IPTC 11199
Reactive Transport Models of Limestone-Dolomite Transitions: Implications for
Reservoir Connectivity
Yitian Xiao, ExxonMobil Upstream Research Company, and Gareth D. Jones,* ExxonMobil Development Company
*Present Address: Chevron Energy Technology Company, San Ramon, California, U.S.A
Copyright 2007, International Petroleum Technology Conference
This paper was prepared for presentation at the International Petroleum Technology
Conference held in Dubai, U.A.E., 4–6 December 2007.
This paper was selected for presentation by an IPTC Programme 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 International Petroleum Technology Conference
and are subject to correction by the author(s). The material, as presented, does not necessarily
reflect any position of the International Petroleum Technology Conference, its officers, or
members. Papers presented at IPTC are subject to publication review by Sponsor Society
Committees of IPTC. Electronic reproduction, distribution, or storage of any part of this paper
for commercial purposes without the written consent of the International Petroleum Technology
Conference 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, IPTC, P.O.
Box 833836, Richardson, TX 75083-3836, U.S.A., fax 01-972-952-9435.
ABSTRACT
Substantial volumes of world hydrocarbon resources occur in
interlayered limestone-dolomite reservoirs. Diagenetic
variations in lithology and primary depositional texture control
the magnitude and spatial distribution of petrophysical
properties. The frequency and nature of limestone-dolomite
transitions that define flow units and baffles/barriers are
critical for understanding reservoir connectivity and
optimizing field development.
Existing subsurface predictions are largely based on
observation and are occasionally linked to sequence
stratigraphy. This approach can be relatively successful at
predicting general trends in limestone and dolomite
occurrence, but there is considerable uncertainty in predicting
and correlating spatial variations in diagenetic styles at the
field scale.
Reactive transport models that explicitly couple fluid flow and
chemical reactions to facilitate quantitative investigations of
limestone-dolomite transitions are a recent addition to the
predictive diagenesis tool kit. The approach is illustrated with
a generic study designed to investigate the fundamentals of
brine reflux dolomitization. Results provide new insights on:
(1) Dynamic propagation of limestone-dolomite fronts by
fingering, (2) The connectivity of dolomite ‘fingers’ to the
‘main body’; (3) Dolomitization and anhydrite occurrence
(primary precipitation and cements); (4) Locating reservoir
‘sweet spots’ in a dolomite geobody and (5) The interaction
and importance of semi-regional versus local brine reflux
flow. This process-based approach, in combination with
available observational data and sequence stratigraphic
paleoenvironmental reconstructions, has significant
implications for predicting limestone-dolomite transitions, the
spatial distribution of petrophysical properties and viable
realizations of reservoir connectivity for flow simulation.
INTRODUCTION
Diagenetic transitions between limestone and dolomite are a
critical control on reservoir quality in many carbonate
reservoirs. The Jurassic Ghawar field in Saudi Arabia and the
Permian-Triassic North Dome field in Qatar and Iran are two
notable prolific Middle East examples. Dolomite can behave
as a reservoir or as a baffle / barrier to flow depending on
original depositional texture, style of dolomitization and
presence of anhydrite cement [1-3]. High permeability, or so-
called “Super K” dolomites with multi-darcy permeability that
sporadically occur in thin zones (typically <5 ft) are vital
contributors to individual well and overall reservoir
performance [2]. Laterally continuous low permeability ‘tight”
dolomites can also have significant impact on subsurface flow
and thus reservoir management strategies. The transition
between limestone and dolomite is readily identified in both
wireline logs and core and is dominantly observed as a sharp
interface described as a “dolomite front” sensu Wilson et al.
[4]. Unfortunately, dolomite fronts in carbonate reservoirs are
often below seismic detection, particularly those associated
with ‘thin’ beds of dolomite on the order of a few to several
tens of feet thick. Predicting the spatial distribution or
connectivity of dolomite and associated petrophysical
properties between wells and away from well control is thus a
fundamental challenge in carbonate reservoir characterization.
In contrast to depositional facies there are no defined rules or
guidelines for correlating dolomite bodies in the subsurface.
Existing correlations of dolomite tend to be lithostratigraphic
and contain considerable uncertainty in the degree of
interpreted reservoir connectivity (Figure 1).
Understanding the distribution and connectivity of
sedimentary bodies has been greatly enhanced from studies of
outcrop analogs [5-6]. A vast number of outcrop studies have
focused on dolomitized carbonates. However, relatively few
have specifically focused on transitions between limestone and
dolomite and / or described the connectivity of dolomite
bodies. Outcrop based studies by Wilson et al [4] and Yao and
Demicco [7] that mapped dolomite fronts associated with
geothermal circulation and ascending hydrothermal fluids,
respectively, are notable exceptions. These analogs are,
however, not appropriate for reservoirs where the majority of
dolomite is interpreted to have formed by the reflux of brines
2 !PTC 11199
from evaporated seawater, for example Ghawar and the North
Dome fields in the Middle East [2-3]. Dolomite fronts mapped
in Plio - Pleistocene outcrops on Bonaire by Lucia and Major
[8] are interpreted to have formed by brine reflux (Figure 2).
The Bonaire study provides insights on the propagation of a
dolomite front and the distribution of petrophysical properties
in a reflux system but is of limited use in helping develop
dolomite correlation strategies in extensive carbonate ramp
reservoirs like Ghawar and the North Dome. Other smaller
scale outcrop examples of limestone dolomite transitions
interpreted to have formed as a consequence of brine reflux
are shown in Figure 3.
Type 1: Dolomite concordant with stratigraphy
Type 2: Dolomite discordant with stratigraphy
0 10 km
Figure 1. Correlations of limestone to dolomite (shaded grey)
transitions from wireline well logs in the Ghawar Field, Saudi
Arabia. Note 1-2 km well spacing and vertical scale in ft. After
Cantrell et al. [2].
depositional dip
0 10 km
Figure 2. Limestone to dolomite transition mapped in a Plio
Pleistocene outcrop on Bonaire, Netherlands, Lesser Antilles.
