skip to main content
OSTI.GOV title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: Seismic velocity estimation from time migration

Abstract

This is concerned with imaging and wave propagation in nonhomogeneous media, and includes a collection of computational techniques, such as level set methods with material transport, Dijkstra-like Hamilton-Jacobi solvers for first arrival Eikonal equations and techniques for data smoothing. The theoretical components include aspects of seismic ray theory, and the results rely on careful comparison with experiment and incorporation as input into large production-style geophysical processing codes. Producing an accurate image of the Earth's interior is a challenging aspect of oil recovery and earthquake analysis. The ultimate computational goal, which is to accurately produce a detailed interior map of the Earth's makeup on the basis of external soundings and measurements, is currently out of reach for several reasons. First, although vast amounts of data have been obtained in some regions, this has not been done uniformly, and the data contain noise and artifacts. Simply sifting through the data is a massive computational job. Second, the fundamental inverse problem, namely to deduce the local sound speeds of the earth that give rise to measured reacted signals, is exceedingly difficult: shadow zones and complex structures can make for ill-posed problems, and require vast computational resources. Nonetheless, seismic imaging is a crucial partmore » of the oil and gas industry. Typically, one makes assumptions about the earth's substructure (such as laterally homogeneous layering), and then uses this model as input to an iterative procedure to build perturbations that more closely satisfy the measured data. Such models often break down when the material substructure is significantly complex: not surprisingly, this is often where the most interesting geological features lie. Data often come in a particular, somewhat non-physical coordinate system, known as time migration coordinates. The construction of substructure models from these data is less and less reliable as the earth becomes horizontally nonconstant. Even mild lateral velocity variations can significantly distort subsurface structures on the time migrated images. Conversely, depth migration provides the potential for more accurate reconstructions, since it can handle significant lateral variations. However, this approach requires good input data, known as a 'velocity model'. We address the problem of estimating seismic velocities inside the earth, i.e., the problem of constructing a velocity model, which is necessary for obtaining seismic images in regular Cartesian coordinates. The main goals are to develop algorithms to convert time-migration velocities to true seismic velocities, and to convert time-migrated images to depth images in regular Cartesian coordinates. Our main results are three-fold. First, we establish a theoretical relation between the true seismic velocities and the 'time migration velocities' using the paraxial ray tracing. Second, we formulate an appropriate inverse problem describing the relation between time migration velocities and depth velocities, and show that this problem is mathematically ill-posed, i.e., unstable to small perturbations. Third, we develop numerical algorithms to solve regularized versions of these equations which can be used to recover smoothed velocity variations. Our algorithms consist of efficient time-to-depth conversion algorithms, based on Dijkstra-like Fast Marching Methods, as well as level set and ray tracing algorithms for transforming Dix velocities into seismic velocities. Our algorithms are applied to both two-dimensional and three-dimensional problems, and we test them on a collection of both synthetic examples and field data.« less

Authors:
 [1]
  1. Univ. of California, Berkeley, CA (United States)
Publication Date:
Research Org.:
Lawrence Berkeley National Lab. (LBNL), Berkeley, CA (United States)
Sponsoring Org.:
USDOE Office of Science (SC), Advanced Scientific Computing Research (ASCR) (SC-21)
OSTI Identifier:
923471
Report Number(s):
LBNL-62726
R&D Project: 619701; BnR: KJ0101010; TRN: US200804%%1178
DOE Contract Number:
AC02-05CH11231
Resource Type:
Thesis/Dissertation
Country of Publication:
United States
Language:
English
Subject:
99 GENERAL AND MISCELLANEOUS; ALGORITHMS; CARTESIAN COORDINATES; CONSTRUCTION; EARTHQUAKES; PROCESSING; SUBSURFACE STRUCTURES; TRANSPORT; VELOCITY; WAVE PROPAGATION

