General Circulation Model

GCMs skillfully simulate observed climate variability on seasonal to decadal timescales, with correlation coefficients of 0.8+ for regional temperature patterns. However, uncertainty increases with projection timescale and at regional scales. By 2100, ensemble spread (range among different GCMs) is often comparable to the forced signal, meaning we can be confident in the direction and broad magnitude of change but uncertain about precise values.

GCMs simulate physical climate (atmosphere, ocean, ice, land) driven by prescribed external forcings (greenhouse gases, aerosols). Earth System Models add interactive modules for the carbon cycle, atmospheric chemistry, and ecosystem dynamics. ESMs are more complex but provide coupled feedbacks; GCMs are simpler and computationally cheaper for ensemble projections.

GCMs differ in spatial resolution, parameterization of subgrid processes (clouds, convection), land-surface schemes, ocean mixing, and parameter choices. These differences lead to different climate feedbacks, particularly cloud feedback, which varies among models. Using multiple GCMs (ensemble multi-model approach) is essential to characterize uncertainty.

GCMs are not designed for weather prediction; they lack the initialization and data assimilation procedures used in weather models. GCMs can sometimes provide seasonal forecast skill (e.g., El Niño forcing can predict temperature/precipitation anomalies 3–6 months ahead), but this is not their primary purpose. Weather prediction uses specialized models (e.g., ECMWF, NOAA GEFS) updated daily with observations.