|
Figure Captions
|
 |
Figure 1.
Schematic distribution of fractured reservoir types. Type I is
fracture-dominated; Type IV is matrix-dominated; Types II and III
are where fractures control permeability and assist permeability,
respectively. |
|
 |
Figure 2.
Quantitative studies of fractures show they are fractal by nature
(from Mattner,
2002). Measurements
of  seismic anisotropy best characterize fracture scales in the
mid-range between 1 meter and 100 meters. |
|
 |
Figure 3.
Converted-wave survey geometry where the P-wave common mid-point
(CMP) is in a different location than the converted PS-wave
common-converted midpoint (CCP). To
measure azimuthal anisotropy reliably, it is important to sample a
full range of source-receiver azimuths over 360 degrees. |
|
 |
Figure 4.
Shear-wave splitting in an anisotropic medium, where the fast-shear
wave is polarized parallel to fractures and is orthogonal to the
slow-shear wave. |
|
 |
Figure 5. Valhall Field in the
North Sea shows a concentric pattern of fast shear waves that
correspond to subsidence around the producing platform (red
triangle).
This is a good example showing that the fast
shear-wave direction is highly sensitive to the maximum horizontal
stress direction. |
|
 |
Figure 6. Diagram of upgoing PS-waves,
showing that the separate fast and slow waves produced by the
initial PS-wave in the first (lower) anisotropic layer encountered
can split again within the next (upper) anisotropic layer above.
The various split S-wave modes are combined
when detected by the two horizontal geophones and must be unraveled
by layer-stripping to estimate the azimuthal anisotropy (fracture
properties). |
|
 |
Figure 7. Madden Field in the Wind River
Basin, Wyoming, shows percent anisotropy in color from zero to 9
percent over the Lower Fort Union (2.2 to 3.3 seconds)
and the fast S-wave orientation by small
vectors. Areas of high percentage of anisotropy may represent sweet
spots of concentrated fracturing or fracture swarms. |
|
 |
Figure 8. Emilio Field in the
Adriatic, offshore Italy, shows the fast S-wave direction in color
to illustrate the bimodal distribution associated with the target
layer (naturally fractured Scaglia carbonates, upper Paleocene).
Note the compartmentalization and apparent
control by faulting (thin black lines).
There is good agreement with the borehole data in wells at the top
of the structure (white points), and based on production, borehole
fracture studies and anisotropy from seismic data, the Emilio Field
has characteristics of a Type II fractured reservoir. |
|
 |
Figure 9. Graph that shows examples from
several reservoirs where percentage of wells are ordered from the
least to the most productive and the vertical axis is cumulative
production. The different fractured
reservoir types correlate nicely with these production
characteristics and the “fracture impact coefficient.”
Our goal is to avoid the scenario of
unproductive wells in the lower left corner of the graph by
characterizing fractures as early as possible. |
Return to top.
Converted waves (PS-waves), created by
traditional downgoing compressional waves (P-waves) that reflect as
shear-waves (S-waves), provide us with a unique ability to measure
anisotropic seismic attributes that are sensitive to fractures.
Solutions that PS-wave anisotropy can bring to fractured reservoir
management are:
-
Sweet-spot detection.
-
Improved models for
reservoir simulation.
-
Production history
matching.
-
Time-lapse behavior of
fracture properties over the life of a field (most important for
dynamic management).
The goal, of course, is to reduce the
total production costs for reservoir depletion by using fracture
information as early as possible.
It is well known that porosity and
permeability are key factors used to describe fractured reservoirs. As a
motivation for the need of azimuthal anisotropy measurements,
Figure 1 shows a schematic distribution of
different reservoir types in terms of percent total porosity and total
permeability (Nelson, 2001).
A Type I fractured reservoir is where
fractures dominate both porosity and permeability. Most of the reserves
are stored in the fractures, and flow is confined within them. These are
very heterogeneous and anisotropic reservoirs.
At the other end of this distribution
are Type IV reservoirs, where fractures provide no additional
permeability or porosity. Ideally this would be a homogeneous "tank"
reservoir when no fractures are present -- but when they are present,
fractures can sometimes be a problem and act as barriers to flow.
Type II and Type III fractured
reservoirs are of an intermediate nature where fractures control
permeability and assist permeability, respectively. In these two cases,
more reserves are stored within the matrix, but fractures still have an
impact and can result in anisotropic permeability and unusual response
to secondary recovery (elliptical drainage).
Bottom line: In going from Type IV to
Type I, there is an increasing effect of fractures.
Fracture properties are fractal by
nature, as illustrated in Figure 2. Cores
and image logs typically provide the small-scale features of the
reservoir and surface-seismic data can provide the largest scale
features like faults with large displacements. Each tool yields a
portion of the total fracture network; however, it is clear that these
end members alone do not control production. If they did, reservoir
models and fluid simulations would be perfect.
