Online climate change projections report Annex 3.1
We can describe the climate system using mathematical equations derived from well established physical laws that capture the evolution of winds, temperatures, ocean currents etc. Computers are used to solve the equations in order to resolve all the complex interactions between components and processes and produce predictions of future climate change (see Chapter 2, Box 2.1 for more information). The core computer code for the atmosphere component of the Met Office climate models is the same as that used to make daily predictions of weather.
The equations of climate are, in the case of the Met Office model, solved by dividing the world up on a grid which follows lines of longitude and latitude and extends above the surface of the Earth and below the oceans (see Fig 2.4). Physical properties such as temperature, rainfall and winds evolve in time on this grid, and these short time scale variations are averaged together to produce climate averages (monthly means, for example). Because the time-variation of atmospheric and oceanic motions is chaotic, it is not possible to reproduce the exact time variation of the real-world weather and climate (it is chaotic behaviour which limits weather forecast accuracy to about a week). Rather the model is representative of one possible trajectory the system may take. This “uncertainty due to natural variability”, is one aspect of the uncertainty captured in the presented in this report.
The size of the grid boxes is limited by the amount of computer power available. Halving the size of the grid boxes in the horizontal and vertical direction makes the model more than 10 times slower to run. A balance must be achieved between resolution and run-time to ensure that enough model experiments can be performed to cover a range of future possibilities. The resulting grid boxes in a global climate model are a few hundreds of kilometres wide in the horizontal. Even in the regional version of the climate model (RCM) they are 25 km, so they cannot resolve all the atmospheric motions and interactions in a single cloud which evolve on much smaller scales. For this reason, small-scale processes must be parameterised i.e. the effect of the small-scale processes on the grid-box scale variables must be simplified in some way.
The critical aspect for climate prediction is that many of the physical processes that are parameterised in climate models are also involved in the physical feedbacks which determine the effect of increasing greenhouse gases on climate, and set some of the regional aspects of climate change. Also important are interactions between the parameterised processes and the coarsely resolved dynamical motions. Parameterisations are necessarily simplified estimates of how the real-world works; hence there is inherent uncertainty in the modelling approach. In UKCP09 we systematically explore these uncertainties by varying parameters in the Met Office Hadley Centre climate model and include information from other climate models in order to quantify the uncertainty in climate predictions arising from parameterised processes.