SPE 68199
Modelling water breakthrough in a fractured carbonate gas reservoir
Alexander Weber and Franz Brauckmann, BEB Erdgas und Erdoel, Hannover, and Lex Rijkels and Stephen J. Bourne,
Shell International Exploration and Production, The Hague
Copyright 2001, Society of Petroleum Engineers Inc.
This paper was prepared for presentation at the 2001 SPE Middle East Oil Show held in
Bahrain, 17–20 March 2001.
This paper was selected for presentation by an SPE Program Committee following review of
information contained in an abstract submitted by the author(s). Contents of the paper, as
presented, have not been reviewed by the Society of Petroleum Engineers and are subject to
correction by the author(s). The material, as presented, does not necessarily reflect any
position of the Society of Petroleum Engineers, its officers, or members. Papers presented at
SPE meetings are subject to publication review by Editorial Committees of the Society of
Petroleum Engineers. Electronic reproduction, distribution, or storage of any part of this paper
for commercial purposes without the written consent of the Society of Petroleum Engineers is
prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300
words; illustrations may not be copied. The abstract must contain conspicuous
acknowledgment of where and by whom the paper was presented. Write Librarian, SPE, P.O.
Box 833836, Richardson, TX 75083-3836, U.S.A., fax 01-972-952-9435.
Abstract
This paper describes a Permian evaporite carbonate gas field
that has been plagued by severe water problems since it was
put on production in the mid-80s. The field is a carbonate
complex consisting of three communicating reservoirs that are
heterogeneously fractured. Matrix permeability is typically
less than 2 mD, except in one highly permeable streak, where
it can be as high as 5 D. The permeability of the natural
fracture network is extremely heterogeneous, and varies by up
to a factor 100 over the field. Complex interaction between
fractures, matrix and the highly permeable streak caused a
surprising pattern of water breakthrough, which can be
explained by a geomechanical model for the heterogeneous
natural fracture network. This predictive, field-wide fracture
model was validated and constrained by both geological and
flow data.
First, the stress distribution around seismically visible faults
was calculated assuming homogeneous, isotropic, linear
elastic rock mechanical properties, and frictionless faults.
Second, the calculated stress field was used to simulate the
growth of discrete fracture networks, which were constrained
by statistically comparing fracture orientation and connectivity
with that derived from core, BHI, PLT, mud loss, and well test
data. Finally, the fracture networks were upscaled dynamically
to the grid of a dual-permeability simulator, enabling field-
scale multi-phase reservoir simulation. The flow model
obtained this way matched historical production data from all
wells. It also explained the source of water breakthrough and
the inflow profile seen on PLTs. Integrating seismic, borehole,
well test and production data to constrain and validate such a
field-wide model considerably reduced the uncertainty in the
final predictions.
This integrated, predictive fracture model is presently used to
investigate future field development scenarios. To this end, the
model is coupled to a surface network simulator, which
comprises the whole infrastructure. The fully coupled surface
and subsurface models offer the flexibility to optimally plan
the position and timing of new wells, the size of compressor
units, additional in-field trunk lines and the gas offtake.
Introduction
The field consists of three communicating, carbonate reser-
voirs that are heterogeneously fractured (Figure 1). Each reser-
voir consists of a dolomite package, sandwiched between two
calcite layers. A layer with low matrix porosity and
permeability is located in the centre of the dolomite package,
and a highly permeable streak of some metres thickness sits
between the top calcite and dolomite in only one part of the
reservoir. Figure 2 shows the matrix layout, and the well
positions. The permeability of the matrix is typically less than
2 mD, except in the streak, where it can be as high as 5 D.
Some wells penetrating this streak experienced water
breakthrough in a most unusual way. Water advanced through
the streak, but rained out into the fractures below before
reaching the wells. Only after some years of production did
this fractured matrix become so saturated that water finally
entered the well perforations. The timing of water
breakthrough in these wells depends on the volume of
fractured rock below the streak. Elsewhere in the reservoir the
highly permeable streak is absent, and water breakthrough
occurs via coning of bottom water, if it occurs at all. Since
water production has a high impact in a sour gas environment,
knowledge of the fracture distribution and their flow
properties is essential for field development.
Fractures are observed in several wells (in cores and
FMIs). Some of these are drilling-induced; some are cemented
or open natural fractures. The open natural fractures influence
flow significantly, as they are more permeable than the matrix
by orders of magnitude. Wherever they form a connected
network, they constitute a dual-permeability system, with the
