You are on page 1of 4

I OIL AND GAS I GEOSCIENCE I

Subsurface Modeling : Back to Geology


By : Jean-Claude Dulac - Executive Vice President and Chief Architect, Paradigm

3D subsurface models are necessary for evaluating reserves, for basin studies,
fault analysis, and reservoir characterization. This article shows the limits of
current modeling technology based on pillar-based 3D grids.

D subsurface models are nec


essary for evaluating reserves,
for basin studies, fault analysis,
and reservoir characterization. In this
article, we will show the limits of
current modeling technology based
on pillar-based 3D grids. We will
introduce our next-generation modeling application called SKUA (Subsurface Knowledge Unified Approach) based on a geo-chronological mathematical model. A SKUA

model is constructed directly from


the interpretation data; no transformation or deformation is applied to
the interpretation, and geological
rules such as fault displacement
constraints, volume preservation
during tectonic deformation, and
erosions are applied to construct a
geologically realistic model.

Current method
Subsurface modeling is the act of

constructing a digital model of the


subsurface to better understand
potential prospects or reservoir
behavior. Today, this digital subsurface model is represented as a
three-dimensional regular grid
where each 3D cell is more or less
cubic. The cells are arranged to represent the geology as best as is
possible. The cells are aligned to fit
the stratigraphy and the faulting.
Figure 1 shows a 2D section of

Figure 1: Section of a simple faulted grid model

Figure 2:Fault linearization to avoid cell intersection in classical regular grids


DEW JOURNAL July 2008

45

Figure 3: Fault rift system where complex faulting occurs

Artifact of the pillars


method

Figure 4: Transformation from the XYZ space


to the Paleo-space

such a grid. The vertical grid lines


(or pillars) are aligned to the faults
and the horizontal lines to the top
and base horizon.
A regular grid representation
of the subsurface has a lot of limitations. A first limitation is shown in
Figure 2. If the cells corners along a
fault are not aligned on a line, the
cells will intersect. To alleviate this
problem the faults are linearized
along so called fault pillars and the
cells are aligned to this fault pillars.
The process to linearize a fault
is straightforward in simple models,
but as soon as multiple faults intersect, (see Figure 3) this process becomes very cumbersome and in
most cases simply impossible. Practitioners must therefore make the
costly decision to modify the fault
network by changing the dip and the
extension of some faults and by
removing others from the data set.

Even when the process of fault linearization succeeds, it can introduce artifacts. To understand these
we need to look at the process of
computing property values, such as
net-to-gross porosity, for every cell
in the grid. These properties are
typically known at the well location,
so to populate these grids we need
to perform extrapolation of log values away from the wells in a geologic way. More explicitly when
extrapolating, we need to work in a
space where folding and faulting is
removed, and which mimics as
close as possible the condition under which the sediment is deposited (i.e. on a pseudo plane). We call
that space the Paleo space. Figure
4 illustrates the transformation from
the current X,Y,Z space to the Paleo-space.
It is very important during this
transformation that XYZ distances
and Paleo distances are uniformly
transformed everywhere in a stratigraphic unit, as it is assumed that
the geological deposition conditions are similar inside one stratigraphic unit. In the case shown in
Figure 5, a 2D grid with pillars parallel to two faults shows clearly the
deformation introduce by aligning
the pillars to the faults. The channel
objects of identical size in the Paleo space are stretched or
squeezed in the XYZ space.

New Paleo-transform
To correct these problems SKUA
uses a new full 3D transformation
from the XYZ space to the Paleospace called the UVT-Transform.
The construction of this transformation is simple. SKUA assigns a
unique geochronological time (T) to
a given horizon (seismic interpretation scattered points and/or well
markers). In the XYZ space the surface defined by that T will be faulted and folded, but in the UVT space
the surface will be a plane (by definition). The UV represents the two
other dimension of the Paleo-space
and defines the paleo-geography of
each T plane.
The UVT Transform is computed everywhere in the volume, honoring the geochronological information given by the horizons (T) and
constructing U and V so that (U,V,T)
are perpendicular (perpendicularity
constraint) and of constant length
(homogeneity constraint) such that
unit cubic cells in XYZ are unit
cubic cells in UVT anywhere in the
volume.

