Climate Models – Simulating Reality

We use weather to refer to the current state of the atmosphere. Weather manifests as wind, rain, clouds or sunshine, which means it’s something everyone can see and feel, and can change in a matter of hours or days. In contrast, climate refers to the long-term state of the atmosphere and ocean over decades, centuries or millennia. To gauge the climate, the statistics of the meteorological parameters also used to determine the weather are calculated for an extended timeframe – most often for the so-called 30-year “normals”. As such, in comparison to weather, climatic changes produced e.g. by greenhouse-gas emissions take place far more slowly and, as a rule, cannot be directly seen or felt.

Today, no-one can say exactly what the future climate will look like. After all, especially when it comes to the human factor, there are still plenty of open questions. How much carbon dioxide and other greenhouse gases will human beings emit? Will emissions climb, remain stable or decline over the next few decades? And to what extent will the planet’s forests be cleared? Consequently, to make forecasts, climate models run through various scenarios using the “what if” principle. In this regard, the IPCC has described a number of scenarios that are taken up by the modelling teams. In news coverage, especially the “business as usual” scenario is often mentioned. This vision of the future, also referred to as the worst-case scenario, shows a global economy that essentially keeps doing things just as it did in the past, consuming massive quantities of fossil fuels and therefore emitting more carbon dioxide.

How can researchers ensure that all climate-relevant processes are accurately reflected – i.e., that their model is a good one? They look to the past. All around the world, there are natural climate archives – the sediments on the ocean floor, ancient ice in the Arctic and Antarctic, wood for tree-ring analyses, and more. These archives contain what is referred to as proxy data, which researchers can use to reconstruct the climate of past millennia – e.g. the temperature or the precipitation distribution. For climate change, the recent past since the beginning of industrialisation is particularly important. For this timeframe there is directly measured meteorological data, which the modellers take advantage of. A good climate model starts its calculations in the past and is able to correctly simulate the past, i.e., it delivers outcomes that can be easily reconstructed and match the actually measured data. After completing successful test runs through the past, the model is ready to calculate the future. In order to gain additional certainty in this regard, too, various modelling teams from around the globe meet on a regular basis to assess their models jointly. To do so, they feed the same initial data into their climate models and compare the simulation outcomes – e.g. the average temperatures for northwest Europe in 2100. The differences between the simulations offer valuable insights into uncertainties in the respective models.

The various “what if” scenarios examined in the models clearly show the potential impacts of climate change in the next several decades if nothing is done to reduce greenhouse-gas emissions. In addition, they show which impacts can still be avoided or mitigated if appropriate measures are taken. This involves not only fairly abstract parameters like the global mean surface temperature, but also very concrete aspects like the intensity of storms, droughts and heavy rains, projected crop yields, and the scale of storm surges and heat waves. This provides important guidance for political decision-makers. On the one hand, politicians have to decide which path humanity should follow, which scenario should become a reality. On the other, they have to roll out adaptation measures to prepare society for what is to come.

In its Assessment Reports and Special Reports, the IPCC regularly summarises the outcomes of numerous climate models. In its sixth Assessment Report (2021/22), the IPCC predicts a global sea-level rise of 28 to 55 centimetres by 2100, based on a scenario with greatly reduced greenhouse-gas emissions. If humanity continues to produce extremely high quantities of CO₂, the global sea-level rise could even be 63 to 101 centimetres. Compared to the preindustrial era, the global mean temperature has risen by 1 degree Celsius. Depending on the scenario, the IPCC projects an increase of 1.5 to 4 degrees Celsius by 2100. In all scenarios, this leads to an increased frequency of extreme weather events like catastrophic droughts and heavy rains. If global warming (in a more favourable scenario) remains below 1.5 degrees Celsius, according to the IPCC the risk of exceeding certain tipping points in the climate system will be lower. This assessment was also the basis for the 1.5-degree target included in the Paris Agreement, in which the Member States of the United Nations committed to limiting global warming by 2100 to 1.5 degrees Celsius.

A tipping point is a critical threshold in the climate system. If this point is exceeded, there is a risk of serious consequences, in some cases even catastrophic and irreversible ones. A good example is the melting of the Greenland Ice Sheet. If a certain global mean temperature is exceeded, the already dwindling ice sheet could become so unstable that its complete loss could no longer be avoided, even if greenhouse-gas emissions were substantially reduced. Were this ice sheet to melt completely, this factor alone would produce a global sea-level rise of up to seven metres. Moreover, the tremendous amounts of meltwater produced could influence ocean currents, sparking massive changes in the climate, especially in the North Atlantic.

How good a given climate model is especially depends on how well the various climate modelling teams can represent the biological, chemical and physical processes in the atmosphere and the oceans in their programming codes. For example, calculating how clouds will change in step with climate change is particularly difficult – and this in turn can produce impacts in both directions. Such changes could significantly intensify global warming – or lessen it. Why? Because clouds can interact with sunlight and the heat radiation given off by the Earth in different ways. On the one hand, they can effectively reflect sunlight, producing cooling effects; on the other, they can intensify global warming by producing effects similar to those of greenhouse gases.

When it comes to climate modelling, collaboration is essential. In this way, individual research institutes can pool their resources and capitalise on their respective strengths. To portray physical processes in the atmosphere, the AWI experts employ specially designed atmospheric models like ECHAM [TS1], which was developed by the Max Planck Institute for Meteorology, and the more recent IFS Model, which was developed by the European Centre for Medium-Range Weather Forecasts. However, these models do not reflect in detail processes in the ocean, which have a major influence on the climate. In response, the AWI has developed a special ocean model (FESOM), which simulates e.g. ocean currents and sea ice. For their climate simulations, AWI researchers have coupled the two atmospheric models and FESOM, yielding the “AWI Climate Model”. This allows them to run climate simulations that consider both the atmosphere and ocean in detail.

Using the AWI Climate Model, simulations can be run at both the global scale and in detail for specific regions. This is possible largely thanks to FESOM’s unique properties: unlike many other ocean models, FESOM does not use cube-shaped grid boxes. Rather, it employs the finite volume method. In simplified terms, it essentially casts a flexible net composed of triangles over the globe, producing prisms instead of cubes. The advantage: the triangles can be used to zoom in or out in order to examine specific areas – like individual ocean currents – in much higher resolution. AWI researchers have used this method e.g. to simulate the climatic phenomenon El Niño in higher resolution and over several centuries. In another case, the researchers calculated how currents under the Antarctic ice shelves could change in the future.

Since various physical formulas and parameters have to be processed for every grid box in each step, the amounts of data involved are immense. Consequently, climate modellers require high-performance supercomputers. For their climate simulations, AWI experts use e.g. the supercomputer at the German Climate Computing Center (DKRZ) in Hamburg, which they can access remotely.