matrix slowly feeding gas to the highly permeable fractures. A
well that taps such a fracture network effectively increases its
perforations to the size of the network.
Isolated fractures may occur outside the connected fracture
networks. These fractures form local shortcuts (they act as
2 A. WEBER, F. BRAUCKMANN, A. RIJKELS, AND S.J. BOURNE
very long pores) and effectively increase the matrix
permeability. In this field they add about 5-10 mD to the
matrix permeability, which is about a factor 5 compared to the
measured φ-k relation based on plugs. Isolated fractures play
an important role for ultimate recovery, but they do not
modify the recovery mechanism.
The objective of the modelling study was to predict where
the isolated fractures turn into a full dual-permeability
connected fracture system with a drastically different flow
pattern. We addressed this challenge using geomechanical
modelling techniques to estimate the reservoir stress field
responsible for the present permeable fracture system.
Fracture prediction
Natural fractures form in response to stress changes that
result from reservoir deformation. Wherever these stress
changes exceed the failure strength of reservoir rock, fractures
initiate and grow. If the stress change is large enough, the
fractures will connect and form a dual-permeability network.
The size of such a network depends on the distribution of
stress and rock strength within the reservoir1,2.
One mechanism that causes stress perturbations is large-
scale faulting. Slip along a fault surface causes rock strain and
hence a stress change. The simplest approach to determine the
stress field within a faulted reservoir is to assume the rocks
behave as a homogeneous, isotropic, and linear-elastic
material, and the faults as surfaces free of shear stress. In this
model large-scale mechanical heterogeneity is represented
only in the form of three-dimensional fault geometries (Figure
3).
We calculate the distribution of slip over the fault network
by loading it according to the remote stress that caused the
faults to slip. The orientation of this remote stress is estimated
from the regional geological history3, and the magnitudes
according to the mean rock strength prior to faulting. A unique
pattern of fault slip exists which releases all shear stress
resolved onto the faults by the remote stress and other nearby
faults.
Numerical solutions for the elastic stress field are obtained
using a three-dimensional boundary element analysis called
Poly3D4. Poly3D solves the equations of linear elasticity by
representing fault surfaces with a series of triangular elements,
each of constant slip5,6. This approximation permits solutions
for complicated three-dimensional fault geometries and slip
distributions by using a large number of triangles to represent
the fault geometries found in seismic data (Figure 4).
Brittle fractures form where the stress field exceeds the
local material strength as characterised by the brittle failure
envelope for both tensile and shear failure. We determine
brittle failure states within the reservoir according to the
Griffith7 stress criterion for tensile failure, and a Coulomb8
stress criterion for shear failure. Whichever criterion is met
first as the stress field changes determines the mode of first
failure.
Figure 5b shows the distribution and tensile failure
associated with the elastic stress field around the seismically
SPE 68199
imaged faults in the field, as shown in Figure 3. The lateral
extent of fractured regions depends on the rock strength
relative to the remote driving stress: increased rock strength
results in fractured areas shrinking towards the faults, and vice
versa. Fracture formation mechanisms such as fluid pressure
increase, diagenesis, and erosion of the overburden appear as
body forces, which produce an isotropic reduction in
compressive stress within the calculations. This has the effect
of shifting all the stress states plotted in Figure 5a towards the
left and consequently increasing the lateral extent of both
tensile and shear failure.
Once the distribution and mode of failure has been
determined the actual geometry of fracture networks
responsible for permeability can be simulated by growing
fractures in the brittle-elastic stress field. During growth,
fracture spacing and interaction are controlled by a forbidden
zone around each fracture - this represents an overall reduction
in local stress due to the presence of the fracture. As a result
other fractures are less likely to nucleate within these zones,
which limits the ultimate spacing of fractures. Furthermore,
when the propagating tips of two fractures approach each
other their mechanical interaction may lead to the fractures
turning towards each other and connecting (fracture hooking)9-
11
. This is an important mechanism affecting fluid flow as it
produces connectivity between essentially parallel fractures.
The discrete fracture networks that are grown this way
(Figure 6 shows an example) were calibrated by comparing
them to the fracture observations on FMIs and cores, and mud
loss data. The statistics from well observations constrain the
predicted distribution and orientation of fractures on a local
scale.
Flow simulation
Calibration of the predictive model on a field scale was
obtained through history matching dynamic data: water cut,
pressure build-up measurements and tubing head pressures.