Fracture properties over the
intermediate range of scales in Figure 2 are
missing. Traditionally this has been filled with paleo-strain fields
that relate to possible fracture directions and intensities, inferred
from geomechanical modeling by palinspastic reconstruction. This method,
however, can be highly non-unique and uncertain in the presence of
unconformities.
Azimuthal anisotropy measurements can
be used for this sub-seismic resolution. Although fractures are smaller
than a seismic wavelength and individual fractures are not directly
observed, we do get an average response. This averaging leads to a
directional dependence; i.e., our velocities are azimuthally
anisotropic.
We can measure anisotropy at the
borehole with vertical seismic profiles (VSPs) and with P-wave surface
seismic data, but I want to focus on the use of PS-waves.
Figure 3 illustrates a typical PS-wave
source-receiver geometry. The most important property is the azimuth or
the propagation direction from source to receive. We need to sample a
full range of azimuths over 360 degrees for azimuthal anisotropy
measurements.
In addition to the P-waves that reflect
at a common midpoint (CMP), we detect PS-waves that convert at
common-conversion points (CCP), using three-component (3C) geophones.
The source-to-detector azimuth controls the direction of polarization of
the created S-wave, but this upgoing S-wave immediately splits and
travels to the surface as two orthogonally polarized S-waves.
Figure 4
shows a more detailed view of S-wave splitting for a single set of
vertical fractures, simulated by a grid that is oriented north-south.
The upgoing converted S-wave travels as a fast and slow component that
is polarized parallel and perpendicular to the fractures, respectively.
The time difference between them depends on the percent S-wave
anisotropy and is proportional to fracture density.
The algorithm used for fracture
characterization is a layer-stripping method that consists of first
finding an optimal rotation of the horizontal components to separate
fast and slow S-waves by Alford rotation. This provides the fast S-wave
direction (fracture orientation). Then correlation of the fast and slow
S-wave provides time delays for estimates of the amount of splitting and
fracture density information.
Figure 5
shows the results from the shallow overburden at the Valhall Field in
the
North Sea. A 3-D ocean bottom cable (OBC) survey was acquired there in
1998 using wide-azimuth source-receiver geometry to provide a full range
of azimuth data.
The small vectors show the orientation
of the fast shear-wave direction, oriented NNW by SSE, as measured along
the receiver lines, and the length of these vectors is proportional to
the time lag or percent anisotropy (maximum is about 3 percent). A
simple interpretation of this display is that the vectors represent a
single set of vertical fractures seen from above.
Note the interesting concentric pattern
centered on the production platform (red triangle). This is a dramatic
example where man-made alterations of the subsurface have induced
horizontal-stress perturbations near the surface.
The pattern of S-wave splitting
correlates precisely with subsidence at the platform due to collapse of
the reservoir. In the center where there has been four meters of
subsidence the anisotropy is relatively small, but as one moves away
from the center, there is an increase in the anisotropy along the flanks
of the subsidence where radial extension is occurring. Here is where the
minimum horizontal stress direction is radial, and the maximum
horizontal stress direction is transverse. This agrees exactly with the
fast S-wave orientation, and it is a good example showing that the fast
S-wave direction is highly sensitive to the maximum horizontal stress
direction.
Advanced Applications
Unfortunately the situation is a bit
more complicated than that described in the Valhall example, and
advanced applications of seismic azimuthal anisotropy are required.
Figure 6 shows that the complexity of S-wave
splitting can increase with the distance of travel. The separate fast
and slow waves produced by the initial PS-wave in the first (lower)
anisotropic layer encountered can split again within the next (upper)
anisotropic layer above.
In addition, each rock layer can have a
different orientation of fractures (coordinate frame) and different
fracture density. The various split S-wave modes are combined when
detected by the two horizontal geophones. In order to estimate the
azimuthal anisotropy (fracture properties) at the target, we need to
unravel the data by layer-stripping in a top-down fashion. As a result,
the overburden anisotropy must be determined and removed first. The
results at Valhall Field (Figure 5)
represent an estimate of this overburden anisotropy.
Several land examples from
Wyoming were acquired to investigate naturally fractured gas sands. Two
of these, from the Green River Basin, show similarities in the
orientation of the fast S-wave and amount of anisotropy in the
overburden, as well as fracture-related anisotropy associated with
faults and lineaments.
Another example is the Madden Field
from the
Wind River Basin (Figure 7). Naturally
fractured tight gas sands in the Tertiary Lower Fort Union formation
produce from depths of 4500 to 9000 feet. A 3-D seismic survey covering
15 square miles over the crest of the field shows the fault trends (bold
east-west lines). The seismic data were acquired using dynamite with 20
pound charges set at a depth of 60 feet.
The important attributes are shown in
Figure 7, the percent anisotropy in color,
from zero to 9 percent over the Lower Fort Union (at 2.2 to 3.3 seconds
reflection time) after correcting for overburden anisotropy by
layer-stripping, and the fast S-wave orientation by small vectors whose
length is proportional to percent anisotropy.
