[CIG-SEISMO] get_attenuation_scale_factor error / PS

Mon Oct 17 09:26:59 PDT 2016

Hi all,

Yes, Brian is right, easy to cut and paste, the routine in 2D and 3D is 
basically the same. Just replace the second call below with the first.

Best regards,
Dimitri.

On 17/10/2016 17:02, Brian Savage wrote:
> Dimitri and Robin,
>
> I assume SolvOpt() would be a replacement to attenuation_invert_by_simplex() routine in src/shared/get_attenuaiton_model.f90 ?
>
> In SF2D:
> SUBROUTINE compute_attenuation_coeffs(N,Qref,f0,f_min,f_max,tau_epsilon,tau_sigma)
>
> In SF3D:
> subroutine attenuation_invert_by_simplex(t2, t1, n, Q_real, tau_s, tau_eps)
>
> The arguments to the routines look similar.
>
> The complicated part, I assume, was getting the actual SolvOpt routine out of SF2D and into SF3D.  Please correct me if I am wrong.
>
> Brian
>
> On Oct 16, 2016, at 6:47 PM, Robin Lee wrote:
>
>> Hi Dimitri,
>>
>> I've attempted to implement the SolvOpt() routine in the specfem3d code but unfortunately was not successful. I followed up on this by talking with some programmers in our research group but they said that it looked quite difficult and seemed like it would require a lot of work to change from 2D to 3D. At the end of the day it seems I am unable to help out with this, my sincerest apologies.
>>
>> Regards
>> Robin
>> ________________________________________
>> From: Dimitri Komatitsch [komatitsch at lma.cnrs-mrs.fr]
>> Sent: Wednesday, September 07, 2016 9:08 AM
>> To: cig-seismo at geodynamics.org; Robin Lee
>> Subject: Re: [CIG-SEISMO] get_attenuation_scale_factor error / PS
>>
>> Hi again Robin,
>>
>> PS: if you implement it please let us know and email us the modified
>> version and we will make it the official version on our Git server (see
>> https://github.com/geodynamics/specfem3d/issues/742 ).
>>
>> Thanks,
>> Best regards,
>> Dimitri.
>>
>> On 05/09/2016 11:55, Dimitri Komatitsch wrote:
>>>
>>> Hi Robin,
>>>
>>> Thanks!
>>> The routine is in src/specfem2D/attenuation_model.f90 .
>>>
>>> Best regards,
>>> Dimitri.
>>>
>>> On 05/09/2016 01:31, Robin Lee wrote:
>>>> Hi Dmitri,
>>>>
>>>> Thank you, I have read the paper and this may very well be the issue I
>>>> am facing. I have looked through specfem2D a bit but can't seem to
>>>> find where the SolverOpt() is implemented. Would you be able to point
>>>> me to where it is?
>>>>
>>>> Regards
>>>> Robin
>>>> ________________________________________
>>>> From: Dimitri Komatitsch [komatitsch at lma.cnrs-mrs.fr]
>>>> Sent: Sunday, August 28, 2016 9:01 AM
>>>> To: Robin Lee; cig-seismo at geodynamics.org
>>>> Subject: Re: [SPAM: 7.000] Re: [CIG-SEISMO]
>>>> get_attenuation_scale_factor error
>>>>
>>>> Hi Robin,
>>>>
>>>> Q = 25 or above is classical and works fine, no problem.
>>>> However there is likely an error in the routine you use to build your
>>>> velocity model, since it gives you mesh points at which Q = 4, and in
>>>> that case yes SPECFEM3D can become unstable (easy to fix, see
>>>> http://komatitsch.free.fr/preprints/GJI_Lombard_2016.pdf , and already
>>>> fixed in the 2D code; we never took the time to fix the 3D code as well
>>>> because such very low Q values never occur in practice).
>>>>
>>>> It is likely your velocity model routine that gives fictitious very low
>>>> Vs or Vp values at at least one location.
>>>>
>>>> Best regards,
>>>> Dimitri.
>>>>
>>>> On 27/08/2016 18:10, Robin Lee wrote:
>>>>> Hi Dimitri,
>>>>>
>>>>> Thank you very much for your answer, and apologies for the (very)
>>>>> late reply, I have been busy with a few other things but am back to
>>>>> addressing this issue with attenuation now.
>>>>>
>>>>> I'm afraid I'm a bit unsure about what this Q value specfem refers to
>>>>> represents. For our velocity model, the lowest Vs is 0.5km/s and we
>>>>> are using the correlations from Graves (2010), Qs = 50Vs and Qp =
>>>>> 2Qs, which for this case is Qs = 25 and Qp = 50. Is this attenuation
>>>>> too high for specfem3D? Or am I misinterpreting something?
>>>>>
>>>>> Regards
>>>>> Robin
>>>>>
>>>>> ________________________________________
>>>>> From: Dimitri Komatitsch [komatitsch at lma.cnrs-mrs.fr]
>>>>> Sent: Friday, July 29, 2016 7:48 PM
>>>>> To: cig-seismo at geodynamics.org; Robin Lee
>>>>> Subject: [SPAM: 7.000] Re: [CIG-SEISMO] get_attenuation_scale_factor
>>>>> error
>>>>>
>>>>> Hi Robin,
>>>>>
>>>>> Q = 4 means attenuation is extremely large. You maybe then need to
>>>>> switch to the "SolvOpt()" routine that we put in the 2D code. See
>>>>> http://komatitsch.free.fr/preprints/GJI_Lombard_2016.pdf and
>>>>> https://github.com/geodynamics/specfem3d/issues/742
>>>>>
>>>>> You should also check your reference frequency
>>>>> (ATTENUATION_f0_REFERENCE) because the correction factor you get (0.37)