Note dolomite ‘fingering’ in direction of depositional dip.
Modified after Lucia and Major [8].
(A) Limestone (B)
Dolomite
Dolomite
Limestone
(C) (D)
Limestone
Dolomite
Figure 3. Outcrop examples of limestone dolomite transitions. (A)
Devonian, Cripple Creek, Canada; (B) Ordovician, Franklin
Mountains, USA, after Tillotson [9]; (C) and (D) Ordovician,
Franklin Mountains, USA (Images courtesy of Jemery Bellian).
In all images Dolomite is tan (light brown) and limestone is grey.
Red arrow shows interpreted direction of brine reflux paleo fluid
flow.
Reactive transport models explicitly couple flow and chemical
reactions facilitate investigations of the distribution of
diagenetic alteration, including dolomitization, in the
subsurface [10-12]. Numerical models have the advantage of
being able to simulate the specific system of interest and to
track the spatial and temporal evolution of diagenesis in
contrast to the terminal diagenetic overprint observed in
outcrop studies [13-15].
In the past three years, we have applied reactive transport
models to investigate several styles of diagenesis in both
carbonate and siliciclastic rocks: A) Early diagenesis in a
freshwater lens and associated hydrological zones in an
isolated carbonate platform; B) Geothermal convection and
burial diagenesis in a salt buried isolated platform; C) Fault
induced hydrothermal fluid flow and illitization; and D)
Formation damage associated with steam (water) injection and
artificial diagenesis (Figure 4). Adopting a sensitivity analysis
approach, we examined how key natural variables, such as
climate, sea level, temperature, salinity and fluid composition
and porosity/permeability heterogeneity impact different styles
of diagenesis and reservoir evolution. We also examined how
formation damage caused by artificial diagenesis in production
can be predicted by reactive transport modeling. Linking
fundamental geochemical processes to predict reservoir
quality has significantly improved our understanding of
several conceptual diagenetic models and paradigms. Our
novel approach suggests that reactive transport models, when
sufficiently integrated with traditional methods and calibrated
with field data, have the potential to significantly improve our
ability to predict carbonate and siliciclastic reservoir quality
[11-12, 15-16].
In this paper we present results from reactive transport models
employed to investigate brine reflux dolomitization with an
emphasis on investigating the origin, distribution and
connectivity of dolomite bodies. We present a brief synthesis
of our previously published 2D reactive transport model study
IPTC 11199 3

of reflux dolomitization [11] followed by preliminary results For example, the reaction of dolomitization can be expressed
from 3D simulations that provide new insights on the as:
distribution of dolomite, anhydrite and implications for
reservoir connectivity and subsurface correlation strategies. ∂Cdolo − a
E
Q
= S A e RT ( − 1) 2.26 (2)
∂t K eq
(A) (B)
25
W 25
30
35 30
k h f i Where S represents the reactive surface area, A is the rate
constant, Ea is the activation energy, Q/Keq is the saturation
40
35
45 40

index, and 2.26 is the reaction order [17].

50 45

Porosity 55 50
45
60 60
65 55 50
70
Vadose
S.L. .2 km
calcite cement, % 50 Due to complex boundary conditions and complicated
Karst 2 km –2 0 2
5
cm/yr coupling between the transport and reaction terms, it is
(C) (D)
Calcite Cement
impossible to provide analytical solutions to equation (1) for
even the simplest geochemical system. Therefore numerical
Illite solutions have to be used. Fortunately, due to the exponential
increase in computational power, realistic reactive transport
models have been developed that are beginning to provide
exploitable insights to reservoir quality prediction. We have
used two commercially available numerical programs, Xt2
Illite kaolinite Kaolinite [18] for the 2D models and ToughReact [19] for the 3D
models. The rest of this paper highlights results from our
recent efforts in applying this emerging technology to the
quantitative prediction of the spatial and temporal distribution
Figure 4. Previous reactive transport modeling applications: (A)
Early diagenesis in an isolated carbonate platform; (B)
of dolomitization and its effect on carbonate reservoir quality
Geothermal convection and burial diagenesis in a salt buried and connectivity.
isolated platform; (C) Fault induced hydrothermal fluid flow and
illitization; and (D) Formation damage associated with water RESULTS
injection. After Xiao and Jones [12].
2D Simulations of Brine Reflux Dolomitization
NUMERICAL MODELS AND APPROACH
Our 2D model was based on the reflux model originally
A reactive transport model relies on a mathematical proposed by Adams and Rhodes [20] to explain dolomitization
formulation to describe geochemical processes involving of Permian carbonate in West Texas, which shows a close
fluid-rock interactions. The general governing equation can be stratigraphic association with platform interior evaporites.
written as: The reflux model has been modified by several researchers
[21-24] to incorporate different geological scenarios recently.
∂ ∂ ⎛ ∂C ⎞ ∂C ⎛ ∂C ⎞ (1) Our 2D simulation domain is 10 km x 500 m (Figure 6). The
(φC i ) = ⎜ φD i ⎟ − φv i + φ ∑ ⎜ i ⎟
∂t ∂x ⎝ ∂x ⎠ ∂x k ⎝ ∂t ⎠ k
simulation domain is divided by 10 x 5 nodes in x- and y-
directions. The upper and lower boundaries are specified as no
Where Ci is the concentration of a specific species in the pore flow (a constraint of current version of the Xt2 program) and
fluid, D is the combined diffusion and dispersion coefficient the left (platform interior) and right (basin ward) boundaries
term, v is the linear fluid flow rate, and ф is the porosity. The are open to flow. In our 3D simulation (discussed in the next
first two terms on the right hand side describe the transport section) the upper boundary is represented as a fluid pressure
process (diffusion, dispersion, and advection) while the last boundary allowing both the recharge and discharge of brines
term describes the effect of geochemical reactions (Figure 5). on a submerged platform-top [25]. Our simulated flow domain
thus represents the section of a carbonate platform where brine
recharge and predominantly horizontal reflux flow occurs as
demonstrated in the Grosmont platform [26]. Reflux flow,
Simulation tracks fluid flow which is from the left (platform interior) to the right (basin
and chemical reactions in wards) was set by specifying injection of brine at a rate of 5
Inflow porous media
Outflow
m3/s in the uppermost left cell (interior) of the flow domain.
