full waveform inversion in migration based travel time

11th World Congress on Computational Mechanics (WCCM XI)
5th European Conference on Computational Mechanics (ECCM V)
6th European Conference on Computational Fluid Dynamics (ECFD VI)
July 20–25, 2014, Barcelona, Spain
Chavent G.1 , Gadylshin K.2,3,∗ and Tcheverda V.3
Inria-Rocquencourt, Domaine de Voluceau,
BP 105, 78153 Le Chesnay Cedex, France
Novosibirsk State University, Mechanics and Mathematics Department
Pirogova 2, 630090 Novosibirsk, Russia
3 Institute of Petroleum Geology and Geophysics, SB RAS
Prosp. Akademika Koptyuga 3, 630090 Novosibirsk, Russia
∗ Email: [email protected]
Key words: seismic imaging, wave propagation, inverse problems, full waveform inversion
Constructing a smooth velocity model (propagator, macro velocity constituent) in the
depth domain, which is responsible for correct travel-times of wave propagation is a key
element of the up-to-date seismic data processing in areas with complex local geology.
Theoretically it could be obtained, along with the subsurface structure, by the Full Waveform Inversion (FWI) technique matching the observed and the synthetic seismograms
(Tarantola, 1984). The L2 norm is usually used for this matching, though other criteria
are also considered. To minimize the misfit function and to find the elastic parameters of
the subsurface, iterative gradient-based algorithms are usually applied. Such approach to
solving seismic inverse problem proposed originally by Tarantola (1984) has been developed and studied in a great number of publications (see Virieux and Operto, 2009, and
the references therein).
However, the straightforward application of FWI reconstructs reliably only the reflectivity
component of the subsurface but fails to provide a smooth velocity (propagator) component of a model. In order to overcome this trouble G.Chavent with colleagues introduced
Full Waveform Inversion in Migration Based Travel-Time formulation (2001). The main
idea of this approach is to decompose model space into two orthogonal subspaces - smooth
propagator and rough reflector with subsequent reformulation of the cost function.
Full Waveform Inversion formally is application of non-linear least squares for seismic
inverse problem treated as a nonlinear operator equation
F[m] = d.
Chavent G., Gadylshin K. and Tcheverda V.
Here the known right-hand side d is multi-source multi-receivers seismic data, F is a
non-linear operator (forward map) which transforms the current model m to synthetic
Instead of regular non-linear least squares formulation of Full Waveform Inversion, when
unknown function c(x) is searched as
c∗ = argmin kF(c) − dk2 ,
MBTT introduces the following decomposition of the model space:
m = p + r = p + Πr M(p) < s > .
Here p ∈ P describes smooth macrovelocity, which does not perturb significantly direction
of waves propagation, but governs their travel times. In contrast the depth reflector r
describes rough perturbations of the model, which send seismic energy back to the surface,
but do not change travel-times. The key moment here is interrelation r = Πr M(p) < s >
where s is unknown time reflectivity, M(p) - a true amplitude prestack migration
operator with linear reweighing W and Πr is the orthogonal projector onto the space of
reflectors (orthogonal to the space of propagators).
In this notations MBTT formulation of FWI with respect to propagator p and time
reflectivity s is as follows:
(p∗ , s∗ ) = argmin kF(p + Πr M(p) < s >) − dk
The numerical examples show that non-linear FWI in MBTT formulation is able to reconstruct the true macrovelocity models, as opposed to the standard L2 FWI.
The computations were performed on the HERMIT supercomputer of the High Performance Computing Center Stuttgart under the PRACE consortium grant 2012071274.
The research was done under financial support of the Russian Foundation for Basic Research grants no. 13-05-12051, 14-05-31257, 14-05-00109, 14-05-00049, 14-05-93090, grant
MK-2909.2014.5 of the President of the Russian Federation, and integration projects of
SB RAS 127 and 130.
[1] F. Clement, G. Chavent, and S. Gomez. Migration-based traveltime waveform inversion of 2-d simple structures: A synthetic example. Geophysics, 66:845–860, 2001.
[2] A. Tarantola. Inversion of seismic reflection data in the acoustic approximation.
Geophysics, 49(8):12591266, 1984.
[3] J. Virieux and S. Operto. An overview of full-waveform inversion in exploration
geophysics. Geophysics, 74(6):WCC1 – WCC26, 2009.