Past events

  • 12:20 Room N106 IFEMA Session Seismic Imaging Theory - New Dire
    ctions. https://www.slim.eos.ubc.ca/content/fast-online-migration-compressi
    ve-sensing
    We present a novel adaptation of a recently developed relativel
    y simple iterative algorithm to solve large-scale sparsity-promoting optimi
    zation problems. Our algorithm is particularly suitable to large-scale geop
    hysical inversion problems, such as sparse least-squares reverse-time migr
    ation or Kirchoff migration since it allows for a tradeoff between parallelcomputations, memory allocation, and turnaround times, by working on su
    bsets of the data with different sizes. Comparison of the proposed method f
    or sparse least-squares imaging shows a performance that rivals and even ex
    ceeds the performance of state-of-the art one-norm solvers that are able tocarry out least-squares migration at the cost of a single migration with a
    ll data.

  • 10:20 Room N112 IFEMA Session Multidimensional Data Reconstruc
    tion.
    https://www.slim.eos.ubc.ca/content/grid-tensor-completion-seismic
    -data-interpolation
    The practical realities of acquiring seismic data in arealistic survey are often at odds with the stringent requirements of Shan
    non-Nyquist-based sampling theory. The unpredictable movement of the ocean`
    s currents can be detrimental in acquiring exactly equally-spaced samples w
    hile sampling at Nyquist-rates are expensive, given the huge dimensionalit
    y and size of the data volume. Recent work in matrix and tensor completion
    for seismic data interpolation aim to alleviate such stringent Nyquist-base
    d sampling requirements but are fundamentally posed on a regularly-spaced g
    rid. In this work, we extend our previous results in using the so-called H
    ierarchical Tucker (HT) tensor format for recovering seismic data to the ir
    regularly sampled case. We introduce an interpolation operator that resampl
    es our tensor from a regular grid (in which we impose our low-rank constrai
    nts) to our irregular sampling grid. Our framework is very flexible and eff
    icient, depending primarily on the computational costs of this operator. W
    e demonstrate the superiority of this approach on a realistic BG data set c
    ompared to using low-rank tensor methods that merely use binning.

  • 11:30 Room N103 IFEMA Session Full Waveform Inversion I.
    http
    s://www.slim.eos.ubc.ca/content/source-estimation-wavefield-reconstruction-
    inversion
    Wavefield reconstruction inversion is a new approach to waveformbased inversion that helps overcome the `cycle skipping' problem. However
    , like most waveform based inversion methods, wavefield reconstruction inv
    ersion also requires good source wavelets. Without correct source wavelets
    , wavefields cannot be reconstructed correctly and the velocity model canno
    t be updated correctly neither. In this work, we propose a source estimati
    on method for wavefield reconstruction inversion based on the variable proj
    ection method. In this method, we reconstruct wavefields and estimate sour
    ce wavelets simultaneously by solving an extended least-squares problem, w
    hich contains source wavelets. This approach does not increase the computat
    ional cost compared to conventional wavefield reconstruction inversion. Num
    erical results illustrates with our source estimation method we are able torecover source wavelets and obtain inversion results that are comparable t
    o results obtained with true source wavelets.

  • Workshop WS05, IFEMA Madrid. Common image gathers are used i
    n building velocity models, inverting for anisotropy parameters, and anal
    yzing reservoir attributes. In this paper, we offer a new perspective on i
    mage gathers, where we glean information from the image volume via efficie
    nt matrix-vector products. The proposed formulation make the computation offull subsurface image volume feasible. We illustrate how this matrix-vecto
    r product can be used to construct objective functions for automatic MVA.

  • 1:30 Session 6 M40-III: OPTIMISING INVERSION MODELS. Universityof Helsinki Fabianinkatu 33. Abstract: Wavefield Reconstruction Inversion
    is a method for PDE-constrained optimization, which revolves around the es
    timation of fields using the PDE as well as the observed data in a least-sq
    uares sense. The method is quadratic penalty based, which offers some inte
    resting possibilities for the construction of algorithms, compared to the
    Lagrangian form. One of the main benefits of the method is when the initialguess is far from the global minimizer. Reduced-space and full-space algor
    ithms are discussed, including illustrative examples. The method was devel
    oped with seismic applications in mind, but applies to other PDE-constrain
    ed optimization problems as well.
    https://www.slim.eos.ubc.ca/content/wave
    field-reconstruction-inversion

  • Rm 1100, 1986 Mathematics Road. Description: Wavelets provide
    a mathematical tool that emerged in the 1980s from a synthesis of ideas in
    mathematics, physics, computer science and engineering. They are now usedin a wide range of mathematical applications, and provide a mathematical
    way to “zoom in” on details, without losing track of the large picture. Th
    e talk will describe some of the essential features of the approach, and i
    llustrate with examples.

