Supercomputer and the climate model

[an error occurred while processing this directive]
[an error occurred while processing this directive]

Supercomputer and the climate model

The climate system and the climate model which simulates it needs complicated and enormous amount of calculations. In the Center for Climate System Research, University of Tokyo, numerical calculations are done using the supercomputer in the Computer Centre, University of Tokyo.

When we calculate the following three equations using the computer,

¡¡C1 = A1 + B1  
¡¡C2 = A2 + B2  
¡¡C3 = A3 + B3
it is faster dealing with them at one time such as 
   A1     A2     A3 
   +      +      +  
   B1     B2     B3
(Of course this needs more memories.) than dealing with them one at a time as listed first. The former is called the scalar calculation and the latter is called the vector calculation. Supercomputer is a calculator which can do the vector calculation with high speed and large memory.


The above graph indicates that the computer has increased its power by leaps and bounds in these 30 years. Its power is more than 10 times bigger than that in 10 years ago.
The next graph depicts the improvement in accuracy of the daily weather forecast by the atmospheric model. Thanks to the progress in meteorology and to the improvement in computing technology, the weather forecast for the next one week becomes fairly reliable.

However, for the climate model which have to calculate for more than 10 to 100 model years, the present computer power is not enough. In the climate model, the atmosphere and the ocean is divided into grids and the change of physical quantities such as wind and temperature are calculated at each grid. As we are simulating the complicated natural phenomena, the finer resolution is the better.

However, as the atmosphere and the ocean extends into three dimension: zonal, meridional, and vertical, if the distance between neighboring grid points is simply reduced to half in every direction, the number of grid points becomes 2 × 2 × 2 = 8 times as many. If the fine resolution is adopted, time step must be set short correspondingly (Upper limit for one time step is a few minutes for the atmospheric model and a few hours for the ocean model.). Thus, for the realization of the climate model with the finer resolution and the higher accuracy, the supercomputer with the more memory and high speed is needed.
The difference between the fine and the coarse resolution is obvious in the illustration below, isn't it?

[an error occurred while processing this directive] [an error occurred while processing this directive]