>>>>> is extremely big (the reference value being 1.).
>>>>>
>>>>> Are you sure the viscoelastic wave equation remains a valid model for
>>>>> such a medium?
>>>>>
>>>>> Best regards,
>>>>> Dimitri.
>>>>>
>>>>> On 29/07/2016 06:31, Robin Lee wrote:
>>>>>> Hi all,
>>>>>>
>>>>>> I've been attempting to implement attenuation in a tomography model
>>>>>> which I have been running. It seems to run fine without attenuation,
>>>>>> but
>>>>>> once I add in Qp and Qs, and set the attenuation to .true. in the
>>>>>> Par_file, I get an error in the gen_databases step which states the
>>>>>> following:
>>>>>>
>>>>>> error : in get_attenuation_scale_factor()
>>>>>>   scale factor:   0.37110551337678144       should be between 0.7
>>>>>> and 1.3
>>>>>>   Q value =    4.3168292045593262       central frequency =
>>>>>> 5.3879655900887585E-002
>>>>>>   please check your reference frequency ATTENUATION_f0_REFERENCE in
>>>>>> constants.h
>>>>>> unreliable correction factor in attenuation model
>>>>>> Error detected, aborting MPI... proc            0
>>>>>> --------------------------------------------------------------------------
>>>>>>
>>>>>> MPI_ABORT was invoked on rank 0 in communicator MPI_COMM_WORLD
>>>>>> with errorcode 30.
>>>>>>
>>>>>> NOTE: invoking MPI_ABORT causes Open MPI to kill all MPI processes.
>>>>>> You may or may not see output from other processes, depending on
>>>>>> exactly when Open MPI kills them.
>>>>>> --------------------------------------------------------------------------
>>>>>>
>>>>>>
>>>>>> I am currently using a simple correlation of Qs=50Vs and Qp=2Qs
>>>>>> where Vs
>>>>>> is in km/s. I have looked in the constants.h file but cannot actually
>>>>>> find the ATTENUATION_f0_REFERENCE parameter. Has anyone experienced
>>>>>> this
>>>>>> error before, know what is going wrong or know how to fix it?
>>>>>>
>>>>>> Regards
>>>>>> Robin
>>>>>>
>>>>>> This email may be confidential and subject to legal privilege, it may
>>>>>> not reflect the views of the University of Canterbury, and it is not
>>>>>> guaranteed to be virus free. If you are not an intended recipient,
>>>>>> please notify the sender immediately and erase all copies of the
>>>>>> message
>>>>>> and any attachments.
>>>>>>
>>>>>> Please refer to http://www.canterbury.ac.nz/emaildisclaimer for more
>>>>>> information.
>>>>>>
>>>>>>
>>>>>>
>>>>>> _______________________________________________
>>>>>> CIG-SEISMO mailing list
>>>>>> CIG-SEISMO at geodynamics.org
>>>>>> http://lists.geodynamics.org/cgi-bin/mailman/listinfo/cig-seismo
>>>>>>
>>>>>
>>>>> --
>>>>> Dimitri Komatitsch
>>>>> CNRS Research Director (DR CNRS), Laboratory of Mechanics and Acoustics,
>>>>> UPR 7051, Marseille, France    http://komatitsch.free.fr
>>>>>
>>>>> This email may be confidential and subject to legal privilege, it may
>>>>> not reflect the views of the University of Canterbury, and it is not
>>>>> guaranteed to be virus free. If you are not an intended recipient,
>>>>> please notify the sender immediately and erase all copies of the message
>>>>> and any attachments.
>>>>>
>>>>> Please refer to http://www.canterbury.ac.nz/emaildisclaimer for more
>>>>> information.
>>>>>
>>>>
>>>> --
>>>> Dimitri Komatitsch
>>>> CNRS Research Director (DR CNRS), Laboratory of Mechanics and Acoustics,
>>>> UPR 7051, Marseille, France    http://komatitsch.free.fr
>>>>
>>>> This email may be confidential and subject to legal privilege, it may
>>>> not reflect the views of the University of Canterbury, and it is not
>>>> guaranteed to be virus free. If you are not an intended recipient,
>>>> please notify the sender immediately and erase all copies of the message
>>>> and any attachments.
>>>>
>>>> Please refer to http://www.canterbury.ac.nz/emaildisclaimer for more
>>>> information.
>>>>
>>>
>>
>> --
>> Dimitri Komatitsch
>> CNRS Research Director (DR CNRS), Laboratory of Mechanics and Acoustics,
>> UPR 7051, Marseille, France    http://komatitsch.free.fr
>>
>> This email may be confidential and subject to legal privilege, it may
>> not reflect the views of the University of Canterbury, and it is not
>> guaranteed to be virus free. If you are not an intended recipient,
>> please notify the sender immediately and erase all copies of the message
>> and any attachments.
>>
>> Please refer to http://www.canterbury.ac.nz/emaildisclaimer for more
>> information.
>> _______________________________________________
>> CIG-SEISMO mailing list
>> CIG-SEISMO at geodynamics.org
>> http://lists.geodynamics.org/cgi-bin/mailman/listinfo/cig-seismo
>

-- 
Dimitri Komatitsch
CNRS Research Director (DR CNRS), Laboratory of Mechanics and Acoustics,
UPR 7051, Marseille, France    http://komatitsch.free.fr