-
107
Quantitative fault seal prediction: a case study from Oseberg
Syd
T. Fristad, A. Groth, G. Yielding and B. Freeman
We describe a case study from Oseberg Syd where fault-seal
behaviour has been predicted from analysis of a detailed depth
model in conjunc-
tion with detailed lithological control.
Juxtaposition seal of reservoir against non-reservoir can be
assessed by fault-plane diagrams. Additional seal may be developed
(at reservoir
juxtapositions) if fault-plane processes increase the capillary
entry pressure. In Oseberg Syd, clay smearing is considered to be
dominant because of
the relatively shaly nature of the Brent Group and the shallow
burial depths during faulting (
-
10
8
T. Fristad
, A. Gro
th, G
. Yie
ldin
g an
d B
. Fre
em
an
t~
o o
. ,..q
o t~
. ,..4
~D
o ;L
.,'4
. ,-~
o 0
.
0
~
0
0
~ ,..~
e~
0
e,i
r
-
Quantitative fault seal prediction 109
Table 1
Simplified stratigraphy of the Brent Group
Formation Depositional environment Permeability characteristics
Sand quality
Draupne Shale Offshore Draupne Sand Turbidite fan Heather shale
Offshore Heather 2 Lower to upper shoreface
Heather 1
Upper Tarbert
Middle Tarbert 2
Middle Tarbert 1 Lower Tarbert Upper Ness Middle Ness
Lower Ness Lower Ness sand
Oseberg/Rannoch/Etive
Lower shoreface to offshore transition zone
Lower to upper shoreface
Coastal lagoon, barriers, inlet channels
Coastal lagoon, swamp Lower shoreface to foreshore Upper to
lower delta plain Upper delta plain (abandoned lobe, lacustrine,
swamp) Upper delta plain Fluvial channel
Marginal marine delta deposits
Non-reservoir Strongly heterogeneous Non-reservoir Strongly
layered, partly calcite cemented, moderate to poor permeability
Strongly layered, partly calcite cemented, poor reservoir quality
Strongly layered, moderate to poor permeability Strongly
heterogeneous but many high permeable intercalations
Strongly heterogeneous Low to very high permeability Strongly
heterogeneous, clay-rich Reservoir quality poor to absent
Strongly heterogeneous, clay-rich Stacked channel sand with high
permeability Not significant reservoir unit
Mudstone Whole range Mudstone Fine to medium
Very fine to fine sand
Fine to medium sand
Whole range, medium to coarse sand forms an important component
Whole range Very fine to coarse sand Whole range Predominantly fine
grained components Whole range Coarse to medium grained sand Whole
range
study area is presented in Table 1. A more compre- hensive
description of the sequence stratigraphy within Block 30/9 can be
found in MOiler and Van der Wel (1997).
Oseberg Syd- structural setting
The Oseberg/Oseberg Syd area lies between the Horda Platform and
the Viking Graben, an area of Mesozoic extension. The study area
comprises some 15-20 elongated fault blocks. Most faults within the
Oseberg/Oseberg Syd region strike N-S to NNW- SSE, subparallel to
the Viking Graben, in an anasto- mosing pattern. The areal extent
of each fault block ranges from 250 km 2 to less than 10 km 2. An
attempt to subdivide the area into structural sub-units outlined by
major faults with offsets in the range 200-1000 m, is shown in Fig.
2.
Recent improvements in the seismic database in the
Oseberg/Oseberg Syd region provided a signifi- cantly improved
seismic interpretability, and a more confident fault interpretation
has greatly enhanced the understanding of the structural framework.
Local ar- eas of excellent seismic data allow for interpretation of
a comparatively large number of seismic reflectors within the
Jurassic succession (Fig. 4). This facilitates a confident mapping
of thickness variations across faults. Even without correction for
differential com- paction, the section demonstrates a spectacular
thick- ness increase of nearly 100% within the Brent, Dunlin and
Statfjord Formations across the major fault between the Gamma and
Omega structures. Similar thickness changes are mapped across
several
major faults in the area, with a stepwise thickness increase
across each of the major faults. There are fairly constant interval
thicknesses within each main fault-bounded compartment (cf.
Yielding et al., 1992).
Tectonic development
The spatial distribution of fault-related growth in the Oseberg
Syd region is shown in Fig. 2. It can be concluded that most of the
main faults (i.e., those that outline the key structural elements)
were subject to substantial differential subsidence even prior to
the main Late Jurassic rifting event. This indicates the existence
of an early Viking Graben, exerting a strong influence on the Early
to Middle Jurassic depositional systems. Furthermore, these faults
were subject to accelerated differential subsidence in the Late
Jurassic, recorded by substantial ex- pansion of the Viking Group
(Heather/Draupne Fms). This Late Jurassic phase of extension and
block rotation caused a collapse along the crests of the major
fault blocks already established in the Lower/Middle Jurassic. A
series of minor fault blocks was thus formed in the Oseberg Syd
region during the Late Jurassic, possibly extending into the Creta-
ceous.
In conclusion, the local thickness variations imply a relatively
shallow depth of burial during the fault activity (below ca. 500m).
Basin modelling and backstripping across the Oseberg area supports
this statement (Roberts et al., 1993, 1995).
