[CIG-SEISMO] Specfem2D : Small Oscillations in Kernel Data

[Carl Tape]

Dear Philip,

I added two kernel examples (with figures) to SPECFEM2D. Please run these
examples and plot the kernels to make sure things look okay. (If you checked
out the tar-ball version, there will be an updated posting soon; otherwise,
"svn update".) There is also a plotting script that uses GMT. If the
examples look okay, then first examine the synthetic seismograms and
corresponding adjoint sources. For your example, there probably should not
be any spurious oscillations -- only the actual seismic phases. If the
synthetics look "clean," then so should the adjoint source, assuming you've
picked a decent window (which it looks like you have). If you still
encounter the effects you sent, let us know.

When testing the plotting, start out by using no interpolation whatsoever --
just the plot the values at the (global) gridpoints exactly using a command
like "scatter" in matlab (or psxy in GMT). That way you can know it's not
related to plotting issues.

Carl

---------------------------------------------------------------
Carl Tape
Assistant Professor
Geophysical Institute
University of Alaska Fairbanks
Phone: 907-474-5456
Email: carltape at gi.alaska.edu
Web: http://www.giseis.alaska.edu/input/carl/
---------------------------------------------------------------

On Mon, Apr 11, 2011 at 5:42 AM, Philip Knaute <philip.knaute at gmail.com>wrote:

> Hello again,
> and thank you for your quick reply. I suspected the crude way I plot the
> Kernels might be the issue. At the moment, I still use the
> "TriScatteredInterp()" Matlab command (this uses a Delauney triangulation to
> interpolate the scattered Data).
> So I had a closer look on the data by fixing the my x coordinate and taking
> a 1D line of the Kernel and plotted this data. To do that, I Imported the
> Kernel File as Matrix into Matlab and used the "sortrows" command, which
> should not mess with the matrix, apart from sorting it.
> Then I just extracted (Number of spectral elements in my model * 5) rows
> from the Kernel at a certain position and plotted the appropriate collumn
> with "plot". That is when I saw the strange zick-zack pattern.
>
> I attached a zip file with pictures of my beta Kernel as a whole (beta.jpg)
> and of a vertical line of the kernel data (beta_fixedX.jpg and
> beat_fixedX_close.jpg). A picture of my rho (from rho kappa mu) kernel
> (rho.jpg) and a closeup (rho-closeup.jpg) to show the checkerboard pattern.
> I also attached my .dat, Par_file and SOURCE file.
>
> Thank you very much,
> Philip
>
>
> 2011/4/9 Dimitri Komatitsch <dimitri.komatitsch at get.obs-mip.fr>
>
>>
>> Dear Philip,
>>
>> Hopefully Carl Tape, or somebody who is more familiar with the 2D kernel /
>> adjoint code than I am, can answer that. Could you maybe email us a few
>> screen captures showing pictures of the problem? I think it would help.
>>
>> Thank you,
>> Best regards,
>> Dimitri.
>>
>>
>> On 04/08/2011 07:47 PM, Philip Knaute wrote:
>>
>>> Dear All
>>>
>>> After getting Specfem2D to run with Scotch (thank you again for your
>>> help, Dimitri) I started to work with the sensitivity-kernel function of
>>> Specfem2D.
>>>
>>> But, when I look at my results (e.g. the Kernels) their values seem to
>>> oscillate (very small deviations) around an average value.
>>>
>>> So, if I have a 2D plot of my sensitivity Kernel, I would take a 1D
>>> slice for a fixed z value.
>>> Looking at that 1D funtcion only depending on the GLL Points in X
>>> Direction it would look like a Triangle wave function ( looking
>>> something like that /\/\/\/\/\ ).
>>> It would oscillate very regularly around an average value, especially at
>>> locations where this average value doesn't change.
>>> The oscillations seem to be too regular to be noise or an instability of
>>> any kind, so I guess it is an intentional feature.
>>> Is it an intentional feature of the code and do I have to post process
>>> the Kernel-data somehow?
>>>
>>> Thank you very much for your help.
>>> Best regards,
>>> Philip
>>>
>>>
>>>
>>> _______________________________________________
>>> CIG-SEISMO mailing list
>>> CIG-SEISMO at geodynamics.org
>>> http://geodynamics.org/cgi-bin/mailman/listinfo/cig-seismo
>>>
>>
>> --
>> Dimitri Komatitsch - dimitri.komatitsch aT get.obs-mip.fr
>> Professor, University of Toulouse and Institut universitaire de France,
>> CNRS UMR 5563 GET, France   http://www.univ-pau.fr/~dkomati1
>>
>
>
> _______________________________________________
> CIG-SEISMO mailing list
> CIG-SEISMO at geodynamics.org
> http://geodynamics.org/cgi-bin/mailman/listinfo/cig-seismo
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://geodynamics.org/pipermail/cig-seismo/attachments/20110411/ae7f8e41/attachment.htm