Reducing fault-related structural uncertainties by ranking models using seismic data misfit functions

in: 2016 RING Meeting, ASGA

Abstract

Fault interpretation from reflection seismic data is often challenging because of complex wave interactions in fault zones and limited seismic bandwidth. In such situations, many fault network models can explain the seismic data equally well; however, these networks may have significantly different behavior with respect to fluid flow. It is therefore risky to interpret a single deterministic fault network; one should rather provide a whole ensemble of fault networks. In this paper, we first look at some geological settings that may result in low resolution seismic images, leading to structural uncertainties, and in particular fault uncertainties. This is achieved by performing forward modeling simulations in realistic velocity models including fine layering, salt bodies and volumetric faults. These models are designed to benchmark quantitative uncertainty assessment methods and test strategies to reduce these uncertainties. We then propose a workflow to select best models from an ensemble of probable models: for a given seismic image, multiple fault interpretations are provided; each fault interpretation then leads to a structural model, which in turn leads to a velocity model. The resulting velocity models are then used to rank the different fault interpretations.

Download / Links

BibTeX Reference

@INPROCEEDINGS{,
    author = { Irakarama, Modeste and Godefroy, Gabriel and Cupillard, Paul and Caumon, Guillaume and Sava, Paul },
     title = { Reducing fault-related structural uncertainties by ranking models using seismic data misfit functions },
 booktitle = { 2016 RING Meeting },
      year = { 2016 },
 publisher = { ASGA },
  abstract = { Fault interpretation from reflection seismic data is often challenging because of complex wave
interactions in fault zones and limited seismic bandwidth. In such situations, many fault network
models can explain the seismic data equally well; however, these networks may have significantly
different behavior with respect to fluid flow. It is therefore risky to interpret a single deterministic
fault network; one should rather provide a whole ensemble of fault networks. In this paper, we first
look at some geological settings that may result in low resolution seismic images, leading to structural uncertainties, and in particular fault uncertainties. This is achieved by performing forward
modeling simulations in realistic velocity models including fine layering, salt bodies and volumetric
faults. These models are designed to benchmark quantitative uncertainty assessment methods and
test strategies to reduce these uncertainties. We then propose a workflow to select best models
from an ensemble of probable models: for a given seismic image, multiple fault interpretations are
provided; each fault interpretation then leads to a structural model, which in turn leads to a velocity model. The resulting velocity models are then used to rank the different fault interpretations. }
}