The Brent Group in general is quite shaly and,
-
1 1
0
T. Fristad
, A. Gro
th, G
. Yie
ldin
g an
d B
. Free
ma
n
d=
o
[.r.,
r.~
.,..q
*-., 0
E~
,.~ ;=
d=
o "~
Z
=
az
~:
a:z b
.,
0 o
~E
s
9
=
E
~m
.'~
o
~
+
=0
o
~
E
~o
.~ .~
.=
o ~
~
~
o
~
=~
.
oo
"~
o
-
Quantitative fault seal prediction 11 1
therefore, we might expect any observed sealing be- haviour to
be a consequence either of juxtaposition or of some mechanism of
clay smearing. Because of the shallow depth of burial during
faulting, the shales would be expected to be ductile. Thin sections
and core fractures from C and J areas show clear indica- tions of
smearing along small-scale faults in shaly sand intervals (Fig. 5).
In more sandy intervals, frac- ture zones are observed where the
fracture porosity is filled with fine-grained material. However,
increased fault throw would probably smear clay along the en- tire
fault surface.
Fault seal mechanisms
Most seals in clastic sequences are membrane seals (Watts,
1987). The dominant control on seal failure is the capillary entry
pressure of the seal-rock, that is, the pressure required for
hydrocarbons to enter the largest interconnected pore throat of the
seal. A num- ber of mechanisms have been recognised whereby fault
planes can act as a membrane seal (e.g., Watts, 1987; Knipe, 1992):
(i) Juxtaposition. Reservoir sands are juxtaposed
against a low permeability unit with a high entry pressure
(e.g., shale).
(ii) Clay smear. Entrainment of clay or shale into the fault
plane, thereby giving the fault itself a high entry pressure.
(iii) Cataclasis. Crushing of sand grains to produce a fault
gouge of finer-grained material, again giv- ing the fault a high
capillary entry pressure.
(iv) Diagenesis. Preferential cementation along an originally
permeable fault plane significantly in- creases the entry
pressure.
Juxtaposition seals can be recognised explicitly by mapping the
juxtaposition of units across the fault plane. To identify or
predict sealing by clay smear, cataclasis or diagenesis requires an
ability to relate these mechanisms to measurable properties of the
subsurface (such as lithology and fault displacement), so that
deterministic estimates of seal potential can be made. These
different mechanisms and methods are discussed below. At present,
significant success has been achieved in developing algorithms for
prediction of seal capacity by clay smear (Fulljames et al., 1997;
Yielding et al., 1997). It seems likely that seal by cataclasis may
be similarly understood in the near future (Fulljames et al.,
1997). Seal by diagenesis, however, will probably be much less
amenable to prediction by simple algorithms.
Clay smear
The classic study of clay smear in a production en-
vironment is that by Bouvier et al. (1989), describing the Nun
River Field in the Niger Delta. They present a predictive method of
assessing whether clay smear is likely to be sufficient to form a
membrane seal along the fault zone. A "clay smear potential" (CSP)
is stated to represent the "relative amount of clay that has been
smeared from individual shale source beds at a certain point along
a fault plane". CSP is not de- fined explicitly by Bouvier et al.,
but is stated to: (i) increase with shale source bed thickness,
(ii) increase with the number of source beds displac-
ed past a particular point along a fault plane, and (iii)
decrease with increased fault throw.
Fulljames et al. (1997) give the algorithm for CSP as
T 2 (1)
where T is the thickness of the source bed, and D is the
distance from the source bed.
Bouvier et al. calibrated their CSP calculations against known
sealing and non-sealing faults, and divided the observed range into
high, medium and low CSP. Low CSP represents little chance for the
presence of continuous clay smear seals that can trap
hydrocarbons.
Lindsay et al. (1993) describe outcrop studies of shale smears
in a carboniferous fluvio-deltaic se- quence in northern England.
As in the study de- scribed above, Lindsay et al. concentrated on
the ef- fects of individual shale beds in the sequence rather than
the bulk properties of the sequence. Smear is observed to be
thickest when derived from thicker source layers and with small
fault throw values; smear thicknesses commonly decrease with
distance from the shale source bed. From a study of 80 faults they
conclude that shale smears may become incom- plete when the ratio
of fault throw to shale layer thickness exceeds 7. Smaller ratios
are more likely to correspond to continuous smears and therefore to
a sealing layer on the fault surface.
Gibson (1994) presents observations from the Ter- tiary
sand-shale sequence of the Columbus Basin, offshore Trinidad. From
an analysis of fault-sealed hydrocarbon columns, he concludes that
the more significant seals are developed where the ratio of fault
throw to shale layer thickness is less than 4 (i.e., the shale bed
is >25% of the displaced section).
The above studies suggest that sealing by clay smear may be
predicted deterministically from a con- sideration of the thickness
and offset of individual shale beds. However, such an approach is
difficult to apply directly in the Brent Group because of the het-
erogeneity of the sequence. It is not feasible to map
-
112 T. Fristad, A. Groth, G. Yielding and B. Freeman
Fig. 5. Thin-section taken from the Ness Fm (2408 m MD) in the
30/9-9 well (J structure). Note the concentration of clay minerals
in the small fault.
Fig. 6. Diagram illustrating the calculation of SGR at a point
on a fault surface. The throw (t) at the point is defined from the
offset horizons. The "throw window" in the hangingwall represents
the thickness of the rock that has slipped past the point. The SGR
at the point is equal to the per- centage of shale in the throw
window. For units composed of "pure" shale and non-shale, SGR is
the sum of the shale thicknesses divided by the throw. For units of
given shale fraction, these fractions are used as weighting factors
in the summation such that the result is the net shale percent- age
within all units in the window.
-
Quantitative fault seal prediction 113
every shale bed and consider its effect at the fault surface.