The initial flow velocity in the majority of the flow domain is
modify porosity/permeability
Mg2+ Ca2+ 3-4 m/yr (at shallow depth) but increases to ~6 m/yr close to
the brine source. The initial flow velocity is slower (< 3 m/yr)
Initial rock property
in a zone that begins 150 m below the brine source and
Dolomitization: 2CaCO3 + Mg2+ = CaMg(CO3)2 + Ca2+ extends 5.5 km basin ward (Figure 6). The flow velocities in
this initial flow field are relatively high but consistent with
Figure 5. Schematic diagram showing the reactive transport previous simulations of reflux in carbonates in the shallow (<
process involving dolomitization. 500 m) subsurface [26].
4 !PTC 11199
Brine injection No Flow
node
Open
Open
Figure 6. Simulation flow domain, hydraulic boundary
conditions, simulation grid and location of brine injection node.
Initial distribution and magnitude of reflux fluid velocities
(contours are of fluid velocity in m/yr with a contour interval of
0.5). After Jones and Xiao [11].
To capture the heterogeneous porosity and permeability of
carbonate sediments in the shallow subsurface (0-500 m), due
to either variations in sediment texture or early diagenesis [27-
28], we incorporated heterogeneous porosity and permeability
distributions in our 2D models to investigate the effect of
heterogeneity on reflux dolomitization. The porosity is treated
as a field variable in the simulation. We set it to a random
distribution about 35%, with a standard deviation of 2.5% (i.e.,
porosity is 0.35 ± 0.25). We placed a top threshold of 35 % on
this porosity distribution to avoid uncharacteristically high
values of permeability that would promote numerical
instability. Sampling randomly from this truncated porosity
distribution generated a heterogeneous porosity distribution
that ranged from 29 to 35 % (Figure 7A). We specified the
Class 1 relationship [11] between porosity and permeability to
generate a corresponding heterogeneous permeability
distribution, which ranged from 12.8 to 126 Darcies (Figure
7B). The permeability anisotropy was held at 100.
To enhance the spatial heterogeneity and investigate smaller
scale features of the reflux process, we doubled the number of
nodes in the flow domain by decreasing the node spacing to
500m in the horizontal direction and 50 m in the vertical
direction (20x10 nodes in x- and y-directions). We reduced the
initial rate of reflux flow to a maximum of 0.9 m/yr (range 0.1
to 0.9 m/yr) to offset the increase in computation time (by a
factor of 8) associated with an increase in the number of grid
nodes (Figure 7C). Our simulated range of porosity and in
particular permeability heterogeneity, which can range up to
five orders of magnitude in shallow carbonate sediments [27-
28], is relatively moderate. No explicit coupling of multiple
deposition and diagenesis stratigraphic sequences
(parasequence stacking) was incorporated in our simulated
heterogeneity. The total simulation time is 2 m.y.
(A) Initial Porosity
(B) Initial Permeability
(C) Initial Fluid Velocity
Figure 7. Initial porosity, permeability and resultant flow field
used for investigating the effect of porosity and permeability
heterogeneity (see Figure 7). (A) Porosity (contours are porosity
[fraction] with a contour interval of 0.01; (B) Permeability
contours are of horizontal permeability [Darcys] with a contour
interval of 20); (C) fluid velocity (contours are of fluid velocity
[m/yr] with a contour interval of 0.5). After Jones and Xiao [11].
The simulation results demonstrated that the initial porosity
and permeability heterogeneity in the precursor sediment had a
significant effect on the pattern of dolomitization (Figure 8).
The applied heterogeneity results in the formation of
pronounced perturbations, in the geometry of the dolomite
front, which extended several km basin ward from the main
dolomite body at 2 My (Figure 8 A-E). These perturbations,
described as dolomite “fingers” by Wilson et al. [4], became
more pronounced with time suggesting a positive feedback
between the initial exploitation of preferential flow paths and
subsequent changes in rock properties by dolomitization that
enhance flow (Figure 8 A-E) as well as negative feedback
associated with anhydrite cementation (Figure 8 F-J).
The simulation results indicate the association of anhydrite
cementation and dolomite distribution was partially and
temporally (Figure 8 F-J). The initial anhydrite cement was
formed due to the interaction of SO42- from the brine with the
excessive Ca2+ released from the limestone due to
dolomitization. However, as the dolomitization completed
“upstream”, no more excessive Ca2+ was released, which led
the brine dipping below gypsum (anhydrite) saturation. This is
the primary reason for the remobilization of the initially
formed anhydrite cementation (Figure 8 F-J). Focusing of
IPTC 11199
reflux flow resulted in enhanced anhydrite dissolution in the
dolomite fingers, whereas anhydrite had a greater preservation
potential adjacent to the fingers (Figure 8 J). Dolomitization
enhances porosity by up to 8 % and anhydrite cements occlude
porosity by up to 14 % (Figure 8 K-O). The dolomite fingers
were porosity “sweet zones” surrounded by non- to partially
dolomitized sediments that were locally cemented with
anhydrite and have a lower porosity (Figure 8 K-O).
Dolomite Anhydrite Porosity
(A) 0.2 My (F) 0.2 My (K) 0.2 My
(B) 0.32 My (G) 0.32 My (L) 0.32 My
(C) 0.56 My (H) 0.56 My (M) 0.56 My
(D) 1 My (I) 1 My (N) 1 My
(E) 2 My (J) 2 My (O) 2 My
Figure 8. Effect of initial porosity and permeability heterogeneity
(see Figure 7) on dolomite, anhydrite and porosity (A-E)
Dolomite (contours are percent dolomite with a contour interval
of 5); (F-J) anhydrite (contours are percent anhydrite with a
contour interval of 2); and (K-O) porosity (contours are porosity
[fraction] with a contour interval of 0.02). After Jones and Xiao
[11].
