How Climate Models Work
Model Architecture
A climate model couples sub-models for atmosphere, ocean, land surface, and sea ice. The atmospheric component solves Navier-Stokes equations on a rotating sphere, calculating wind, temperature, and moisture at each grid cell every 15 to 30 minutes. The ocean models three-dimensional flow, heat transport, and salinity. Land surface models track soil moisture, vegetation, and carbon. Modern Earth System Models add chemistry, carbon cycle, and dynamic vegetation.
Resolution has improved from 500-kilometer cells in the 1970s to 25-50 kilometers today, resolving tropical cyclones and frontal systems. Components exchange information at interfaces: atmosphere passes wind stress and heat to ocean, ocean returns surface temperature. Each advance in resolution reveals new processes but demands exponentially more computing power.
Parameterization
Sub-grid processes like cloud formation, turbulent mixing, and convection must be parameterized through simplified equations. Cloud parameterizations are the largest uncertainty source, as different approaches yield different climate sensitivity estimates. Machine learning trained on high-resolution simulations is improving these representations. Convective parameterizations approximate thunderstorm effects at 1-10 km scales.
Validation
Models are tested against observations, paleoclimate, and natural experiments. Key tests: reproducing warming trends, spatial patterns, seasonal cycles, and ENSO variability. When Pinatubo erupted in 1991, models correctly predicted 0.5 degrees cooling and recovery timing. Paleoclimate simulations of the Last Glacial Maximum and Pliocene build confidence across different climate states.
Scenarios and Projections
Shared Socioeconomic Pathways (SSPs) describe different emission trajectories. SSP1-1.9 limits warming to about 1.5 degrees; SSP2-4.5 reaches about 2.7 degrees by 2100; SSP5-8.5 reaches about 4.4 degrees. These are if-then scenarios showing consequences of different choices. Climate sensitivity is constrained to 2.5-4.0 degrees per CO2 doubling through combined evidence.
Climate models are physics-based simulations validated against observations. They successfully reproduce past changes and project futures under different scenarios. Main uncertainties come from cloud processes, not fundamental physics.