Welcome to the Climate Code: A blog at the intersection of climate and computer science with a hint of graduate student career advice

Last year, our research group at the University of Minnesota was awarded a $10 million grant from the U.S. National Science Foundation to leverage the advances in computer science to study and quantify climate change.

The project is meant to complement climate model simulations, which are the main tool used by climate scientists to study climate change. But what is a climate model?

A cartoon cross-section of the earth. This just an example and representations are not necessarily accurate. Image from Go Green Blog

Conceptually, you can think of the earth as a completely interconnected system such as a tennis ball. Our planet can be divided into three compartments: the atmosphere (i.e. all the air that is above the surface), the ocean (which makes up the majority of the earth's surface), and land (continents, mountains, islands, etc.) Following the tennis ball metaphor, the atmosphere would be the tennis ball's outer green fibres. The second layer of (white) fibers the ocean, and the rubber ball underneath is land.

Theoretically speaking, a change in any of these three components (atmosphere, ocean, and land) could affect the others. This is phenomena is known as feedback. So, if you put pressure on the outer layer of a tennis ball, as a result, the shape of the inner rubber ball morphs until the pressure is released. A climate model, tries to capture the relationship between the earth's components using mathematical equations (e.g. Newton's second law of motion F(orce) = m(ass) x a(cceleration)). The more realistic you want the model to be, the more complicated it gets. Especially because of the ocean-air-land feedbacks, many of which are not fully understood. Furthermore, there are several additional factors that current models account for such as the cryosphere (frozen portion of the Earth's surface), solar radiation, human impact, etc.

While climate modeling has its unique challenges for climate scientists, we as outsiders, have our own reservations. First, climate models are difficult to reproduce because they depend on numerous parameters (the physics equations the model solves and their variables). In computer science, if your results depend on model parameters, it raises questions as to whether the results are because of a good algorithm or because of random parameter selection*. Second, climate models take a very long time to run and generate tremendous amounts of data (in the order of tera- or petabytes). Moreover, it is anticipated that soon we won't be able to load a full climate dataset into memory for post-processing because of its size. Finally, we still lack the understating of the physics and thermodynamics that drive certain aspects of our climate. For instance, you might have an established theory that works well in the tropical region but breaks once you move to higher latitudes.

This blog isn't about climate models, rather it is about how computer scientists can work together with climate scientists to address some the biggest challenges we are facing today: extreme climate events, food security, global land-cover change, etc. As observational climate data (from satellites and other measurements) become more abundant and as we continue to develop scalable algorithms that can handle terabytes of data in mere minutes, we have an unprecedented opportunity to bring these two disciplines together! But first, we must start talking.

*: scroll down to the slide "Parameters (are bad)"