Concern about potential human impacts on the climate system has generated interest in predicting future climate and focused public attention on climate models—the tools used to make such predictions. Climate is the result of complex interactions among a number of factors: solar radiation, greenhouse gases, land cover, etc. The interactions between these factors are usually non-linear; in mathematical terms, climate is a chaotic system. This means that small changes in inputs (e.g. solar radiation) can have large and unpredictable effects on the system’s output (e.g. temperature). However, chaotic systems can be predictable over a limited range of change. Based on our understanding of the climate of the last century, it is reasonable to assume that climate may be predictable for a few decades into the future, but unpredictable centuries into the future.
Climate models are a mathematical representation of the physical and chemical processes occurring in the climate system. Because our understanding of these processes is incomplete, current climate models do not accurately represent the climate system. Some climate models have been adjusted, or calibrated, to provide a reasonable simulation of some aspects of recent climate. However, calibrating a model to make its output look more like the real world does not provide a basis for assuming it will generate realistic predictions of future climate. Realistic predictions of future climate are assured only if the climate model is validated and run with an accurate set of inputs.