A Brief History of Climate Models

Read...

Climate models have grown remarkably complex over the past few decades as both computing power and our understanding of the Earth system have advanced. The graphic below shows how climate models have evolved, and as you can see, the earliest models from the 1970s were quite basic—they included only essential atmospheric processes and a few greenhouse gases. But with each passing year, we gain new insights, and our computers get more powerful. Today’s models (shown toward the bottom of the graphic) are far more sophisticated. They incorporate detailed representations of the land surface, oceans, and ice coverage, and they can simulate complex exchanges of carbon and water between the surface and atmosphere.

The earliest GCMs were quite crude in their depiction of the climate system, but GCMs have become increasingly sophisticated and realistic in recent decades. However, even the most sophisticated GCMs still can't fully match the complexity of the real climate system.

Credit: The World in Global Climate Models by the Intergovernmental Panel on Climate Change (IPCC)

By the way, the acronyms, “FAR,” “SAR,” “TAR,” and “AR4” in the image above refer to the state of GCMs at the times of the first, second, third, and fourth assessment reports (AR) of the Intergovernmental Panel on Climate Change, respectively. We'll talk about the IPCC on and off through the rest of this class. For perspective, the first assessment report (FAR) was published in 1990, and the fourth (AR4) in 2007. There’s also been an AR5 and an AR6 since this figure was generated – over this time, climate models have become even more sophisticated. But even with the increasing sophistication of these models, the latest and greatest ones still can't match the true complexity of the real climate system. I’m not sure any of us will be alive to see every molecule of the atmosphere accurately predicted for the next 100 years!

The most advanced models today are “fully coupled,” meaning the atmosphere, land, ocean, and ice components all interact with each other within the model. Changes in the atmosphere impact the ocean and vice versa, creating a more realistic and interconnected simulation than simulating the two independently.

Want an analogy for "fully coupled?" Think of it like a football play -- yes, each wide receiver can run their own route and go off and do their own thing. However, the play is much more likely to succeed if all the players respond to each other. For example, we don't need three linemen blocking the same pass rusher; they need to interact and respond accordingly! While very simple, this ability to interact and respond is why it is important for the components of a climate model to be coupled, i.e., talk to each other within a model.

A scientist showcases how a National Science Foundation-sponsored supercomputer for climate modeling uses water-based cooling to manage its temperature.

Credit: "Derecho." NCAR. 2024.

These incredibly sophisticated models demand enormous computing power to handle the sheer number of calculations they require. While some simpler climate models can actually run on your smartphone (yes, really!), the models used for studying global climate and informing policy decisions need vastly more power than what’s available to everyday consumers. To meet this demand, these models rely on the world’s fastest supercomputers—machines capable of performing trillions of calculations per second. Think of a supercomputer as thousands of laptops stitched together with cables, all working in perfect sync. Each grid cell in a climate model—each piece of its "skeleton"—has its own set of equations for energy, moisture, and momentum. Multiply this by thousands of cells spanning the globe, and the computational load becomes staggering. This isn’t something you can just run at home! Supercomputers break the problem into smaller chunks, assigning each piece to a different part of the system so they can all work simultaneously. To give you an idea of scale, the largest supercomputers boast more than a million times the processing power of the fastest consumer laptop.

The image above shows a worker at the NCAR-Wyoming Supercomputing Center inspecting a section of the water-cooling system for one of these supercomputers. Notice she’s working with just one "cabinet"—a tiny fraction of the full machine used for climate modeling. Each of those red and blue tubes carries water to cool a specific part of the supercomputer. To put it in perspective, each "small piece" of the supercomputer contains 128 "cores." For comparison, the laptop, tablet, or phone you’re using right now probably has between 1 and 4 cores. This immense power and scale are what make modern climate modeling possible.