Our 2D reactive transport simulations demonstrate the
complex and dynamic nature of dolomitization associated
anhydrite cementation and porosity evolution in a simplified
reflux system. Incorporation of initial porosity and
permeability heterogeneity resulted in a complex dolomite
geometry with pronounced dolomite “fingers” extending
ahead of the main dolomite body. The simulated dolomite
distributions seem to change from thick and extensive to thin
and separated in the direction of the reflux flow, consistent
with outcrop observations [Figures 2 and 3]. The predicted
distribution of limestone to dolomitization transition and
potential location of reservoir “sweet” spots can have
important implications on reservoir connectivity study. For
example, the extensive main dolomite body in the platform
interior (Figure 8 E) suggests good lateral and vertical
reservoir connectivity, whereas in the limestone to dolomite
transition zone the layered dolomite “fingers” and anhydrite
cement indicates complex reservoir connectivity, with good
reservoir intervals (uncemented dolomite) sandwiched by tight
zone with extensive anhydrite cemented (Figure 8 J, O).
3D Simulations of Brine Reflux Dolomitization
Geological models developed for simulating production of
hydrocarbons and reservoir management requires 3D
5
distribution of petrophysical properties. 2D reactive transport
models of dolomitization are limited by their ability to
incorporate spatial variations in natural variables that control
brine reflux and thus extrapolation to the distribution of
dolomite in 3D. We employed 3D reactive transport models to
investigate the effect of permeability and spatial distributions
in the magnitude and distribution of platform-top brines that
source subsurface brine reflux. Salinity variations observed in
modern environments, for example Lake Macleod [29] in
addition to those interpreted from the thickness and type of
evaporite deposits in the rock record were used to condition
model boundary conditions and initial rock and fluid
properties. We specified a flow domain 5 km x 5km x 100 m
with a grid node spacing of 200 m horizontally and 2 m
vertically. Reflux was simulated from three platform-top brine
ponds approximately 500 m in diameter with different
maximum salinities of 210, 175, and 140 ppt (Figure 9) The
basal boundary was specified as a no flow boundary (refluxing
brines never reached this depth in the geological time
simulated) and the sides of our flow domain cube were
specified as hydrostatic and thus open to flow. All other
parameters and boundary conditions were specified as those
used in our 2D reactive transport models of brine reflux that
incorporated a heterogeneous distribution of permeability
described above in this paper and [11].
Model results showing the 3D evolution of brine reflux up to
5000 years are depicted in Figure 9. Individual reflux brine
plumes sourced by the three different ponds, evident after 500
yrs of flow, eventually coalesce to form a single brine plume
(Figure 9). At 5000 years the mesosaline to hypersaline brine
plume extends over the 25 km2 area modeled to a maximum
depth of approximately 50 m but with considerable variation
in vertical extent related to brine pond proximity and local
permeability heterogeneity. Modeled maximum rates of brine
reflux were on the order of few meters per year with spatial
variations controlled by the proximity to the different brine
ponds and permeability. The resultant flow field is the net
product of both reinforcing and opposing flows from the three
separate ponds. In contrast to uni-directional flow in our 2D
results, 3D simulations demonstrate that refluxing bines, at
least locally in the vicinity of brine ponds, have the potential
to flow up and down depositional dip and along depositional
strike. After approximately 5000 yrs the brine plume appeared
to reach a quasi steady state relative to the effective hydraulic
head in the brine ponds that drives flow. Beyond 5000 years
advection is limited and the slower process of diffusion
controls brine plume spreading. It is unlikely that steady state
reflux would be obtained in nature due to variations in climate,
sea level and geomorphology that act at shorter time scales to
change the distribution of platform-top salinity that drives
flow.
The pattern of brine reflux observed in our 3D simulations
generated more complex distributions of dolomite and
anhydrite than observed in our 2D study and has significant
implications for understanding connectivity in carbonate
reservoirs (Figures 10 and 11).
6 !PTC 11199
50 yr 100 yr
• 5 km x 5 km x 100 m
• Three brine ponds
• Regional brine flow
• Salinity 4~5 x seawater
Salinity 5000 yr
200 yr
distribution
500 yr
Figure 9. 3D Simulation flow domain and boundary conditions
and distribution of brine ponds and RTM 3D distribution of
salinity and evolution of reflux brines at 50, 100, 200, 500, and
5000 yr. Sea water in black, hypersalinity brines in yellow and
orange, and mixed fluids in blue.
3D Evolution of Dolomite - ‘Main Body’ and ‘Fingers’
The evolution of dolomite, simulated over a period of 100 ky
for the reflux brine plume depicted in Figure 9, demonstrates
how initially small isolated dolomite bodies located beneath
and adjacent to the brine ponds grow and amalgamate to form
a single dolomite body that extends over the entire 25 km2
(Figure 10). Results show dolomite bodies grow at different
rates. In a similar manner to our 2D simulations the rate of
dolomite body propagation is a function of the Mg2+
concentration of the brine and the flow rate. In a reflux system
this is controlled by the salinity of the brine ponds, distance
from the pond and the magnitude of permeability (assuming a
uniform temperature and reactive surface area). Simulated
dolomite body growth occurs by the propagation and eventual
amalgamation of dolomite ‘fingers’ sense Wilson [4]. The
reaction front from limestone to dolomite is sharp.
Visualization of the 3D dolomite body after 100 kyr of reflux
shows the significant variation in the dimensions and shape of
dolomite “fingers” that extend form the “main” dolomite
body, here described as dolomite that is laterally continuous
over the areal extent of the flow domain (Figure 10).
Dolomite fingers thin in the direction of flow and tend to
become more pronounced with time depending on the
specified feedback of both dolomization and anhydrite
cementation on permeability and the initial permeability of
limestone the dolomitization front intersects along a reflux
flow path. Examination of model results using 2D slices show
dolomite bodies that are apparently isolated from the main
body (Figure 10). However, these bodies are connected to the
main body in 3D and were generated by the transport of Mg2+
along a flow path out of the 2D plane visualized (Figure 10).