Therefore, we take here a simpler approach of considering only the
bulk properties of the se- quence at the scale of the reservoir
mapping (later we show the equivalence of the two approaches by
sen- sitivity analysis). We define a fault-surface attribute called
the shale gouge ratio (SGR) which is simply the percentage of shale
or clay in the slipped interval. Fig. 6 illustrates how this would
be calculated at a point on a fault surface:
SGR - E (V~l Az) t
x 100% (2)
Vc~ is the clay or shale volume fraction in each inter- val of
thickness Az and t is the fault throw at that point. The interval
thicknesses are measured in a "window" with a height equal to the
throw; this win- dow therefore represents the column of rock that
has slid past this point on the fault. The SGR represents, in a
general way, the proportion of shale that might be entrained in the
fault zone. The more shaly the wall rocks, the greater the
proportion of shale in the fault zone and therefore the higher the
capillary entry pressure. Whilst this is undoubtedly an
oversimplifi- cation of the detailed processes occurring in the
fault zone, it represents a tractable "up-scaling" of the
lithological diversity at the fault surface; the required
information is simply fault displacement and shale fraction through
the sequence. SGR is approximately the reciprocal of "shale smear
factor" (SSF) defined by Lindsay et al. (1993).
Direct observations of sub-surface pressure allow a calibration
to be made between the SGR and seal ca- pacity. Ideally, an in situ
measurement of the pore- pressure in the reservoir and that inside
the fault zone would allow the capillary entry pressure of the
fault to be calculated. However, fault-zone pressures are rarely
available. Instead, the pressure difference be- tween the two walls
of the fault is a more general parameter that can be derived from
pressure meas- urements in pairs of wells across the fault. Fig. 7a
shows one such calibration, based on the Nun River dataset of
Bouvier et al. (1989). From their strike projections of Fault "K",
values of SGR have been calculated on a dense grid across the fault
surface. On the same grid, minimum across-fault pressure differ-
ences have also been derived, using the proven distri- bution of
hydrocarbons in the footwall sands to cal- culate buoyancy
pressures. Fig. 7a shows a cross-plot of these two parameters for
the areas of sand-sand contact at the fault surface. The dashed
line indicates the inferred relationship between SGR and seal ca-
pacity. At SGR < 20%, no fault-sealed hydrocarbons are observed;
the shale content of the slipped interval
is too low. Above 20%, the maximum observed pres- sure
difference progressively increases, reaching ca. 7 bars at a SGR of
ca. 60% (for gas). The large cloud of points lying below the dashed
line indicates that many points on the fault do not achieve their
full seal capacity, because they lie at lower elevations in the
structurally-controlled hydrocarbon columns. The line is important
as a calibration in describing the maximum pressure difference
supportable by that part of the fault, if other factors are
favourable.
Interfacial tension (and, therefore, the entry pres- sure) for
the gas-water system is typically as much as twice that for the
oil-water system over a wide range of conditions (Schowalter,
1979). This suggests that the likelihood for seal is greater for
gas than for oil. Therefore a given SGR value might be expected to
sustain a greater pressure difference for gas than for oil. In Fig.
7a, the higher values of pressure differ- ence (> 1 bar) are
from gas caps, whereas the smaller pressure differences are
generated by oil.
Fig. 7b shows a similar cross-plot, using data pro- vided by
Gibson (1994; his Fig. 8) from the Colum- bus Basin. In this plot,
each data-point represents one reservoir top, with observations
from many different faults. All reservoirs contain oil, and no gas.
The dis- tribution of points is similar to that in Fig. 7a, al-
though with a slightly different position for the bounding line.
The similarity of the plots is encour- aging, in that they
represent data from different se- quences in different areas. This
implies that the SGR might be a useful predictive attribute across
a range of environments. Detailed differences in the calibra- tions
in different areas might possibly be due to fac- tors such as shale
lithology, degree of consolidation, fluid type, etc.
Cataclasis
Cataclasis is the brittle deformation of material in a fault
zone, and typically involves grain breakage and comminution (often
associated with improved pack- ing). This results in a
significantly reduced grain size in the fault zone which can
therefore support a pres- sure difference because of the increased
capillary en- try pressure. Knipe (1992) reviews microstructural
studies of fault-zone rocks and notes that cataclastic fault gouge
may have pore throat radii of
-
114 T. Fristad, A. Groth, G. Yielding and B. Freeman
oo k_
133 i
(1) o r (1) t
@ tl= s (1)
oo d~
13. m
:3
!
0 o <
10
8
6
+ gas
9 oil
seal capacity/
/ I l
I I ""
/ /
' i t . . . .
0 20 40 60 80 Shale gouge ratio (%)
(a)
100
03 L ,_
co i
(D 0 t-"
t_
@ tl= s 9 t~
O3 OO (D k_
n
I
co
0 t_
0 <
10
8
6
4
2
0
seal capacityt
/0
0 100
/
/
/ ~ ;OOi 9 I O0 9 9 I
j 9 ' I ' ' I
2O 40 60 80 Shale gouge ratio (%)
(b)
Fig. 7. Examples of calibration of SGR against across-fault
pressure difference. (a) Data from the Nun River field (Fault K of
Bouvier et al., 1989). Each point represents one point (grid-node)
on the fault surface. SGR was calculated using the sand-shale
sequence shown by Bouvier et al. Across-fault pressure difference
was calculated using densities of 1.0, 0.83 and 0.3 for water, oil
and gas, respectively. The dashed line labelled "seal capacity"
represents the maximum pressure difference that can be supported by
a given value of SGR. (b) Data from the Columbus Basin, offshore
Trinidad, based on Fig. 8 of Gib- son (1994). Each point represents
one reservoir top, with observations from many different faults
(all reservoirs are oil-beating). Data points falling well to the
right of the "seal capacity" line are faults bounding relatively
small dip closures (i.e., the seal capacity is not realised).
gouge should be more likely at greater depth, and during reverse
and strike-slip faulting rather than ex- tensional faulting.