    Ingrid Daubechies, one of the world's leading
    mathematicians, is a member of the United States' National Academy of Scie
    nces, was a MacArthur Fellow, and is President of the International Mathe
    matical Union.

    Professor Daubechies was born and educated in Belgium. Sh
    e moved to the United States in 1987 where she first worked for Bell Labora
    tories and then at Princeton University where she was full Professor of Mat
    hematics from 1993-2011. She is best known for her discovery and mathematic
    al analysis of compactly supported wavelets, which are used in image compr
    ession, for example in JPEG 2000 for both both lossless and lossy compress
    ion. She was awarded the Steele Prize for mathematical exposition in 1994 f
    or her book, Ten Lectures on Wavelets.

    One focus of Daubechies' currentresearch is the development of analytic and geometric tools for the compar
    ison of surfaces. Her new approach, developed with Yaron Lipmon [1] uses c
    onformal mapping to define a metric between surfaces. Comparison of surface
    s plays a central role in many scientific disciplines and in the constructi
    on of video animations, and it is also a crucial step in many medical and
    biological applications. In an earlier collaboration, she worked with pale
    ontologists to develop a quantitative method to characterize the complexityof molar tooth surfaces, in an effort to reconstruct the diet of various
    extinct taxa [2].

    A particular interest of Professor Daubechies is the i
    mprovement of secondary mathematics education in the US and worldwide, andthe stimulation of mathematics, science and technology in developing coun
    tries. In 2009 she spent part of her sabbatical in Madagascar; she continu
    es to work with Malagasy mathematicians and scientists towards fostering a
    richer and more stimulating environment for students interested in developi
    ng a career in research and higher education.

  • PIMS Workshop on Advances in Seismic Imaging and Inversion
    M
    ay 20 to 22, 2015
    Centennial Centre for Interdisciplinary Sciences, Univ
    ersity of Alberta

  • 8:30 - 9:00, Centennial Centre, University of Alberta

  • 3:50 PM, Geoconvention 2015, Telus Convention Centre, Glen R
    oom. Time-lapse images void of acquisition and processing artifacts can pro
    vide more useful information about subsurface changes compared to those wit
    h acquisition footprints and other unwanted anomalies. Although, several p
    re-processing techniques are being developed and used to mitigate these unw
    anted artifacts, these operations can be very expensive, challenging and
    data dependent. Migration, as a processing tool, using a sparsity constra
    int has been shown to reduce artifacts drastically but little is known abou
    t the significance for compressed time-lapse seismic data. Leveraging ideasfrom distributed compressed sensing, and motivated by our earlier work onrecovery of densely sampled time-lapse data from compressively sampled mea
    surements, we present a sparsity-constrained migration for time-lapse datathat uses a common component shared by the baseline and monitor data. Our
    algorithm tested on a synthetic example highlights the advantages of exploi
    ting the common information, compared to ad hoc methods that involve paral
    lel processing of the time-lapse data before differencing.

  • Room 5106, 2207 Main Mall. Abstract
    Surface-related multipleshave been utilized in the reverse time migration (RTM) procedure and provi
    de extra illumination for subsurface. Many artifacts, however, are simult
    aneously generated due to the undesired cross correlations of forward- and
    backward-propagated seismic-events (i.e. primaries and different-order mult
    iples), which are hard to be attenuated and seriously contaminate the trueimages. We propose an approach denoted as RTM of controlled-order multiple
    s with least square sense, which can avoid most undesired cross correlatio
    ns and produce similar true-image compared with RTM of multiples without or
    der controlled. In our method, primaries are forward propagated and cross
    correlated with backward propagated controlled-order multiples. Primaries a
    nd multiples are separated by utilizing surface-related multiple-eliminatio
    n (SRME) and/or radon-based multiple-elimination methods during the seismic
    -data processing, where different-order multiples can be predicated by a m
    odified SRME procedure. With numerical experiments on two synthetic data se
    ts and the field data set of the South China Sea, it can be demonstrated t
    hat the proposed least square RTM of controlled-order multiples is a promis
    ing choice for imaging multiples with high signal-to-noise(S/N) ratio.