[aspect-devel] Aspect Questions

[Wolfgang Bangerth]

Hi Eric,

> We've started using Aspect for our tutorial and we've had good
> success compiling then running some of the example files.  During the
> course of our work we came up with a list of questions/comments we're
> hoping you can answer.

Let me start by saying that these are all excellent questions. Some we 
have thought about, some we haven't quite yet, some show our lack of 
geophysical knowledge, and some we know about but haven't quite gotten 
around to implementing solutions. I appreciate you bringing all of these up!


> - Other codes like CitcomS are non-dimensional (i.e. they used a
> temperature between 0 and 1 rather than explicit degrees in Kelvin).
> We were wondering why you chose to develop Aspect to have dimensional
> units for all values.  It doesn't seem this is mentioned in the
> manual so it might be good to add an explanation.

Yes, this is a good point and related to another one where we have 
gotten critical comments on the GJI article we are hopefully able to get 
published soon: why we choose to compute the total pressure, not just 
the dynamic (non-adiabatic) part.

Rather than trying to answer things here, I thought it would be useful 
to just discuss this in the manual, and you can find new sections 2.3 
and 2.4 here:
   http://dealii.org/aspect/doc/manual.pdf

These are my opinions on how to write modern codes that are geared 
towards complex models. I know that the geodynamics community usually 
does this differently. I would actually very much like to hear what you 
think about the reasons outlined in these sections -- including, if you 
think so, to hear that this is stupid because I haven't though about 
important reasons A, B and C. Please don't spare with criticism.

I will admit that my personal opinion is that we should get away from 
non-dimensional equations. I think they're a wonderful tool for 
equations with constant coefficients, but this is not our goal with 
Aspect, and I think it is time to move into real-world models with 
real-world units :-)


> - In the parameters, it is not clear what the "locally conservative
> discretization" and "stabilization parameters" refer to.  Is there a
> reference or explanation we can add to the manual for these?

Good point too. This is something that's discussed in detail in the 
paper Martin, Timo and I have been writing, but since it's not out yet 
I've given a brief summary in the manual. See section 5.6.


I'll get back to your other points in the next few days!
Best
  W.
------------------------------------------------------------------------
Wolfgang Bangerth               email:            bangerth at math.tamu.edu
                                 www: http://www.math.tamu.edu/~bangerth/