linked in logo | IS-ENES | Contact

Exploring climate model data

Uncertainties related to the climate models

The current generation of climate models are able to faithfully represent may aspects of the climate in a reliable way. However, because the full global climate system is very complex and involves processes at many spatial and temporal scales all climate models will by necessity have to include simplifications. And these simplifications will lead to uncertainties in the projections of future climate.

Main challenges when modelling the climate system

Measurements of the atmosphere and the ocean — or weather observations for short — are made at manual or automatic stations, with satellites or balloons, and by means of other technical systems. These observations, which represent weather parameters at a specific point in time at a specific place, are collected and combined to provide an integrated picture of the weather. Despite the wealth of data there are limitations to this picture because not all weather parameters are measured at every point in time and space.

In a similar way, climate models cannot represent every point in the atmosphere and the oceans, but make use of a grid and simulate mean values of weather parameters in each grid box. Most global climate models have resolutions of 100-300 km. This means it is difficult to compare model results directly with observations. For example one station might observe a heavy rainfall, when nearby stations observe just small amounts or nothing at all. If the climate model simulates the same amount of precipitation it will distribute the precipitation equally within the current grid box. The precipitations amount is the same in the model as in reality, but the intensity is much lower in the model. The topography is also described as a mean value inside a grid box of a climate model.

Climate change simulations gives 'scenarios' and not 'forecasts' (read more about the differences here), partly because of simulations is based on assumptions of how the world will develop and partly because of the temporal resolution of climate models. A climate model gives a probable realization of the weather, with realistic characteristics, but cannot be implemented as a forecast. A weather forecast gives information on a specific place at a specific time. The non-linearity of the climate system limits the length of a useful weather forecast but makes it possible to calculate the development of a climate system over a long period of time. 30 years is often used as a minimum period to use when analyzing climate.

In the same way as we cannot expect the model to be in phase with reality we cannot expect two different models to be in phase with each other. This is called natural natural variability and is a result of the non-linearity of the climate system. A way of sampling the uncertainties is by using an ensemble approach.

Key sources of uncertainty in climate models

To better understand the uncertainties that are related to the climate models it is useful to divide them into different categories.

  • Structural modelling uncertainty due to limitations in the scientific understanding of the processes. One example concerns the future ability of land and sea to absorb carbon dioxide. The understanding of this ability is limited, mostly due to gaps in the knowledge about the long-term response of biological processes to changes in carbon dioxide concentration and climate.
  • Uncertainty in parameterisations of sub-grid-scale processes. For example at the micro-scale there is substantial scientific knowledge of cloud processes, but how to represent these processes at the macro-scale of climate models is still scientifically challenging.
  • Stochastic uncertainties arising from coupling between unresolved sub-grid-scale variability and the resolved grid-scale flow (i.e. the “butterfly effect”.)
  • Available computer resources impose limits to the spatial and temporal resolution that can be implemented in the model simulations.
  • The initial model state is not fully known. For the starting point of a typical GCM ‘historic’ simulation this could be 1850 or 1860, and for that time period the knowledge of the state of the atmosphere and ocean is not fully known, in particular knowledge about ocean temperature and currents are limited.

Related links

Model calendar and climate variability



The IS-ENES project has received funding from the European Union's Seventh Framework Programme for research, technological development and demonstration.