These data were used to constrain rock strength and stress
orientation on the inter-well scale. By simulating the historical
gas production from the field, and monitoring how well the
simulated associated production data fit historical
measurements, the quality of each fracture model was
assessed. If a model based on one particular set of rock
properties and remote stress state did not match historical data,
even if other reservoir properties were varied within their
uncertainty range, the model was discarded as invalid. The
requirement that the field-wide model matches production data
from all wells, makes this route a powerful filter.
To simulate flow in the field, we upscaled the discrete
fracture network model dynamically to a dual-permeability
simulator. Using MaficOil12, a finite-element simulator for
single-phase flow, we determined the effective permeability of
the fracture network within each simulator grid block. Within
each gridblock-sized segment of the three-dimensional
fracture network, the matrix is discretised by three-
dimensional tetrahedral elements and discrete fractures are
represented by a parallel plate model at the faces of these
SPE 68199 MODELLING WATER BREAKTHROUGH IN A FRACTURED CARBONATE GAS RESERVOIR
elements. No-flow boundary conditions are specified for the
faces of grid block whose normal vector is perpendicular to
the flow direction being calculated (Figure 7).
In addition to the effective permeability, several other
static, parameters are computed: the number of matrix blocks
in each grid block, the imbibition length, the permeability
tensor etc. These parameters are used to check alignment of
the simulator grid and fracture system, and to model water
imbibition from fractures into the matrix. The imbibition rate
was an important match parameter in the field, as it
determined how fast the fractures below the highly permeable
streak can take up the water raining from this streak. By
relating the fracture intensity to the calculated stress field, we
computed the fracture spacing grid depicted in Figure 8, which
gave a match in water breakthrough in the Western wells.
Breakthrough in other wells, the tubing head pressures and
reservoir pressure communication were subsequently matched
by varying the aperture of the fractures (fracture aperture was
treated as a single, field-wide, constant parameter). Figure 9
shows the results for a typical well.
The requirement to match all geological and production
data with a single geomechanical fracture model constrained
the match parameters in the predictive fracture model very
effectively. For a given fault network and remote stress
orientation these parameters are:
- relative rock strength and body force (2 parameters),
- fracture aperture (1 parameter), and
- fracture spacing versus stress intensity (1 parameter).
Sensitivity studies demonstrated that he ranges of these
parameters that are consistent with all data is very narrow, and
as a result the uncertainty in production forecasts can be
reduced considerably.
Field development and forecasts
The history matched simulation model now forms the basis
for extensive field development planning. To this end, the
subsurface simulation model was coupled to a surface network
simulator, describing the field's piping and facilities, such as
separators, gathering centres, compressors etc. For each
timestep in forecasting mode, this integrated Hydrocarbon
Field Planning Tool (HFPT)13 derives a solution for the
pressure distribution within the surface network, and
calculates the corresponding (multiphase) flow from the
reservoir to the delivery point.
A snapshot of the surface network's pressure state is given
in Figure 10. Apart from the wells producing from the
carbonate reservoir, two other satellite fields feed gas to the
delivery point. It is this point that puts a constraint on the
network; be it a fixed delivery pressure, a contracted delivery
rate or a certain gas quality requirement. The simulated off-
take from the reservoir is controlled by opening or beaning
back wells such that the network constraints and delivery
requirements are fulfilled.
In our example, the delivery point constitutes a sour gas
purification plant, which has a fixed inlet pressure. The last
pipeline before the plant is a bottleneck in the system, since its
3
throughput capacity is restricted to 110000 m³/h maximally.
The goal is now to assure optimal pipeline usage by putting
new wells on stream, re-perforating and re-completing existing
wells and determining the optimal point in time for
compression. In fact, the HFPT model does this itself by
introducing relevant capacity enhancement measures from a
given development list. Figure 11 displays a possible solution:
initially, until 2002, the system's capacity is under-used, which
triggers the re-completion of well W3 in early 2001, and the
putting on stream of well W1 and W5 in early 2002. These
measures suffice to exhaust the pipeline's capacity until 2006,
when W6 is put on stream to fill the emerging ullage. Lastly, a
compressor is activated mid 2009, and the whole system is
capable to keep its plateau rate until early 2011, when natural
decline sets in.
Because of its flexible surface network capabilities and the
tight link to the subsurface reservoir simulator, HFPT offers a
platform to quickly model various offtake scenarios and to