Thus reflux associated with local brine ponds has the potential
to generate complex distributions of dolomite with ‘finger’
orientations that reflect multi directional fluid flow. This has
three important implications for subsurface correlation and
geological modeling of reflux dolomites: 1) Understanding the
processes of reflux and paleo fluid flow can provide a
framework for correlation scenarios and guidelines for
geological modelers 2) Reflux flow can be locally multi
directional with respect to depositional dip and strike and 3)
Dolomite “fingers” tend to connect to the “main body” in a
single reflux episode and are rarely isolated “pods”. The
occurrence of isolated dolomite pods is restricted to the early
termination of reflux as illustrated by the distribution of
dolomite at times less than 60 k.y. (Figure 10), multiple
episodes of brine reflux separated in time and space and
dolomitization by other mechanisms [12].
(A) (B) Dolomite
main body
20 ky 40 ky
Dolomite 60 ky
geobody Dolomite
fingers
100 ky
80 ky
Dolomite
main body
(D) (C)
Dolomite
fingers
Figure 10. RTM 3D distribution of dolomite: (A) Dolomite
evolution from 20 to 100 k.y.; (B)-(C) 2D slices showing the
dolomite main body and fingers in x- and y-directions; (D) 2D
slices showing dolomite in z-direction. Dolomite in red and
limestone in blue.
3D Evolution of Anhydrite Distribution (Bedded and
Cements).
Dolomites generated in a reflux system have a multi faceted
relationship with evaporites that are depositional and / or
occur as cements [11]. Our 3D simulations of brine reflux
dolomitization show two types of anhydrite: 1) Anhydrite
cements that precipitate ahead of a migrating dolomite front
due to the release of calcium as explained in our 2D study [11]
and 2) the direct precipitation of “bedded” anhydrite from
evaporated seawater (Figure 11). Bedded anhydrite is
distributed as a laterally continuous body (Figure 11) with rare
isolated “holes” that may act as potential fluid conduits
(Figure 11 D). Anhydrite cement has a more complex
distribution that is related to the distribution of dolomite.
Greater preservation of anhydrite cement is evident in the 3D
models of dolomitization (compare Figures 8 and 11). This is
due to the shorter duration of brine reflux simulated and the
precipitation of bedded anhydrites that reduce permeability,
IPTC 11199
retard flow and slow rates of anhydrite dissolution. Future
simulations will explore the relationship between dolomization
and anhydrite and the implications for identifying dolomite
reservoir sweet spots where anhydrite cements are rare /
absent.
(A) (B)
Anhydrite
bedding
20 ky 40 ky
60 ky
Anhydrite
geobody
Anhydrite
cement
100 ky
80 ky
Anhydrite
bedding
(D) (C)
Anhydrite
cement
Figure 11. RTM 3D distribution of anhydrite (bedded and
cement): (A) Anhydrite evolution from 20 to 100 k.y.; (B)-(C) 2D
slices showing the bedded anhydrite (red) and anhydrite segment
(green) in x- and y-directions; (D) 2D slices showing anhydrite in
z-direction. Thin section photos courtesy of Jim Markello.
High Permeability Dolomite Distribution
Understanding the distribution and connectivity of dolomite
bodies in the subsurface is the first step towards building more
accurate reservoir models. The magnitude and distribution of
permeability in a dolomite body is an equal if not more,
important consideration. Contrasts in dolomite permeability
can be extreme and occur over short distances. The problem is
exemplified by the occurrence of high permeability (> 100 mD
to several D) dolomites that are often thin (a few feet) such as
those in Ghawar which account for a significant portion of
total production [2]. Reactive transport models provide a
process based method of investigating the origin and spatial
distribution of dolomite permeability (Figure 8) [11]. We
interrogated our 3D model to illustrate reservoir sweet spot
identification applying the following illustrative criteria:
Dolomite with > 15 % porosity and < 5% anhydrite (Figure
12). Dolomite “sweet spots” (shown in red) are located close
to the dolomite front, have a variable thickness and are
generally well connected in this case (Figure 12). These sweet
spots occur in both the dolomite “main body” and “fingers”
(Figure 12). This general pattern is similar, but not exclusive
to several dolomite reservoirs, although our model is a
simplified hypothetical example (authors’ personal
observation). Permeability reduction via the
overdolomitization process was not observed in our 3D model,
perhaps due to the relatively short simulation time. Chemical
compaction, hydrothermal dolomitization and dolomite
recrystallization could cause overdolomitization. These
diagenetic processes were not included in our simulations.
7
(A)
Dolomite
(B) Anhydrite
main body
bedding
Isolate
Anhydrite
dolomite cement
fingers
(C)
High
Permeability
Dolomite
Figure 12. RTM 3D distribution of (A) dolomite; (B) Anhydrite;
(C) High-permeability dolomite (dolomite with < 5% anhydrite
and > 15 % porosity). All properties are shown in 2D slices in x-
direction. Porosity intervals are shown in contours and “high-K”
dolomite is shaded in red. Thin section photo courtesy of Jim
Markello.
DISCUSSION
Reactive transport models have improved our fundamental
understanding of reflux dolomitization. Results provide a
process-based framework for understanding the evolution of
dolomite, associated anhydrite cements and petrophysical
properties. Insights from these generic simulations can be
applied to reservoir examples to help constrain and develop
scenarios for subsurface correlations of dolomite bodies and
their connectivity.
Our 2D and 3D modeling results provide unique insight about
the dynamic propagation of limestone-dolomite fronts by
fingering, the connectivity of dolomite ‘fingers’ to the ‘main
body’, dolomitization and anhydrite occurrence (primary
precipitation and cements), and locating reservoir ‘sweet
spots’ in a dolomite geobody. The predicted pattern of
limestone to dolomite transition is consistent with outcrop
observations (Figures 2 and 3), yet the modeling results
offered theoretical basis for new interpretations.
