Climate models are representations of physical processes within and in between the atmosphere, the land surface, the oceans and the sea ice. In some models, also lakes and yet additional components are included. There is a hierarchy of climate models. The so-called General Circulation Models (also known as Global Climate Models), the modelled atmosphere, land surfaces and ocean is divided into a three-dimensional computation grid that extends upwards in the atmosphere, down into the ocean and also covers the surface of the globe.
When the climate model is then run over some past, present or future period, in every point in the grid, the development over time of meteorological, hydrological, and ocean parameters are calculated. Comprehensive climate simulations indeed need to be done with global models, as the climate system is coupled in space and time. What takes place somewhere on the globe soon affects the conditions somewhere else.
A comprehensive climate model requires lots of computing power. Consequently, in global models, the grid has to be relatively sparse. This results in very little detail on regional and even more so on local scales. Regional climate models are used to study specific areas in more detail. The grid can be put over e.g. Europe. The smaller area allows for a denser grid and thus more detailed results.
What happens outside a regional model domain is given by results from an earlier global climate model simulation, used as so-called boundary conditions in the regional simulation.
At SMHI, and in particular for the scenarios shown here, the regional climate model of choice is the Rossby Centre regional climate model, RCA3 (Kjellström et al. 2005). The boundary conditions are fetched from simulations made with the ECHAM4 and ECHAM5 global climate models, kindly provided by the Max-Planck-Institute for Meteorology, Hamburg (Roeckner et al. 1999).
The RCA3 model covers the region of Europe. The regional horizontal resolution is approximately 50x50 km.
When climate models are applied in simulations of some specific period, information is needed on possible changing conditions (e.g. land use, emissions, solar variability etc.). When it comes to man-made climate change scenarios, IPCC (Intergovernmental Panel on Climate Change, Nakicenovic, N. & Swart, R. (ed.), 2000), has worked out scenarios of the development of the world up until year 2100. These make different assumptions of population, economic growth, technological change etc. in the world. These assumptions lead to scenarios of greenhouse gas emissions over time. These emissions in turn give rise to atmospheric composition changes, such as the amount of carbon dioxide in the air. It is important to remember that future scenario calculations are based on assumptions on the future world development.
The regional climate change simulations here are for the period from 1961 to 2100. The models simulate possible climate response to the man-made emissions, but also the internal (unforced) variability, from 1961 on, in every scenario. The results can therefore differ some already in the early decades. Later on, differences in the assumptions underlying the atmospheric composition scenarios of course also lead to differences in the climate scenarios.
One way to evaluate climate models is to compare model simulations and observed data over past and present periods. For this, the so-called climate normal period of 1961-1990 is often used. Evaluation measures the capability of the model to reproduce average conditions, but also statistical properties of variability and extremes. The period 1961-1990 is also often used as a reference when looking at climate changes. A scenario result might consequently be expressed as a difference from the 1961-1990 climate conditions.
Observations are made at stations, with satellites, weather balloons etc. Observations represent conditions in a certain point or for some smaller or larger area. Climate models have a regular grid, the model simulating average conditions in each grid square. This introduces complications in comparing observations and model simulations. Consider, for example precipitation. It can fall in large amounts at and in the vicinity of a single observation station. At the same time little or no precipitation might fall in the surroundings. In a climate model, precipitation is typically spread evenly in a grid square. Even if the amount is the same in the observation and in the model, the intensity becomes much less in the model than what in reality was measured in one station.
Scenarios are not forecasts
Climate change simulations lead to "scenarios", instead of "forecasts". This is both because the simulations are based on assumptions about the future world. Climate models don't either reproduce the exact weather in a specific location at a specific time. Rather, a good quality model provides a plausible realisation of the weather, with realistic statistical properties. A weather forecast on the other hand aims to provide information about what will happen in a specific location at a specific time. The non-linearity of the climate system limits the length of a useful weather forecast, but still makes it possible to calculate realistic climate system evolution over much longer times. This is the basis of climate simulations.