evaluate the impact of future in-field measures, like timing of
a drilling sequence, pipeline debottlenecking (loops in the
figure), design and timing of compressor units, economic
screening of different gas contracts etc. In combination with
the fractured reservoir simulation model, its relevance for
BEB's integrated asset development has been increasing
steadily.
Conclusions
The fracture model for the fractured carbonate gas field
presented here uses geomechanical methods to predict the
field-scale distribution of fractures that affect flow with
reservoir simulations. Structural geometry governs the
distribution of stress responsible for fracturing and can be used
to constrain stress calculations. The field of stress determines
the likely geometry and distribution of fracture networks
which can be determined from models of fracture initiation,
growth, and interaction. The lateral extent and permeability of
fracture clusters are sensitive to uncertainties in both rock
strength and hydraulic fracture aperture. These uncertainties
can be significantly reduced using well test data and
production history to validate and constrain flow simulations.
By modelling the physical processes responsible for
fractures and flow, more meaningful and realistic fracture
systems can be predicted. Moreover, uncertainty can be
minimised by integrating all the available static and dynamic
data. As the model parameters are field-scale (i.e. mean rock
strength, remote stress, etc.), information from each well
constrains the whole fracture model and not just the areas
close to wells. This makes the model suitable for fracture
prediction and flow forecasting in all parts of the reservoir and
not just those parts around existing wells.
Access to a predictive field-scale fracture model allows
enhanced field development of naturally fractured reservoirs.
Understanding and quantification of the fracture system helps
to mitigate risks via improved flow forecasting, and optimal
well placement and design.
4 A. WEBER, F. BRAUCKMANN, A. RIJKELS, AND S.J. BOURNE
In combination with the Hydrocarbon Field Planning Tool
HFPT, the model offers a platform to carefully plan and
evaluate possible future development opportunities and as
such is an important ingredient for modern gas development
planning.
Acknowledgements
This work benefited greatly through the involvement and
ideas of Jean Borgomano, Tony Cortis, Arnout-Jan Everts,
Karen Foster, Joel Ita, Bettina Kampman, Robin Leinster,
Steve Livera, Thomas Mauduit, Dick Nieuwland, Steve Oates,
Pascal Richard, Roeland Roeterdink. We thank Stanford
University Rock Fracture Project for access to Poly3D.
References
1. Bourne, S.J., F. Brauckmann, A. Rijkels, B.J. Stephenson, A.
Weber and E.J.M. Willemse. "Predictive modelling of naturally
fractured reservoirs using geomechanics and flow simulation",
paper ADIPEC-0911, presented at the 9th ADIPEC, held in Abu
Dhabi, 14-18 October, 2000.
2. Bourne, S.J., Rijkels, A., Stephenson, B.J., and Willemse,
E.J.M. "Predictive modelling of naturally fractured reservoirs
using geomechanics and flow simulation", GeoArabia, 6 (2001)
p.87-102.
3. Grote, R. "Maximum Horizontal Principal Stress Direction in
Rotliegendes and Upper Carboniferous in North Germany."
Erdöl, Erdgas, Kohle, 10 (1998).
4. Thomas, A.L. "Poly3D: A three-dimensional, polygonal ele-
ment, displacement discontinuity boundary element computer
program with applications to fractures, faults, and cavities in the
Figure 1. Depth map of the fractured, carbonate field. The three
communicating gas reservoirs are delineated by NW-SE trending
faults (indicated by black lines).
SPE 68199
Earth’s crust." Masters dissertation, 1993, Stanford University,
California 94305-2115.
5. Comninou, M.A., J. Dunders. "The angular dislocation in a half-
space." Journal of Elasticity, 5 (1975) p. 216.
6. Jeyakumaran, M., J.W. Rudnicki, L.M. Keer. "Modeling slip
zones with triangular dislocation elements." Seismic Society of
America Bulletin, 82 (1992) p. 153-169.
7. Griffith, A.A. "The phenomena of rupture and flow in solids."
Philosophical Transactions of the Royal Society, London, A 221
(1921) p. 163-198.
8. Coulomb, C.A. "Sur une application des règles de Maximis et
Minimis a quelques problèmes de stratique relatifs à
l’Architecture." Académie Royal Des Sciences Memoires de
mathématique et de physique par divers savans, 7 (1773) p. 343-
382.
9. Cruikshank, K.M., G. Zhao, A.M. Johnson. "Analysis of minor
fractures associated with joints and faulted joints." Journal of
Structural Geology, 13 (1991) no. 8, p. 865-886.
10. Olson, J.E., D.D. Pollard. "The initiation and growth of en
échelon veins." Journal of Structural Geology, 13 (1990) no. 5,
p. 595-608.
11. Thomas, A.L., and D.D. Pollard. "The geometry of echelon
fractures in rock: implications from laboratory and numerical
experiments." Journal of Structural Geology, 15 (1993) no. 3-5,
p. 323-334.
12. MaficOil. A special purpose version of Mafic. See:
www.golder.com for a detailed description.
13. Beliakova, N., J.T. van Berkel, , G.J. Kulawski, A.M. Schulte,
A.J. Weisenborn. "Hydrocarbon Field Planning Tool for me-
dium to long term production forecasting from oil and gas fields
using integrated subsurface-surface models". SPE 65160,
presented at SPE EUROPEC 2000, 24-25 October, Paris,
France.
SPE 68199 MODELLING WATER BREAKTHROUGH IN A FRACTURED CARBONATE GAS RESERVOIR 5

