Special Topic : Earth Simulator

In this March, the super gigantic computer named the Earth Simulator will be build in the same campus of the Yokohama Institute forEarth Sciences as FRSGC. In this edition, we would like to introduce the FRSGC's activities to be co-operated with the Earth Simulator.

Dr. Matsuno will explain the background and expected outputs of the Earth Simulator.

Toward more Accurate Global Warming Prediction Using the Earth Simulator

Dr. Taroh Matsuno (Director-General, FRSGC)

The Kyoto Protocol for reductionof greenhouse gas emissions tomitigate global warming is nowabout to be ratified by Japan and willenter into force soon. Thus, society(governments) begins to takeactions against global warming before any damages become reality,relying upon what scientists tellabout the future state of the earth.

The scientific grounds on which thedecisions rest are computer simulations of the future climate by use ofmodels. Climate models are, however,not sufficiently accurate at present.Origins of insufficiencies are that current treatments of individual physical processes such as reflection of the solar insolation by clouds or generation and development of convective clouds are not accurate enough, partly by the shortage of our knowledge and also by insufficiency of resolution of climate models. Climate models 10 years ago were of 500km resolution, by which large-scale clouds such as those covering extratropical cyclones were represented but narrow frontal clouds were not.

The resolutions of climate models in recent days are 100km-200km which may be sufficient to resolve frontal clouds, but still insufficient to represent convective cloud systems, having a scale of several tens of km which cause heavy rainfalls. They are also crucial to transport heat and water upward to affect the whole global circulations.

So far, by use of 100km or larger meshes, occurrence and strength of these convective systems have been determined from the vertical stability, calculated from 100km-scale temperature and moisture distributions on the basis of empirical rules. The method is called "parameterization" and it has been a greatest source of inaccuracy and ambiguity in climate modeling for a long time.

The performance of the coming Earth Simulator is expected to be several hundreds times greater than current supercomputers, which may enable us to use mesh sizes of about 10km. So the research group at the Meteorological Research Institute has begun development of a 20km-resolution global atmosphere model. At this resolution, however, the parameterization of convective systems should still be retained. At the FRSGC, we decided to jump to development of a 5-km mesh global atmosphere model, which can marginally represent meso-scale convective systems directly. Even with the Earth Simulator, running climate models with such ultra-high resolution for a long period (a few years) is not possible. However, considering ever increasing computer power, we have judged that it is now the time to switch to new kind of models or next-generation models. The difference of current and new models can be understood from the figure.

At the same time, we are now developing world ocean circulation models toward the final target of 10km-mesh. With this mesh size, meso-scale ocean eddies can be directly represented, and their effects of transporting heat and salinity which play crucial roles can be automatically reproduced correctly without uncertainly arising from parame-terization. By use of such models, as the ocean component of a coupled model with a medium resolution atmosphere model and conducting global warming experiments, we may be able to give a robust answer to the question under debate, whether the sinking in the north North Atlantic ocean will weaken or not under global warming conditions.

Back Cover Next

FRSGC Website