However, experimental studies (Mandl et al., 1977) show that some
grain breakage
can occur at low overburden pressures, corresponding to only 100
m of burial.
Knipe (1992) makes the pertinent point that the nature of a
fault zone will vary across the fault sur- face, depending on which
lithotypes are being dis- placed. Clay smear may be significant
over some parts of the fault surface, but cataclastic gouge may be
developed where shale beds are absent. We should expect that the
sealing capacity of a fault will often be highly variable over
different parts of its surface, and simple whole-fault descriptions
such as "sealing" and "non-sealing" may often be misleading. Any
pressure difference across the fault would have to be supported by
its weakest point.
Fault-seal methodology
Our approach in this paper has been to examine juxtaposition
relationships and compute fault seal attributes on strike
projections of fault surfaces. The analysis was carried out using
FAPS software (Freeman et al., 1989; Needham et al., 1996).
The overall procedure was as follows: (i) produce depth
cross-sections from mapped hori-
zon depth grids, incorporating reconstructed fault "sticks";
(ii) generate gridded representations of the individual fault
surfaces, modelling their three-dimensional shape, displacement
variation and horizon inter- sections;
(iii) construct a simplified geological layer model for the
Brent Group, and interpolate this zonation into the mapped horizon
intervals at the fault sur- faces;
(iv) establish shale-volume fractions within each of the layers
(reservoir zones) at each fault;
(v) calculate values of SGR over each fault surface, using fault
displacement and layer shale-volume fractions;
(vi) compare SGR with pressure (RFT) data where available for
well pairs across a fault.
Each of these steps is described in more detail be- low.
Depth sections All fault and fault-seal analysis was performed
in
the depth domain, since this allows: (a) direct comparison with
fluid contact levels, and (b) incorporation of additional
geological informa-
tion such as zone isochores. Primary mapping of the subsurface,
however, was performed on seismic data, in the time domain. The
seismic interpretation was depth-converted by ap- plying
appropriate velocity models to the TWT grids. The faults are now
represented as polygons on each
-
Quantitative fault seal prediction 115
Horizon grids with fault polygons Cross-sections with
reconstructed fault segments Fig. 8. Diagram illustrating how
vertical depth sections of a fault can be reconstructed from fault
polygons and horizon grids. The points defining the fault segments
on the sections correspond to the centre-lines of the polygons on
the depth grids.
of the horizon surfaces, and information about the vertical
correlation of faults between horizons is lost. An important part
of the analysis was therefore the reconstruction of the fault
planes in depth. Depth grids for the primary mapped horizons (Base
Creta- ceous, Top Lower Tarbert, Base Brent, Top Cook and Top
Statfjord) were available at a 50 x 50 m grid node spacing. Most of
the faults in the study area trend approximately N-S and therefore
the strategy adopted was to sample the depth grids on E-W rows to
create a suite of sections at 50 m spacing (486 in total). Taking
each horizon grid in turn, in conjunc- tion with its fault-polygon
file, an automated search of the grid rows was made in order to
locate the posi- tions of the horizon cutoffs at the faults. Fault
"sticks" on the cross-sections were generated by joining the
centre-line points of corresponding fault polygons on successive
horizons: the position of each point is defined by the x,y
information from the poly- gon and the z data from the grid (see
Fig. 8). These fault segments were then labelled according to the
fault surface to which they belong.
Gridded fault surfaces The x,y,z information of a group of fault
segments
is used to construct a grid that accurately matches the
three-dimensional shape of the fault plane. This grid is the base
on which all the calculations of fault at- tributes are performed.
In this study, the larger faults were gridded at 100 x 100 m, the
smaller faults at 50 x 50 m. The primary information for
displacement and stratigraphic computations is the geometry of the
horizon/fault intersections. Gaps between horizons and faults are
corrected for by applying a "snapping" procedure (see Needham et
al. (1996) for further dis- cussion). Taking the depth difference
between the upthrown and downthrown cutoffs of the same hori- zon
gives the throw at that point on the fault. These measurements are
used as the control points for pro-
ducing a grid of throw variation over the entire fault
surface.
Geological layer model In addition to the mapped horizons (i.e.,
those im-
ported from the grids), additional horizons such as the
intra-Brent zonation were interpolated into the fault models by
reference to the primary, mapped horizons. Detailed isochore maps
for the study area were con- structed on the basis of well data and
seismic charac- ter mapping. Five to seven zones were recognised
within the Brent Group, with an additional overlying sand in the
Heather Formation. The additional hori- zons are posted onto each
fault grid in one of two ways: firstly, as a fixed distance
(thickness) above or below a primary horizon, or secondly, at a
fixed per- centage of the interval between two primary horizons.
Posting of these horizons for both the footwall and hangingwall
side of the fault results in a detailed and geometrically-robust
juxtaposition plot.
Shale-volume data Petrophysical analysis of the well data was
used to
define the shale fraction in each stratigraphic unit. CPI logs
were used to derive explicit shale percent- ages within both
"sandstone" units (e.g., 5% shale in the Lower Ness Sandstone) and
"shale" units (e.g., 65% shale in the upper part of the Dunlin
Group). This information was then compiled geographically to
estimate likely compositions between the wells, i.e., at the fault
locations. Example profiles of shale- volume fractions are shown in
Fig. 9.
