[aspect-devel] Problem with iterative advection solver
l.f.j.schuurmans at students.uu.nl
Wed Nov 29 07:28:47 PST 2017
First of all, thanks a lot for your reply! Since I was not subscribed to
the mailing list, initially I did not receive your email, but now I
subscribed to the list. I discussed your proposed solution with Menno, and
tried to run the model with all heating models turned off.
For the screenshots I sent to you in my previous email, I used an
incompressible model. However, since I am using the extended Boussinesq
approximation, both shear heating and adiabatic heating were turned on. In
my previous mail I already showed that turning off the shear heating did
not solve the problem, and unfortunately switching off the adiabatic
heating also did not have any effect on the problem. The strange low
temperature peaks still are present, but like the model without shear
heating, the model run did not crash. What is remarkable, is that the
errors in the temperature seem to occur at or around the same (random)
timesteps, e.g. 18 (for model 1, 2 and 3) and 194 (model 3), 195 (model 2)
and 196 (model 1). To clarify, model 1 corresponds to the model with both
adiabatic and shear heating on, model 2 to the one with only adiabatic
heating, and model 3 to the one with both heating models switched off.
When the value of the composition field is analyzed, it becomes clear that
again values slightly larger than 1 are obtained for model 3 around the
location of the erroneous temperature, which still is the boundary between
the subducting and overriding plate. The value is definitely highest (e.g.
1.09) at the problematic gridpoint, but also at the ones where there seems
to be no temperature problem the composition value can be slightly larger
than 1 (e.g. 1.003).
As a conclusion, we still do not know the exact cause of the problem.
Hopefully you can make something of this additional information. If you
need any further information, please let me know!
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Aspect-devel