# Climate Models Climate models are computer simulations of Earth's climate system, built from the established laws of Physics — conservation of energy, mass, and momentum, and the radiative properties of gases. The planet is divided into a three-dimensional grid of cells covering atmosphere and ocean, and the equations governing energy and fluid flow are solved step by step. The most comprehensive are general circulation models (GCMs) and their successors, Earth system models, which add the [[Carbon Cycle]], vegetation, and ice. Models are not where climate science's core conclusions come from — those rest on basic radiative physics and the paleoclimate record — but they are how scientists quantify *how much* and *how fast*, and they are the part of the field deniers most relentlessly attack. ## How They Are Tested A model is only as trustworthy as its track record against reality, and climate models are tested in several independent ways. They are run backward over the past century to see whether they reproduce observed warming (they do, including the mid-century plateau caused by industrial aerosols). They are tested against natural experiments such as the 1991 eruption of Mount Pinatubo, whose cooling effect models predicted well in advance. They are validated against the deep-time record from Paleoclimatology and Ice Cores. And they are checked against the specific spatial fingerprints of greenhouse warming: stratospheric cooling above tropospheric warming, polar amplification, greater warming over land than ocean, and the narrowing of the day-night temperature range. Models reproduce all of these — patterns that solar or natural-cycle explanations cannot. ## The Track Record The "models always run too hot" claim collapses under examination. [[James Hansen]]'s 1988 projections to Congress have proven remarkably accurate when judged against the emissions that actually occurred. A 2019 study by Hausfather and colleagues in *Geophysical Research Letters* evaluated 17 model projections published between 1970 and 2007 and found 14 of 17 consistent with subsequent observations once actual greenhouse-gas emissions were accounted for. The key qualifier is that models project *conditional* futures: given a particular emissions path. When a projection is compared against the emissions that genuinely happened rather than a hypothetical, the agreement is strong. Comparing a model's high-emissions scenario against a lower-emissions reality and declaring the model "wrong" is a sleight of hand, not a finding. ## Why the Attack Persists Models are an attractive target precisely because they are complex and because "it's just a computer model" sounds like a devastating critique to a lay audience. The rhetorical move exploits the gap between how scientists use models — as tools for understanding processes and bounding uncertainty — and how the public imagines them, as crystal balls that either get the future exactly right or are worthless. The genuine uncertainties that remain, chiefly in cloud feedbacks and therefore in [[Climate Sensitivity]], are real and openly published, which makes them easy to quote-mine as if they were admissions of fundamental failure. ## Connections Models translate [[Emissions Scenarios]] into projected warming, and their spread is dominated by uncertainty in [[Climate Sensitivity]]. Their validation draws on Paleoclimatology and Ice Cores and underpins the [[The Scientific Consensus on Climate Change]]. For the field overview, see [[Climate Science]].