Climate models have become a key instrument in climate research since the 1970s. The most comprehensive and common type of climate models are so-called “General Circulation Models (GCM),” which evolved from numerical weather prediction models developed in the 1950s.These models are computer programs based on the physical laws which simulate major features of the behaviour of the atmosphere and other earth systems. In other words, climate models are a simplified representation of the climate system in a purely mathematical language. They serve to compute the state and the change of the climate system in terms of properties such as temperature, pressure, wind, moisture etc. For doing so, the atmosphere and the oceans are broken up into a three dimensional grid, for each of which the atmospheric properties are computed.

Versions of climate models can be used to study and better understand atmospheric and climatic processes and their interactions [see “heuristic use of climate modelling”] Versions of climate models can, alternatively, be dedicated to compute projections of  future (or past) climates [see “predictive use of climate models”].

As climate models simulate the earth’s climate system, they are used for many virtual experiments. One example is the investigation of the rise of future CO2 concentrations (along with other potential changes) and the consequences of these changes in the virtual climate system. This is a most important research strategy, because such experiments cannot be pursued in the laboratory or in the real world. Hence, climate models represent a virtual laboratory.

While climate models are very powerful research tools, they also suffer from significant limitations. They do not represent the climate system in a realistic way. The spatial resolution of global climate models is usually limited to several hundred kilometres. Smaller scale processes have to be represented in a simplified and generalized manner called parameterization. Hence, scientists have to test and validate climate models and their behavior very carefully by comparing data sets based on simulations with data sets based on observation and by comparing model behavior with other models. A large part of research activities are hence invested into the improvement of the models.