Citation Formats

Cameron, Maria Kourkina. Seismic velocity estimation from time migration. United States: N. p., 2007. Web. doi:10.2172/923471.
Cameron, Maria Kourkina. Seismic velocity estimation from time migration. United States. doi:10.2172/923471.
Cameron, Maria Kourkina. Mon . "Seismic velocity estimation from time migration". United States. doi:10.2172/923471. https://www.osti.gov/servlets/purl/923471.
@article{osti_923471,
title = {Seismic velocity estimation from time migration},
author = {Cameron, Maria Kourkina},
abstractNote = {This is concerned with imaging and wave propagation in nonhomogeneous media, and includes a collection of computational techniques, such as level set methods with material transport, Dijkstra-like Hamilton-Jacobi solvers for first arrival Eikonal equations and techniques for data smoothing. The theoretical components include aspects of seismic ray theory, and the results rely on careful comparison with experiment and incorporation as input into large production-style geophysical processing codes. Producing an accurate image of the Earth's interior is a challenging aspect of oil recovery and earthquake analysis. The ultimate computational goal, which is to accurately produce a detailed interior map of the Earth's makeup on the basis of external soundings and measurements, is currently out of reach for several reasons. First, although vast amounts of data have been obtained in some regions, this has not been done uniformly, and the data contain noise and artifacts. Simply sifting through the data is a massive computational job. Second, the fundamental inverse problem, namely to deduce the local sound speeds of the earth that give rise to measured reacted signals, is exceedingly difficult: shadow zones and complex structures can make for ill-posed problems, and require vast computational resources. Nonetheless, seismic imaging is a crucial part of the oil and gas industry. Typically, one makes assumptions about the earth's substructure (such as laterally homogeneous layering), and then uses this model as input to an iterative procedure to build perturbations that more closely satisfy the measured data. Such models often break down when the material substructure is significantly complex: not surprisingly, this is often where the most interesting geological features lie. Data often come in a particular, somewhat non-physical coordinate system, known as time migration coordinates. The construction of substructure models from these data is less and less reliable as the earth becomes horizontally nonconstant. Even mild lateral velocity variations can significantly distort subsurface structures on the time migrated images. Conversely, depth migration provides the potential for more accurate reconstructions, since it can handle significant lateral variations. However, this approach requires good input data, known as a 'velocity model'. We address the problem of estimating seismic velocities inside the earth, i.e., the problem of constructing a velocity model, which is necessary for obtaining seismic images in regular Cartesian coordinates. The main goals are to develop algorithms to convert time-migration velocities to true seismic velocities, and to convert time-migrated images to depth images in regular Cartesian coordinates. Our main results are three-fold. First, we establish a theoretical relation between the true seismic velocities and the 'time migration velocities' using the paraxial ray tracing. Second, we formulate an appropriate inverse problem describing the relation between time migration velocities and depth velocities, and show that this problem is mathematically ill-posed, i.e., unstable to small perturbations. Third, we develop numerical algorithms to solve regularized versions of these equations which can be used to recover smoothed velocity variations. Our algorithms consist of efficient time-to-depth conversion algorithms, based on Dijkstra-like Fast Marching Methods, as well as level set and ray tracing algorithms for transforming Dix velocities into seismic velocities. Our algorithms are applied to both two-dimensional and three-dimensional problems, and we test them on a collection of both synthetic examples and field data.},
doi = {10.2172/923471},
journal = {},
number = ,
volume = ,
place = {United States},
year = {Mon Jan 01 00:00:00 EST 2007},
month = {Mon Jan 01 00:00:00 EST 2007}
}

Thesis/Dissertation:
Other availability
Please see Document Availability for additional information on obtaining the full-text document. Library patrons may search WorldCat to identify libraries that hold this thesis or dissertation.

