Download as pdf or txt
Download as pdf or txt
You are on page 1of 31

Carbonate petrophysical rock typing: integrating geological

attributes and petrophysical properties while linking with


dynamic behaviour
MARK SKALINSKI1* & JEROEN A. M. KENTER1,2
1
Chevron Energy Technology Company, 6001 Bollinger Canyon Road,
San Ramon, CA 94583-2324, USA
2
Present address: ConocoPhillips, 600 N Dairy Ashford, Houston, TX 77079, USA
*Corresponding author (e-mail: jeroen.kenter@conocophillips.com)

Abstract: Carbonate rock typing provides a vehicle to propagate petrophysical properties through
association with geological attributes and, therefore, is critical for distributing reservoir properties,
such as permeability and water saturation, in the reservoir model. The conventional approaches to
rock typing have significant gaps in incorporating diagenetic processes, transferring rock types
from core to log domain, accounting for fractures and using appropriate methodology to realisti-
cally distribute rock types in the static reservoir model. The workflow proposed in this paper
addresses these issues in a comprehensive way by determination of petrophysical rock types
(PRTs), which control static properties and dynamic behaviour of the reservoir, while optimally
linking to geological attributes (depositional and diagenetic) and their spatial interrelationships
and trends. This approach is novel for the fact that it: (1) integrates geological processes, petrophy-
sics and Earth modelling aspects of rock typing; (2) integrates core and log scales; and (3) provides
a flexible ‘road map’ from core to 3D model for variable data scenarios that can be updated with
progressive changes in data quality and quantity during the life cycle of an asset. This paper intro-
duces the rationale behind this workflow, and demonstrates its workings and agility through
deployment in two large carbonate fields.

Carbonate rock typing drives the quality of the dis- knowledge and expertise at the interface of
tribution of petrophysical parameters in three- geology and petrophysics
dimensional (3D) Earth models and is fundamental † Lack of a proper definition of the spatial pro-
to reservoir characterization. Despite its recognized perties of petrophysical rock types, resulting in
importance, the oil industry is lacking a common obscuring the spatial definition of rock types
definition and standards for carbonate rock typing. and, hence, their distribution in the interwell
If the goal of carbonate rock typing is to properly space. The relationship between petrophysical
and realistically distribute log-derived reservoir properties (such as permeability and water sat-
properties in 3D models, and to generate a spatial uration) and geological attributes (such as depo-
distribution of appropriate rock types that control sitional texture, diagenetic modification, and
oil-in-place and fluid flow, then most of the existing their spatial trends and juxtaposition) are poorly
rock typing definitions are not meeting this goal. documented and/or understood, while no ade-
The main reasons for this gap are as follows: quate diagenetic analogues from outcrop or sub-
surface are available in the public domain.
† Depositional bias, which assumes that deposi- † Biased focus on reservoir properties while ignor-
tional facies are adequately representing reser- ing dynamic data. Most of the rock typing studies
voir properties, while, in fact, most carbonate are focused only on reservoir rock, ignoring per-
reservoirs are strongly influenced by diagenetic meability barriers/baffles that are important for
modification. Even when diagenetic modifi- the flow modelling and definition of the flow
cation is properly noted and included in the rock- units in reservoir models.
typing step, it is often ignored for reasons of
complexity in the spatial distribution stage, Another consideration is that of scale. Rock types
which is usually driven by depositional trends. are scale dependent and can be defined at: (1) the
† Lack of integration between rock type defini- pore scale using dual-energy computerized tom-
tions in the core domain and determinations in ography (CT) scan data (Walls et al. 2013); (2) the
the log domain, which are required to popu- core-plug scale using mercury injection capillary
late static models. The current modelling prac- pressure (MICP) data, routine core analysis (RCA
tice is, still, suffering from a common lack of or conventional core analysis, CCAL) data and

From: Agar, S. M. & Geiger, S. (eds) 2015. Fundamental Controls on Fluid Flow in Carbonates: Current
Workflows to Emerging Technologies. Geological Society, London, Special Publications, 406, 229– 259.
First published online May 30, 2014, http://dx.doi.org/10.1144/SP406.6
# The Geological Society of London 2015. Publishing disclaimer: www.geolsoc.org.uk/pub_ethics
230 M. SKALINSKI & J. A. M. KENTER

petrographical thin sections (Amaefule et al. 1993; dominated by depositional or early diagenetic poros-
Marzouk et al. 1998; Cantrell & Hagerty 2003; ity or where porosity modification resulted in mostly
Masalmeh & Jing 2004; Gomes et al. 2008; Hollis fabric selective pore systems. However, in reservoirs
et al. 2010); (3) the log scale (Cuddy 1998; Ye & with complex and multimodal pore systems, the
Rabiller 2000; Sagaaf & Nebrija 2003; Skalinski classification fails to resolve the defined fabric
et al. 2009); or (4) the flow-unit scale (Ebanks elements, and, as a result, fails to adequately char-
et al. 1992; Gunter et al. 1997; Corbett & Potter acterize porosity–permeability relationships (e.g.
2004; Cortez & Corbett 2005; Kazemi et al. Kenter et al. 2006; Johnson et al. 2010), hydraulic
2012). Several papers have made a serious attempt flow units (Bust et al. 2011) or reservoir heterogen-
to link different scales and data types (i.e. Vahren- eity (Abbaszadeh et al. 2000). Although there are
kamp et al. 2008; Zang et al. 2009); however, if examples of successful prediction of such fabric-
the ultimate goal is to populate 3D static and dyna- based rock types in the log domain (i.e. Bagheri
mic models, then there is a scale challenge that has et al. 2005), this approach commonly lacks proper
not, yet, been addressed in a rigorous manner. The prediction in the log domain in uncored wells/
workflow presented in this paper is integrating intervals. One of the underlying reasons for the
core and log scale by defining rock types in the Lucia classification falling short in complex pore
core domain, and predicting/integrating in the log systems is the absence of capillary pressure in its
domain. Further work is needed to develop rigor- definition (Johnson et al. 2010; Chehrazi et al. 2011).
ous multiscale approaches that also integrate the
interwell/seismic scale. Rock typing is situated Rock types based on pore typing
right at the interface between the disciplines of
geology, petrophysics and engineering but gener- Pore types defined using pore size and/or pore
ally is published mostly in petrophysical journals throat size distributions are better tied to the flow
and out of sight of the petroleum geologist. There- units and matrix permeability, but they might not
fore, prior to introducing the new petrophysical have the spatial guidance or trends for proper
rock type (PRT) workflow, it is critical to review spatial distribution in 3D models. Examples are
the existing status quo in carbonate rock typing the Winland R35 method (Kolodzie 1980), the par-
practices and highlight those issues that the new titioning model by Ramakrishnan et al. (1999,
workflow attempts to mitigate. The present status 2001), the pore throat classification by Marzouk
of carbonate rock typing can be summarized in et al. (1995) and Skalinski et al. (2005), or the
four different schemes. Table 1 summarizes the dif- ‘Rosetta Stone’ approach by Clerke et al. (2008).
ferent schemes in terms of data needs, scale, and The Winland R35 is based on the pore throat
suggested advantages and disadvantages that are radius corresponding to 35% of mercury (non-
core for the rationale of developing the new rock wetting phase) saturation in mercury injection capil-
typing ‘road map’ that we present in this work. lary pressure (MICP) measurements as an indicator
of effective flow. Marzouk et al. (1998) partitioned
Rock types based on mostly geological the carbonate pore system into three components
(depositional) facies (micropores, mesopores and macropores). The
three components were adopted for porosity parti-
Rock types here are defined as lithofacies and tioning in carbonates and are defined by their pore
associated attributes, what is commonly referred to throat radius as measured by MICP or air–water
as depositional rock types (DRTs). This includes centrifuge. However, the pore radii to define parti-
pure textural classes (Dunham 1962; Embry & tioning cutoffs for pore classes are not consistent
Klovan 1971), generic pore types (Choquette & between authors. For example, microporosity was
Pray 1970) or combinations of Dunham/Embry – defined as pore throats below 0.1 (Chekani &
Klovan classes with grain/pore size, which is rep- Kharrat 2009), 0.2 (Porras & Campos 2001), 0.3
resented by the popular Lucia classification (Lucia (Marzouk et al. 1998; Skalinski et al. 2009), 0.5
1995). Such classifications relate to geologists’ (Arfi et al. 2006) or 2 mm (Hulea & Nicholls
classification of geological attributes, and are rela- 2012). Marzouk et al. (1998) related pore types to
tively simple as they can be linked to depositional the Dunham textures for a Middle East reservoir
environments and associated spatial distributions, dominated by depositional attributes. However,
but are often not relevant to fluid flow as a result of Waravur et al. (2005) did not observe a satisfactory
diagenetic modification. Rock fabric and inferred link between MICP-derived pore types and lithofa-
pore types (Lucia 1983, 1995, 2007) can provide a cies due to diagenetic modification. Clerke et al.
link between petrophysics and spatial geological (2008) utilized the relationship between perme-
trends only if pore systems confirm the fabric ability and Thomeer parameters (Thomeer 1983)
element definitions and are relatively uniform. This and demonstrated that permeability is most strongly
is only valid for a subset of carbonate reservoirs controlled by the largest pore modes (so-called
CARBONATE ROCK TYPING 231

Table 1. Summary and comparison of current rock typing schemes including the scheme proposed in this paper

Method/author Data* Scale Plus Minus

DRT based
Dunham (1962) Core description Core Defines depositional Ignores diagenesis and pore types
texture
Lucia (1995, 2007) Core observations Core Link to depositional Ignores later diagenetic events and
trends and microporosity, large uncertainty
permeability in perm prediction, inferred pore
types

Pore type based


Choquette & Pray Core Core Genetic pore types No link to petrophysical properties
(1970) and spatial trends
MICP based (PTD, MICP Core Link to permeability and Prediction of pore type in log
R35, Pd) dynamic data domain, ignores pore throat
diameter .140 mm, no link to
spatial trends
NMR based NMR Log Log domain Pore size limitations, T2 might not
represent pore body size
Lønøy (2006) Thin sections Core Pore size and texture, No link to geological trends,
simple semi-quantitative porosity
Ahr (2008) Thin sections Core Link to genetic processes Core domain only

Integrated
Archie (1952) Core observations Core Practical well site Based on ‘visible’ porosity, no link
integration of geologic to spatial trends
and petrophysical
attributes
Hollis et al. (2010) DRT + DM + PT Core Integration of: DRT, DM Transfer to log domain, does not
and pore types. SCAL include fractures and the 3D
validation. model
Salman & Sameer DRT+ MICP Core Link to depositional Diagenetic modification not
(2009) trends and pore types included as defining factors

Partitioning – flow units


Amaefule et al. Core FZI Core Link to porosity – No link to spatial trends, transfer to
(1993) permeability log scale, fails when K and phi
transforms do not conform
Cortez & Corbett Core GHE Core Link to flow units in Poor link to geological trends,
(2005) larger scale challenging prediction from logs
Gunter et al. (1997) Core or log Core or Link to flow units in Need continuous core profiles,
porosity and log larger scale does not include fractures
permeability
Wibowo & Permadi Core RCA Core Link to permeability Core domain, assumes
(2013) conformance of porosity and
permeability

Partitioning – log clustering or electrofacies


Serra & Abbot Logs Log Easy to apply, link to No link to geological trends, driven
(1980) petrophysics by input logs

Dynamic rock types


Ghedan (2007) SSRT and Core Better dynamic control Need wettability model, not
wettability practical to acquire dynamic
SCAL data for rock typing
Gomes et al. (2008) DRT, DM and Core Better control of the fluid Only core domain
SCAL flow

Petrophysical Rock DRT, DM, PT, Core and Integration of DRT, DM, Not easy to apply
Types (this paper) barrier log PT and barrier in log
domain, includes 3D
Earth model, adaptable
to various data
scenarios

