer scientist Rachel Robey simulates the atmosphere’s interactions with land and sea to improve wind energy generation and climate modeling. In both areas, she captures vastly different scales in a single model. “If we had infinite computing power, a lot of issues could be solved by doing high resolution everywhere,” she says. “Because that’s infeasible, we have to come up with clever ways to account for these different scales.”

The Department of Energy (DOE) Computational Science Graduate Fellowship recent alum began working on her Ph.D. at the University of Colorado Boulder in 2019, with department chair Keith Julien as her advisor.

She reconnected with a CU-Boulder researcher she had worked with as an undergrad, Julie Lundquist, professor of atmospheric and oceanic sciences and a scientist at DOE’s National Renewable Energy Laboratory (NREL) in Golden, Colorado. Lundquist and her colleagues combine field observations and lidar measurements with computational methods to study air motion in and around wind plants.

Lundquist had Robey explore how simulations could sort out error causes in lidar airflow studies. Lidar devices emit pulses of laser light along a beam, then capture the light that backscatters from particles and aerosols in air currents. Doppler shifts reveal the speed and direction of each current. Errors arise from elusive sources such as instrument error, atmospheric conditions and the distribution of aerosols. Researchers lump the errors together statistically.

Following a concept by Lundquist, Robey developed a virtual lidar that operates with large eddy simulations (LES) of atmospheric winds in different scenarios using the Weather Research and Forecasting (WRF) model.

Her WRF-LES lidar model can simulate one or more lidar devices working in various conditions, from a flat, clear landscape to a real wind plant on complex terrain. For example, the group modeled airflows over a flat landscape and found that turbulence alone introduced errors. “There are a lot of things baked into how those instruments measure the winds,” Robey says. “You’re always going to have to grapple with it in some sense.”

Robey also collaborated with Ian Grooms, associate professor of applied mathematics at CU-Boulder, to develop a new vertical grid to represent the ocean in regional and climate models. The ocean surface boundary layer transfers heat, mass and momentum between the ocean and the atmosphere. As horizontal resolution increases in ocean modeling, the need to capture critical, depth-varying complexity of turbulent eddies also grows. The new grid promises to be computationally less expensive than other grids because it divides the ocean into fewer, strategically placed layers.

However, the puzzle of ocean modeling is far from solved. “Ocean models have been around forever, and they’ve always had a vertical grid,” Robey says. “Even with existing vertical grids and ocean models, we see differences in how they behave, depending on what grid is used. This is an active area of trying to understand what we should be doing, but it’s not necessarily obvious what the vertical grid needs to capture, and what the right answer is when there’s so much variation.”

Robey appreciates the opportunity to pursue applied math from two angles in independent projects. “There’s a commonality between the two in terms of thinking about the ranges of scales,” she says. “But one is more about how we collect measurements in multiscale flow and the other is more about how we model it.”

She says the CSGF program gave her rich experiences. In 2021, she did her practicum remotely, due to the pandemic, working with Matt Norman at Oak Ridge National Laboratory to better integrate cloud effects into a global climate model. The challenge was to filter fast-moving sound waves from simulated plumes of moist, rising air. The acoustics play a negligible role in climate outcomes, but they complicate the model, limiting it to small, computationally expensive timesteps.

“It turned out to be quite simple after all of our sweat and tears,” Robey says. They implemented an approximation to leave out sound waves altogether. “We could do it with a single solve on the grid that we were using.”

For as long as she can remember, a career in science and mathematics has felt natural. Growing up in Los Alamos, New Mexico, with a parent working as a scientist at Los Alamos National Laboratory (LANL), she was “inundated by that culture,” she says.

She credits her father for encouraging her through her first efforts in computer modeling. “Having someone there when you’re first learning to code — that quick feedback of someone looking over your shoulder and helping you get through those first snags — was really helpful,” she says.

After a high-school internship at LANL spent modeling wildfires and wind turbines, she headed to CU-Boulder. She coded for coursework, for more LANL internships and in Lundquist’s group.

During a post-graduate job at LANL, she delved into climate modeling, helping to simulate mixing processes in the ocean surface boundary layer. Soon, she headed back to Boulder for her Ph.D., wanting to dig deeper into applied math. “And honestly, the geography was a big draw for me,” she says. “I do a lot of climbing, whitewater kayaking, skiing, hiking, and it’s a fantastic spot for it.”

Robey defended her dissertation April 30, 2024, and is considering what career direction to take. “One of the beauties of having come from the applied math and computational background is I get to play in all these different application spaces,” she says. “There’s a lot of room to explore new projects and ideas, or potentially to continue with some of the work I’ve been doing.”

Whatever she does, she’d like to do it in Boulder. She loves the area for the National Center for Atmospheric Research, the National Oceanic and Atmospheric Administration, CU-Boulder — and the outdoors.