[CIG-SEISMO] integrated inversion tool for SPECFEM3D (will be released before the summer)
Dimitri Komatitsch
komatitsch at lma.cnrs-mrs.fr
Thu Jun 22 09:18:55 PDT 2017
Hi all,
Done!
(added to specfem3d/src/inverse_problem_for_source;
Vadim's unified FWI framework for SPECFEM3D will go in
specfem3d/src/inverse_problem_for_model, we will commit it on Monday).
Best regards,
Thanks again to Brad for the information on how to add git submodules!
Dimitri.
On 06/20/2017 03:40 PM, Dimitri Komatitsch wrote:
>
> Hi Federica and Emanuele, Hi all,
>
> Thanks for your answers. Very useful. I am going to cut and paste them
> in the manual of SPECFEM and put a copy of cmt3d and of pycmt3d in the
> Git version of SPECFEM (they are compatible in terms of license, pycmt3d
> is LGPLv3 and SPECFEM is GPLv2+, which makes the combination GPLv3
> according to Wikipedia).
>
> best wishes,
> Dimitri.
>
> On 06/20/2017 10:54 AM, Federica Magnoni wrote:
>> Hi all,
>>
>> Initially we used the fortran code cmt3d
>> (https://github.com/UTCompSeismo/GRD_CMT3D/tree/master/cmt3d) and
>> recently, as Emanuele said, we switched to the python code pycmt3d.
>>
>> Best regards,
>> Federica
>>
>> 2017-06-19 19:53 GMT+02:00 emanuele casarotti
>> <emanuele.casarotti at gmail.com <mailto:emanuele.casarotti at gmail.com>>:
>>
>> You should consider pycmt3d https://github.com/wjlei1990/pycmt3d
>> <https://github.com/wjlei1990/pycmt3d>
>>
>>
>> Il giorno lun 19 giu 2017 alle 19:48 Dimitri Komatitsch
>> <komatitsch at lma.cnrs-mrs.fr <mailto:komatitsch at lma.cnrs-mrs.fr>> ha
>> scritto:
>>
>>
>> Hi Surendra, Hi all,
>>
>> Vadim's tool currently inverts for the model only, mostly
>> because we
>> developed it for oil industry problems, in which case the
>> location of
>> the active source is known. Here is what SPECFEM3D users
>> classically do
>> to relocate earthquake sources (paragraph taken from a text
>> written by
>> Federica Magnoni and Emanuele Casarotti for Italy):
>>
>> "As initial source parameters of the chosen events we use Time
>> Domain
>> Moment Tensor solutions (TDMT, Dreger and Helmberger, 1993;
>> Scognamiglio
>> et al. 2009) that are calculated by inverting full
>> three-component
>> traces of regional broad-band stations using 1D wave speed
>> models.
>> Before starting the tomographic inversion, we then exploit the
>> initial
>> 3D model of Italy of Di Stefano and Ciaccio (2014) to
>> recalculate the
>> source parameters of the considered events. The TDMT solutions,
>> based on
>> 1D wave speed models, are considered as initial solutions and we
>> invert
>> for the six moment tensor components and the event location by
>> using the
>> technique presented in Liu et al. (2004). We use SPECFEM3D to
>> numerically calculate the Fréchet derivatives with respect to
>> the source
>> parameters based on 3D models and we minimize a waveform misfit
>> function
>> to invert for source solutions. If the variance reduction of the
>> solution with the 3D model is significant, this is considered as
>> the new
>> moment tensor solution of the event."
>>
>> which comes from this paper by Qinya Liu:
>>
>> http://komatitsch.free.fr/preprints/bssa_Qinya_inversion_2004.pdf
>> <http://komatitsch.free.fr/preprints/bssa_Qinya_inversion_2004.pdf>
>> and in
>> which the derivatives are computed using finite-differences.
>>
>> Emanuele and Federica (and Qinya), do you have scripts to do
>> that?
>> if so, are they in the package somewhere (or should we consider
>> committing them?); that would be very useful.
>>
>> Thanks,
>> Best regards,
>> Dimitri.
>>
>> On 06/06/2017 06:31 PM, Surendra Nadh Somala wrote:
>> > Dear Dimitri,
>> >
>> > Are these tomographic inversions or can SPECFEM3D also be
>> used for
>> > source inversions?