*MICP, mercury injection capillary pressure; NMR, nuclear magnetic resonance; DRT, depositional rock type; DM, diagenetic modifi-
cation; PT, pore type; PTD, pore type distribution; FZI, flow zone indicator (Amaefule et al. 1993); GHE, generalized hydraulic elements
(Cortez & Corbett 2005); RCA, routine core analysis; CCAL, conventional core analysis; SSRT, static reservoir rock types (Ghedan 2007);
SCAL, special core analysis.
232 M. SKALINSKI & J. A. M. KENTER

porositons) corresponding to the entry pressure, Basioni et al. 2008; Salman & Sameer 2009) inte-
which were used as rock typing classification grated depositional lithofacies with MICP-derived
thresholds. Lønøy (2006) defined rock types by esti- pore types. However, the distribution in the result-
mation of conventional pore size and texture from ing models is generally guided by depositional
porosity distributions, but a significant mismatch trends, distributed within a sequence stratigraphic
is observed when comparing resulting rock types framework (SSF), but without any clear explanation
with porositon-based classification on the same as to what extent diagenetic modification controls
dataset (O. Karoussi, 2013, pers. comm.). Some the pore throat distributions and rock types.
studies used MICP-calibrated nuclear magnetic res- Several studies (Francesconi et al. 2009; Skalinski
onance (NMR) T2 cutoffs to define pore types et al. 2009) reported the lack of a significant
(Frank et al. 2005; Arfi et al. 2006; Ramamoorthy relationship between lithofacies and petrophysical
et al. 2008). NMR is a relatively new (1990s) tech- properties and defined rock types based on diage-
nique that utilizes short-duration magnetic pulses to netic features. Rock type population in the latter
excite and measure the relaxation time of hydrogen model was guided by conceptual diagenetic trends
atoms, which largely depends on the pore size distri- and multiple point statistics (MPS) pixel-based
bution in porous media. T2 is the transverse relax- modelling. The widely known Archie (Archie 1952)
ation time typically measured in boreholes and classification scheme is a hybrid approach that com-
generally plotted, following inversion, v. the bines textural (grain size) and petrophysical (pore
signal distribution, while the area under the curve size) parameters. Archie made a first and successful
represents the associated pore volume (T1 is the attempt to relate texture and pore types to petrophy-
longitudinal relation time). T2 cutoffs are user sical properties in carbonates. His classification
defined and are assigned to ranges in pore size primarily estimates porosity, but it also estimates
and/or type (isolated v. connected). However, permeability and capillary properties by subdivid-
NMR has limitations related to pore size (up to ing pore space into (invisible) matrix and visible
100 mm) and the assumption of lack of diffusion porosity. Archie classes proved to be useful in
effects. If diffusion effects are strong and pore well site classification of cuttings, but they are
size is larger than 100 mm, T2 distributions are no lacking a link to depositional or diagenetic pro-
longer considered representative of the pore body cesses that have more complex pore networks.
distribution. Prediction of pore types in the log Hollis et al. (2010) defined rock types by partition-
domain and uncored intervals can create quite a ing depositional lithofacies on MICP-derived pore
challenge without a strong link to geological trends. types and validated those, basically, core-defined
Indeed, the observed pore typing approach in the rock types with relative permeability measurements.
literature is generally lacking spatial guidance and This integrated and comprehensive definition of
has scaling problems when integrating secon- rock types in core domain, however, lacks an ade-
dary pore types (vugs and fractures). Despite those quate prediction in the log domain.
observed challenges, pore types are the most critical
element of rock typing due to their strong control on
the fluid flow in carbonates (Sung et al. 2013). Rock types based on petrophysical partitioning
of core or log data
Rock types based on integration of This rock typing approach is using log clusters
depositional facies, diagenetic modification (combinations) known also as electrofacies (Serra
and pore types & Abbott 1980; Wolff & Pellissier-Combescure
1982) or core porosity –permeability partitioning on
Here the goal is to define rock types by integrat- flow units (Amaefule et al. 1993). These methods
ing depositional lithofacies with diagenetic modifi- are linked to petrophysical properties but are lack-
cations and pore types. In reality, most of the rock ing the critical link to geology and spatial rules.
typing is heavily biased towards depositional Serra & Abbott (1980) coined the term ‘electrofa-
facies (or lithofacies) that are integrated with pore cies’, which basically captures the set of log res-
types from MICP. Diagenetic modifications are ponses that uniquely characterizes a rock unit. In
therefore captured indirectly since there is a strong reality, it is ‘log typing’ and was proven success-
relationship between diagenetic processes and pore ful in several siliciclastic reservoir studies where
types. Spatial trends and modelling are driven by ‘natural ordering’ in lithofacies successions was
depositional trends from analogues and core data correlated with petrophysical ordering observed in
observations, and, in general, a properly and system- the log domain. Using reservoir quality index
atically documented effort to distribute ‘diagenetic (RQI) and flow zone indicators (FZI), rock types
bodies or trends’ is lacking. Numerous studies are classified through artificial binning of the poros-
(e.g. Rebelle et al. 2005; Lehmann et al. 2008; ity –permeability space (Amaefule et al. 1993).
CARBONATE ROCK TYPING 233

A similar approach, also based on the Kozeny– mixed wettability rocks under water injection. Der-
Carman model (Carman 1939), was proposed naika et al. (2012) reported results from relative per-
by Wibowo & Permadi (2013). Corbett & Potter meability drainage and imbibition curves measured
(2004) and Corbett (2010) to expand the FZI con- for static rock types and found coherent results vali-
cept, by introducing generalized hydraulic ele- dating initial rock types. Ghedan (2007) and Gomes
ments (GHEs) that improve the link to geological et al. (2008) reconciled differences between static
facies. Although such approaches are more practi- rock types and dynamic rock types. They proposed
cal and efficient in carbonates with complex pore an approach of generating dynamic rock types by
systems and strong diagenetic modification, the imposing wettability or fluid model on the static
link between log response and geology is generally rock types. This superposition can be performed
rather weak and leads to inaccurate spatial predic- on the upscaled simulation model. Wettability dis-
tion and proportions of resulting rock types. This tributions can be controlled by a position in the
group of methods, which includes multi-resolution hydrocarbon column (Ghedan 2007; Al Jenaibi
graph-based clustering (MRGC) (Ye & Rabiller et al. 2008) or pore type (Marzouk et al. 1998). Gen-
2000), is also strongly dependent on the selection erating dynamic rock type based on relative per-
of input logs for the partitioning process. Rock meability data is impractical since SCAL data on
typing becomes, in essence, an exercise of ‘log preserved samples are costly, sparse and not fully
typing’ with an arbitrary selection of logs with the representative. SCAL measurements can be per-
risk of ignoring meaningful geological influence. formed to define/validate relative permeability
FZI or electrofacies concepts were applied with and saturation profiles for final rock types in the
variable degrees of success in reservoir models dynamic model.
(Guo et al. 2005). The measure of success was This paper describes a new workflow that intends
related to the goodness of prediction of core-derived to address – and mitigate – the gaps specified above
FZI from logs and/or the strength of the link and define petrophysical rock types (PRTs) which
between FZI, GHE or electrofacies and geological account for both depositional and diagenetic pro-
trends. Those links and therefore the resulting pre- cesses, are predictable from logs in uncored wells,
dictions are stronger in siliciclastic reservoirs and and can be distributed in a geologically realistic
carbonates with predominant depositional control. way in 3D Earth models. This workflow is designed
These methods remain to be challenged by carbon- as a general road map for carbonate rock typing and
ate reservoirs with strong diagenetic modification. does not pretend to solve all of the technical chal-
lenges related to this process.
Static or dynamic rock types
A final consideration in the rock typing method- Petrophysical rock type (PRT) – workflow
ology is related to the goal of rock typing. If the
goal is to properly populate static Earth models The workflow consists of eight composite and
with porosity, permeability and water saturation, sequential steps, which are represented by the
the static rock types based on the integration of diagram shown in Figure 1. With progressing field
depositional facies, diagenetic modifications and maturity and data scenarios, multiple loops are
pore types will be adequate. If the goal is to populate required to capture reservoir heterogeneity and to
relative permeability and capillary pressure curves, optimize the representation of the subsurface data
then they might not be appropriate for all reser- (see Fig. 1). The final product is the result of Step
voir types (i.e. Hamon 2003; Masalmeh & Jing 8; that is, the 3D static PRT model or series of
2004; Ghedan 2007). Gomes et al. (2008) extended models capturing recorded uncertainties. Below
the definition of rock types to include fluid –rock follows a brief description of each of the workflow
interaction/wettability from special core analysis steps.
(SCAL). His final reservoir rock types, therefore,
are not only a function of reservoir properties but Step 1: Data scenario
also of fluid content and the effect of fluid –rock
interactions. Hamon (2003) and Masalmeh & Jing PRT workflows are designed to be adaptable to all
(2004) examined static and dynamic parameters data scenarios that are driven by (in order of rel-
derived from core plugs and concluded that sim- evance): (1) well density; (2) logging surveys (vin-
ple static rock types based on RCA or CCAL data tage and completeness); (3) available core data;
(porosity, permeability and water saturation) and and (4) dynamic data. Representativity is another
drainage capillary pressure might be not adequate critical parameter linked to each of these data
to capture dynamic impact resulting in different types. Representativity is typically an uncertainty
residual oil saturation (Sor) and relative permea- that decreases with data collection during the life
bility. Static rock types are not adequate for the of a field but is very difficult to estimate, in
234 M. SKALINSKI & J. A. M. KENTER

Fig. 1. Schematic diagram showing the PRT workflow, which consists of eight composite and sequential steps that
gradually build to the definition, determination and distribution of PRTs in a static model. The workflow represents a
road map that can handle different data scenarios (data quality and quantity, and representativity) ranging from prospect
to appraisal to development. The second, empty loop illustrates that with progressing field maturity and data scenarios
multiple loops are required to capture reservoir heterogeneity and optimize the representation of the subsurface data.

particular near the start of a project, and its pro- that is exclusively based on geological attributes
gression very much depends on the understanding while ignoring any petrophysical information from
of the reservoir quality distribution. Generally, with RCA/CCAL or logs, as inclusion of such infor-
field development, the data scenario will change. mation will bias the DRTs towards petrophysical
This progression is shown in Figure 1, where differ- rock typing and obscure the primary purpose of
ent loops of the workflow represent different data this test. These geological attributes allow the gen-
scenarios and reservoir maturity. Data scenarios eration of a set of DRTs, which represent categories
include those for prospects, field appraisal and pro- of non-overlapping lithofacies as well as a sepa-
ducing fields, although the lack of core (prospects) rate set of diagenetic attributes that may be used
will change the workflow towards a log-driven later in the workflow to explain the disconfor-
approach. mity between reservoir properties and DRTs. The
DRT catalogue summary classification (Step 2.2;
Step 2: Depositional rock typing Fig. 2a) represents the DRT elements of a deposi-
tional model based on core observations and con-
The depositional rock type (DRT) Step is designed cepts from literature and/or analogues. The DRT
to define a DRT catalogue in the core domain and catalogue includes one or more alternate scenarios
to predict lumped DRTs in the log domain. If the that combine DRTs according to geological criteria
subsequent Step 3 indicates a strong control of such as, for example, depositional regions and facies
flow by DRTs, the predicted DRTs will be used as belts to a statistically acceptable number for predic-
a primary input into the PRT definition. Step 2 con- tion using logs. Generally, no more than 15 DRTs
sists of three substeps. The DRT determination are lumped into a maximum of five DRT associ-
(Step 2.1; Fig. 2a) requires an unbiased determi- ations. DRT prediction from logs (Step 2.3) requires
nation of depositional and diagenetic attributes as iterative ‘lumping’ of the DRTs determined from
a function of depth along the cored intervals. core (Fig. 2b). Lumping should follow geological
Unbiased refers to a DRT determination approach associations, similarity in petrophysical space and
CARBONATE ROCK TYPING 235

Fig. 2. Set of diagrams illustrating workflow Step 2. (a) Simplified workflow showing, when core is available,
DRT determination, classification and lumping, and prediction steps. (b) Progression of three successive lumping
steps on the DRT set shown on the left. (c) Graph displaying the progression of prediction scores corresponding to
successive DRT lumping steps shown in (b). (d) Cross-plots showing the general improvement of the prediction score as
a function of the number of logs used for the prediction (left), and the influence of the quality and type of logs used on the
prediction score (right). This clearly warns that care should be taken not to push for a high prediction score but for a
geologically acceptable number of representative DRTs.

