Some of the earliest attempts to model Earth’s climate, like so many of the first stirrings of science, can be traced to the ancient Greeks. They made the connection between climate, the planet’s spherical shape and its angle to the sun [Edwards et al., 2010]. By the 19th century, atmospheric circulation, the effects of what would come to be called greenhouse gases, and the climatic role of carbon dioxide were being modeled conceptually. Fluid-filled globes and other physical models arrived in the early 20th century, with the first low-speed computer simulations appearing by the mid 1950s.

Today these mathematical chains of computer code provide representations of critical climate processes—the atmosphere, ice and oceans—across timescales, with increasing sophistication and accuracy. Modeling ensembles, such as the Coupled Model Intercomparison Project (CMIP), grow ever closer to the goal of interlinking disparate models to simulate the entire climate system [Flato et al., 2013, Edwards et al., 2010].

Challenges remain, however, in accurately modeling such factors as clouds, precipitation and extreme events [Flato et al., 2013]. The coupling of land ice models to ocean models has so far been achieved only in limited circumstances. And while land ice and ocean models can be run separately, with their results later combined, the combined results do not currently account for potential feedback between land ice and oceans.