[CIG-LONG] April 2 CIG-LTT Conference Call - summary
dharry at warnercnr.colostate.edu
Sun Apr 6 10:22:51 PDT 2008
Please see below for my draft summary of our conference call. Please
add/edit as necessary. Also, please consider prioritizing the future
work on GALE that we discussed. We agreed at the end of the conference
call to have the tasks prioritized prior to the CIG workshop that will
be held in Davis, CA in July.
By the way, I have added some things not discussed in the conference
call but that were delivered via email from those who were unable to
join in the call.
Minutes from April 2 2008 Meeting of the CIG-LTT Working Group
The meeting was convened as a conference call. Those attending were:
Jolante van Wijke
The discussion centered on improvements to GALE and consideration of
accuracy and efficiency of GALE. The working group makes the following
recommendations (not in order or priority):
1) The reasons for GALE's lack of convergence when running the extension
benchmark need to be understood.
2) Further work needs to be done to determine how GALE scales as compute
nodes are added.
3) An interface needs to be built to allow GALE to import pre-meshed
4) The ability to consider pressure and temperature dependence of
physical properties needs to be added. Density and thermal properties
were specifically mentioned. Others?
5) Some benchmark testing needs to be done to assess if GALE is properly
behaving according to the plastic material behavior that is selected.
Can we identify some benchmark tests?
6) Implementation of boundary conditions that can vary along the sides
of the model domain is desired.
7) Elasticity should be incorporated. Is this the solution to getting
the extension benchmark to work?
8) The Working Group discussed the desire to have different element
constructs (e.g., triangular elements). Walter indicated that this
would require a fundamental re-write since GALE inherits it's element
constructs from Underworld. Is that correct, Walter? And Noah, does this
capture your comments re: irregular mesh, or was there more to that?
9) Understanding the cause of and eliminating the pressure oscillations
that appear in some of the cookbook examples is required.
10) Convergence behavior of the Uzawa solver seems odd compared to other
codes, and needs to be understood. Fixing this may fix the pressure
oscillation problem. Or is it the other way around? We should think
about this when we prioritize tasks. Also, there was some further
discussion of solver issues that went a bit over my head. Could someone
try to summarize that discussion if I have not fully captured it here?
Via email, Todd Ehlers also added the following:
11) Coupling with a landform (surface process) evolution model is
desirable. Two mature codes in the geomorphology/geodynamics
communities that are well suited for this task are CASCADE and CHILD.
12) GALE output should include prediction of geobarometer and
thermochronometer observables at the surface. There are mature and free
codes available that use time-temperature histories of rocks and predict
observables (e.g. Cooling ages) that could be used to test models.
Todd, I think GALE already provides the P-T history as output. Do we
just need a post-processing filter to convert this into the proper input
format for the other codes you refer to? Also, could the working group
try to identify specific codes that we need to make GALE get cozy with?
Walter's email announcing the GALE 1.2.2 release identified these future
* Cleanup friction boundary conditions
* Better normal stress boundary conditions
* Deformed lower boundaries
* Interface with custom surface process models in parallel
I seem to recall that Patrice had also contributed comments via email,
but I can't find them. Patrice, please insert them if they are not
included in the issues I summarized above.
Dennis L. Harry
Edward M. Warner Professor of Geophysics
Department of Geosciences
Colorado State University
Fort Collins, CO 80523-1482
phone: (970) 491-2714 fax: (970) 491-6307
More information about the CIG-LONG