statistical representativity of DRTs in the core set on the left. Prediction of DRTs from logs is gen-
descriptions. The diagram to the right in Figure 2b erally performed using multivariate statistical tools
shows three successive lumping steps on the DRT (such as stepwise discriminant analysis or principle
236 M. SKALINSKI & J. A. M. KENTER

component analysis), nearest neighbour method, in workflow Step 5 (i.e. a tie to depositional and/
supervised clustering or neural networks combined or diagenetic processes that may require further
with deterministic methods. Figure 2c presents pro- study, and potential spatial rules and trends for dis-
gression of prediction scores corresponding to suc- tribution in the static model).
cessive DRT lumping steps shown in Figure 2b.
The diagram in Figure 2d shows the general
improvement of the prediction score as a function Step 3: Reservoir typing – the effect of
of the number of logs used for the prediction. diagenesis
The prediction score will also depend on the
quality and type of logs used, and, therefore, the Carbonate reservoirs are highly susceptible to dia-
determination of the optimal combination of pre- genetic processes that alter their original depo-
diction logs is critical. These two observed trends sitional fabric and petrophysical properties. In
also signal a warning that while the predictions addition, diagenetic, tectonic and depositional frac-
score may increase with progressive lumping and tures can overprint both type of systems. The pro-
inclusion of more logs (when available), this does cess of reservoir typing (RT) is the determination
not necessarily mean that the resulting DRT set is of the relative contribution on fluid flow by the
still maintaining integrity in terms of their geologi- degree of diagenetic modification and high-K zones
cal (depositional and diagenetic) definition. As a (diagenetic, depositional and tectonic fractures,
rule, it is critical that lumping maintains the critical and well-connected high-permeability stratigraphic
geological definition, since if that is lost the test will zones) on the original DRTs. The outcome of this
lose its significance. The resulting degrees of suc- step would be classification of reservoir using four
cess for the final set of lumped DRTs provide impor- RT categories: (1) RT1, when PRTs and flow are
tant information for the PRT definition described controlled by DRTs; (2) RT2, when flow is

Fig. 3. Set of diagrams illustrating workflow Step 3. (a) Flow diagram with four steps leading to the assessment of
reservoir type (RT). Note that not all data types are necessarily available to reach an RT definition and, with progression
of data availability, the RT may even change. However, for the current workflow, it is critical to have a ‘best’ assessment
to define and determine a first set of PRTs. The four resulting RTs represent combinations of the relative influence of
depositional, diagenetic control on flow and that of high-K features such as conductive fractures or karst horizons.
CARBONATE ROCK TYPING 237

Fig. 3. Continued. (b) Schematic diagram visualizing the definitions from (a) in a ‘qualitative’ ternary diagram that
shows the relative contribution to flow in PRTs of diagenetic modification, depositional attributes and high-K pathways.
The stars represent plotted positions for the First Eocene and Tengiz platform and margin (modified after Ahr 2008).
(c) Porosity– permeability (phi– K) cross-plot for 15 Tengiz platform wells with textures displayed as colours. The
dominant texture in the platform is grain-supported skeletal sand with minor contributions of mud-supported
(wackestone –mudstone). Texture has a weak control on porosity–permeability trends; maximum permeability spans
over one order of magnitude, which is in the range of the uncertainty in the permeability prediction.
238 M. SKALINSKI & J. A. M. KENTER

Fig. 3. Continued. (d) Petrographical images confirming the predominance (..33%) of diagenetically altered pores
over primary depositional pores, as well as the complex multimodal pore system (modified after Kenter et al. 2006). (A)
Overpacked ooid grainstone with minor blocky calcite and bitumen occluding solution-enhanced interparticle porosity.
(B) Moldic porosity showing evidence for enhanced dissolution shown by corrosion of the mold boundaries, nearly
breaking through at contacts and corroding earlier cement. (C) Moldic porosity in its early phase: micrite grains are
gradually being dissolved following partial fill of interparticle porosity by equant to rhombic spar cement. (D) Advanced
stage of enhanced dissolution of moldic and interparticle porosity (and minor intraparticle) leading to the destruction of
cement bridges and the development of vugs. (E) Skeletal and coated grain grainstone dominated by interparticle
porosity with minor cement. (F) Example of microporosity, pore size smaller than 25 mm, developed in grains. (G)
Intraparticle porosity developed in benthic foraminifera. (H) Skeletal and peloid grainstone with rims of equant spar
lining interparticle pores and minor moldic porosity. Width of photomicrographs is 4.20 (A, B, E, F and G) and 1.58 mm
(C, D and H).
CARBONATE ROCK TYPING 239

controlled by diagenesis and truncates DRT trends; the scope of this work. Figure 3d shows examples
(3) RT3, when flow is controlled by the fracture of the highly complex and heterogeneous, multimo-
system; and (4) a hybrid class, RT4, when flow dal, pore system in the Tengiz platform that is the
has mixed controls (Fig. 3a). The triangle in result of modifying mineralogy and pores through
Figure 3b visualizes this classification with the pos- multistage and stacked diagenetic dissolution and
ition of reservoirs discussed in the ‘Examples’ cementation (Kenter et al. 2006; Dickson pers.
section. This ternary graphic is similar to the comm. 2012). As a consequence, pore typing attem-
genetic pore type classification of Ahr (2008), but pts following Lucia (1995, 2007) and Lønøy
with a different objective (flow control and reservoir (2006) failed to identify single dominant fabrics
typing v. pore types) and with a different definition and associated pore types, but, instead, confirmed
of the hybrid case. There is a link between RT and the dominance (..33%), intimate juxtaposition,
pore types captured by predominant pore systems and mixing of diagenetically altered pores (Kenter
in each reservoir class. Considering the variable et al. 2006). Similar observations have been noted
data scenarios (data quality, quantity and represen- in other studies (e.g. Johnson et al. 2010). Although
tativity), an assessment of RT at this stage of the biased by the individual observer, such estimates
workflow (and, in many cases, field development) can be used with caution as yet another method
is a best guess only. However, even a best guess at to define RT. Finally, when available, PLT profiles
this point in the process is important for concep- and DST build-ups can be examined to identify frac-
tualizing the first-order controls and trends on the ture flow behaviour, while borehole image logs (i.e.
reservoir quality distribution (depositional v. diage- formation micro-imager, FMI) or drilling data (fluid
netic and role of high-K zones), and the subsequent losses, rate of penetration (ROP)) identify fractures.
planning of further investigations to converge on Even though each of those methods has issues with
these controlling parameters. Several methods are data scenarios, they do have merits when used
available to provide estimates for RT, each with together as an (best guess) estimate of the RT for
particular uncertainties that mostly depend on the reasons explained earlier. Figure 3a shows the inte-
data scenario: (1) divergence– convergence of poro- gration of the methods above in a flow diagram that
sity– permeability transforms; (2) predictability of provides a first estimate of the RT, which directly
DRTs using logs; (3) contribution of diagenetically influences further attempts to delineate reservoir
altered pores; (4) flow regime from production quality trends and patterns, and steps to define rock
logging tools (PLT) and/or the drill stem test types and their controls. To achieve the RT, five
(DST); and (5) fractures from borehole image logs steps are required in an ideal data scenario that
and drilling data. Each of these methods is dependent includes application of the four methods discussed
on the data scenario and has particular limitations. above and the determination of the RT assignment
Porosity– permeability regression lines by DRTs based on the relative contribution of the RT indi-
are generally closer spaced when controlled by dia- cator methods above, see Figure 3b. The RT step is
genesis but are highly dependent on the representa- included in the workflow for two main reasons.
tivity of core data within each DRT. Figure 3c First, to prioritize flow control and define the nest-
shows an example of conformity of porosity– per- ing scheme for the PRT definition in Step 5, and,
meability transforms in a cross-plot using data second, to frame the PRT distribution in the 3D
from 15 wells in the Tengiz platform. Transforms model using methods described in Step 8. The first
for different textures are very closely spaced and step will allow, when present, the definition of PRTs
all fall within the range of the uncertainty in the per- that host certain high-K phenomena, like deposi-
meability prediction, suggesting a very weak control tional fractures in tight marginal microbial bound-
by texture (or DRTs). The prediction score of stone or high-K stratigraphically bound zones with
(unbiased) DRTs using logs will generally connected vugs resulting from extensive meteoric
improve with a reduction (lumping) in the number leaching.
of DRTs (Fig. 2d), the quality and quantity of log-
ging suites, and the prediction method used (neural Step 4: Pore typing
networks or nearest neighbour over linear discri-
minant analysis). Ahr (2008) emphasized the rel- Carbonate petrophysical heterogeneity (flow prop-
evance of estimating the diagenetic alteration of erties) is generally the result of complex and multi-
the pore system from petrographical thin sections modal pore systems including vugs and fractures.
as an estimate for the contribution of diagenesis Identification and prediction of pore types is there-
on reservoir quality. Ahr (2008) provides a detailed fore essential for reliable rock typing in carbon-
overview with excellent examples of the various ate. The pore typing workflow (Fig. 4) accounts
diagenetic processes, their products, and criteria for different data scenarios depending on the avail-
for recognition and estimation that can be used in ability of core, MICP data, digital image analysis
this context rendering a thorough synthesis from (Weger et al. 2009) or other pore characterization
240 M. SKALINSKI & J. A. M. KENTER

Fig. 4. Diagram showing the pore typing elements required for the determination and definition of the PRTs. Similar to
other steps in the workflow, the pore typing step is flexible and accounts for different data scenarios and integrates pore
types from different scales and measurements, such as pore types derived from MICP, digital image analysis (DIA),
NMR, gas absorption, nuclear, high-resolution micro-CT or industrial whole core CT, and speciality logs such as NMR,
acoustic borehole images (BHI) or formation microscanner (FM).

measurements (e.g. NMR, nuclear, gas adsorption), Step 5: PRT Definition


and speciality logs such as nuclear magnetic reson-
ance (NMR), ultrasonic borehole imager (BHI) In the context of this workflow, PRTs are defined
or FMI (Fig. 4). Appropriate pore type identifi- as: (1) the category of rocks that are characterized
cation can be best performed using MICP data, by specific ranges of petrophysical properties (e.g.
which provides accurate information on pore throat porosity, permeability and water saturation); (2)
distributions controlling flow in the reservoir. MICP exhibiting distinct petrophysical relationships rel-
derived initial pore throat types (IPT) have to be evant for the reservoir modelling (e.g. porosity v.
combined with larger scale observations such as permeability, water saturation v. capillary pressure);
vugs and fractures from specialty logs. In the case (3) identified by wireline or logging while drilling
of data scenarios lacking MICP data but including (LWD) logging surveys; and (4) linked to geologi-
available representative core, an alternative path cal attributes such as primary texture or diagenetic
(B) is to be followed that produces core pore types modifications. In this critical step, PRTs are defined
(CPT). CPT can be defined using other data types according to the relative influence of the following
(i.e. digital image analysis (DIA)) or conventional attributes: permeability barriers, DRTs, diagenetic
classification such as Choquette & Pray (1970), modifications and pore types. The permeability
but are limited to the pore type and size, and do barrier(s) are either non-reservoir rocks or low-
not provide information on the pore throats. When permeability rocks that act as flow barriers or
NMR and/or BHI/FMI data are not available, baffles, as indicated by dynamic data. They can be
pore types could be defined from IPT or from CPT a product of either depositional or diagenetic pro-
data. In the case when neither core nor NMR/ cesses. In carbonates they can be represented by
BHI/FMI data are available, the determination of hard grounds, evaporite layers, zones with dispersed
relevant pore types is less reliable. In that scenario, volcanic ash, bitumen-tarmac layers, cemented
pore types can be derived from cuttings (MICP, pet- zones or interbeds or laminations with a high shale
rographic thin section image analysis or micro-CT or volcanic ash contribution. In summary, PRTs
scans) or from petrophysical analogues such as are defined by combining elements such as the pre-
those available in public or commercial carbonate dicted DRT, PT, barriers and other diagenetic modi-
rock catalogues. fiers affecting the log response. The combination
CARBONATE ROCK TYPING 241

