Parallel computation of sensitivity analysis data for the Danish Eulerian model

Abstract

Sensitivity Analysis of the Danish Eulerian Model requires an extensive amount of output data from computationally expensive numerical experiments with a specially adapted for the purpose version of the model, called SA-DEM. It has been successfully implemented and run on the most powerful parallel supercomputer in Bulgaria - IBM BlueGene/P. A new enhanced version, capable of using efficiently the full capacity of the mashine, has recently been developed. It will be described in this paper together with some performance analysis and numerical results. The output results are used to construct some mesh-functions of ozone concentrations ratios to be used further in sensitivity analysis of the model by using Monte Carlo algorithms. © 2012 Springer-Verlag.