A Schwarz domain decomposition method for solution of transient unsaturated water flow on parallel computers
A Schwarz domain decomposition method for solution of transient unsaturated water flow on parallel computers
A parallel Schwarz domain decomposition method for the numerical solution of water flow in soils and aquifers is presented. This method offers the possibility to solve large numerical problems occurring by calculations of water flow in heterogeneous porous media on parallel computer systems. Strongl...
Saved in:
Personal Name(s): | Vereecken, H. (Corresponding Author) |
---|---|
Neuendorf, O. / Lindenmayr, G. / Basermann, A. | |
Contributing Institute: |
Agrosphäre; ICG-4 Agrosphäre; IBG-3 Jülich Supercomputing Center; JSC Zentralinstitut für Angewandte Mathematik; ZAM |
Published in: | Ecological modelling, 93 (1996) 1-3, S. 275 - 289 |
Imprint: |
Amsterdam [u.a.]
Elsevier Science
1996
|
DOI: |
10.1016/0304-3800(95)00224-3 |
Document Type: |
Journal Article |
Research Program: |
ohne Topic |
Publikationsportal JuSER |
A parallel Schwarz domain decomposition method for the numerical solution of water flow in soils and aquifers is presented. This method offers the possibility to solve large numerical problems occurring by calculations of water flow in heterogeneous porous media on parallel computer systems. Strongly nonlinear problems requiring fine discretizations in time and space can be handled. The method may easily be incorporated in existing models with only slight changes in the computer code. Communication routines are implemented in such a way that they can be portated on various parallel computer systems, ranging from workstation clusters to massively parallel systems with distributed memory or vector computers with several processors. The method was tested using 3D homogeneous and heterogeneous flow domains with 9216 nodal points. A 3D finite element method was used to solve the Richards equation in an unsaturated flow domain. For runs on an Intel iPSC/860 and Paragon XP/S10 high speedups and considerably reduced execution times compared to the sequential version were achieved. |