The present Masters Thesis had for objective the study of the seismic inversion problem
based on flat reflectors for common-source (SC) and common- mid-point (CMP) gathers. The
forward model is described by homogeneous, isotropic, plane horizontal layers. The problem is
related to the NMO stack based on the optimization of the semblance function, for CMP sections
corrected for moveout time. The study was based on two principles. The first principle adopted was the combination
of two groups of inversion methods: A global and a local method. The second principle adopted
was stripping according to the Wichert-Herglotz-Bateman theory, that establishes that to know a
lower layer it is necessary to know first the upper layer.
The application of the study aims at the seismic simulation of the terrestrial Solimões
and marine sedimentary basins to obtain a 1D distribution of velocity and layer thicknesses of the
subsurface of target horizons. In this sense, we limited the inversion experiments to 4 to 11
reflectors, once in practice the industry limits the interpretations to be to about one same number
of 3 to 4 main reflectors. Stands out that this model is applicable as initial condition to the
imaging of seismic sections in geologically complex regions wit h slow lateral variation of
velocities. The synthetic data was produced based on geological information that corresponds to
strong a priori information for the inversion model. For the construction of models related to the projects in progress, we analyzed the
following relevant subjects: (1) Geology of sedimentary basins terrestrial Solimões and marine
(stratigraphy, structural, tectonics and petroliferous); (2) Physics of the vertical and horizontal
seismic resolution; and (3) Temporal-spatial discretization of the multi-coverage cube. The
inversion process is dependent on the discretization of the wave field in time-space, on the
physical parameters of the seismic survey, and of further on the resampling in the multiple
coverage cube. The direct model us ed corresponds to the case of the NMO (1D) stack operator,
considering a flat observation topography. The basic criterion taken as reference for the inversion
and curve fit is the norm 2 (quadratic). The inversion using the simple present model is computational attractive for being fast,
and convenient for allowing several other techniques be included with a logical physical
interpretation; e.g., the Fresnel projected zone (ZFP), the direct calculation of the spherical
divergence, Dix inversion, linear inversion by reparametrization, a priori information, and
regularization. The ZFP shows to be a useful concept to establish the aperture of the spatial inversion
window in the time-distance section. The ZFP represents the influence of the data in the
horizontal resolution. The estimative of the ZFP indicates a minimum aperture based on an
adopted model. The spherical divergence is a smooth function, and it has physical basis to be used in the
definition of a data weight matrix for tomographic inversion methods. The necessity of robustness in the inversion can be analyzed in seismic sections (CS and
CMP) submitted to filtering (corners frequencies: 5; 15; 75; 85; pass-band trapezoidal), where
one can identify, compare and interpret the information contained. From the sections, we
conclude that the data are contaminated with isolated points, what proposes methods in the class
considered as robust having as reference the norm 2 (least-square) of curve fitting. The development of the algorithms used the FORTRAN 90/95 programming language,
the program MATLAB for presentation of results; and the package CWP/SU for synthetic
seismic modeling, picking of events and presentation of results.
|