For example, the 2D simulated dolomite distributions (Figure
8) seem to change from thick and extensive to thin and
separated in the direction of the reflux flow. The predicted
distribution of limestone to dolomitization transition and
potential location of reservoir “sweet” spot revealed some
unique features. In particular, the extensive dolomite main
body in the platform interior (Figure 8 E) suggests good lateral
and vertical reservoir connectivity, whereas in the limestone to
dolomite transition zone the layered dolomite “fingers” and
anhydrite cement indicates complex reservoir connectivity,
with good reservoir intervals (uncemented dolomite)
8 !PTC 11199
sandwiched by tight zone with extensive anhydrite cemented
(Figure 8 J, O).
Our 3D modeling results suggest that the extensive dolomite
main bodies and the underline dolomite fingers (which are
often a few feet thick) are not necessarily associated with
different depositional and diagenetic events. They could be
associated with the same diagenetic events and in fact, they
could well be interconnected. The distribution of primary
anhydrite precipitation (bedded anhydrite) and in particular the
anhydrite cement can be complex. For example, the anhydrite
cement can be located either at the dolomite front or at close
approximity to the brine source and may subject to
remobilization by subsequent reflux events. The predicted
high-K dolomites were likely to be distributed at the dolomite-
limestone transition zone in a hypersalinity reflux
environment. They tend to be away from the source of the
hypersalinity brine so little or no anhydrite occurred. They are
represented by a complex 3D interconnected underground
network that could significantly affect reservoir fluid flow.
ASSUMPTIONS AND LIMITATIONS
The validity of a reactive transport model depends on how the
(geologically-based) conceptual model is numerically
represented, how the results are affected by the assumptions
and limitations [11-16, 30-35].
Our 2D and 3D reflux models only predict reservoir quality
after early reflux diagenesis. The present day reservoir has
gone through additional alteration including compaction and
possible late burial diagenesis such as hydrothermal
dolomitization. Burial alteration likely affects an early
dolomite generation, dolomite textures are not incompletely
altered, and grainstones with pre-existing
porosity/permeability are most likely affected [2, 36-37]. It is
likely that “high permeability dolograinstones” are ultimately
the product of burial alteration (e.g., recrystallization)
superimposed on early dolomites.
Uncertainty in the quantitative description of some key
parameters, particularly initial porosity and permeability,
limits the precision of reservoir quality predictions using
reactive transport models [11-16]. Incorporating different
scales of permeability heterogeneity can result in predictions
of diagenetic geobodies with a greater resemblance to those
observed in nature [11-12]. Another key uncertainty is the
specified porosity to permeability feedback relationships for
different mineral reactions in the variety of carbonate sediment
types modeled [11]. Scaling up permeability to “effective”
values for the dimensions of simulation grids is a further
challenge, particularly when the rocks are fractured [38-39]. In
addition to rock properties, pore fluid composition is also a
significant control on the pattern and magnitude of
groundwater flow and associated mineral reactions [11-12].
CONCLUSIONS
We utilized reactive transport models (RTM) to investigate
key controls on reflux dolomitization and reservoir quality.
The models predicted the spatial and temporal distribution of
key geological parameters, such as limestone, dolomite,
anhydrite and porosity, and how these parameters were
affected by depositional and diagenetic variables (initial
lithology, porosity and permeability, diagenetic fluid
compositions, temperature and pressure, etc). The 2D and 3D
modeling results were consistent with regional and reservoir
trends in dolomite/anhydrite distribution published in the
literature and provided a better understanding of key
diagenesis processes controlling reservoir quality in
dolomitized carbonate reservoirs. The modeling results,
together with well-based observations, generated predictive
diagenetic rules that can be used to guide porosity and
permeability distribution in 3D reservoir connectivity models.
The key conclusions and new insights from the reactive
transport model include:
1. Reactive Transport Models have the capability to simulate
the process of brine reflux dolomitization and the resultant
distribution of dolomite, anhydrite and petrophysical
properties in 2 and 3D.
2. Sensitivity analyses revealed the hierarchical controls of
natural variables (salinity, initial lithology, porosity,
permeability and heterogeneity, temperature, and residence
time) on the distribution and evolution of the reflux system.
3. Simulations show the propagation of dolomite fronts by
fingering generating patterns of dolomite consistent with those
observed in the limited number of outcrop studies.
4. Dolomite “fingers” tend to connect to the “main body” in a
single reflux episode and are rarely isolated “pods”.
5. The salinity of the reflux brine controls anhydrite primary
precipitation and cementation, leading to the formation of
important diagenetic features such as bedded anhydrite and
anhydrite cement, with the latter subject to remobilization.
6. Dolomite “sweet spots” are located close to the dolomite
front, have a variable thickness and are generally well
connected in 3D.
The predicted distribution of dolomite (main body and fingers)
anhydrite (bedded anhydrite and anhydrite cement), and high-
K dolomite provided important guidance to develop 3D
reservoir connectivity models with different scenarios for
reservoir simulation studies. Future case studies will utilize
well and field data to test the reactive transport models.
ACKNOWLEDGEMENTS
We thank contributions from ExxonMobil colleagues Sean
Guidry, Jeremy Jameson, Jon Kauffman, Bob Alway, Scott
Parker, Mike Weaver, and Jim Anderson. We also thank Drs.
Tianfu Xu and Karsten Pruess from Lawrence Berkeley
National Laboratory for their generous support in using
ToughReact program. ExxonMobil’s permission to publish
this paper is greatly appreciated.
IPTC 11199
APPENDICES: SOFTWARE DESCRIPTION
X2T (University of Illinois)
The X2t program models couples transport and reaction in
geochemical systems open to groundwater flow in one or two
dimensions. The package includes the full scope of
geochemical modeling features of The Geochemist's
Workbench¨ (GWB) as well as complete capabilities for
modeling mass transport and heat transfer in one radial or one
or two linear dimensions. X2t program can be obtained from
RockWare Inc. (www.rockware.com)
Tough2/ToughReact (Lawrence Berkeley National
Laboratory)
Tough2 (transport of unsaturated groundwater and heat) is a
numerical simulator for nonisothermal flows of
multicomponent, multiphase fluids in one, two, and three-
dimensional porous and fractured media. Tough2 uses an
integral finite difference (IFD) method for space
discretization, and first-order fully implicit time differencing.
