How can we predict how the climate will change?

How can we predict the climate when weather forecasts become inaccurate after just a few days?

Weather forecasting is an initial value problem. You start with details of the currently observed weather situation, and by applying the laws of physics, project forward how weather systems will develop over the next few days. Using these methods useful predictions of up to 8 days (in rare cases, up to 4 weeks) are possible.

Climate forecasts are produced in a different fashion, as here the problem is fundamentally a boundary value one. The circulation of atmosphere and ocean in such a climate model is not dependent on the initial state of the model but rather on the boundary conditions like the input of solar energy and the chemical composition of the Earth’s atmosphere (e.g. greenhouse gases). You cannot predict the weather for individual days with a climate forecast (for example, the question of the temperature in Hamburg on the 23.12.2005 is meaningless), but you can say something about the average conditions for an area (e.g. the average January temperature between 2010 and 2020), as well as the probability and magnitudes of deviations from this average.

The same basic computer models are used for both weather and climate forecasts, but they are adapted to their specific purpose. In addition, whilst weather forecast models often concentrate on the atmosphere, climate models require extra features to be coupled in, for example a three-dimensional model of the world ocean.

Climate models are calibrated so that they reproduce both modern day and past climate. The relevant boundary conditions are set for the time period in question, such as the sun’s output of radiation and concentrations of greenhouse gases, and the model computes the climate that results. The output from the model can then be compared against observations of the real climate. The fact that climate models can simulate both current and past climate to reasonable degrees of accuracy gives us some confidence that their predictions of future climate also have some merit.

Assumptions must be made when predicting future climate, like how the world’s population will change, how much energy will be used and how the use of the various fuel sources will develop in the future. These assumptions allow one to calculate different greenhouse gas concentrations for different future energy usage patterns, which then serve as input parameters for climate models. The most commonly used scenario is known as ‘business as usual’ (that we continue increasing emissions of greenhouse gases into the future with the same trend as today) because it supposes an increase in the concentration of long-lived greenhouse gases of about 1% a year - the same rate of increase as has been measured in the past decade.

It is worth taking some care when interpreting predictions of future climate change if the future emissions scenario used in the forecast is not specified along with the results.