Shale gouge ratio Having constructed the fault grid, with
detailed
juxtapositions and compositional data for all layers, we
calculate a SGR. It was stated earlier that the fault surfaces were
gridded at 100 x 100m or 50 x 50 m. Whilst this is adequate for
analysis of displacement
-
1 16 T. Fristad, A. Groth, G. Yielding and B. Freeman
(a) (b) Fig. 9. Schematic illustration of the shale-fractions
for (a) Fault 1 (G structures) and (b) Fault 9 (J structures). Note
that the lowest fraction of shale for Fault 1 is within the upper
part of the interval (Tarbert Fm), whereas for Fault 9 it is found
in the lower half. The total Brent thickness in (a) is
approximately four times the thickness in (b).
variation, it cannot capture the detailed stratigraphic
variation that affects the SGR calculation. A grid re- finement was
therefore applied when calculating SGR, replacing each original
grid node by 5 x 5 new nodes. At each node, the local throw value
defines the height of a "search window" in the hangingwall (cf.
Fig. 6). Within the search window, the program measures the
thickness of each unit (down the fault plane) and combines these
with the units' shale frac- tions to calculate the net shale
percentage in the search window. By definition (Eq. (2)) this is
equal to the local SGR at that point on the fault.
The window over which shale values are summed could be in the
footwall or hangingwall. In the ab- sence of sedimentary growth
across the fault these will be identical. If growth has occurred
then an aver- age of the two is more appropriate. In the Oseberg
Syd dataset there is some growth across some of the faults; however
it is often not possible to use data from the footwall "window"
because there has been erosion. Therefore, SGR has always been
calculated using the hangingwall "window".
ment is static (pre-production) it is possible to project these
pressure profiles to adjacent faults. With a cross-fault well pair,
the pressure profile can be con- structed in both the footwall and
hangingwall of the fault, on the same refined grid as that used for
the SGR calculation. Comparison of across-fault pressure
differences with SGR at every point on the fault al- lows any
relationship between the two to be exam- ined. This relationship
can then be extrapolated to those faults where well control is
lacking, i.e., SGR can be used to constrain predictions of
potential pres- sures (and hence hydrocarbon columns) in untested
compartments.
Fault descriptions
The poor seismic data quality in the C and J structures limits
the reliability of the fault seal analy- sis in these areas.
Consequently, the description be- low is focused upon the good
quality seismic of the western area of Block 30/9 (Omega, B and G).
The eastern area is covered in more general terms.
Comparison with pressure data Where wells are present on both
sides of a fault,
we calibrate the SGR attribute with pressure data. Detailed RFT
data permit the construction of pressure profiles at the wells, and
since the pressure environ-
Fault I
Fault 1 is located in the south-west part of the study area in
the G area (see Fig. 3). The maximum throw observed on this fault
is about 175 m, dimin-
-
Quantitative fault seal prediction 117
a
b
Fig. 10. Strike projections of Fault 1, viewed from the
downthrown (west) side; vertical exaggeration x5. (a) Juxtaposition
plot. Upthrown Brent zones are shown with coloured fill (see
legend); downthrown zones are shown in black outline, labelled at
each end of the fault. Footwall hydro- carbon contacts are shown in
black, hangingwall contacts in blue. (b) SGR for the area of
Brent-Brent overlap. Upthrown zones outlined in blue, downthrown
zones in black; contacts as in (a). SGR is colour-coded in the
ranges 0-15%, 15-20%, 20-30% and >30%. Note the area of slightly
lower SGR on the upper part of the overlap zone; this is the
critical area for fault seal.
ishing to zero displacement towards the south. The G East and G
Central aquifers are therefore in commu- nication around the
southern end of the fault. The pattern of juxtaposed Brent zones on
the fault is shown in Fig. 10a. Note the considerable area of
Brent-Brent overlap: maximum fault offset is about half the Brent
thickness. Wells are located both in the
footwall (30/9-13S) and hangingwall (30/9-14) and have different
hydrocarbon columns, and so this fault provides a good calibration
point with respect to the SGR calculation. The hangingwall
oil-water contact is probably controlled by a structural
spill-point (saddle) along the southern part of the fault. The
fault is therefore probably not at seal capacity, and the
-
1 18 T Fristad, A. Groth, G. Yielding and B. Freeman
E v
c -
E3
O!m 1t
Pore pressure (bar) 305 310 315 320 325 330 335
l l l , l , , , , l l l l l l l l l l l , , , , I , , , , F~/
9.5 bar HW A 30/9-13s
1 . . , " ' m --~ 30/9-14 Footwall GOC
Foo~,all
~gas gradient 03 g/cc
'+
_Hanging wall GOC .
/dlLHa_ngin_.g w_all OWC
Fig. 11. Graph illustrating the pore-pressure profile through
the Brent Group on each side of Fault 1. FW, footwall; HW;
hangingwall. Note the upwards increase in pressure difference
through the hydrocarbon columns. The two aquifer gradients are
believed to be coincident (within the uncertainty of the tool
measurements) since the reservoir is continuous around the southern
end of the fault (see Fig. 3).
calibration below represents a minimum potential for seal on
this fault surface.
The display of SGR on the fault surface (Fig. 10b) uses the
shale fractions observed in the ad jacent wells. Since the fault
displacements are generally greater than the zone thicknesses, the
calculated SGR values are relatively homogeneous. However, a sig-
nificant point is the area of lower values (in yellow,
-
Quantitative fault seal prediction 119
Brent overlap zone. For fault seal, the significant question
here is: which points on the fault are capable of holding back a
large pressure difference for a rela- tively small SGR? Such points
represent the critical areas for seal, and in Fig. 12 they will lie
to the upper left of the cross-plot. A SGR of ca. 18% is capable of
sustaining a pressure difference of almost 8 bar, and slightly
higher SGRs (23%) can sustain 9.5 bar. These points correspond to
the uppermost part of the reservoir-reservoir overlap zone: in Fig.