The interesting point here is that
variations in percent anisotropy appear to be controlled by the faults;
the orientation of the fast S-wave is usually oblique to them. Areas of
high percentage of anisotropy may represent sweet spots of concentrated
fracturing or fracture swarms.
Although fracture properties have not
been directly calibrated with anisotropy measurements from borehole data
in the survey area, a VSP outside the area showed changes in anisotropy
(S-wave orientation) between the overburden and Lower Fort Union that
are similar to the PS-wave anisotropy.
The next example (Figure
8) is from the
Adriatic Sea, offshore Italy, where the target is the naturally
fractured upper Paleocene Scaglia carbonate. Significant east-west
tectonic compression creates north-south anticlinal structures where
commercial quantities of gas have accumulated in fractured zones.
The operators (Agip) acquired an ocean
bottom cable (OBC) seismic survey to help them position two horizontal
wells for optimal recovery. The fast S-wave direction shown in color
illustrates the bimodal distribution associated with the target layer.
Yellows and oranges are oriented roughly east-west, and blues and greens
north-south.
Note the compartmentalization and
apparent control by faulting (thin black lines). Where faults and
anticlinal structure (thick red arrows) change direction in the south,
there is also a change in the fast S-wave direction (browns and dark
blues).
The most important result is the good
agreement with the borehole data in wells at the top of the structure
(white points). From breakout analysis and induced fracture studies, the
maximum horizontal stress is consistently about N70E. This agrees with
P-wave fast directions determined from AVO analyses as a function of
azimuth.
Based on production, borehole fracture
studies and anisotropy from seismic data, the Emilio Field has
characteristics of a Type II fractured reservoir. Out of the small
number of wells drilled, only a few are highly productive. Although
there may be some secondary matrix or vuggy porosity, it appears that
fractures control the permeability and have a significant impact on the
production.
Historically the classification of Type
I (fracture-dominated) to Type IV (matrix-dominated) reservoirs has
proved to be quite useful. Figure 9 is a
graph, also from Nelson (2001), showing examples from several reservoirs
where the percentage of wells are ordered from the least to the most
productive, nd the vertical axis is cumulative production. The different
fractured reservoirs correlate nicely with these production
characteristics.
For the Type I, fracture-dominated
heterogeneous reservoirs, a small percentage of wells contribute to most
of the production, and there are many dry and marginal wells. As we
transition through the other types, the curves become straighter, and
more wells contribute equally to the total production. The 45-degree
line corresponds to a homogeneous-isotropic, matrix-dominated reservoir
where all wells contribute equally.
Nelson has quantified these fractured
reservoir types by a “Fracture Impact Coefficient.” He points out that
this is not necessarily a physical property of the reservoir, but is
instead a result of drilling fields on regular grids without exploiting
the presence of fractures -- something he calls “fracture denial.”
Consequently, it might be more appropriate to call this quantity the
“Fracture Denial Coefficient,” because it appears to be directly
proportional with fractured reservoir type and ranges between 0.28 --
0.73.
Ultimately our goal is to avoid the
scenario of unproductive wells in the lower left corner of the graph in
Figure 9 by using every tool at our disposal
to characterize fractures as early as possible for efficient reservoir
depletion. One of these tools can be PS-wave data for measuring
azimuthal anisotropy and the heterogeneity related to fractures.
Conclusion
The examples presented in these
articles suggest that azimuthal anisotropy can be measured with
wide-azimuth PS-wave surveys and that S-wave splitting is highly
sensitive to the maximum horizontal stress direction. Knowing these
maximum stress directions, which are aligned with open fractures when
the differential stress is large enough, provides valuable information
about preferred reservoir flow directions.
Potentially, PS-wave data could become
an integral part of fracture sweet-spot detection, reservoir model
building/simulation,and dynamic reservoir management through the use of
time-lapse surveys. However, to utilize this technology optimally, it is
important to calibrate results with ground truth for incorporating into
reservoir models. One approach is VSP data to acquire azimuthal S-wave
information at the same scale as surface-seismic data. Dipole sonic and
FMI logs are also valuable for characterizing small-scale fracture
properties that can be related to larger scale features.
It also is important to improve our
resolution with smaller seismic time windows and more accurate
anisotropy models that include dipping fracture properties. However,
these will have to be the subject of future research.
The author
thanks Rich Van Dok, Richard Walters and Bjorn Olofsson from WesternGeco,
for their expertise in data processing of the Madden, Emilio, and Valhall
studies, respectively; and also Lynn Inc., Eni/Agip division, BP, and
WesternGeco for their support and permission to publish this material.
References
Nelson, R.A.,
2001, Geologic analysis of naturally occurring fractured reservoirs (2nd
edition): Gulf Professional Publishing, Boston,
332 p.
Mattner, Joerg, 2002, Fractured reservoir characterization from
collecting data to dynamic modeling : Course, GeoTech Consulting, slide
(as part of presentation).
Return to top.
|
Print this page
Click to view article in PDF format.