Computer models are used in many aspects of life, testing the structural integrity of buildings, controlling traffic lights, directing airplanes and spacecraft, and in the form of video games. The term “model” can also represent different concepts in different science disciplines. In Climate Science, and the associated subjects of meteorology and oceanography (the studies of weather and oceans, respectively), computer models serve to substitute reality for a range of scenarios.


Changes in the climate are large-scale relative to the many complex interactions in small-scale physical systems. While lots of significant changes in the climate can be investigated with basic climate models, the modelling community are continually developing models to better understand every aspect and scale of the Earth’s climate system.


How are climate models constructed?

Earth’s climate varies from place to place, therefore the surface of the planet needs to be divided up into boxes (“grid cells”) to accurately represent every part of the planet. Each grid cell contains climate-related physical information about that individual location. Tens of thousands of grid cells cover the surface in any climate model. The model becomes 3-dimensional when information is then included for grid cells above and below the Earth’s surface, for example, different levels of the oceans and atmosphere.


The inputted data is represented by mathematical equations based on process-based physical laws. Fundamental principles (including the conservation of energy, momentum and mass) and processes (such as orbital mechanics) form the core physical inputs. Theoretical physics which is well known (for example, the transfer of radiation and equations of fluid motion) has to be approximated in practice. Meanwhile other empirically known (observed) physics (such as that of evaporation as a function of wind speed and humidity) are also included in all models.


Data is averaged across smaller scales to provide a summary of the processes, a practice known as “parameterisation”. Models can only account for a limited number of infinite real processes, and the computerised processes are often simplified compared to reality. Climate models are also closed, meaning that they have a limited number of specific inputs and cannot be influenced by unaccounted for factors.


Grid resolution of climate models. Many levels in the atmosphere and ocean are incorporated in models. Shown here is the progression of Met Office’s Hadley Centre climate models to include more information.


What information is included in the models?

Knowledge of the climate comes from modelling the interaction of the global climate with regional physical variables, included in a model as a grid cell. Climate models use both palaeo-data  and observational data. Every aspect of the Earth needs to be incorporated into the models. Each interconnected component has its own independent modelling and varies over time:

  • Atmosphere – contains many altitudinal levels of different air pressure, temperature, water moisture and gas concentrations;
  • Oceans – have a larger inertia (long reaction time) than that of the atmosphere, and contain many levels of differing processes;
  • Cryosphere (ice sheets and sea ice) – has its own models which relate to sea level, temperature and atmospheric conditions;
  • Biochemical cycles – for example the carbon, oxygen, nitrogen and water cycles, which affect the radiation balance;
  • Biosphere (as vegetation) – occurs in a range of influential forms across vast areas of the Earth’s surface with a strong control on the biochemical cycles;
  • Humans – have the ability to modify land surfaces and affect the biochemical cycles;
  • External – influence from the sun (through orbit and sun-spot cycles) and other cosmic processes, and volcanic activity (through CO2 and aerosols release).


How are climate models tested?

Models are assessed at both the small-scale, testing the specifics of parameterisation, and the large-scale, where prediction scenarios can be evaluated. Firstly, climate is modelled and tested for the present era, as this is the period of observed and monitored data, the most reliable information we have. The models can then be compared to palaeo data, before finally investigating probable future scenarios.


Mount Pinatubo, in the Philippines, erupted in 1991 and was the second largest land-based eruption in the world during the last century. Volcanic ash can accumulate in the atmosphere and encircle the Earth, causing short-term (over a few years) changes in the global climate. This eruption therefore provided a great test for climate models. A subsequent cooling of 0.5 °C was correctly forecast and the radiation, water-vapour and dynamical feedbacks included in the climate models were accurately verified.


Climate models are developed and tested by teams of scientists worldwide, specialising in mathematics, physics and climate disciplines. The models are continually being improved to account for more and more observations, better understood dynamics and complex variability. The scientific climate modelling community frequently perform comparisons of different models to understand any issues with simulations. The largest collaborative project occurs for the International Panel on Climate Change (IPCC) reports, providing comprehensive reviews of climate models every few years.


In order to further develop current climate models and reduce the uncertainty of future climate projections, some aspects of the models require improvement. The turbulent behaviour of the near-surface atmosphere, the effects of ocean eddies and the microphysics of clouds and aerosols need to be better incorporated into climate models.


Models aim to replicate and predict reality correctly, however it is not always possible to test that the results are “correct”. Therefore a model is judged on how well it reproduces reality, not whether it is right or wrong.


How are climate models used to further understand climate?

The outputs of climate models are frequently used by farmers, town planners, insurance companies, the energy sector, those concerned with water management, as well as scientists. The main goal for Climate Science and climate models is a prediction of future climate. While key physical inputs such as the Earth’s position relative to the sun can be calculated accurately for the future, we do not know how much carbon dioxide and methane will be released into the atmosphere. Therefore climate models produce results for different scenarios. The last IPCC report included 4 projected temperature scenarios (see figure below), and these will be updated later this year (2013) with the publication of the next IPCC report.


Modelled scenarios of surface temperature, averaged from multiple climate models. The models have been validated against palaeo-data and observational data for the past. Simulations are then projected into the future based on 4 different scenarios. The shading either side of the simulations is the uncertainty of outputs (1 standard deviation) between the different climate models.


Repeated tests carried out and comparisons made between models often lead to new research questions. For example, a certain area of the Earth (like the tropical Pacific) may show a lot of unexpected temperature variation from year-to-year, or the West Antarctic Ice Sheet might have suddenly collapsed at a particular time in the past. The scientific community can then address these questions by focusing collection and analysis of data to obtain an improved understanding.


Below is an animation by the Met Office explaining how climate is modelled and why modelling it is useful.