Figure 2. Perspective view of the field showing matrix porosity of the top layer. Well positions are indicated by circles and fractures observed
on borehole images in these wells are indicated in stereoplots. The inset shows a highly permeable streak in the western part of the reservoir,
occurring between the top calcite layer and the dolomite (colours of the inset indicate streak permeability, which varies from 0 to 5 D).

Figure 3. Perspective view of the faults used to compute the stress perturbations related to fault slip. Fault surfaces were extruded from
traces imaged in seismic data to yield vertical faults each of equal height.
6 A. WEBER, F. BRAUCKMANN, A. RIJKELS, AND S.J. BOURNE SPE 68199

Figure 4. The field of elastic stress computed in relation to the fault network geometry in Figure 3. In this example the mean value of the least
and most principal compressive stresses is shown – red and green colours denote areas of increased and decreased compressive stress
respectively.

Figure 5. One example of brittle failure computed from the elastic stress field shown in Figure 4. Failure is determined according to a
combined Griffith7—Coulomb8 criterion which depends solely on the magnitude of the most and least compressive stresses (σ1 and σ3
respectively) as shown in (a). The lateral extent of brittle failure (b) depends on the magnitude of reservoir rock strength as compared to the
remote stress load – increased rock strength yields less extensive fracturing localised closer to faults.
SPE 68199 MODELLING WATER BREAKTHROUGH IN A FRACTURED CARBONATE GAS RESERVOIR 7

Figure 6. One example of a discrete fracture model grown within the brittle-elastic stress field shown in Figures 4 and 5. The enlargement
illustrates effects due to the fracture interaction rules used to simulate fracture initiation and growth: (i) an approximate regular spacing
commensurate with bed thickness due to a forbidden zone around each fracture, and (ii) the oblique hooking between closely spaced
propagating fracture tips.

Figure 7. Flow-based upscaling is used to translate the discrete fracture network into a dual-porosity grid. For each grid block, the
corresponding section of the fracture network is subjected to a single-phase, finite-element flow simulation in three orthogonal directions.
The resulting values for effective permeability of the fracture network are allocated to the dual-permeability grid of the reservoir simulator.
8 A. WEBER, F. BRAUCKMANN, A. RIJKELS, AND S.J. BOURNE SPE 68199

Figure 8. Calculated fracture permeability (top) and fracture spacing (bottom), based on the fracture model in Figure 6. The aperture of
fractures is one of the match parameters in the simulation of field performance. Fracture spacing determines the imbibition rate of water, and
its relation with the computed stress field is another match parameter.

Figure 9. Simulated and historical water production rates of one well (top); pressures (bottom). The cumulative gas production was used as a
history match constraint. The associated water production consists of vapour (initial rise in cumulative water), and formation water (steep
increase in slope). The cross section shows how this formation water cones towards the perforation via the fracture system. Historical bottom
hole pressure data are based on yearly build-ups.
SPE 68199 MODELLING WATER BREAKTHROUGH IN A FRACTURED CARBONATE GAS RESERVOIR 9

Figure 10. Snapshot of pressure state in coupled surface-subsurface network. Circles indicate wells in the field (top), or supply from satellite
fields (bottom). Each well has its own free water knock-out unit (FWKO). Several gathering centres (GCs) lead the total gas stream to the
delivery point, which sets the constraint on the total simulation model.

Figure 11. One possible scenario, as simulated by the hydrocarbon field planning tool. Initially, until 2002, the system's capacity is under-
used, which triggers the re-completion of well 3, and the putting on stream of well W1 and W5 in early 2002. These measures suffice to
exhaust the pipe-line's capacity until 2006, when W6 is put on stream to fill the emerging ullage. Lastly, a compressor is activated mid 2009,
and the whole system is capable to keep its plateau rate until early 2011, when natural decline sets in.