Up
Introduction
Strategies
Output
Prospects
Acknowledgments

Prospects and Discussion

The demonstration shows the ability of the MC2 mode to simulate various local meteorological processes at high resolution (up to 1 km), and demonstrates the practical possibility of applying the model into real-time analysis or short-term forecasting of local air pollution meteorology. 

The Run-Time

Since we run the MC2 model using the the multi-node parallel CPU on NEC SX-4 vector supercomputer and select 4 CPUs (maximum 16) in the multi-tasking runs,  the real user time (wall-clock time) is significantly less than the CPU time. The CPU time and real time in this case are:

Resolution
CPU Time (sec)
Real Time (sec)
50 km 500 200
15 km 4,030 1,500
4 km 10,000 3,600
1 km 6,200 2,300
Total
20,730 7,600

The real run-time is about 2 hours 10 mins or 1/3 of CPU time.  It is expected that the real run-time can be reduced to 1-1.5 hour if we were to use 8 or 16 CPUs rather than 4.

About Memory and Output Data Size

The maximum memory used in this case is 1030Mb which was required for the 1 km resolution run. The 4 km, 15 km and 50 km runs took about 350Mb ~ 560Mb memory only. So the memory demand in the case is far from exceeding the memory limit on NEC SX-4 (up to 8Gb).

The output files from the all nesting runs with FST format occupy about 840Mb hard disk space. Moreover, the output data size is changeable since we can control the output variables according to demand.     

Resolution

The highest resolution used in this case is 1 km. The MC2 model can run well at this resolution with a couple of explicit convective and condensation schemes and boundary-layer schemes. However, the high resolution run gives rise to the demand of high resolution geophysical data. Our current climate has a resolution of 25 -100km and geographic data has a resolution of 5 km. The coarse resolution geophysical data makes some flaws such as noises and waves in the 1 km run (click here to see figures in detail).  We expect to obtain the appropriate geophysical data to solve this problem and get more improvement on the highest resolution run in the near future.    

Back to Top

 

For problems or questions regarding this web site contact Xin Qiu [xinqiu@yorku.ca].
Last updated: April 06, 1999.