>> >
>> > Thanks,
>> > Surendra
>> >
>> >
>> > ___________
>> > *Surendra Nadh Somala*
>> > Assistant Professor
>> > Department of Civil Engineering (Block-E, 208)
>> > Indian Institute of Technology (IIT) Hyderabad
>> > Kandi, Sangareddy
>> > Telangana, India - 502285
>> > Phone : +91 (0)40 2301-8457 <tel:+91%2040%202301%208457>
>> > http://www.iith.ac.in/~surendra/
>> <http://www.iith.ac.in/~surendra/>
>> >
>> > On Tue, Jun 6, 2017 at 9:56 PM, Dimitri Komatitsch
>> > <komatitsch at lma.cnrs-mrs.fr
>> <mailto:komatitsch at lma.cnrs-mrs.fr>
>> <mailto:komatitsch at lma.cnrs-mrs.fr
>> <mailto:komatitsch at lma.cnrs-mrs.fr>>> wrote:
>> >
>> >
>> > Hi all,
>> >
>> > Vadim Monteiller here in our group has created a nice
>> integrated FWI
>> > tool for SPECFEM3D (based on L-BFGS and following
>> > http://komatitsch.free.fr/preprints/GJI2_Vadim_2015.pdf
>> <http://komatitsch.free.fr/preprints/GJI2_Vadim_2015.pdf>
>> > <http://komatitsch.free.fr/preprints/GJI2_Vadim_2015.pdf
>> <http://komatitsch.free.fr/preprints/GJI2_Vadim_2015.pdf>>
>> Sections 3
>> > and 4). The advantage is that it uses no I/Os to disk,
>> all the
>> > inversion routines (computing sensitivity kernels,
>> smoothing them,
>> > summing, iterating) are called from a single (parallel)
>> calling
>> > program and thus only memory is used.
>> > We have designed a set of examples, which we will put in
>> BuildBot.
>> >
>> > We will release it before the summer, Vadim is finishing
>> a few final
>> > tests.
>> >
>> > Best regards,
>> > Dimitri.
>> >
>> > --
>> > Dimitri Komatitsch, CNRS Research Director (DR CNRS)
>> > Laboratory of Mechanics and Acoustics, Marseille, France
>> > http://komatitsch.free.fr
>> > _______________________________________________
>> > CIG-SEISMO mailing list
>> > CIG-SEISMO at geodynamics.org
>> <mailto:CIG-SEISMO at geodynamics.org>
>> <mailto:CIG-SEISMO at geodynamics.org
>> <mailto:CIG-SEISMO at geodynamics.org>>
>> >
>> http://lists.geodynamics.org/cgi-bin/mailman/listinfo/cig-seismo
>>
>> <http://lists.geodynamics.org/cgi-bin/mailman/listinfo/cig-seismo>
>> >
>> <http://lists.geodynamics.org/cgi-bin/mailman/listinfo/cig-seismo
>> <http://lists.geodynamics.org/cgi-bin/mailman/listinfo/cig-seismo>>
>> >
>> >
>> >
>> >
>> > _______________________________________________
>> > CIG-SEISMO mailing list
>> > CIG-SEISMO at geodynamics.org <mailto:CIG-SEISMO at geodynamics.org>
>> >
>> http://lists.geodynamics.org/cgi-bin/mailman/listinfo/cig-seismo
>>
>> <http://lists.geodynamics.org/cgi-bin/mailman/listinfo/cig-seismo>
>> >
>>
>> --
>> Dimitri Komatitsch, CNRS Research Director (DR CNRS)
>> Laboratory of Mechanics and Acoustics, Marseille, France
>> http://komatitsch.free.fr
>>
>>
>>
>>
>> --
>> Federica Magnoni
>> Istituto Nazionale di Geofisica e Vulcanologia (INGV)
>> Via di Vigna Murata, 605
>> 00143 ROMA, Italy
>> e-mail: federica.magnoni at ingv.it <mailto:federica.magnoni at ingv.it>
>>
>
--
Dimitri Komatitsch, CNRS Research Director (DR CNRS)
Laboratory of Mechanics and Acoustics, Marseille, France
http://komatitsch.free.fr
More information about the CIG-SEISMO
mailing list