Fig. 5. PRTs are defined by combining elements including the predicted DRT, pore type (PT), barriers and other
diagenetic modifiers affecting the log response. Their combination order is mapped to the four reservoir types that were
defined in Figure 3a. See the text for more discussion.

order is defined by RT (see Fig. 5), which is defined to designing the distribution approach in the static
in Step 3 (see Fig. 3a). The definition of PRTs has to model (Step 8). This data interrogation step is extre-
take into account the logging data scenario; that is, mely important to constrain PRT patterns and
the ability to predict PRTs from logs in the majority trends, and juxtaposition rules that need careful
of wells. The PRT definition at the beginning of this contrast and comparison with the geological obser-
subsection contains the primary criteria used to vations from Steps 1 –4. This step is especially criti-
define PRT; the final PRTs have to conform to all cal when RT is greater than 1 and dominated by
four segments of the definition. diagenetic modification since little or no informa-
tion on spatial rules of diagenetic processes is avail-
Step 6: PRT determination in a multi-well able in the literature, and the user may be limited
to forward modelling of diagenetic processes (for
setting and quality control using maps references, see Step 8). The PRT determination
This step includes the determination of PRTs in all follows three steps: (6.1) development of algorithms
wells using algorithms for PRT prediction from or models for PRT prediction; (6.2) PRT determi-
logs developed in the previous steps (see Step 2 nation in all wells; and (6.3) quality control (and
and Fig. 2c, d). These algorithms are designed to trend mapping and extraction) (see Fig. 6).
predict lumped DRTs (Step 2), pore types (Step
4), permeability barriers (Step 5) or diagenetic Step 7: Dynamic validation of PRTs
modifications (Step 3). Usually a master program
is designed to predict PRTs in a logical succession A quantitative validation of the link between
as defined in Step 5. For example, if we have RT1, PRTs and available flow indicators is performed
the PRTs could be determined by predicting DRTs by: (1) comparing PRTs with core RCA data (such
first and distributing pore type based groups nested as porosity–permeability cross-plots and modified
within the DRTs (see the Wafra Field example in Lorenz plots (Gunter et al. 1997); and (2) the
the next section on ‘Examples’). The resulting comparison of PRTs with dynamic data such as
PRT logs are mapped across the field in all (cored PLTs, DSTs, wireline formation test (WFT) data
and uncored) wells using kriging (or comparable) or injection profiles. The goal of this step is to
techniques. Following an assessment (and possible confirm that PRTs are linked to flow profiles
removal) of outliers, spatial trends and relationships observed in dynamic data (Fig. 7). In particular, bar-
between PRTs are extracted and used as input riers and flow zones should be positively correlated
242 M. SKALINSKI & J. A. M. KENTER

Fig. 6. The PRT determination is represented by three simple steps. The first step builds on algorithms developed during
Step 2 (DRT prediction) and Step 4 (pore typing), and supplemented with information on permeability barriers (Step 2)
or diagenetic modifications (Step 3). In general, a tailor-made set of algorithms is captured in a master program designed
to predict PRTs in a logical succession, as defined in Step 5.

to appropriate PRTs. In case the linkage is poor, a other preferred diagenetic zonation scheme), one
loopback to Step 5 is required. Definition of rejec- can choose to generate PRT maps per selected inter-
tion criteria is field specific and depends on the val. Obviously, data density may limit the options
type and reliability of dynamic data. If dynamic during this phase of investigation. The resulting
data are available, the workflow includes the follow- maps serve as input to a (so-called) facies prob-
ing three steps: (7.1) validation of PRTs with core ability cube or for soft conditioning during the
data; (7.2) validation of PRTs with dynamic data; static modelling step. The distribution method is
and (7.3) PRT conformance assessment. This step also a function of RT and data scenario. In other
might be skipped if reliable dynamic data are words, spatial patterns of – and interrelationships
not available. between – PRTs are controlled by the relative con-
tribution of depositional v. diagenetic processes
(Fig. 2). In case RT ¼ 1, PRTs and flow are con-
Step 8: 3D PRT realization and spatial trolled by DRTs. In addition to RT, the data
validation density also controls the choice of methods to distri-
bute PRTs in the static model. With increas-
This step comprises the completion of the spatial ing density of data, the spatial trends are driven by
trends and interrelation rules for PRTs extracted in well control rather than concepts. Multiple point
Step 6. Those rules and trends can be used as soft statistics (MPS) is a pixel-based technique and is
constraints (probability maps) and/or to design one of the more powerful geostatistical tools to
training images or variograms for the distribution honour and control proportions and trends of –
of PRTs in 3D static models. Trend maps can be and spatial interrelationships among – PRTs (i.e.
generated in many of the existing geomodelling Strebelle 2000, 2002; Strebelle & Zhang 2004;
platforms and are generally labelled as facies pro- Caers 2005). Examples of PRT definitions and dis-
portion maps. These PRT proportion maps can be tribution using MPS are shown in the section on
manually modified to fit patterns and trends from ‘Examples’, later in this paper. High-resolution
concepts or forward models in case the data den- and quality 3D seismic (attribute) data can be used
sity is low. Depending on the sequence stratigraphic as a soft constraint for PRTs, especially for green
character of the grid layering (i.e. packages of layers fields or intermediate fields where well density is
that represent transgressive systems tracts (TST), not capturing spatial heterogeneity. Seismic attri-
highstand systems tracts (HST), lowstand systems butes can be built into the PRT definition follow-
tracts (LST) or regressive wedges (RST), or any ing careful validation with log-derived acoustic
CARBONATE ROCK TYPING 243

Fig. 7. (a) In case reliable dynamic data are available, a quantitative validation of the link between PRTs and available
flow indicators is performed. This should include a comparison of PRTs with core RCA data and modified Lorenz plots,
and PRTs with dynamic data. The ultimate purpose of this step is to evaluate the conformance between PRTs and flow
profiles observed from dynamic data. (b) Diagram showing an example of the application of Step 7 to Tengiz. Here a
quick visual inspection confirmed that non-reservoir PRT1–PRT3 correspond to non-flow zones, while reservoir
PRT4– PRT6 correspond to the inflow zones (green bars). CGR/SGR, total gamma ray minus uranium contribution and
total gamma ray (CGR, when available, preferred over SGR); PHIE, effective porosity PRT, petrophysical rock type;
PLT, production logging tools; PERM, permeability.
244 M. SKALINSKI & J. A. M. KENTER

properties. In poor data scenarios, many other tech- function of diagenesis at the sub-micron scale (e.g.
niques can serve to reduce uncertainty in the final Algive et al. 2012; van der Land et al. 2013). Dis-
distribution of PRTs. For example, analogues from crete fracture networks can be distributed using con-
outcrops or the literature or forward stratigraphic cepts derived from analogues (e.g. Narr & Flodin
modelling tools can assist to reduce uncertainty 2012) or predicted using mechanical Earth model-
and to generate trends and juxtaposition patterns ling (MEM) or geomechanical forward modelling.
(or variogram design) in RT1 cases when data These mechanical modelling techniques provide
density is poor and not capturing heterogeneity deformation and stress information that are used
and/or core is not representative of (anticipated) to model fracture distributions (e.g. Xi et al. 2011;
reservoir quality. Forward stratigraphic modell- Abul Khair et al. 2013), as well as to allow esti-
ing platforms include SEDPAK (Kendall et al. mation of the fracture behaviour (and flow) during
1989), STRATA (Flemings & Grotzinger 1996), the development of the field. Fracture flow proper-
CARBONATE 3D (Bosence & Waltham 1990), ties can be embedded in the PRT definition by
REPRO (Hüssner & Roessler 1996) and DIONISOS reconciling the combined effect of matrix and frac-
(Granjeon 1997). In data scenarios where diagenetic tures on porosity and permeability. However, such a
modification is dominant (RT2 – RT3), diagenetic decision is complex and depends on the observed
forward modelling tools can be used to generate contribution to flow from dynamic data. Alternati-
trends and patterns for the associated PRTs. Such vely, fractures can be distributed as discrete objects
software platforms have been designed and pub- and co-located with PRTs (if desired) in the static
lished and are available to the user. Predictive or dynamic model (i.e. Dershowitz et al. 1998;
diagenesis modelling techniques in carbonates are Bourbiaux et al. 2002; Jonoud et al. 2013). Since
in their infancy. CARB3D+ is a forward simula- the modelling and distribution approach of fractures
tion model for carbonate platforms developed by in the static and/or dynamic model is very complex
the University of Bristol, which incorporates near- and receiving much discussion, such an extensive
surface (eogenic) diagenetic processes and predicts review is excluded from this paper.
the distribution of diagenetic products (Whitaker Figure 8 summarizes the discussion above on the
et al. 1997). Reactive transport modelling (RTM) relationship between data scenario, RT, and tools
code, such as TOUGHREACT developed by the to improve or optimize the distribution of PRT pat-
Lawrence Berkeley National Laboratory for terns and juxtaposition. In addition, it shows the
environmental studies, has been used in a variety positions of the fields that are used in the following
of carbonate studies both on outcrops and reservoirs, ‘Examples’ section of this paper. The following
but requires a separately built ‘stratigraphic’ geocel- four steps are included in the PRT distribution and
lular model as input to the reactor (Pruess et al. spatial validation process: (8.1) determination of
1999). Recent publications have clearly dem- PRT trends and spatial interrelationships assisted
onstrated their value in assessing the spatial dis- by forward modelling techniques and/or ana-
tribution of diagenetic products, and, hence, in logues; (8.2) determination of optimal geostatistical
designing trends and patterns (e.g. Whitaker et al. method; (8.3.1) when applying stochastic geomo-
1997; Jones et al. 2003; Jones & Xiao 2006; Pater- delling algorithms, analyse variogram lengths, gen-
son et al. 2006; Garcia-Fresca 2009; Barbier et al. erate the PRT probability cube and run the PRT
2012; Frazer et al. 2012). Ahr (2008) not only distribution; (8.3.2) in the case of MPS, design train-
provides a comprehensive workflow for assessing ing image(s) (TIs), generate the PRT probability
associated pore types and systems, as well as the cube and run the PRT distribution; and (8.4)
genetic models of diagenetic modification, the quality control and evaluation of the resulting PRT
workflow also provides suggestions on their spa- distribution and, if required, loop back.
tial trends. The application and use of forward
modelling platforms varies with the data scenario
and specific geological demands, and cannot be Examples
further detailed in this workflow. What is critical,
however, is that certain PRTs have been designed The workflow described above was applied to the
to contain high-permeability (high-K ) elements in supergiant Tengiz Field in Kazakhstan, as well as
their definition and, hence, distribution. This is rel- the high-porosity First Eocene reservoir interval of
evant for further upscaling and design of PRTs the Wafra Field in the Partitioned Zone (Kuwait –
and flow features in the simulation model. In Saudi Arabia).
addition to forward modelling depositional and
diagenetic spatial trends at the metre to hundreds Tengiz Field
of metres scales, promising results of 3D pore archi-
tecture models have been obtained for modelling The Tengiz Field is located near the NE shore
the evolution of porosity and permeability as a of the Caspian Sea in Kazakhstan (Fig. 9). It is
CARBONATE ROCK TYPING 245

Fig. 8. Schematic diagram summarizing the relationship between data scenario, RT and tools to improve or optimize
the distribution of PRT patterns and juxtaposition following the discussion in the text. In addition, it shows positions
of the fields that are used in the ‘Examples’ section of this paper (see also Figure 3b). The diagram assists the
integrated reservoir management team by offering the appropriate combinations of PRT trends, data scenario and tools,
and/or concepts that may help to optimize a realistic set of spatial distributions of PRTs to capture the uncertainty in
reservoir quality and flow. Clearly, this is part of the road map and many deviations to those solutions shown above
will exist.