Save / Share:
  • I. Amplitudes of short-period teleseismic P waves from underground explosions at the Nevada Test Site (NTS) exhibit coherent azimuthal variations that are positively correlated with travel time residuals. Both the Pahute Mesa and Yucca Flat subsites have similar patterns, indicating focusing and defocusing by deep (>400 km) heterogeneity. Variations are inverted via a thin lens procedure for warped velocity boundaries at depths of 25 and 160 km. Synthetic seismograms computed using the Kirchhoff integral show that these models explain much of the amplitude and some of the travel time behavior. Early P coda from NTS tests also shows coherent azimuthalmore » variations due to deep heterogeneity. Inversion for isotropic point scatterers near the source suggests that topography may play a key role in scattering surface waves to teleseismic P waves. II. The source-time functions of two great intraplate earthquakes the 1975 North Atlantic (M{sub w} = 7.8) and the 1077 Sumba (M{sub w} = 8.3) events, are deconvolved from long-period P waves. The 1975 earthquake changes focal mechanism from strike-slip to oblique normal, but no spatial variations of moment release can be resolved from the body waves. The normal-faulting 1977 Sumba earthquake has a central region with 9 m displacement, and 3 m displacement in surrounding regions. The rupture nucleated at 35 km depth near the edge of the high-displacement region and propagated 30 km west with little moment release before the main pulse, suggesting the rupture of an asperity. Discrepancies between the depth range and source duration implied by the P waves (0-35 km and 40 s) and long-period surface waves (0-80 km and 80 s) can be reconciled by a mode of rupture in the lower lithosphere that preferentially radiates very low frequencies.« less
  • The method of progressive hypocenter-velocity inversion has been extended to incorporate S-wave arrival time data and to estimate S-wave velocities in addition to P-wave velocities. Results of a P and S-wave progressive hypocenter-velocity inversion at The Geysers show that the top of the steam reservoir is clearly defined by a large decrease of V/sub p/ + V/sub s/ at the condensation zone-production zone contact. The depth interval of maximum steam production coincides with minimum observed V/sub p/ + V/sub s/, and V/sub p/ + V/sub s/ increases below the shallow primary production zone suggesting that reservoir rock becomes more fluidmore » saturated. The moment tensor inversion method was applied to three microearthquakes at The Geysers. Estimated principal stress orientations were comparable to those estimated using P-wave first motions as constraints. Well constrained principal stress orientations were obtained for one event for which the 17 P-first motions could not distinguish between normal-slip and strike-slip mechanisms. The moment tensor estimates of principal stress orientations were obtained using far fewer stations than required for first-motion focal mechanism solutions. The three focal mechanisms obtained here support the hypothesis that focal mechanisms are a function of depth at The Geysers. Progressive inversion as developed here and the moment tensor inversion method provide a complete approach for determining earthquake locations, P and S-wave velocity structure, and earthquake source mechanisms.« less
  • To determine to the extent possible the three dimensional velocity structure of the Los Angeles basin, and to make use of that in the relocation of earthquakes, a study was undertaken to invert for that velocity structure, making use of existing earthquake data and incorporating a ray tracing routine which allowed use of the program for the larger area of the current study.
  • In control centers, state-estimation processes a set of redundant measurements to obtain an estimate of the current states that are bus-voltage magnitudes and phase angles in the power network. Different algorithms for solving the state-estimation problems are discussed in first. Implementation of the algorithms that handle zero injections explicitly, such as normal equations with equality constraints and Hachtel's augmented-matrix method, is studied. In the second portion, network-topology errors are discussed. Modeling of network-topology errors, such as removal of branches, bus split, and disconnection of shunt capacitors, is developed. In addition, conditions for detectable or non-detectable topology errors are presented inmore » mathematical terms and physical terms. A sequential testing is proposed to distinguish topology errors and bad data. The final portion proposes a feasible algorithm to estimate the parameters. The model developed for topology error is used for analyzing the problem of network parameter error. In addition, the convergence of the algorithm is studied.« less
  • The experimental demonstration of a method for obtaining online, real-time particle-size distribution (PSD) measurements in a fluidized bed is detailed. In this method, a continuously flowing stream of sample particles drains from a well-mixed laboratory fluidized bed. The sample particles are pneumatically entrained, blown through the object plane of an optical array spectrometer particle-size analyzer, and channeled back to the freeboard region of the fluidized bed. The analyzer automatically sizes individual particles as they cut the object plane and transmits the data to a process computer system. The computer collects the in-coming data and resolves it to yield overall bedmore » PSD measurements once every 60 seconds. This method produces PSD measurements that are representative of actual conditions in the bed. The measurements are somewhat corrupted with random error (noise), however, and they are available to process computer only at discrete intervals in time. Advanced monitoring and control methods require that measures of important process parameters, such as fluidized bed PSD, be not only on-line and real-time but also continuous in time and well-behaved (little random error). Optimal estimation theory provides a mathematical framework for obtaining well-behaved, virtually continuous estimates from noisy, discrete-time measurements.« less