The results presented here are based on different emission scenarios and different global climate models, downscaled into more detail with a regional climate model. These results are based on scientific scenarios, but cover only few of many conceivable calculations. Similarities across different scenarios is a measure of the robustness of the results, while differences imply uncertainties.
There are many common features in the results, for example the warming during all seasons and increased precipitation during winter. Sources of uncertainties are:
- Which emission scenarios that are used
- Climate models
- Natural variability
An emission scenario is an assumption about how future emissions of greenhouse gases will develop. Results from calculations are based on these assumptions. The emission scenarios used here are both consistent and conceivable, but none of them can be regarded as more probable than the others.
Studies where different climate models are compared reveals a relatively large spread in model results for some climate variables. This is because different models describe atmospheric processes differently. That can also affect how much the climate is projected to change due to changes in the amounts of greenhouse gases. The scientific question is how sensitive to changes the climate system is, and about the so called feedbacks. Changes in cloud cover as a result of a warmer climate is one example of feedback.
The third source of uncertainty is natural variability. The climate in the model is not expected to be in phase with the real climate. But a good quality model should be able to calculate good averages and a climate with representative, characteristic variability, e.g. the right number of cold and warm winters in a 30-year period. The cold and warm winters could still appear in another order than in the observed climate.
These uncertainties are studied by doing several computations with different emission scenarios, climate models and initial conditions. To look at results from several model simulations gives the opportunity to deal with uncertainties, but also means to judge which results that are robust and which that are not. Results that appear to be contradicting might seem confusing, but one should use the extra information given in this fact. If the models give different results, it is simply an uncertain results. If the models on the other hand give similar results, then it is a more certain result.
Besides looking at scenarios one by one, statistical methods and special analyses can be used to combine several simulations to get a result that on one hand is better than any individual simulation, and on the other hand is a combined but still complete data set.
In the regional climate model RCA, a rotated coordinate system is used. This means that the north and south pole in the model's coordinate system are not the geographic north and south pole, but shifted in a certain way. This is a technical choice made to enable an as even spacing of the model's longitudes and latitudes as possible, to achieve an even resolution in the regional model domain. When using the results in analysis or graphics in a regular grid, the data need first be transformed to the appropriate coordinate system.
Used models and scenarios
Data are available from the simulations with the following models and scenarios.
ECHAM4: A2 and B2
ECHAM5_r1: A1B, A2 and B1
For scenario A1B the model ECHAM5 has been run three time with different initial conditions, these three runs are numbered r1, r2 and r3.
New file format
File formats and units are adjusted according to the international standard for climate data, CF (http://cf-pcmdi.llnl.gov/). The variables have changed names and in some cases units. Some variables are aggregated in new ways, leading to differences between the new and the old files that is not due to unit changes (see list on changes in file names, variable names, units, aggregation here).
Kjellström, E., Bärring, L., Hansson, U., Jones, C., Samuelsson, P., Rummukainen, M., Ullerstig, A., Willén, U. and Wyser, K., 2005. A 140-year simulation of European climate with the new version of the Rossby Centre regional atmospheric climate model (RCA3). SMHI Reports Meteorology and Climatology No. 108, SMHI, SE-60176 Norrköping, Sweden, 54 pp.
Lind, P. and Kjellström, E., 2009. Temperature and precipitation changes in Sweden, a wide range of model-based projections for the 21st century. SMHI Reports Meteorology and Climatology No. 113, SMHI, SE-60176 Norrköping, Sweden, 50 pp.
Nakicenovic, N. & Swart, R. (ed.), 2000. Special report on emissions scenarios. A special report of Working Group III of the Intergovernmental Panel on Climate Change. Cambridge University Press. 612 pp.
Roeckner, E., Bengtsson, L., Feichter, J., Lelieveld, J. and Rodhe H., 1999. Transient climate change simulations with a coupled atmosphere-ocean GCM including the tropospheric sulphur cycle. J. Climate, 12, 3004-3032.