A choice of a sparse direct solver or various preconditioned
conjugate gradient algorithms is available for linear equation
solution. The program provides options for specifying
injection or withdrawal of heat and fluids. Double-porosity,
dual-permeability, and multiple interacting continua (MINC)
methods are available for modeling flow in fractured porous
media.
ToughReact is a non-isothermal reactive geochemical
transport model by introducing reactive geochemistry into the
framework of the existing Tough2 program. ToughReact uses
a sequential iteration approach, which solves the transport and
the reaction equations separately. An implicit time-weighting
scheme is used for individual component of flow, transport,
and geochemical reaction. The chemical transport equations
are solved independently for each component, whereas the
reaction equations are solved on a grid block basis using
Newton-Raphson iteration. An improved equilibrium-kinetics
speciation model for simulating water-rock-gas interaction is
employed. Quasi-stationary approximation and an automatic
time stepping scheme are implemented in ToughReact.
The methods are applicable to one-, two-, or three-dimensional
geologic domains with physical and chemical heterogeneity.
Transport of aqueous and gaseous species by advection and
molecular diffusion is considered in both liquid and gas
phases. Any number of chemical species in liquid, gas and
solid phases can be accommodated. Aqueous complexation,
acid-base, redox, gas dissolution/exsolution, and cation
exchange, are considered under the local equilibrium
assumption. Mineral dissolution and precipitation can proceed
either subject to local equilibrium or kinetic conditions.
Tough2 and ToughReact programs can be obtained by
contacting Drs. Tianfu Xu and Karsten Pruess from Lawrence
Berkeley National Laboratory (www.lbl.gov).
9
REFERENCES
1. Power, R.W. 1962. Arabian Upper Jurassic carbonate
reservoir rocks. In HAM, W. E. (ed.). Classification of
carbonate rocks. Memoir of the American Association of
Petroleum Geologists, 1, 279 pp.
2. Cantrell, D. Swart, P. and Hagerty, R. 2004. Genesis and
characterization of dolomite Arab-D reservoir, Ghawar Field,
Saudi Arabia: Geo-Arabia, v. 9, p. 11–36.
3. Insalaco, E,, Virgone, A., Courme, B., Gaillot, J., Kamali,
M., Moallemi, A., Lotfpour, M. and Monibi, S. 2006. Title:
Upper Dalan Member and Kangan Formation between the
Zagros Mountains and offshore fars, Iran: depositional
system, biostratigraphy and stratigraphic architecture.
Geoarabia, 11 (2): p. 75-176.
4. Wilson, E. N., Hardie, L.A. and Phillips, O.M. 1990.
Dolomitization front geometry, fluid flow patterns, and the
origin of dolomite: the Triassic Latemar Buildup, northern
Italy: American Journal of Science, v. 290, p. 741-796.
5. Posamentier, H.W., Allen, G.P., James, D.P. and Tesson, M.
1992. Forced regressions in a sequence stratigraphic
framework: Concepts, examples, and exploration
significance: American Association of Petroleum Geologists
Bulletin, v. 76, p. 1687-1709.
6. Van Wagoner, J.C., Posamentier, H.W., Mitchum, R.M.,
Vail, P.R., Sarg, J.F., Loutit, T.S. and Hardenbol, J. 1988. An
overview of the fundamentals of sequence stratigraphy and
key definitions. In C.K. Wilgus, B.S. Hastings, C.G.St.C.
Kendall, H.W. Posamentier, C.A. Ross, J.C. Van Wagoner,
eds., Sea-level changes: an integrated approach. Society of
Economic Paleontologists and Mineralogists Special
Publication No. 42, p. 39-45.
7. Yao, Q. and Demicco, R.V. 1997. Dolomitization of the
Cambrian-Ordovician carbonate platform, southern Canadian
Rocky Mountains: dolomite front geometry, fluid inclusion
geochemistry, isotopic signature, and hydrogeological
modelling studies: American Journal of Science, v. 297, p.
892-938.
8. Lucia, F.J. and Major, R.P. 1994. Porosity evolution through
hypersaline reflux dolomitization, in B. Purser, M. Tucker
and D. Zenger, eds., Dolomites: International Association of
Sedimentologists Special Publication 21, p. 345-360.
9. Tillotson, B. 2003. Bed and facies scale selectivity during
late-stage dolomitization: Lower Ordovician El Paso Group,
Franklin Mountains, West Texas. AAPG Annual
Convention, Salt Lake City, Utah.
10. Wilson, A. M., Sanford, W.E., Whitaker, F.F. and Smart,
P.L. 2001. Spatial patterns of diagenesis during geothermal
circulation in carbonate platforms: American Journal of
Science, v. 301, p. 727-752.
11. Jones, G.D. and Xiao, Y. 2005. Dolomitization, anhydrite
cementation, and porosity evolution in a reflux system:
Insights from reactive transport models. AAPG Bulletin;
May 2005; v. 89; no. 5; p. 577-601.
12. Xiao, Y. and Jones, D.J. 2006. Reactive transport modeling
of carbonate and siliciclastic diagenesis and reservoir quality
prediction. SPE paper #101669, p. 1-10.
13. Lichtner, P. C. 1996. Continuum formulation of
multicomponent-multiphase reactive transport, in Lichtner,
10
P. C., Steefel, C. I. and Oelkers, E. H. (eds.), Reactive
transport in porous media, Reviews in Mineralogy, Mineral
Society of America, v. 34, p. 1-79.
14. Steefel, C.I, DePaolo, D. and Lichtner, P.C. 2005. Reactive
transport modeling: An essential tool and a new research
approach for the Earth sciences, Earth and Planetary Science
Letters 240: p. 539-558.