one of several large hydrocarbon accumulations Tengiz is a stratigraphic trap consisting of a car-
in the carbonate formations found around the edge bonate build-up with a large central platform area
of the Pricaspian Basin (Lisovsky et al. 1992). (10 × 15 km), a raised rim area (up to 1 –2 km

Fig. 9. Simplified map showing the locations of the Tengiz Field and Wafra (First Eocene) Field.
246 M. SKALINSKI & J. A. M. KENTER

wide), and is surrounded by an upper-slope micro- gamma, array induction, neutron, density, dipole
bial boundstone sector and a lower-slope debris sonic, FMI, and NMR tools. Pressure mea-
apron. The field is divided into platform and slope surements from the well modular formation
elements (Weber et al. 2003; Collins et al. 2006; dynamics tester tool (MDT), PLT and vertical
Kenter et al. 2006). The platform deposits are divi- sonic profile (VSP) data were also available for
ded into Unit 1 (Late Viséan–Bashkirian), Unit 2 every well. As a result, the data scenario for
(Tournaisian– Late Viséan) and Unit 3 (Devonian). the platform is borderline to capturing reservoir
The Unit 1 platform is the best reservoir with an heterogeneity, while that for the margin, with
average porosity of 9% and target of this example. fewer wells but similar data quality, is con-
sidered moderate (see Fig. 8).
PRT workflow application. The Tengiz Field was † Step 2: Depositional rock typing. The Unit 1
used as a first application of the petrophysical rock platform area is generally flat with parallel
typing (PRT) workflow. The following is a short depositional cycles, which are made up of a suc-
description of the workflow steps outlined earlier cession of generally shoaling DRTs overlying a
in this paper: sharp base with evidence for subaerial exposure
† Step 1: Data scenario. The PRT definition for the and/or flooding. Cycle boundaries are gener-
platform was based on 14 cored wells with a total ally associated with early cementation and the
of 2650 m of core, which were assigned DRTs, presence of solid and dispersed volcanic ash,
and plugged every 1 ft for permeability, porosity and are well defined by log expression and
and water-saturation analyses (CCAL or RCA). relatively easy to correlate across the central
Plug trims were subject to total organic carbon platform. The outer platform is the gently basin-
(TOC) determination to quantify solid bitumen. ward-dipping (up to 98) sector, several hundreds
More than 350 MICP measurements were used of metres in width, connecting the central plat-
for matrix pore typing. Modern logging suites form with the upper slope. Here, cycles are gen-
were available for 15 wells and included spectral erally poorly defined as a result of increasing

Fig. 10. Diagrams illustrating steps 2 and 3 in the Tengiz PRT workflow. (a) Diagram showing the lumped DRT set
consisting of a generally shoaling-upward cycle of DRTs; DRT7 is only present in the deeper outermost platform.
(b) Diagram showing four major porosity enhancing/occluding episodes, each of which had a particular process and
associated spatial trend (modified after Kenter et al. 2010).
CARBONATE ROCK TYPING 247

diagenetic overprint and the presence of tight was used for the PRT definition. The resulting
microbial boundstone intervals deposited SSF was used for further PRT distribution in
during early flooding of the platform (Kenter Step 6.
et al. 2006). The ‘reservoir’ in between such † Step 3: Reservoir typing. The study of the diage-
baffled cycle boundaries was overprinted by netic modification of the original DRTs revealed
meteoric and (to a lesser extend) late burial dia- nine steps in a complete paragenetic sequence
genesis increasing porosity by dissolution in the that collapsed to four major porosity enhancing
centre of the platform, and reduced porosity and/or occluding episodes (Fig. 10b). Each
through compaction, calcite cementation and of these diagenetic episodes had a particular
pyrobitumen cementation towards the margins. process and associated spatial trend, which
Geological descriptions identified a set of 15 were used to construct diagenetic trend maps.
DRTs using texture, depositional setting and Figures 3c, d illustrate, respectively, the confor-
grain types. Those DRTs were reduced to a mity of porosity –permeability trends and the
final set of seven DRTs (Fig. 10a), which were predominance of diagenetically altered pores in
correlated across the platform using sequence the Tengiz platform, while limited conformity
stratigraphic principles. However, due to pre- was observed between DRTs and logs. The
dominantly diagenetic control (see below), observed high-K features, strata-bound zones
only the DRT corresponding to volcanic layers of connected vugs and high-permeability matrix,

Fig. 11. Diagrams explaining the PRT definition step in the Tengiz Platform. (a) PRT1 corresponds to the volcanic tuff
recognized from spectral gamma (thorium), and is related to DRT2 and linked to the early cementation stage. PRT2 was
defined as a rock with solid bitumen volume exceeding an effective porosity (occupying more than 40% of the pore
space), determined from Multimin analysis. (b) PRT3 corresponds to tight cemented rock as a result of late cementation
and was defined as a rock dominated by microporosity, with PHIE (effective porosity) below 5%. (c) The PRT4–PRT6
definition was derived from clustering NMR T2 distributions, and combining them with the effective porosity scale (30
clusters or bins of relaxation times) and cross-plotted against porosity. (d) PRT4– PRT6 correspond to an increasing
degree of the corrosion and represent the reservoir quality rock (modified after Skalinski et al. 2009).
248 M. SKALINSKI & J. A. M. KENTER

Fig. 12. (a) T2 distributions combined over the approximately 500 m reservoir interval partitioned by PRT. The
signal distribution on the vertical axis reflects the volumetric proportions, normalized to 1, corresponding to the pore
volumes associated with each pore size shown on the horizontal axis. (b) Well plot confirming that non-reservoir
PRT1–PRT3 have a large spread of T2 distribution mostly below 200 ms, while reservoir PRT4–PRT6 are
characterized by higher T2, indicating increased amount of vuggy porosity (modified after Skalinski et al. 2009).

and conductive fractures from BHI and FMI, Institute common scale units from a US test pit).
were confirmed by well tests and production This PRT is related to DRT2 and linked to the
logs. As a result, the RT was defined as a early cementation stage (Fig. 10a, b). PRT2
Hybrid RT4 for the margin, while close to diage- was defined as a rock with solid bitumen satur-
netic RT2 for the platform (see Fig. 3b). ation exceeding 40%. Bitumen saturation was
† Step 4: Pore typing. MICP-based pore throat dis- defined as a percentage of bitumen volume in
tributions led to the definition of five matrix pore reference to the original porosity (Fig. 11a),
types; that is, nano-, micro-, meso-, macro- and determined from multi-log mineralogical analy-
megaporosity (Skalinski et al. 2009). Pore types sis (i.e. using Multimin in Geolog from Paradigm
were used to validate NMR-derived PRT4– formation evaluation software; Quanti.Elan in
PRT6 from the clustering of T2 distributions. Techlog from Schlumberger). The Multimin
Vuggy porosity was captured by NMR and model was calibrated to core data to assure reli-
FMI logs. able quantification of both parameters. Bitumen
† Step 5: PRT definition. PRTs were defined at the was formed during late burial cementation (Fig.
log scale, as indicated in Figure 9 (Skalinski 10b). PRT3 corresponds to tight-cemented rock
et al. 2009). PRT1 corresponds to the volcanic as a result of late cementation. This PRT was
tuff recognized from spectral gamma (thorium) defined as a rock dominated by microporosity
logs, using a 15 API cutoff (American Petroleum (pore throat ,0.3 mm) and porosities below
CARBONATE ROCK TYPING 249

Fig. 12. Continued.

5% (Fig. 11b). PRT4–PRT6 correspond to an large spread of T2 distribution mostly below


increasing degree of the corrosion and represent 200 ms, while reservoir PRT4–PRT6 are charac-
the reservoir quality rock. Those PRTs were terized by higher T2, indicating an increased
defined from clustering NMR T2 distribu- amount of vuggy porosity.
tions and combining them with the effective † Step 6: Determination in a multi-well setting and
porosity scale. This process allowed the capture quality control using maps. PRTs defined in Step
of both pore types (degree of ‘vugginess’) and 5 were determined in all platform wells with
porosity. Figure 11c illustrates the definition modern well suits following specific cutoffs
of PRT4–PRT6 from T2 clusters (groupings) or clustering incorporated into the master petro-
cross-plotted against log porosity. Figure 11d physical evaluation code (i.e. Loglan in Geolog;
shows a depth plot with (from left to right): spec- Paradigm formation evaluation software; sim-
tral gamma ray, depth, core and log porosity, ilar routines available in Techlog from Schlum-
PRT and T2 distribution. Figure 12a contains T2 berger). The quality control step consisted of
distributions combined over the approximately comparing PRT maps with diagenetic and depo-
500 m reservoir interval partitioned by PRT. sitional trends. Care was taken to eliminate out-
As expected, non-reservoir PRT1–PRT3 have a liers and minimize overlap. The PRT maps were
250 M. SKALINSKI & J. A. M. KENTER

Fig. 13. (a) Illustration showing the mapped relationship between DRTs and resulting PRTs. (b) Simplified
explanation of the PRTs in terms of relative depositional and diagenetic control, where DRT1– DRT3 are non-reservoir,
and PRT4– PRT6 represent fair to very good reservoir.

used to define lateral trends used for the creation juxtaposition rules (Fig. 13a, b) required for
of MPS training images in Step 8. building appropriate training images, as well as
† Step 7: Dynamic validation of PRTs. In Tengiz, the soft probability cube for PRT conditioning.
the dynamic validation was performed by com- Some of the PRTs, although diagenetically
paring PRTs with PLT profiles (see Fig. 7b). modified, still mostly resemble the original
This exercise confirmed that non-reservoir DRTs and conform to primary depositional
PRT1–PRT 3 correspond to non-flow zones, spatial trends. Other PRTs have trends that are
while reservoir PRT4–PRT6 correspond to the the result of diagenetic processes, truncate the
inflow zones (green bars in Fig. 7b). This confor- SSF and change within cycles (Fig. 13a, b).
mance was successfully assessed in all wells with The interesting observation is that the timing of
PLTs. those diagenetic processes played a crucial
† Step 8: PRT realization and spatial validation. role: PRT1 and PRT3 were modified early and
During steps 5 –7 the relationship between formed baffles with reduced porosity and per-
PRTs and DRTs was carefully monitored to meability bordering cycles or ‘containers’.
maintain optimal relationships for spatial distri- These containers were later overprinted by cor-
bution in the model. The resulting mapping of rosion, mostly in the centre of the platform,
processes and spatial characteristics of DTs in and cementation towards the NE and eastern
PRTs is shown in Figure 13a. Following the sectors in the field where calcite and bitumen
prediction of the carefully defined – and vali- cements occluded porosity. Key elements of
dated by dynamic data – set of PRTs in all the static model workflow included training
wells, kriged PRT maps were generated for all images, vertical proportion curves (VPC) and
SSF zones. Those maps, along with the earlier facies probability cubes. The static geomodel-
observations on geological attributes (deposi- ling process utilizes MPS, since this handles
tional and diagenetic), resulted in a linkage complex juxtaposition while preserving pro-
between DRTs and PRTs that provided portions better than most other available model-
additional information on spatial trends and ling platforms. The MPS modelling workflow
CARBONATE ROCK TYPING 251

Fig. 14. Simplified diagram summarizing the geomodelling steps using multiple point statistics (MPS). Only five PRTs
were distributed in this realization as the pre-Joint Venture (Russian) well group was not capable of predicting PRT6.
The key in the lower-left corner shows the ‘lumping’, while Figure 13b provides the key for the PRTs. The modelling
steps run from (a) the generation of training images capturing relative proportions and juxtaposition rules, to (b) the
calculation (and, if required, manual modification) of vertical proportion curves, to (c & d) kriged PRT maps (from the
determined well PRTs) that were manually adjusted to generate more conceptually realistic trends, to (e) the soft
probability cube. Cross-sections of the final (base case) realization are shown in the centre of the diagram (modified after
Kenter et al. 2010).

