Generating variable shapes of salt geobodies from seismic images

in: 2018 Ring Meeting, ASGA

Abstract

Modeling 3D salt geobodies from subsurface data usually requires lots of time and expertise. The encountered difficulties stem from the assumptions underlying geomodeling techniques. These techniques are designed for almost vertical single-valued surfaces, and generally minimize the surface curvature and the layer width variations. These assumptions are not verified for salt envelopes, which are thus often manually modeled with a deterministic expert-driven approach. In this paper, we present an implicit methodology to generate automatically several possible models of salt boundary surfaces with varying geometries and topologies. We take as input a seismic image that we segment into three regions: salt, sediments and uncertain. The uncertain region is assumed to contain the salt boundary. We generate a normalized monotonic scalar field in this region, which can be seen as the cumulated probability for any point of being outside the salt boundary, i.e., to be sediments. This scalar field represents, in a way, a reference scenario. Each model is then obtained by perturbing this reference scenario. We illustrate the work ow and the impact of simulation parameters on 3D synthetic salt diapir data, and discuss some key elements that we consider as essential perspectives of this work.

Download / Links

BibTeX Reference

@INPROCEEDINGS{,
    author = { Clausolles, Nicolas and Collon, Pauline and Caumon, Guillaume },
     title = { Generating variable shapes of salt geobodies from seismic images },
 booktitle = { 2018 Ring Meeting },
      year = { 2018 },
 publisher = { ASGA },
  abstract = { Modeling 3D salt geobodies from subsurface data usually requires lots of time and expertise.
The encountered difficulties stem from the assumptions underlying geomodeling techniques. These
techniques are designed for almost vertical single-valued surfaces, and generally minimize the surface
curvature and the layer width variations. These assumptions are not verified for salt envelopes,
which are thus often manually modeled with a deterministic expert-driven approach.
In this paper, we present an implicit methodology to generate automatically several possible
models of salt boundary surfaces with varying geometries and topologies. We take as input a
seismic image that we segment into three regions: salt, sediments and uncertain. The uncertain
region is assumed to contain the salt boundary. We generate a normalized monotonic scalar field in
this region, which can be seen as the cumulated probability for any point of being outside the salt
boundary, i.e., to be sediments. This scalar field represents, in a way, a reference scenario. Each
model is then obtained by perturbing this reference scenario. We illustrate the work
ow and the
impact of simulation parameters on 3D synthetic salt diapir data, and discuss some key elements
that we consider as essential perspectives of this work. }
}