Advertisement

Parallel Implementations of the Euros Model: The Algorithm and Some Preliminary Results

  • K. Georgiev
  • C. Mensink
Conference paper
Part of the NATO Science Series book series (NAIV, volume 30)

Abstract

A short description of the EUROS model as well as a parallel algorithm for its implementation on high performance computer platforms is presented. Different parallelization strategies (chemical decomposition, horizontal decomposition and vertical decomposition) are described and analysed. Taking into account the advantages and disadvantages of each of the three approaches some conclusions about the optimal strategy for an implementation of the EUROS model on the existing at VITO parallel computer architecture are done. The first parallel version of EUROS is based on the domain decomposition approach. Different decompositions of the model domain are used for the different submodels obtained after a splitting procedure: (i) with an overlap for the horizontal advection and diffusion and (ii) nonoveralpping for the submodels describing the other physical and chemical processes. The grid refinement procedure used in EUROS leads to some difficulties in the load balancing. They are discussed in the paper and an algorithm to avoid them is presented.

Keywords

Model Domain Atmospheric Boundary Layer Domain Decomposition Parallelization Strategy Local Refinement 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. 1.
    Delobbe, L., Kinnaer, L. and Mensink, C. (2000) Optimization of chemical and advection modules in EUROS, Scientific Report on OSTC Project AS/DD/10, VITO, Belgium.Google Scholar
  2. Dimov, I., Georgiev, K., Ostromsky, Tz., van der Pas, R. and Zlatev, Z. (2002) Computational challenges in the numerical treatment of large air pollution models, SIAM Review, (to appear).Google Scholar
  3. 3.
    Georgiev, K. (2002) Optimal strategies for parallelization of the EUROS model: design and implementation, Scientific Report 2002/TAP/R/011, VITO, Belgium.Google Scholar
  4. 4.
    Gropp, W., Lusk, E. and Skjellum, A. (1994) Using MPI: Portable programming with the message passing interface, MIT Press, Cambridge.Google Scholar
  5. 5.
    Endlich, R. M. (1967) An iterative method for altering the kinematic properties of wind fields, J. of Appl. Meteorology 6, 837–844.CrossRefGoogle Scholar
  6. 6.
    Van Loon, M. (1996) Numerical methods in smog prediction, PhD Thesis, University of Amsterdam, The Netherlands.Google Scholar
  7. 7.
    McRae, G. J., Goodin, W. R. and Seinfeld, J. H. (1982) Numerical solution of the atmospheric diffusion equation for chemistry reacting flows, J. Comp. Phys. 45, 1–42.CrossRefGoogle Scholar
  8. 8.
    Owsczarz, W. and Zlatev, Z. (2002) Running a large air pollution model on an IBM SMP computer, Parallel Computing 28 (2), 355–368.CrossRefGoogle Scholar
  9. 9.
    Strang, G. (1968) On the construction and comparison of difference schemes, SIAM J. Numer. Anal. 5, 506–517.CrossRefGoogle Scholar
  10. 10.
    Verver, J. G. and Simpson, D. (1995) Explicit methods for stiff ODE’s from atmospheric chemistry, Appl.Numer. Math. 18, 413–430.CrossRefGoogle Scholar

Copyright information

© Springer Science+Business Media Dordrecht 2003

Authors and Affiliations

  • K. Georgiev
    • 1
  • C. Mensink
    • 1
  1. 1.VITO — Centre for Remote Sensing and Atmospheric ProcessesBelgium

Personalised recommendations