used for the Tengiz platform has been published in the interwell areas, and, next, convolved with
by Kenter et al. (2010) and is briefly summarized the VPCs to generate a soft probability cube
below. The training images (Fig. 14a) depict (Fig. 14e). Following a declustering step to
PRT body shapes, relative dimensions, size– account for clustering of pre-Joint Venture and
frequency distribution and (in this scenario) 3D modern wells, for each sequence the soft prob-
spatial interrelationships from core and pre- ability cubes were convolved with the training
dicted well observations, as well as from geo- images. Finally, porosity, water saturation and
logical concepts supplemented by information permeability were distributed in the PRT model
from forward models and outcrops analogues using standard stochastic methods.
(for more detailed information, see Kenter et al.
2010). Vertical proportion curves (Fig. 14b) Wafra Field
display the proportion of facies per layer in the
geocellular model, while facies probability The First Eocene reservoir at Wafra Field, located in
cubes integrate information from the wells the onshore area of the Partitioned Zone between
through kriged PRT maps, seismic or geological Saudi Arabia and Kuwait (Fig. 9), is a heavy oil car-
interpretation, and the VPC into facies probabil- bonate reservoir with an average porosity of 37%,
ities for each cell. Kriged PRT maps (Fig. 14c) which is exceptionally high for carbonates (Med-
were manually adjusted to generate more con- daugh et al. 2013). The total onshore oil production
ceptually realistic trends (Fig. 14d), especially in the Partitioned Zone reached 3 billion (3 × 109)
252 M. SKALINSKI & J. A. M. KENTER

Fig. 15. Graph showing the pore throat radius histograms (from MICP) groupings that were used for further lumping.

barrels in late 2004, with production from four heterogeneity, it also has a high degree of clus-
fields. In 2009, a pilot steam injection project tering. Since logging suites are generally poor,
began at the carbonate First Eocene reservoir at the data scenario is considered moderate (see
the Wafra Field. Steam flooding involves injecting Fig. 8).
steam into heavy oil reservoirs to heat the crude † Step 2: Depositional rock typing. DRTs were
oil underground, reducing its viscosity and allowing described in 10 cored wells with eight initial
its extraction through wells. groups, which were lumped to three facies
assemblages (Fig. 16). The heavy oil reservoir
PRT workflow application. The rock typing was per- consists of dolomitized subtidal packstone and
formed as part of the model to support the steam grainstone deposited under arid or semi-arid
injection project. The eight-step PRT workflow conditions in a shallow, very gently dipping
was applied to predict and distribute the dominant restricted ramp environment. The shallowing-
petrophysical rock types: upward cycles are capped by mud-dominated
† Step 1: Data scenario. The resulting PRT set was tidal flat facies, which are followed or replaced
defined using a subset of 10 cored wells and by evaporites indicating occasionally hypersa-
applied to 113 wells in the model area. Nearly line lagoons and sabkhas. DRTs were further
120 MICP samples were used to define throat- lumped to three to allow reliable prediction from
based pore types. The 56 wells in the steam logs: (1) evaporites (anhydrite and gypsum);
injection pilot area include 15 wells with triple (2) tidal flat capping facies including algal-
combo (standard set of measurements used in dominated grainstone, and dolomitic wackes-
formation evaluation and wireline logging: tone and packstone; and (3) subtidal (dolomitic)
gamma ray, porosity and resistivity), and NMR packstone. The lumping honoured both deposi-
and FMI. However, for the sake of consistency, tional patterns and similarity in logging space.
PRTs were derived from the triple combo log The three lumped DRTs were predicted in a
suite present in all wells. Although the well reliable way from logs using Multimin for eva-
density is high locally and capturing reservoir porites and multivariate stepwise discriminant
CARBONATE ROCK TYPING 253

Fig. 16. (a) Porosity– permeability cross-plot with lumped pore types from seven MICP analysis to three lumped pore
types that were predictable from core and log data. (b) For the purpose of further prediction from logs, pore types were
further lumped to two groups: macro- (macro-lumped with meso-) and microporosity. (c) As a result of the RT1
assessment, PRTs were defined by first considering DRT and then by incorporating diagenetic pore types. Hence, both
the tidal flat DRT2 and subtidal DRT3 were subdivided into two groups based on pore types. See the text for discussion.

analysis for the remaining DRTs, with a predic- of further prediction from logs, pore types were
tion score of 80% (Fig. 2c). further lumped to two groups: macro- (PT1 and
† Step 3: Reservoir typing. The determination PT2) and microporosity (PT3, see Fig. 16b).
of RT was based on the conformance of Examination of NMR and FMI logs helped to
DRT prediction from logs and examinations identify moldic porosity, which was lumped
of porosity–permeability cross-plots by DRTs. with macroporosity as it did not stand out on
The strong diagenetic modification by dolomiti- the porosity–permeability (phi–K ) cross-plot
zation did not cross-cut depositional trends and as a distinct group.
therefore the reservoir type was considered † Step 5: PRT definition. The PRTs were first
as RT1 (Fig. 3b). defined as a combination of predicted DRTs
† Step 4: Pore typing. MICP data allowed the defi- and pore types. Since the reservoir type is close
nition of seven PT groups, which were further to depositional, the PRTs were defined by con-
lumped into three lumped pore types predictable sidering first DRT and incorporating diagenetic
from core and log data. Figure 15 shows the first- pore types in the second step in which, both,
generation MICP groupings, and Figure 16a the tidal flat DRT2 and subtidal DRT3 were sub-
gives a porosity –permeability cross-plot with divided into two groups based on pore types (see
lumped pore types as colours. Lumped pore Fig. 16c). The five resulting PRTs defined this
types were: PT1, dominated by macroporosity; way were predicted from logs using stepwise dis-
PT2, dominated by mesoporosity; and PT3, criminant analysis with a prediction score of
dominated by microporosity. For the purpose 77% using conventional logs only. In order to
254 M. SKALINSKI & J. A. M. KENTER

Fig. 17. Diagram showing the strategy of nesting PRTs within DRTs for the top stratigraphic interval as an example of
the areal quality control check of the predicted PRTs. Nesting of PRTs followed the initial distribution of DRTs using
dedicated DRT training images defined using ramp analogues and available core descriptions. Note that TF, ST and E
stand for, respectively, tidal flat facies, subtidal facies and evaporite, see the text for discussion. Figure is courtesy of
M. Andres and M. Levy.

Fig. 18. Cross-section over the resulting sector model showing the distribution of PRTs, which were simulated using
‘PRTs nested in DRTs’ in MPS. Note that TF, ST and E stand for, respectively, tidal flat facies, subtidal facies and
evaporite, see the text for discussion. Figure courtesy of M. Andres & M. Levy.
CARBONATE ROCK TYPING 255

preserve low-permeability baffles in the model, workflow and demonstrates its workings and
PRT2 was further subdivided into two PRTs, agility through deployment in two large carbonate
based on an evaporite threshold of 35%. fields. This workflow is novel in several ways: (1)
† Step 6: Determination in a multi-well setting and it combines geological processes, petrophysics and
quality control using maps. The PRTs defined in Earth modelling aspects of rock typing in one com-
Step 5 were determined in all wells from the prehensive approach; (2) it integrates core and log
model area. The PRT maps helped to define scales, and provides consistent input into reservoir
the dimension and shape of the depositional models in the log domain; and (3) it provides a flex-
‘bodies’ and their spatial trends. ible ‘road map’ from core to 3D model for variable
† Step 7: Dynamic validation of PRTs. The data scenarios such as prospects, appraisal and pro-
modular formation dynamics tester tool pressure ducing fields that can be updated with progressive
profiles were successfully used to validate changes in data quality and quantity during the life
evaporite barriers. cycle of an asset.
† Step 8: PRT realization and spatial validation.
The resulting PRTs at the wells were sub- The authors would like to thank the Kingdom of Saudi
Arabia Ministry of Petroleum, Saudi Arabian Chevron,
sequently analysed in map and in cross-section the Kuwait Gulf Oil Company, the Kuwait Ministry of
view to unravel their spatial trends and juxta- Oil and TengizChevroil for permission to publish the
position relationships, necessary to construct paper. Discussions with, and support and encouragement
training images and distribute the PRTs in the from, many peers and colleagues significantly improved
3D static model. PRTs were distributed in the the workflow and contributed to this paper. In particular,
model using MPS. First, DRT training images we would like to acknowledge M. Andres and M. Levy
were defined using ramp analogues and available for their contribution to the PRT modelling in the (First
core descriptions. Next, PRT distribution and Eocene) Wafra example, M. Harris for continuously
testing the underlying principles and concepts, and
training images were nested within correspond-
A. Latham and F. Harris for trust and committing funds
ing DRTs as per definition (Fig. 17). Figure 18 to the research project. Lastly, we would like to thank
shows a cross-section over the sector model, the reviewers, J. Lucia, P. Corbett and S. Geiger, for con-
with PRTs simulated using MPS. structive and positive critique, which greatly improved the
quality of this paper.

Conclusions
References
Conventional carbonate rock typing workflows are
based either on textural properties, pore type clas- Abbaszadeh, M., Koide, N. & Murahashi, Y. 2000.
sification, the application of electrofacies or Integrated characterization and flow modeling of a
flow zone indicators (FZI), and are biased either Heterogeneous Carbonate Reservoir in Daleel Field,
Oman. (SPE paper 62514.) SPE Reservoir Evaluation
towards depositional or towards diagenetic attri- & Engineering, 3, 150–159.
butes. The resulting proportions and spatial distri- Abul Khair, H., Cooke, D. & Hand, M. 2013. Natural
butions of rock types and estimated flow may, fracture networks enhancing unconventional reservoirs
therefore, be poor estimates of the real subsurface producibility. Paper presented at the AAPG Annual
distributions. This workflow addresses this bias in Convention and Exhibition, Pittsburgh, Pennsylvania,
a comprehensive way by determining petrophysical 19–22 May 2013. http://www.searchanddiscovery.
rock types, which control static properties and com/documents/2013/41182abulkhair/ndx_abulk
dynamic behaviour of the reservoir while optimally hair.pdf.
Ahr, W. M. 2008. Geology of Carbonate Reservoirs.
linking the geological attributes (depositional and
Wiley, New York.
diagenetic as well as their hybrid), and their spatial Algive, L., Békri, S., Nader, F. H., Lerat, O. & Vizika,
interrelationships and trends. The integrated work- O. 2012. Impact of diagenetic alterations on the petro-
flow consists of eight composite and sequential physical and multiphase flow properties of carbonate
steps, which account for progressing field maturity rocks using a reactive pore network modeling
and data scenarios, including matrix and secondary approach. Oil & Gas Science and Technology –
pore types such as vugs and fractures. Our new Revue d’IFP Energies nouvelles, 67, 147–160.
workflow validates to dynamic data and optimizes Al-Jenaibi, F., Hammadi, K. & Salameh, L. 2008. New
the representation of geological attributes (deposi- methodology for optimizing field development plan,
tional and diagenetic), as well as their spatial trends why do we need to introduce dynamic rock typing.
(SPE paper 117894.) Paper presented at the Abu
and juxtaposition rules. Initial tests on Tengiz Field Dhabi International Petroleum Exhibition and Confer-
and Wafra Field were highly successful and sup- ence, held in Abu Dhabi, UAE, 3– 6 November 2008.
port the applicability of the workflow in different Amaefule, J. O., Altunbay, M., Tiab, D., Kersey, D. G.
data scenarios and reservoir types. This paper intro- & Keelan, D. K. 1993. Enhanced reservoir descrip-
duces the rationale behind this integrated tion: Using core and log data to identify hydraulic
256 M. SKALINSKI & J. A. M. KENTER

