A new cloud imaging program uses satellite modeling to study cloud formation and transport, helping to improve climate models by optimizing the collection of 3D cloud data. David Stanley’s passion for climate change inspired him to create a program aimed at improving data collection to study inside clouds. The program simulated multiple satellites taking simultaneous images of the cloud from different angles, providing a more detailed understanding of the processes occurring inside.

“We can usually only see the outer elements of the cloud,” Stanley said. “Cloud computed tomography is named after computed tomography, which is similar to CT. Instead of X-rays, satellites take pictures of the cloud from as many angles as possible and in the shortest possible time.”

Study of convection and cloud growth

Stanley said one of the unknowns in climate modeling is how much convective convection affects the rebirth of new clouds. Convection is the upward and downward movement of heat and moisture in the atmosphere, especially in unstable conditions.

“By creating multiple time passes at the center of a cloud, you can see how convection changes over time and affects the growth of other clouds in the future. And the growth of clouds can increase the greenhouse effect.”

After earning a master’s degree in aerospace engineering from the University of Illinois at Urbana-Champaign, Stanley said he reapplied for a Ph.D. in Illinois.

“I talked about my general interest in engineering and aerospace engineering and how important it is for us to better understand climate change and work towards solutions,” he said. “Robin Woollands noticed this interest in me and asked me to join his research group. He put me in touch with Federico Rossi and Amir Rahmani of the Multi-Agent Autonomy Group at NASA’s Jet Propulsion Laboratory, and they introduced me to JPL scientists Changrak Choi and Anthony Davis, experts in cloud tomography, atmospheric clouds, and aerosols. “This overlapped with some of my interests and was something Robin saw as an interesting mission proposition: using multi-agent systems to support earth science missions.”

Modeling development

For the simulation, Stanley used the mixed integer linear programming solver used in many different types of programs. Stanley wrote the code to develop a scheduler that would optimize the timing and camera angles of the satellite swarm to capture as many images of the cloud as possible.

“What was interesting about this was how we used a mixed integer linear programmer to automatically determine the most efficient guidance pattern to create the satellites. All satellites had to be aimed at a single target at the same time. But there can be dozens of different targets under each satellite, and some targets can be missed if not timed correctly.” ”

The goal was to maximize the number of times the satellites saw different targets throughout orbit.

“We ran two different simulations. One simulation of clouds forming on the Earth’s surface over a certain lifetime. On a computer, these are simply coordinates on a sphere. The second simulation emits a swarm of satellites. This can be done simply or with more complex, more accurate models.

“When we combine the data from these two simulations, the program calculates information about where the satellites are at different points in orbit and where the clouds are at the point in orbit, and then decides what the best-fitting picture is between those satellites… and the clouds on Earth.”

Problems in data management

He said that during his research he had different ideas about the best way to model the data and pass the data to the solver.

“Maybe you just want one array for each time step and each satellite, or you could have one array for different regions of the Earth. At first I tried using different parts of the world as reference coordinates, dividing everything by brute force. But there are many areas on Earth. And you get millions and millions and millions of indexes that can’t be deciphered on a desktop computer.”

Ultimately, Stanley said he was inspired by Woollands’ previous work. He developed a method for a constellation of satellites orbiting Mars to collect as many observations of Martian dust devils as possible; Here, instead of dividing the entire Earth, they divided the sections below the satellites, allowing them to only need a few indexes at a time. .

“So in addition to that, I was able to realize that I could actually just use the clouds themselves as an index,” Stanley added. “It worked well and went from millions of directories to about 100 directories at a time, which is much easier to resolve.”

“We’ve made some assumptions about where clouds form and where they go, so there’s a lot of room to improve on this work and look at more real data rather than creating our own data. More importantly, we’ve found that we can significantly improve the way we collect 3D cloud data by looking at cloud dynamics within clouds, and therefore We have developed a new method that could lead to a better understanding of long-term climate impacts.