New Subsurface
representation
Along with the UVT Transform,
SKUA introduces a new type of grid
to represent the subsurface called a
Geologic Grid. Its hexahedral cells
are constrained not to cross layer
boundaries; however they are split
by the faults, at the exact location

Figure 5: Deformation of objects between the top and bottom of the reservoir induced by the pillar gridding

46

DEW JOURNAL July 2008

Fig 6: Split cells in Geologic Grid (a) sediment volume is deposited,


(b) faulting occurs, (c) displacement occurs along fault plane.

Figure 7: Section of a Geologic Grid with erosion and a growth fault with (a) grid lines and
stratigraphic units and (b) global simulation of a continuous property.

Figure 8: Channel object simulation in a Geologic Grid. All channels correctly have the same width (only one out of ten gridlines is displayed for visual clarity).

of the fault, not at cell boundaries


as with pillar grids. The offsets of
the possibly multiple split parts are
equal to the displacement that occurrs along the faults. This is illustrated in Figure 6 where a sediment
volume is deposited and faulting
occurs, splitting the volume and
forming a displacement along the
fault plane.
Columns of cells are orthogonal
to their respective stratigraphic unit
boundaries and not parallel to the
faults (see Figure 7a); layers of cells
show the correct throw offset
across fault cuts even in the case of
reverse faulting.
The UVT Transform minimizes
the distortions of stratigraphic distances and volumes (orthogonality
and homogeneity constraints). Figure 8 shows the same channel objects realization as Figure 5 in the
Geologic Grid. The simulation correctly displays all channels at the
same width, exhibiting the net volume and connectivity imposed by
the model parameters. Grid line
definitions and faults should not
have an influence on the property
model. The simple grid examples in
Figures 5 (pillar grid) and 8 (Geologic Grid) exhibit a 5% difference in
net (channel) rock volume inside
the central fault block for the same
48

DEW JOURNAL July 2008

bulk rock volume.

Advantages of the method


SKUA is a step change in modeling.
There are no horizon surfaces to
create, no pillars to create between
a top and a bottom horizon, no pillars to align when faults are in contact. The user constructs a faulted
volume, and all horizons and 3D
geologic grids are constructed simultaneously as the UVT Transform
is computed. All the data is used,
no faults need to be removed or
transformed to fit a pillar model.
By representing faults independently from the grid, the paleo
transformation necessary to compute rock properties in the paleo-

space is more accurate than with


pillar models leading to more accurate property distribution and volume computations.
By suppressing many of the
interactive steps necessary to fit
the data to the modeling tool, the
batch modeling process improves
repeatability, updatability, and auditability.
Because of the elimination of
time-consuming steps, models
constructed with SKUA can be
constructed in hours versus weeks
or months. Finally, SKUA models
more accurately represent geology
by eliminating geometrical and
rock property distribution
distortions. dewjournal.com

ABOUT THE AUTHOR


Jean-Claude Dulac,, Executive Vice President and Chief
Architect, Before joining the Paradigm team, Dulac was the
founder and chief architect of Earth Decision, a leading
provider of fully integrated shared-earth modeling for asset
teams, before the company was acquired by Paradigm in
2006. As CEO of Earth Decision, Dulac grew the company to
$33 million in revenue. Prior to creating Earth Decision,
Dulac held positions at Unocal, including development manager and senior
research geophysicist. In these roles, he managed the industrialization of
Gocad, was responsible for most of the software architecture and design
and developed new techniques and programs for velocity analysis and
velocity correction before or after depth migration. Prior to his work at Unocal,
Dulac was a developer at Total. Dulac earned an M.S. degree in geophysics
from Stanford University and an M.S. degree in geology from Ecole Nationale
Suprieure de Gologie in France.

You might also like