15. Jones, G.D. and Xiao, Y. 2006. Geothermal convection in the
Tengiz carbonate platform, Kazakhstan: Reactive transport
models of diagenesis and reservoir quality. AAPG Bulletin;
August 2006; v. 90; no. 8; p. 1251-1272.
16. Xiao, Y. and Jones, D.J. 2006. Testing carbonate diagenetic
paradigms using reactive transport models. AAPG Annual
Convention, Houston, Programs 11C.
17. Arvidson, R. S. and Mackenzie, F.T. 1999. The dolomite
problem: control of precipitation kinetics by temperature and
saturation state: American Journal of Science, v. 299, p. 257-
288.
18. Bethke, C.M. 1996. Geochemical Reaction Modeling,
Concepts and Applications. Oxford University Press, New
York, 397 pp.
19. Xu, T., Sonnenthal, E.L., Spycher, N. and Pruess, K. 2006.
TOURGHREACT: A simulation program for non-isothermal
multiphase reactive geochemical transport in variably
saturated geologic media, Computer & Geoscience,
Lawrence Berkeley National Laboratory.
20. Adams, J. E. and Rhodes, M.L. 1960. Dolomitization by
seepage reflux: AAPG Bulletin, v. 44, p. 1912-1920.
21. Sun, S.Q. 1995. Dolomite reservoirs: porosity evolution and
reservoir characteristics: AAPG Bulletin, v. 79, p. 249-257
22. Zenger, D.H., Dunham, J.B. and Ethington, R.L. (eds) 1980.
Concepts and models of dolomitization: SEPM special
publication, 28, 320 p.
23. Moore, C.H. 2001. Carbonate reservoirs: porosity evolution
and diagenesis in a sequence stratigraphic framework:
Developments in Sedimentology 55, Elsevier, Netherlands,
444 p.
24. Whitaker, F.F., Smart, P.L. and Jones, G.D. 2004.
Dolomitization: from conceptual to numerical models, in C.
Braithwaite, G. Rizzi and G. Darke, eds., The Geometry and
Petrogenesis of dolomite hydrocarbon reservoirs: Special
Publication of the Geological Society. Vol. 235, p. 99-139.
25. Jones, G. D. and Rostron, B.J. 2000. Analysis of fluid flow
constraints in regional-scale reflux dolomitization: constant
versus variable-flux hydrogeological models: Bulletin of
Canadian Petroleum Geology, v. 48, p. 230-245.
26. Jones, G. D., Smart, P.L., Whitaker, F.F., Rostron, B.J. and
Machel, H.G. 2003. Numerical Modeling of reflux
dolomitzation in the Grosmont Platform complex (Upper
Devonian), Western Canada Sedimentary Basin. AAPG
Bulletin, v. 87, p. 1273-1298.
27. Enos, P. and Sawatsky, L.H. 1981. Pore networks in
Holocene carbonate sediments: Journal of Sedimentary
Petrology, v. 51, p. 961-985.
28. Melim, L.A., Westphal, H., Swart, P.K., Eberli, G.P. and
Munnecke, A. 2002. Questioning carbonate diagenetic
paradigms: evidence from the Neogene of the Bahamas.
Marine Geology 185, p. 27 - 53.
!PTC 11199
29. Logan, B.W. 1987. The MacLeod evaporite basin, Western
Australia: AAPG Memoirs, v. 44, 140 p.
30. Lee. A. 2003. 3-D numerical modeling of freshwater lens on
atoll islands. Proceedings, Tough Symposium 2003.
Lawrence Berkeley National Laboratory, Berkeley,
California, 1-7.
31. Whitaker, F. F. and Smart, P.L. 1997. Geochemistry of
meteoric waters and porosity generation in carbonate islands
of the Bahamas. Geofluids II, 2nd International Conference
on Fluid Evolution, Migration and Interaction in Sedimentary
Basins and Orogenic Belts, University of Belfast, 415-418.
32. Whitaker, F.F., Hague, Y., Smart, P.L, Waltham, D.A. and
Bosence, D. 1999. Coupled modeling of carbonate diagenesis
and sedimentology: structure and function of a coupled 2-
dimensional diagenetic and sedimentological model of
carbonate platform evolution, Special Publication of the
Society of Economic Palaeontologists and Mineralogists 62,
p. 69-84.
33. Sanford, W.E., Whitaker, F.F., Smart, P.L. and Jones, G.D.
1998. Numerical Analysis of seawater circulation in
carbonate platforms: I geothermal circulation: American
Journal of Science, v. 298, p. 801-828.
34. Kaufman, J.K. 1994. Numerical models of fluid flow in
carbonate platforms: implications for dolomitization: Journal
of Sedimentary Research, v. A64, p. 128-139.
35. Kohout, F. A., Henry, H.R. and Banks, J.E. 1977. Hydrology
related to geothermal conditions of the Floridan Plateau, in
K. L. Smith and G. M. Griffin (eds) The geothermal nature
of the Floridan Plateau: Florida Department of Natural
Resources Bureau Geology Special Publication, v. 21, p. 1-
34.
36. Corbella, M., Ayora, C., Cardellach, E., and Soler, A. 2006.
Reactive transport modeling and hydrothermal karst genesis:
The example of the Rocabruna barite deposit (Eastern
Pyrenees). Chemical Geology, Volume 233, Issues 1-2, 30,
p. 113-125.
37. Garven G., Appold, M.S., Toptygina, V.I. and Hazlett, T.J.
1999. Hydrogeologic modelling of the genesis of carbonate-
hosted lead-zinc ores: Hydrogeology Journal v. 7, p. 108-
126.
38. Thawer, R., Alhendi, A., Al Mazroui, Y., Boyd, D.,
Masuzawa, T., Sugawara, Y., Hollis, C. and Lowden, B.
2000. Controls on vertical and horizontal flow in a carbonate
reservoir that impact gasflooding and waterflooding: Society
of Petroleum Engineers, SPE87237, 10 p.
39. Ewing, R.E. 1997. Aspects of upscaling in simulation of flow
in porous media: Advances in Water Resources, v. 20, p.
349-358.