Innovative Methods for Solving Multicomponent Biogeochemical Groundwater Transport on Supercomputers

To evaluate the performance of the Jacobian-Free Newton-Krylov method, PARTRAN was used to simulate enhanced in situ bioremediation of subsurface NAPL (non-aqueous phase liquid) contamination on the ASCI Q supercomputer at Los Alamos National Laboratory, on which a study of parallel efficiency/scaling using up to 512 processors was conducted. Recently, researchers have been investigating emplacement of carbon sources to create anaerobic conditions and enhance in situ reductive dechlorination of halogenated NAPLs. For the test problems presented, it is shown that Jacobian-Free Newton-Krylov saved up to 33% in memory compared to conventional methods where the entire Jacobian matrix is stored. The Jacobian-Split preconditioner performed superior to conventional preconditioners (i.e. faster simulation time and better parallel scalability) for scenarios exhibiting little solute retardation. However, Matrix-Lite preconditioning was required for scenarios with severely retarded reaction fronts. Although this method scaled well, it failed to surpass the performance of conventional preconditioners due to excessive simulation time.

Name thesis:degree_name

Ph.D.

Level thesis:degree_level

Dissertation

Discipline thesis:degree_discipline

Civl and Environmental Engineering

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2015

Author dc:creator

• Hammond, Glenn Edward

Contributors dc:contributor

• Valocchi, Albert J.

• Biogeochemistry

Language dc:language

eng

Identifier

(MiAaPQ)AAI3086072

OAI identifier oai:identifier

oai:www.ideals.illinois.edu:2142/83204