Appraising structural interpretations using surface seismic data: where we stand and the road ahead

Modeste Irakarama and Paul Cupillard and Guillaume Caumon and Paul Sava. ( 2019 )
in: 2019 Ring Meeting, ASGA

Abstract

A seismic image will often support multiple structural interpretations and, consequently, multiple structural models. The problem of using seismic data to determine which of these models are more consistent than others is referred to as structural model appraisal. The general solution proposed to tackle this problem is to generate synthetic data for each structural modeland compare them to observed data; this allows to assign a data misfit value for each structural interpretation. It has been suggested that appropriate data misfit functions for structural model appraisal should be localized in space. Such misfit functions are made of two main components: a residual operator which computes differences between observed and synthetic data, and a projection operator which maps data residuals from the data-domain into the image-domain. Earlier theoretical studies investigated the use of phase-shift data residuals operators and linear projection operators. Numerical experiments, however, have revealed that a more practical approach would be to use L1-based data residual operators and nonlinear projection operators. We investigate the latter strategy in this paper and argue that it relaxes some theoretical and practical constraints inherit in the former strategy.

Download / Links

BibTeX Reference

@INPROCEEDINGS{IrakaramaRM2019,
    author = { Irakarama, Modeste and Cupillard, Paul and Caumon, Guillaume and Sava, Paul },
     title = { Appraising structural interpretations using surface seismic data: where we stand and the road ahead },
 booktitle = { 2019 Ring Meeting },
      year = { 2019 },
 publisher = { ASGA },
  abstract = { A seismic image will often support multiple structural interpretations and, consequently, multiple structural models. The problem of using seismic data to determine which of these models are more consistent than others is referred to as structural model appraisal. The general solution proposed to tackle this problem is to generate synthetic data for each structural modeland compare them to observed data; this allows to assign a data misfit value for each structural interpretation. It has been suggested that appropriate data misfit functions for structural model appraisal should be localized in space. Such misfit functions are made of two main components: a residual operator which computes differences between observed and synthetic data, and a projection operator which maps data residuals from the data-domain into the image-domain. Earlier theoretical studies investigated the use of phase-shift data residuals operators and linear projection operators. Numerical experiments, however, have revealed that a more practical approach would be to use L1-based data residual operators and nonlinear projection operators. We investigate the latter strategy in this paper and argue that it relaxes some theoretical and practical constraints inherit in the former strategy. }
}