10b they occur at and just above the yellow area ca. 1 km from the
north end of the fault. This part of the fault is holding back the
higher-pressured gas column in the hangingwall.
As with Fig. 7, Fig. 12a shows many data points that represent
smaller pressure differences than the maximum, for a given SGR.
These points correspond to structurally lower parts of the fault
surface, for example near the hangingwall fluid contacts. Here seal
is probably well-developed, but the in situ pres- sure difference
is small.
To test the sensitivity of the analysis a detailed template
consisting of 65 layers within the Heather Fm and the Brent Gp was
generated (Fig. 13a). The resulting SGR plot (Fig. 13b), shows only
minor dif- ferences compared to the coarse model (Fig. 10b). This
is because Eq. (2) tends to reduce the effect of stratigraphic
complexity as the throw increases. As a rule of thumb, the detail
of the stratigraphic template needs to be at the same order of
scale as the minimum throw in the area of interest. The plot of
pressure dif- ferences versus SGR (Fig. 12b) reveals more or less
the same trend as observed in Fig. 12a, but with a SGR of 17%
sustaining an 8.3 bar pressure differ- ence.
The fault demonstrates static sealing since it sepa- rates two
hydrocarbon columns with a maximum pressure difference of 9.5 bar.
Accordingly, this fault can be used for calibration of the
calculated SGR val- ues. The lowest calculated SGR value above the
HC contacts is slightly below 20% (ca. 18%), indicating that for
other faults having a SGR value in the same range, static sealing
up to ca. 8 bar differential pres- sure could be anticipated (for
gas as the high pressure phase). Eight bars differential pressure
corresponds to about 240 m difference in OWC or 106 m difference in
GWC, for single-phase hydrocarbon columns and typical densities
(gas 0.25 g/cm 3, oil 0.66 g/cm 3, wa- ter 1 g/cm3).
Fault 2
Prior to the drilling of well 30/9-14, the ca. E-W fault located
350-450 m to the south was regarded as a block-bounding fault (see
Fig. 3 for location). The
fault has a minimum displacement of about 15-20 m at its centre
and consequently the different units within the Tarbert Fm are
juxtaposed against them- selves (self-juxtaposed) (Fig. 14a).
The SGR values within the Tarbert Fm juxtaposi- tion are close
to 15% (Fig. 14b). DST testing of well 30/9-14 indicated the fault
to be open, as the closest barrier to flow was interpreted to be
810 m away.
Implications from this fault and Fault 1 therefore suggest that
a SGR below or close to 15% corre- sponds to no seal and SGR
above-18-20% corre- sponds to significant seal. This is a very
tight range, but it remains quite consistent throughout the
dataset.
Fault 3
This fault was selected to investigate the segmen- tation of the
B structures and for calibration purposes with respect to Fault 1
(see Fig. 3 for location).
The maximum displacement lies between the branch lines with
Faults 1 and 7, and the displace- ment decreases southwards. Just
north of the southern branch with Fault 4, the uppermost part of
the Tarbert and Heather Fms (oil and gas) are juxtaposed against
the lower parts of the Tarbert Fm (water), with about 2 bar
pressure difference. In addition, the SGR values are just below 20%
or higher, indicating by analogy that the observations from Fault 1
can be applied to this fault as well.
Between B South and B North, the SGR is above 20%, which agrees
well with the different fluid con- tacts and pressure regimes
observed in wells 30/9-7 and 30/9-4S (ca. 5 bar pressure
difference). In the central part of the fault (between G Central
and B South), over 8 bar pressure difference is observed, at a SGR
of ca. 28%.
Fault 4
Just north of the southern branch line with Fault 3, the
displacement on Fault 4 is at a minimum (Fig. 3). The throw is in
the order of 15-30 m, and a large area with SGR of 15-20% is
observed, implying that there is a likelihood of having no seal, or
a slight static seal across the fault. In the area of low SGR
values, the Tarbert Fm is juxtaposed above the OWC in a re-
stricted area only. This could explain the slight differ- ences in
OWC between the two compartments (B South and Omega South). With
respect to the aquifer, it is likely that the B structure is in
communication with the Omega South structure, because the area of
juxtaposition is increased. The water gradient in well 30/9-7 is
almost equal to the gradients in well 30/9-8 and 30/9-10.
The gas in the B North compartment is separated
-
120 T. Fristad, A. Groth, G. Yielding and B. Freeman
El
b
Fig. 13. Strike projections of Fault 1, using a detailed
(65-layer) stratigraphic template (cf. Fig. 10). (a)
Juxtapositions. Upthrown units are colour- filled, downthrown units
shown outlined. The colour-coding is: yellow, 40% shale. (b) SGR.
In com- parison with Fig. 10b, note low values on upper part of
fault, and increased variability at south (fight) end of the fault
where the throw decreases to zero.
from the Omega North structure by a 5 bar pressure difference,
and on this part of the fault the SGR is about 24%.
Fault zone 5 and 6
This fault zone most likely explains a difference in OWC of
about 30 m between Omega North and
Omega South (Fig. 3). Unfortunately the seismic data quality is
poor, and the definition of the details of the zone is difficult to
elaborate. Consequently the two mapped faults separating the
structures were analysed together. Where the throw on one fault
decreases, the throw on the other increases accordingly. This
obser- vation strongly suggests that this pair of faults devel-
oped simultaneously and partitioned the displacement
-
Quantitative fault seal prediction 121
between them. It might therefore be expected that the overlap
zone has distributed strain, and may have unresolved small faults
linking the two main faults.