(flow) units and predict permeability in uncored inter- Chekani, M. & Kharrat, R. 2009. Reservoir rock typing
vals/wells. (SPE paper 26436.) Paper presented at the in a carbonate reservoir – Cooperation of core and log
68th Annual Technical Conference and Exhibition of data: Case study. (SPE paper 123703.) Paper presented
the Society of Petroleum Engineers, Houston, TX, at the SPE/EAGE Reservoir Characterization and
3 –6 October 1993. Simulation Conference in Abu Dhabi, UAE, 19–21
Archie, G. E. 1952. Classification of carbonate reservoir October 2009.
rocks and petrophysical considerations. American Choquette, P. W. & Pray, L. C. 1970. Geologic nomen-
Association of Petroleum Geologists Bulletin, 36, clature and classification of porosity in sedimentary
278– 298. carbonates. American Association of Petroleum Geol-
Arfi, A. S., Heliot, D., Zhan, J. L. & Allen, D. 2006. A ogists Bulletin, 54, 207–250.
new porosity partitioning-based methodology for per- Clerke, E. A., Mueller, H. W., Phillips, E. C., Eyvaz-
meability and texture analysis in Abu Dhabi carbon- zadeh, R. Y., Jones, D. H., Ramamoorthy, R. & Sri-
ates. (SPE paper 101176.) Paper presented at the Abu vastava, A. 2008. Application of Thomeer Hyperbolas
Dhabi International Petroleum Exhibition and Confer- to decode the pore systems, facies and reservoir prop-
ence, 5 –8 November 2006. erties of the Upper Jurassic Arab D Limestone, Ghawar
Bagheri, A. M., Biranvand, B., Rezazadeh, S., field, Saudi Arabia: A ‘Rosetta Stone’ approach.
Fasih, M. & Bahtiari, H. 2005. Integrated analysis GeoArabia, 13, 113 –160.
of core and log data to determine reservoir rock Collins, J. F., Kenter, J. A. M., Harris, P. M., Kuany-
types and extrapolation to uncored wells in a hetero- sheva, G., Fischer, D. J. & Steffen, K. L. 2006.
geneous clastic and carbonate reservoir. Paper pre- Facies and reservoir quality variations in the Late
sented at the International Symposium of the Visean to Bashkirian outer platform, rim, and flank
Society of Core Analysts, Toronto, Canada, 21– 25 of the Tengiz buildup, Precaspian Basin, Kazakhstan.
August 2005. In: Harris, P. M. & Weber, L. J. (eds) Giant Hydro-
Barbier, M., Hamon, Y., Doligez, B., Callot, J. P., carbon Reservoirs of the World: From Rocks to Reser-
Floquet, M. & Daniel, J. M. 2012. Stochastic joint voir Characterization and Modeling. American
simulation of facies and diagenesis: A case study on Association of Petroleum Geologists, Memoirs, 88,
early diagenesis of the Madison formation (Wyoming, 55–95.
USA). Oil & Gas Science and Technology – Revue Corbett, P. W. M. 2010. Petrotype-based sampling
d’IFP Energies nouvelles, 67, 123–145. applied in a saturation exponent screening study,
Basioni, M. R. E., Negahban, S., Dawood, A., El Nubian Sandstone Formation, Sirt Basin, Libya. Petro-
Mahdi, A. & Bahamaish, J. 2008. Reservoir rock physics, 51, 264 –270.
typing from crest to flank is there a link? (SPE paper Corbett, P. W. M. & Potter, D. 2004. Petrotyping:
11.) Paper presented at the 2008 Abu Dhabi Inter- A basemap and atlas for navigating through per-
national Petroleum Exhibition and Conference, 5 –6 meability and porosity data for reservoir comparison
November. and permeability prediction. Paper presented at the
Bosence, D. W. J. & Waltham, D. 1990. Computer mod- SCA Annual Conference, Abu Dhabi, 5– 9 October
eling the internal architecture of carbonate platforms. 2004.
Geology, 18, 26–30. Cortez, O. D. & Corbett, P. W. M. 2005. Time lapse
Bourbiaux, B., Basquet, R., Cacas, M., Daniel, J. & production logging and the concept of flowing units.
Sarda, S. 2002. An integrated workflow to account (SPE paper 94436.) Paper presented at the European
for multi-scale fractures in reservoir simulation EAGE Annual Conference, Madrid, Spain, 13–16
models: Implementation and benefits. (SPE paper June 2005.
78489.) Paper presented at the Abu Dhabi International Cuddy, S. J. 1998. Litho-facies and permeability predic-
Petroleum Exhibition and Conference, Abu Dhabi, tion from electrical logs using fuzzy logic. Paper pre-
UAE, 13– 16 October 2002. sented at the Eighth Abu Dhabi International
Bust, V. K., Majid, A. A., Oleto, J. U. & Worthington, Petroleum Exhibition and Conference, Abu Dhabi,
P. F. 2011. The petrophysics of shale gas reservoirs: UAE, 11–14 October 1998a.
Challenges and Pragmatic solutions. (IPTC 14631.) Dernaika, M. R., Basioni, M. A., Dawoud, A. & Kalam,
In: International Petroleum Technology Conference, M. Z. 2012. Variations in bounding and scanning rela-
Volume 2. Society of Petroleum Engineers, Richard- tive permeability curves with different carbonate rock
son, TX, 1440–1454. types. (SPE paper 162265.) Paper presented at the Abu
Caers, J. 2005. Petroleum Statistics. Society of Petroleum Dhabi International Petroleum Exhibition and Confer-
Engineers, Richardson, TX. ence, 11–14 November 2012.
Cantrell, D. L. & Hagerty, R. M. 2003. Reservoir rock Dershowitz, B., Lapointe, P., Eiben, T. & Wei, L. 1998.
classification Arab D reservoir, Ghawar field, Saudi Integration of discrete feature network methods with
Arabia. GeoArabia, 8, 435– 462. conventional simulator approaches. (SPE paper
Carman, P. C. 1939. Permeability of saturated sands, 49069.) Paper presented at the SPE Annual Technical
soils and clays. Journal of Agricultural Sciences, 29, Conference and Exhibition, New Orleans, LA, 27–30
57–68. September 1998.
Chehrazi, A., Rezaee, M. R. & Rahimpour, H. Dunham, R. J. 1962. Classification of carbonate rocks
2011. Pore-facies as a tool for incorporation of small according to depositional texture. In: Ham, W. E.
scale dynamic information in integrated reservoir (ed.) Classification of Carbonate Rocks: A Symposium.
studies. Journal of Geophysics and Engineering, 8, American Association of Petroleum Geologists,
202– 224. Memoirs, 1, 108–121.
CARBONATE ROCK TYPING 257

Ebanks, W. J., Jr, Scheihing, M. H. & Atkinson, C. D. Hollis, C., Vahrenkamp, V., Tull, S., Mookerjee, A.,
1992. Flow units for reservoir characterization. In: Taberner, C. & Huang, Y. 2010. Pore system charac-
Morton–Thompson, D. & Woods, A. M. (eds) terization in heterogeneous carbonates: An alternative
Development Geology Reference Manual. American approach to widely-used rock typing methodologies.
Association of Petroleum Geologists, Methods in Marine and Petroleum Geology, 17, 772–793.
Exploration Series, 10, 282– 285. Hulea, I. & Nicholls, C. A. 2012. Carbonate rock
Embry, A. F. & Klovan, J. E. 1971. A late devonian reef characterization and modeling: Capillary pressure
tract on Northeastern Banks Island, NWT. Canadian and permeability in multimodal rocks – a look
Petroleum Geology Bulletin, 19, 730– 781. beyond sample specific heterogeneity. American
Flemings, P. B. & Grotzinger, J. P. 1996. STRATA – Association of Petroleum Geologists Bulletin, 96,
freeware for analyzing classic stratigraphic problems. 1627– 1642.
Geological Society of America Today, 6, 1 –7. Hüssner, H. & Roessler, J. 1996. Modeling of reef
Francesconi, A., Bigoni, F., Balossino, P., Bona, N., growth in a 3-dimensional space. In: Reitner, J., Neu-
Martchini, F. & Cozzi, M. 2009. Reservoir rock weiler, F. & Gunkel, F. (eds) Global and Regional
types application – Kashagan. (SPE paper 125342.) Controls on Biogenic Sedimentation. Goettinger
Paper presented at the SPE/EAGE Reservoir Charac- Arbeiten Geologischen Specialisten, 2, 397 –404.
terization Conference, Abu Dhabi, UAE, 19– 21 Johnson, K., Barnett, A. & Wright, V. P. 2010. An
October 2009. evaluation of existing carbonate pore system classifi-
Frank, S., Narayanan, R., Hansen, P. M. & Allen, D. cations and rock-typing approaches. Paper presented
2005. Carbonate rock typing using NMR data: A case at the AAPG Annual Convention and Exhibition,
study from Al Shaheen Field. (IPTC paper 10889.) New Orleans, LA, 11–14 April 2010.
Paper presented at the Offshore Qatar, International Jones, G. D. & Xiao, Y. 2006. Geothermal convection in
Petroleum Technology Conference, Doha, Qatar, 21– the Tengiz carbonate platform, Kazakhstan: Reactive
23 November 2005. transport models of diagenesis and reservoir quality.
Frazer, M., Hollis, C. & Whitaker, F. 2012. Numerical American Association of Petroleum Geologists Bulle-
reconstructions of post-rift burial diagenesis, lower tin, 90, 1251–1272.
carboniferous, Derbyshire Platform, UK. Paper pre- Jones, G. D., Smart, P. L., Whitaker, F., Rostrom, B. J.
sented at the AAPG Annual Convention, Long & Machel, H. G. 2003. Numerical modeling of reflux
Beach, CA, 22–25 April 2012. dolomitization in the Grosmont platform complex
Garcia-Fresca, B. 2009. Outcrop-constrained flow and (Upper Devonian), Western Canada sedimentary
transport models of reflux dolomitization. PhD thesis, basin. American Association of Petroleum Geologists
University of Texas at Austin, Austin, TX. Bulletin, 87, 1273–1298.
Ghedan, S. G. 2007. Dynamic rock types for generating Jonoud, S., Wennberg, O. P., Casini, G. & Larsen, J. A.
reliable and consistent saturation functions for simu- 2013. Capturing the effect of fracture heterogeneity on
lation models. Paper presented at the SPE/EAGE multiphase flow during fluid injection. SPE Journal,
Reservoir Characterization and Simulation Confer- 16, 194 –208.
ence, Abu Dhabi, UAE, 28–31 October 2007. Kazemi, A., Corbett, P. W. M. & Wood, R. 2012. New
Gomes, J. S., Ribeiro, T., Strohmenger, C. J., Negah- approach for geomodeling and dynamic calibration of
ban, S. & Kalam, M. Z. 2008. Carbonate reservoir carbonate reservoirs using porosity determined
rock typing – the link between geology and SCAL. system (PODS). Paper presented at the 74th EAGE
(SPE paper 118284.) Paper presented at the Abu Conference and Exhibition, Copenhagen, Denmark,
Dhabi International Petroleum Exhibition and Confer- 4– 7 June 2012.
ence, held in Abu Dhabi, UAE, 3 –6 November 2008. Kendall, C. G. St. C., Strobel, J., Tang, J., Moore, P.,
Granjeon, D. 1997. Modélisation stratigraphique déter- Cannon, R., Bezdek, J. & Giswas, G. 1989. Simu-
ministe, conception et applications d’un modèle diffu- lation of the West Texas, new Mexican Permian Gua-
sif 3D multilithologique. Mémoires de Géosciences dalupian basin margin – a response to eustatic change,
Rennes, 78. an example of SEDPAK. In: Harris, P. M. & Grover,
Guo, G., Diaz, M. A., Paz, F., Smalley, L. & Wanniger, G. A. (eds) Subsurface and Outcrop Examination of the
E. A. 2005. Rock typing as an effective tool for per- Capitan Shelf Margin. Society of Economic Paleontol-
meability and water saturation modeling: A case ogists and Mineralogists, Core Workshop, 13,
study in a clastic reservoir in the Oriente Basin. (SPE 423– 426.
paper 97033-MS.) Paper presented at the SPE Annual Kenter, J. A. M., Harris, P. M., Collins, J. F., Weber,
Conference and Exhibition, Dallas, TX, 9– 12 L. J., Kuanysheva, G. & Fisher, D. J. 2006. Late
October 2005. Visean to Bashkirian platform cyclicity in the central
Gunter, G. W., Finneran, J. M., Hartmann, D. J. & Tengiz buildup, Pricaspian basin, Kazakhstan: deposi-
Miller, D. J. 1997. Early determination of reservoir tional evolution and reservoir development. In:
flow units and integrated petrophysical method. (SPE Harris, P. M. & Weber, L. J. (eds) Giant Hydro-
paper 38679.) Paper presented at the SPE Annual carbon Reservoirs of the World: From Rocks to Reser-
Technical Conference and Exhibition, San Antonio, voir Characterization and Modeling. American
TX, 5– 8 October 1997. Association of Petroleum Geologists, Memoirs, 88,
Hamon, G. 2003. Two phase flow rock typing: another 7– 54.
approach. (SPE paper 84035.) Paper presented at the Kenter, J. A. M., Tankersley, T., Skalinski, M. T.,
SPE Annual Technical Conference and Exhibition, Harris, P. M., Levy, M., Dickson, T. & Jacobs, G.
Denver, CO, 5 –8 October 2003. 2010. Tengiz Field (Republic of Kazakhstan) Unit 1
258 M. SKALINSKI & J. A. M. KENTER

