We didn’t start the fire, or did we? This summer, the Dixie fire in northern California burned more than 900,000 acres. Climate change results in drying of trees and plants and has doubled the number of large fires in the past 30 years.

Wouldn’t it be wonderful if we could come up with a strategy to mitigate these effects? We can, with simulations and supercomputers. Simulating the real world helps us predict which actions lead to a better or worse climate so we can adjust our behaviors to lower global temperatures and the number of devastating fires.

Simulations are based on finding solutions for mathematical equations. The equations that represent climate behavior lead to 100s of millions of data points. Forty years ago we couldn’t imagine handling these massive datasets accurately, back then we could represent global changes in Western Europe, but didn’t have enough detail to show the Alps. Now, with modern supercomputers, we can distinguish between changes in the mountain ranges and river valleys. 

But why do we need supercomputers? A typical laptop would need years to handle these massive datasets. We don’t have that time, the fires are burning now. Supercomputers are powerful, but only in combination with specialized software. My team creates software that effectively uses modern supercomputers to accurately solve these large problems.

Supercomputers comprise thousands of smaller parts, called nodes, that are interconnected. Our software breaks up large datasets into smaller parts so that each node can process a part of the equation. But the data parts aren’t all the same size, some are really big, so we need to “compress” them.

You might know compression from photographs. With a compression format such as a JPEG file, you'll fit more files onto a memory card, but you'll also sacrifice quality which doesn’t help your dating profile.

In the same way, we need to find the right compression balance to accurately solve our equations. This is where I come in. I combine two methods, butterfly compression and block low rank compression. This novel combination advances our software efficiently. Butterfly is a very complex method well suited for compressing climate data. It is great for bigger parts because it compresses a lot, which saves storage. However, its complexity makes it too time consuming to use for smaller parts. Block Low Rank compression is a classic compression technique well suited for the small parts. 

Combining both methods gives us the best of both worlds, and allows us to harness the power of supercomputers to solve some of the biggest problems in science. And, we are one step closer in predicting accurately, which measures we can all take to prevent devastating fires.