Geoscience Reference
In-Depth Information
including sequential indicator simulation
(SIS), indicator kriging and various facies
trend or facies belt methods which attempt to
capture gradational lateral facies changes. The
most common approach is the SIS method.
-
Texture-based methods
use training images to
recreate the desired architecture. Although
this has been experimented with since the
early days of reservoir modelling this has
only recently 'come of age' through the devel-
opment of multi-point
2.7
Algorithm Choice and Control
The preceding sections presented the basis for
the design of the rock modelling process:
Form geological concepts and decide whether
rock modelling is required;
Select the model elements;
Set the balance between determinism and
probability;
Intuitively set parameters to guide the
geostatistical modelling process, consistent
with the architectural concepts.
The next step is to select an algorithm and
decide what controls are required to move
beyond the default settings that all software
packages offer. Algorithms can be broadly
grouped into three classes:
-
Object modelling
places bodies with discrete
shapes in 3D space for which another model
element, or group of elements, has been
defined as the background.
-
Pixel-based methods
use indicator variograms
to create the model architecture by assigning
the model element type on a cell-by-cell basis.
The indicator variable is simply a variogram
that has been adapted for discrete variables.
There are several variants of pixel modelling
statistical
(MPS)
algorithms (Strebelle
2002
).
The pros and cons of these algorithms, includ-
ing some common pitfalls, are explored below.
2.7.1 Object Modelling
Object modelling uses various adaptations of the
'marked point process' (Holden et al.
1998
). A
position in the 3D volume, the marked point, is
selected at random. To this point the geometry of
the object (ellipse, half moon, channel etc.) is
assigned. The main inputs for object modelling
are an upscaled element log, a shape template
and a set of geometrical parameter distributions
such as width, orientation and body thickness,
derived from outcrop data (e.g. Fig.
2.32
).
Fig. 2.32
An early
example of outcrop-
derived data used to define
geometries in object
models (Fielding and Crane
1987
) (Redrawn from
Fielding and Crane
1987
,
#
SEPM Society for
Sedimentary Geology
[1987], reproduced with
permission)
