September 30, 2011 - by Simone Ulmer
Mr. Schär, as Chairman of the C2SM Competence Centre (Center for Climate Systems Modeling) you were significantly involved in creating the report “Swiss Climate Change Scenarios” (CH2011). Did this cooperation make the CH2011 report possible?
C2SM certainly played a decisive role and brought together researchers from ETH Zurich and MeteoSchweiz. However, one should also mention other institutions including the National Centre of Competence in Climate Research (NCCR Climate). C2SM’s key task is to develop and use climate models. Climate models are becoming ever more complex. Developing them further, maintaining them and using them in an optimum way on supercomputers is beyond the resources of a single research group. That’s why there is a need for cooperations like C2SM that bring together the entire know-how from all the relevant areas.
What was the aim of the study?
Our aim was to analyse the latest climate simulations and present the results in a suitable form. In this respect, we had two target groups in mind. Firstly, the wide-ranging scientific community that deals with the consequences of climate change and is very interested in scenario data. For example, this involves effects on agriculture, hydrology, hydroelectric power, permafrost or glaciers. Secondly, administrations, politicians and economies can use the report as a basis for planning. Many governmental agencies in the Swiss as well as major industrial corporations are increasingly interested in the consequences of climate change in their area.
As a result of CH2011, do you know more about the future of Switzerland’s climate than you did when a previous report was published four years ago?
In particular we were able to estimate and quantify the uncertainties more accurately by using new climate simulations and new statistical methods. Better models with improved modelling components and finer computational resolution were available to us for this report compared to the OcCC Report published in 2007. We also used more than twenty new climate simulations for the scenarios. The input into the 2007 report consisted of models with a resolution of 50 to 100 kilometres, whereas in CH2011 we took into account data from models with a grid spacing of only 25 kilometres. In the current report, we also distinguish between different emission scenarios, including an intervention scenario which specifies a halving of global greenhouse gas emissions. This scenario is new and was not considered in the 2007 IPCC report. We can use it to visualise the effects of a global climate policy on Switzerland’s climate. We were unable to make any specific statements about that until now.
The Report says it will become warmer and dryer and we must expect more extreme events and heat waves. Exactly which scenarios do the models depict?
The warming has already become distinctly visible in the observations and will manifest itself with increasing clarity. In the report, we assessed the resulting changes in relation to extreme events mainly based on the available literature. The clearest signal emerges with regard to heat waves. Within Switzerland, the effects of heat waves and droughts are greatest in Ticino, but not as strong as expected for parts of the Mediterranean region and the Po Valley. The picture there will change: the summery “dolce vita” will probably need to move out of the street cafés and into air-conditioned rooms.
The scenarios also show that there will be significant changes in precipitation as well, especially an increase in rain at the expense of snowfall. That will affect the runoff of many rivers and streams. However, this change in precipitation will not take place continuously. Natural variability is a decisive factor until approximately the middle of the century. That is, natural fluctuations may hide or intensify man-made signals. In the second half of the century, however, the climate change will become clearly noticeable in even more climate variables. Even if by 2050 we succeed in halving global greenhouse gas emissions compared to 2000, the Swiss climate will by then become warmer by 1.2 to 1.8 degrees Celsius compared to the period between 1980 and 2009.
Are there serious differences between the OcCC Report in 2007 and the new CH2011 Report?
Uncertainties are quantified for each of the two scenarios and within the framework of these uncertainties the old and new scenarios are consistent. This shows that the earlier research work was already on the right track. The current research results contain three major advances. Firstly, it was possible to estimate the uncertainties more objectively. Secondly, we can now submit detailed scenario data, for example artificial climate series for the present and the future which are available for the study of climate impacts. Thirdly, we have now taken different greenhouse gas scenarios into account. In addition to a scenario that assumes regionally-oriented industrial development and a scenario based on the balanced use of fossil and non-fossil fuels, we took what is known as an intervention scenario into account.
In the new report, you included models with twice the resolution. Why are models with increasingly improved resolution so important?
In summer, for example, a large part of the precipitation is of a thundery nature. A thunderstorm has a horizontal scale of a few kilometres and thus falls through the computational mesh of the current models, which use a grid spacing of 25 km. As the next step, in a globally coordinated project, we will carry out calculations using 10 km and the associated simulations will start soon. We are also already working on the next-but-one step and then we want to model the climate at a resolution of two kilometres. Only then can we take the physical processes in thunderclouds explicitly into account, thus eliminating key uncertainties in our models.
What is behind the difficulties to run higher-resolution models, is this due to the limited computing capacities?
Partly, but not only that. Increasing the resolution alone is not enough. Refining the resolution from 20 to 2 kilometres, to a certain extent, fundamentally also requires new models.
Higher resolution means more computing time. How has the demand for computing time developed in the past four years?
Doubling the resolution results in an eight-fold increase in computing time. That’s why high-resolution simulations are very expensive and laborious. However, an almost even greater problem than the increasing computing time is the amount of data generated during the process. Our own climate simulation took 20 weeks on the Swiss National Supercomputing Centre (CSCS) computer, at that time the fastest in Switzerland, and produced 9.5 terabytes of data. We need to keep this data available for at least five years because it is used and needed by various different groups. Data storage has become an enormous cost driver and is a big burden on these projects’ budgets.
C2SM has a project at the Swiss Platform for High Performance and High Productivity Computing (HP2C). The aim is to optimise the existing computer codes for future computer architectures. That means, among other things, adapting the codes and algorithms. Did the CH2011 models already include the initial results?
No, because we had already started four years ago with the model runs for the scenarios that are now available. However, with the new supercomputers, for example with graphical processor units (GPU), we are confronted with fundamental changes. We need to make our models run on these new architectures. That’s why we are extremely pleased about the collaboration with the CSCS in the HP2C project. Thanks to this collaboration, we are among the first climate groups internationally to adapt their codes and algorithms to these new architectures. More specifically, we want to rewrite the core of our regional weather and climate code accordingly and adapt them to the new GPUs.
Are climate scientists able to achieve all of that?
Two C2SM staff from the supercomputing area are working in the cooperative project. A single scientist cannot fully cover the whole area from computing and programming right through to the science. That’s why we must work in teams in which all the competences are covered. On the present scale, that is a new development for us.
Does that mean that close collaboration between specialists in high performance computing, hardware and software developers and climate scientists will become increasingly important?
Undoubtedly. The collaboration with the CSCS is a stroke of luck for us and it has all developed very well. However, it will take time before the new models are reflected in better climate scenarios.
In a project lasting several years, the Center for Climate Systems Modeling (C2SM), MeteoSwiss, ETH Zurich, NCCR Climate (the National Centre of Competence in Research – Climate) and OcCC (the Advisory Body on Climate Change) have worked out scenarios for the future development of temperature and precipitation in Switzerland. The latest scientific knowledge resulting from this was presented yesterday at a major event in the Auditorium Maximum of ETH Zurich. The climate scenarios are based on refined climate simulations and new statistical methods. Detailed scenario data in digital form is also available for the first time. This data should provide impetus for research into the consequences of climate change, and should enable political and industrial decision-makers to have access to comprehensive information about the development of the climate in Switzerland in the 21st century.