Updated: November 1, 2004

After One Hundred Years from V. Bjerknes:
The Virtual Atmosphere and Ocean in the Earth Simulator

W. Ohfuchi, Atmosphere and Ocean Simulation Research Group Leader

This work has been carried out under the collaboration between the Earth Simulator Center, Frontier Research Center for Global Change, and universities worldwide.


Figure 1: Global view (top) and magnified view over Japan-East Asia region (bottom) of three-hourly averaged precipitation from a 10-km mesh global simulation of the atmospheric circulation. (Click to enlarge the figures.)

Figure 2: Global view (top) and magnified view in the North Atlantic (bottom) of a snapshot of ocean current speed at 100-m depth from a 0.1-degree quasi-global simulation of the ocean circulation. Blue represents very slow speed, green slow, yellow fast, and red very fast. (Click to enlarge the figures.)

This year, 2004, is the centennial anniversary of weather forecast. In 1904, a Norwegian meteorologist, V. Bjerknes, published “Das Problem von der Wettervorhersage, betrachtet vom Standpunkt der Mechanik und der Physik” (The problem of the weather forecast from the point of view of the mechanics and physics). In this paper with no mathematical equations, he argues that weather forecast is possible if 1) one knows the condition of the atmosphere with sufficient accuracy at a certain time, and 2) one knows the laws that govern the evolution of the atmospheric condition with sufficient accuracy. This statement is still valid, and operational weather forecast centers are trying to improve the accuracy.

The first attempt of weather forecast was carried out by L. F. Richardson during World War I (1914–1918), well before the advent of electronic computers, some important theories of numerical integration, and initialization techniques of atmospheric data. Richardson's attempt failed completely but he published the results in 1922 and stated that weather prediction would be possible one day. His perspective has been called “Richardson's dream.”

The first successful numerical weather forecast was carried out in 1950 by J. G. Charney, R. Fjörtoft and J. von Neumann with an electronic computer called “ENIAC” (Electronic Numerical Integrator and Computer). Then, N. A. Phillips obtained the first numerical realization of the atmospheric general circulation in a computer in 1956. Since then, “numerical experiments” have been integral part of modern meteorology and contributed to better understanding of the weather and climate systems.


Now with the Earth Simulator, what can we learn better about the atmospheric and oceanic circulations? We have been running our general circulation models, AFES (atmosphere), OFES (ocean) and CFES (coupled atmosphere–ocean) with ultra-high resolution. Probably our 10-km mesh global simulations of the atmospheric circulation are first in the history of computational meteorology. Although somewhat higher resolution simulations for several model years have been carried out by other groups, our 0.1-degree mesh quasi-global simulations of the oceanic circulation for about 100 model years are distinguished by their longterm integrations. With these long-term numerical experiments, it is at last possible to study climate or statistical state of near-surface oceanic circulations with higher confidence.

Higher resolution may not necessarily lead to better simulations or better understandings. Looking at figures from the results of our ultra-high resolution simulations, many people have said; “OK. Figures look fancy. But are these realistic?” Actually, we have been asking ourselves the same questions. Verifying our results against observations and analyzing data to understand what is happening in our simulations have been painstaking and lengthy work.

But through this work, we have realized that our ultra-high resolution simulations may have more information than observations. This is much truer for the ocean. Figures of our oceanic simulations reveal fine structures of circulations. It is very difficult to tell whether or not these are true as four-dimensional (longitude, latitude, depth and time) observations of oceanic flows are very difficult, if not impossible, especially for subsurface. We need to note that these figures are from our simulations that have not been fully verified by observations, yet. However, we argue that simulation is probably the only way to obtain detailed four-dimensional structures of oceanic circulations at this moment.

Some figures of our atmospheric simulations may look like satellite images. Note again that these are from our ultra-high resolution simulations. However, from these simulations, we can obtain detailed four-dimensional structures of atmospheric circulations, including variables that are very difficult to observe in the reality, such as vertical velocity and diabatic heating. Even with observations by satellites, we may not be able to get this much information.

We argue that our ultra-high atmospheric and oceanic simulations may be considered as “the virtual atmosphere and ocean.” The virtual atmosphere and ocean in the Earth Simulator may give more information than observations. While verifying our simulations with observations, we are also trying to suggest what kind of observations are necessary to obtain better understandings of the atmospheric and oceanic circulations from our ultra-high resolution simulations. We believe that this feedback process of information between observations and simulations shall lead to a breakthrough in atmospheric and oceanic sciences.


Atmospheric and oceanic simulation has a hundred-year history. With the Earth Simulator, we are now in an era when simulations may give equal or more information than observations. We argue that we can now build the virtual atmosphere and ocean in the Earth Simulator. Information exchange between observations and simulations shall lead to better understandings of the weather and climate systems. The Earth Simulator has been opening a door to next generation atmospheric and oceanic sciences.