The SGR variations along-strike where units within Tarbert Fm
overlap are between 15 and 20% and minor faults linking the two
main faults can be expected to have a SGR profile similar to that
seen in the tip regions of the mapped faults (between 15 and 17%).
The observed kinematic linkage of the large faults, in combination
with a consideration of the SGR, therefore leads to an
interpretation of the fault zone as being able to support a small
differential pressure (less than 1 bar) in the area of Tarbert
juxta- position. This is sufficient to explain the differences in
OWC observed in wells 30/9-8 and 30/9-10.
The faults described above all have wells located on either side
of the fault. In Fig. 15, a summary of the SGR versus across-fault
pressure difference is plotted for critical oil and gas values
along the differ- ent analysed faults. One fault can have several
values depending on how many compartments are present on each side
of the fault. For example the points for Fault 4 represent the
segments where Omega North is juxtaposed against B North and B
South, respec- tively. The purpose of compiling this essential
infor-
133 10
8 - C
- ~
a 6 - -
r 4"~
&. -
~ 2- - -
o o
o
1
4-
3m rq
34
3s
r---] 4s 5/6
2 ~
' I "" I '
--F gas
Q oil
I--I water I
10 20 30 Sha le gouge rat io (%)
40
Fault 1. G-Central (30/9-14) against G-East (30/9-13S) Fault 2.
intra-G-Central (-14 DST) Fault 3. B-North (-4S) against B-South
(7) Fault 3m. G-Central (-14) against B-South (7) Fault 3s. G-East
(-13S) against B-South (7) Fault 4. B-North (-4S) against Omega
North (-3,-3A) Fault 4s. B-South (-7) against Omega North (-3,-3A)
Faults 5/6. Omega North (-8) against Omega South (-10)
Fig. 15. Summary diagram indicating the observed relationship
be- tween SGR and across-fault pressure difference for all analysed
faults in the study area. Each point represents the "critical" part
of the fault surface, i.e., maximum pressure difference for small
SGR values.
mation in one figure is to make predictions for faults where
sufficient well control points are lacking. For the faults
described below, SGR distributions were used to predict likely seal
capacities and therefore constrain the occurrence of hydrocarbons
in undrilled compartments.
Fault 7
An E-W syncline defines a separate closure north of the 30/9-14
well. Analysis of Fault 7 was conse- quently performed in order to
conclude whether a HC-column could be trapped in the hangingwall to
the B West structure or not (Fig. 3).
The fault has its minimum displacement (ca. 75 m) where it
branches with Fault 3. In this area the SGR is just below 20% or
higher, and by analogy with Fault 1, the potential for having a
trapped HC-column at the extension of G Central is good. In
addition, a gas column is more likely to be present rather than an
oil column, increasing the possibilities for a static seal.
The B West structure itself has not been tested by wells, but by
analogy with Fault 3 (separating wells 30/9-4S and 30/9-7), a
differential pressure across the fault between B West and B North
could be antici- pated as the throw is of the same order of
magnitude.
Fault 8
This fault is an example of a number of cross- faults
intersecting Omega South (Fig. 3). The throw decreases towards the
NW, and along-strike the dif- ferent units within the Tarbert Fm,
except Lower Tarbert Fm, are self-juxtaposed. Where Middle Tar-
bert 2 and Upper Tarbert Fms are juxtaposed, the SGR varies between
15 and 20% possibly introducing small pressure differences across
the fault. Because the area south of well 30/9-10 is intersected by
sev- eral of these NW-SE striking faults, it is likely that
different HC-contacts could be present. The differ- ence in OWC
across each separate fault, however, is probably not more than
10-15 m (or about 0.5 bar).
Fault 9
In light of the poor seismic data quality on the C and J
structures the SGR calculations in these areas should be treated as
guidelines rather than an exact definition of the individual
faults. Fault 9 was se- lected in order to focus upon internal
segmentation in the J structures (Fig. 3).
The R2A unit of the Lower Ness Fm consists of a sheet of
relatively homogeneous and clean channel sands. The thickness
variation of the sheet is around
-
122 T. Fristad, A. Groth, G. Yielding and B. Freeman
Fig. 14. (a) The juxtaposition profile of Fault 2 (upper left)
shows that the lowest value of throw is located at the centre of
the fault (ca. 15 m). (b) SGR (lower fight) below 15% are found
where the Middle Tarbert 2 unit is self-juxtaposed. This area of
low SGR is not likely to behave as a pres- sure barrier. Note that
the low SGR is found in the upper part of the interval as
illustrated in Fig. 9a.
Fig. 16. The SGR for Fault 9 reveals that the lowest values are
found in the lower part of the Brent Group. The weakest point with
respect to leak would consequently be expected to be found in the
lower one third of the Brent Gp, whereas the upper two thirds would
be expected to seal well.
-
Quantitative fault seal prediction 123
10-20 m where present. In addition, the shale per- centage is as
low as 8%. Consequently, where the sands are juxtaposed against
themselves, clay smear- ing is probably absent and here, cataclasis
is more likely than in the western area.
The calculated SGR is generally above 20%, ex- cept for the R2A
unit, where it is less than 15% in areas of small throw. If the SGR
thresholds that we have obtained on the G structures are
representative for the C structure, the upper two-thirds of the
Brent Group juxtaposition are expected to seal well, and the lower
part would be open to flow (Fig. 16). Note that this is in contrast
with the faults in the western area, where seal is poorest on the
upper parts of the faults (Tarbert Fro).
