2005 Community Finite Element Models for Fault Systems and Tectonic Modeling

Workshop Discussion on CIG Priorities

Discussion on CIG priorities for short-term geodynamics led by Brad Aagaard.

• Brad Aagaard has a number of slides that will be posted as a PDF on the website.
• Workflow diagram. Which portion of the workflow is highest priority?
• Summary of short-term priorities. The order of these may be revised. Items in bold are applicable to multiple CIG working groups.
  • PyLith development.
  • CIG build procedure.
  • Mesh preprocessing.
  • Benchmark archiving.
    • Mesh interrogator.
• Summary of long-term priorities.
  • Multiple earthquake cycles.
  • Code coupling between working groups.
  • Adaptive mesh refinement.
  • Sensitivity modeling (adjoint).
  • Data assimilation.
  • Large deformation.
  • Improve solver efficiency (multigrid).

PyLith

• LithoMop and EqSim (8 months).
  • Stable code, Pyrized, parallel.
  • Use sieve for mesh topology.
  • Implement faults with cohesive elements.
  • Linear and parabolic tet and hex elements.
  • Most widely used constitutive models.
• PyLith (4 months).
  • Integration of LithoMop and EqSim.
  • Pyrized, parallel, modular.
  • Use FIAT for basis functions and quadrature.
• Requests
  • Documentation.
  • Regression testing.
  • Simulation Controller.
  • Ease of use.
    • Setting up boundary conditions, etc.
    • More automatic setting of parameters (convergence criteria, etc.).

Okada Code

Pyrized version of Okada dislocation code (2 weeks?).

• Serve as example of Pyrizing code.
• Become module used in other codes?

Mesh Preprocessing (mesh topology)

Common denominator between mesh generators and physics codes.

• Holds mesh topology.
  • Nodes.
  • Elements.
  • Lists/groups of nodes and elements for BC, etc.
• Tools.
  • Create "faults" (code dependent).

Benchmark archiving

Tools to facilitate benchmarking.

• Repository archiving.
  • Simulation metadata.
    • Software name.
    • Software version.
    • Benchmark name.
    • ...
  • Simulation output data files.
  • Discretization information.
    • Basis functions.
    • Element shape.

Spatialdata: Boundary conditions and material properties

Tool for specifying field over some spatial domain.

• Need standard interface for plug-n-play.
• Common functionality with mesh interrogator.
  • Search for point in space.
  • Interpolation of field.
• Simple spatialdata package for simple fields.
  • Current status.
    • Efficient for data with only a few control points (< 100).
    • Doesn't support units.
    • No documentation.
  • TO DO
    • Needs documentation.
    • Support for units (few hours' work).
    • More efficient search algorithm (get from interrogator).

Revised year 1 priorities

Revised priorities after workshop.

1. Workflow infrastructure. These are general CIG priorities.
   • Archiving facility. Change name to Benchmark Infrastructure.
   • Mesh preprocessing. We probably have some short-term fixes, but need a long-term strategy. This is considered to include a mesh partitioning and refinement package being worked on by Matt Knepley.
2. Build procedure. This is a general CIG priority.
3. PyLith development. We want to make sure that PyLith has nearly all of the proposed functionality.

Revised year 2 priorities

Revised priorities after workshops. Some of these have been moved from previous short-term priorities. These are not ranked.

• Spatialdata interface for material properties and boundary conditions.
• Okada/layered-space/layered viscoelastic analytical solution package.
• Adaptive mesh refinement.
• Improve solvers (multigrid).

Working Group Action Items

• Organization of working group.
  • Leader(s).
  • Use of CIG website.
    • cig-short listserv.
    • Work space.
• Proposals to CIG SSC. These proposals should be submitted as coming from the entire cig-short list.
  • Benchmark archiving. Charles Williams has written a draft of this proposal and posted it on the website. Other workshop attendees/cig-short members should look over this and provide comments.
  • Mesh preprocessing. Brad Aagaard, Charles Williams, and Oliver Boyd will work on a draft of this proposal.
  • Mesh postprocessing?
  • Spatialdata?