Conclusions
A fault-surface attribute called the shale gouge ra- tio (SGR)
has been defined for calculation of clay smearing in the
heterogeneous Brent Group sequence. The attribute corresponds
simply to the percentage of shale in the slipped interval.
Furthermore a method- ology for incorporating this fault related
attribute into the evaluation of sealing properties has been imple-
mented in the Oseberg Syd area. The SGR is variable over the fault
surface, varying as the displacement changes and depending on the
lithology of the wall rocks. Therefore the predicted sealing
properties vary over the fault surface. The following observations
are seen.
Western area
Throughout the western part of Block 30/9, clay smearing and
sealing by juxtaposition seem to be the main contributors to static
seal. In light of the growth observed across most of the faults in
this region, such a conclusion seems appropriate. However, the ob-
served range of SGR, from non-seal to considerable static seal, is
extremely tight, but remains quite con- sistent in light of the
fluid contacts and pressure data in the 30/9 wells. Seal capacities
for the individual faults are plotted in Fig. 15. Note that the
highest seal capacities are observed where the gas is the higher
pressure phase. The following is observed:
SGR < 15% 15% < SGR < 18%
SGR > 18%
no seal expected slight seal expected (
-
124 T. Fristad, A. Groth, G. Yielding and B. Freeman
terpretation and fault sealing investigations, Nun River Field,
Ni- geria. Am. Assoc. Pet. Geol. Bull., 73: 1397-1414.
Engelder, J.T. 1974. Cataclasis and the generation of fault
gouge. Geol. Soc. Am. Bull., 85: 1515-1522.
Freeman, B., Yielding, G. and Badley, M.E. 1989. Fault
correlation during seismic interpretation. First Break, 8:
87-95.
Fulljames, J.R., Zijerveld, L.J.J., Franssen, R.C.M.W., Ingram,
G.M. and Richard, P.D. 1997. Fault seal processes. In: P. M~ller-
Pedersen and A.G. Koeslter (Editors), Hydrocarbon Seals: Im-
portance for Exploration and Production, NPF Special Publication 7.
Elsevier, Singapore, pp. 51-59.
Gibson, R.G. 1994. Fault-zone seals in siliciclastic strata of
the Co- lumbus Basin, offshore Trinidad. Am. Assoc. Pet. Geol.
Bull., 78: 1372-1385.
Knipe, R.J. 1992. Faulting processes and fault seal. In: R.M.
Larsen, H. Brekke, B.T. Larsen and E. Talleraas (Editors),
Structural and Tectonic Modelling and its Application to Petoleum
Geology. NPF Special Publication 1, pp. 325-342.
Lindsay, N.G., Murphy, F.C., Walsh, J.J. and Watterson, J. 1993.
Outcrop studies of shale smear on fault surfaces, Special Publica-
tion 15. Int. Assoc. Sediment, pp. 113-123.
Mandl, G., de Jong, L.N.J. and Maltha, A. 1977. Shear zones in
granular material. Rock Mech., 9:95-144.
M~ller, N.K. and van der Wel, D. 1997. High resolution sequence
stratigraphy as a basis for 3D reservoir modelling: a case
study
from the southern Oseberg area. In: K.O. Sandvik, F. Gradstein
and N. Milton (Editors), Predictive High Resolution Sequence
Stratigraphy, NPF Special Publication, in press.
Needham, D.T., Yielding, G. and Freeman, B. 1996. Analysis of
fault geometry and displacement patterns, Special Publication 99.
Geol. Soc. London, pp. 189-199.
Roberts, A.M., Yielding, G., Kusznir, N.J., Walker, I. and Dorn-
Lopez, D. 1993. Mesozoic extension in the North Sea: constraints
from flexural backstripping, forward modelling and fault popula-
tions. In: J.R. Parker (Editor), Petroleum Geology of Northwest
Europe: Proc. 4th Conf., Geol. Soc. London, pp. 1123-1136.
Roberts, A.M., Yielding, G., Kusznir, N.J., Walker, I. and Dorn-
Lopez, D. 1995. Quantitative analysis of Triassic extension in the
northern Viking Graben. J. Geol. Soc., 152: 15-26.
Schowalter, T.T. 1979. Mechanics of secondary hydrocarbon migra-
tion and entrapment. Am. Assoc. Pet. Geol. Bull., 63: 723-760.
Watts, N. 1987. Theoretical aspects of cap-rock and fault seals
for single- and two-phase hydrocarbon columns. Mar. Pet. Geol., 4:
274-307.
Yielding, G., Badley, M.E. and Roberts, A.M. 1992. The
structural evolution of the Brent province. In: A.C. Morton, R.S.
Haszeldine, M.R. Giles and S. Brown (Editors), Geology of the Brent
Group. Geol. Soc. Special Publication 61. pp. 27-43.
Yielding, G., Freeman, B. and Needham, D.T. 1997. Quantitative
fault seal prediction. Am. Assoc. Pet. Geol. Bull., 81:897-917.
T. FRISTAD A. GROTH G. YIELDING B. FREEMAN
Norsk Hydro, P.O. Box 200, N-1321 Stabekk, Norway Norsk Hydro,
P.O. Box 200, N-1321 Stabekk, Norway Badley Earth Sciences Ltd.,
North Beck House, North Beck Lane, Hundleby, Spilsby, Lincolnshire,
PE23 5NB, UK Badley Earth Sciences Ltd., North Beck House, North
Beck Lane, Hundleby, Spilsby, Lincolnshire, PE23 5NB, UK