platform static model: Using a hybrid depositional – Petroleum Geologists, Annual Convention and Exhibi-
diagenetic approach. Paper presented at the SPE tion, Long Beach, California, April 22– 25, 2012.
Caspian Carbonate Technology Conference, Atyrau, Paterson, R. J., Whitaker, F. F., Jones, G. D., Smart, P.
Republic of Kazakhstan, 8– 10 November 2010. L., Waltham, D. & Felce, G. 2006. Accommodation
Kolodzie, S., Jr . 1980. Analysis of pore throat size and and sedimentary architecture of isolated icehouse car-
use of the Waxman– Smits equation to determine bonate platforms: insights from forward modeling with
OOIP in Spindle Field, Colorado (SPE paper 9382.) CARB3D + . Journal of Sedimentary Research, 76,
Paper presented at the SPE Annual Technical Confer- 1162– 1182.
ence and Exhibition, Dallas, TX, 21– 24 September Porras, J. G. & Campos, O. 2001. Rock typing: A key
1980. approach for petrophysical characterization and defi-
Lehmann, C., Al Hosany, K., Cobb, D. O. & Al-Hend, nition of flow units, Santa Barbara Field, Eastern Vene-
A. 2008. Rock typing of upper Jurassic (Arab) carbon- zuela Basin. (SPE paper 69458.) Paper presented at the
ates, offshore Abu Dhabi. (SPE paper 117889.) Paper SPE Latin American and Caribbean Petroleum Engin-
presented at the International Petroleum Exhibition eering Conference, Buenos Aires, Argentina, 25–28
and Conference, Abu Dhabi, UAE, 5 –6 November March 2001.
2008. Pruess, K., Oldenburg, C. & Moridis, G. 1999.
Lisovsky, N. N., Gogonenkov, G. N. & Petzoukha, TOUGH2 User’s Guide, Version 2.0. Report LBNL-
YU. A. 1992. The Tengiz oil field in the Precaspian 43134, Lawrence Berkeley Nat. Lab, Berkeley, CA.
Basin of Kazakhstan (former USSR) – supergiant of Ramakrishnan, T. S., Poosala, V. & Suel, T. 1999. On
the 1980s. In: Halbouty, M. T. (ed.) Giant Oil and rectangular partitionings in two dimensions: algor-
Gas Fields of Decade 1978– 1988. American Associ- ithms, complexity, and applications. In: Beeri, C. &
ation of Petroleum Geologists, Memoirs, 54, 101–122. Buneman, P. (eds) Proceedings of 7th International
Lønøy, A. 2006. Making sense of carbonate pore systems. Conference on Database Theory (ICDT ‘99). Springer,
American Association of Petroleum Geologists Bulle- New York, 236–256.
tin, 90, 1381– 1405. Ramakrishnan, T. S., Ramamoorthy, R., Fordham, E.,
Lucia, F. J. 1983. Petrophysical parameters estimated Schwartz, L., Herron, M., Saito, N. & Rabaute,
from visual descriptions of carbonate rocks: A field A. 2001. A model-based interpretation methodology
classification of carbonate pore space. Journal of Pet- for evaluating carbonate reservoirs. (SPE paper
roleum Petrology, 35, 629– 637. 71704.) Paper presented at the SPE Annual Technical
Lucia, F. J. 1995. Rock-fabric/petrophysical classifi- Conference and Exhibition held in New Orleans, LA,
cation of carbonate pore space for reservoir character- 30 September–3 October 2001.
ization. American Association of Petroleum Geologists Ramamoorthy, R., Boyd, A., Neville, T. J., Seleznev,
Bulletin, 79, 1275– 1300. N., Sun, H., Flaum, C. & Ma, J. 2008. A new work-
Lucia, F. J. 2007. Carbonate Reservoir Characterization: flow for petrophysical and textural evaluation of
An Integrated Approach, 2nd edn. Springer, carbonate reservoirs. Paper presented at the SPWLA
New York. 49th Annual Logging Symposium, 25– 28 May 2008.
Marzouk, I., Takezaki, H. & Miwa, M. 1995. Geologic Rebelle, M., Al-Neaimi, M. A., Ribeiro, M. T.,
controls on wettability of carbonate reservoirs, Abu Gottlib-Zeh, S., Valsardieu, B. & Moss, B. 2005.
Dhabi, U.A.E. Paper presented at the SPE Middle Quantitative and statistical approach for a new
East Oil Show, Bahrain, Manama, 11– 14, March, rock and log-typing model: Example of onshore Abu
1995. Dhabi Upper Thamama. (IPTC paper 10273.) Inter-
Marzouk, I, Takezaki, H. & Suzuki, M. 1998. national Petroleum Technology Conference, Doha,
New classification of carbonate rocks for reservoir Qatar, 21–23 November 2005.
characterization. (SPE paper 29883.) Paper presented Sagaaf, M. M. & Nebrija, E. L. 2003. A fuzzy logic
at the Eighth Abu Dhabi International Petroleum Exhi- approach for the estimation of facies from wire-line
bition and Conference, Abu Dhabi, UAE, 11– 14 logs. American Association of Petroleum Geologists
October 1998. Bulletin, 87, 1223–1240.
Masalmeh, S. K. & Jing, X. D. 2004. Carbonate SCAL: Salman, M. S. & Sameer, B. 2009. Rock typing: An inte-
Characterization of carbonate rock types for deter- grated reservoir characterization tool to construct a
mination of saturation functions and residual oil robust geological model in Abu Dhabi carbonate oil
saturations. Paper presented at the International Sym- field. Paper presented at the SPE/EAGE Reservoir
posium of the Society of Core Analysts, Abu Dhabi, Characterization and Simulation Conference in Abu
UAE, 5–9 October 2004. Dhabi, UAE, 19–21 October 2009.
Meddaugh, W. S., Osterloh, W. T. et al. 2013. The Serra, O. & Abbott, H. 1980. The contribution of
Wafra Field first eocene carbonate reservoir steam- logging data to sedimentology and stratigraphy. (SPE
flood pilots: Geology, heterogeneity, steam/rock inter- paper 9270.) Paper presented at the 55th Annual Fall
action and reservoir response. (SPE paper 158324.) Technical Conference and Exhibition of the Society
Paper presented at the SPE Annual Technical Con- of Petroleum Engineers of AIME, Dallas, TX, 21–24
ference and Exhibition, San Antonio, TX, 8– 10, September 1980.
October, 2012. Skalinski, M. T., Gottlib-Zen, S. & Moss, B. 2005.
Narr, W. & Flodin, E. 2012. Fractures in steep-rimmed Defining and predicting rock types in carbonates: An
carbonate platforms: Comparison of Tengiz reservoir, integrated approach using core and log data in
Kazakhstan, and outcrops in canning basin, NW Aus- Tengiz field. Paper presented at the SPWLA 46th
tralia. Paper presented at American Association of Annual Logging Symposium, 26– 29 June 2005.
CARBONATE ROCK TYPING 259

Skalinski, M. T., Kenter, J. A. M. & Jenkins, S. 2009. northern Oman. 8th Middle East Geosciences Confer-
Rock type definition and pore type classification of a ence, GEO 2008. GeoArabia, 13, 248.
carbonate platform, Tengiz Field, Republic of Kazakh- Walls, J., Morcote, A. & Devito, J. 2013. Shale reservoir
stan. Paper presented at the SPWLA 50th Annual evaluation improved by dual energy X-Ray CT imag-
Logging Symposium, 21–24 June 2009. ing. Paper presented at the International Symposium
Strebelle, S. 2000. Sequential simulation drawing struc- of the SCA, Napa Valley, 16– 19 September 2013.
tures from training images. PhD thesis, Department of Waravur, S., Shebl, H., Salman, S. M., Shibasaki, T. &
Geological and Environmental Sciences, Stanford Uni- Dabbouk, C. 2005. Reservoir rock typing in a giant
versity, Stanford, CT. carbonate. (SPE paper 93477.) Paper presented at the
Strebelle, S. 2002. Conditional simulation of complex 14th SPE Middle East Oil & Gas Show and Confer-
geological structures using multiple-point statistics. ence, Bahrain, 12– 16 March 2005.
Mathematical Geology, 34, (1), 1 –21. Whitaker, F., Smart, P., Hague, Y., Waltham, D. &
Strebelle, S. & Zhang, T. 2004. Non-stationary Bosence, D. 1997. Coupled two-dimensional diage-
multiple-point geostatistical models. In: Leuangth- netic and sedimentological modeling of carbonate plat-
ong, O. & Deutsch, C. V. (eds) Geostatistics Banff form evolution. Geology, 25, 175– 178.
2004, Volume 1. Springer, New York, 235–244. Weber, L. J., Francis, B. P., Harris, P. M. & Clark, M.
Sung, R. R., Clerke, E. A., Buiting, J. J. & Ettajer, A. 2003. Stratigraphy, lithofacies, and reservoir distri-
T. 2013. Integrated geology, sedimentology and petro- bution, Tengiz Field, Kazakhstan. In: Ahr, W. M.,
physics application technology for multi-modal car- Harris, P. M., Morgan, W. A. & Somerville, I. D.
bonate reservoirs. (IPTC paper 16988.) Paper (eds) Permo-Carboniferous Carbonate Platforms and
presented at the IPTC International Conference, Reefs. SEPM Special Publication 78 and AAPG
Beijing, China, 26–28 March 2013. Memoir 83, 351–394.
Thomeer, J. H. M. 1983. Air permeability as a function of Weger, R. J., Eberli, G. P., Baechle, G. T., Massa-
three pore-network parameters. Journal of Petroleum ferro, J. L. & Sun, Y. F. 2009. Quantification of
Technology, 35, 809 –814. pore structure and its effect on sonic velocity and per-
Zang, T., Hurley, N. F. & Zhao, W. 2009. Numerical meability in carbonates. American Association of Pet-
modeling of heterogeneous carbonates and multi-scale roleum Geologists Bulletin, 93, 1297–1317.
dynamics. Paper presented at the SPWLA 50th Annual Wibowo, A. S. & Permadi, P. 2013. A type curve for car-
Logging Symposium, TX. bonate rock typing. (IPTC paper 16663.) International
Ye, S. J. & Rabiller, P. 2000. A new toll for electro-facies Petroleum Technology Conference, Beijing, China,
analysis: multi-resolution graph-based clustering. 26–28 March 2013.
Paper presented at the SPWLA 41st Annual Logging Wolff, M. & Pellissier-Combescure, J. 1982. FACIO-
Symposium, 4– 7 June 2000. LOG – Automatic electrofacies determination. Paper
van Der Land, C., Wood, R., Wu, K., van Dijke, presented at the SPWLA Annual Logging Symposium,
M. I. J., Jiang, Z., Corbett, C. W. M. & Couples, Corpus Christi, 6– 9 July 1982.
G. 2013. Modelling the permeability evolution of car- Xi, G. F., Koutsabeloulis, N., Mohamad Hussein, A. &
bonate rocks. Marine and Petroleum Geology, 48, Ni, Q. 2011. Fracture modeling associated with geo-
1–7. logical events. Paper presented at the EAGE Workshop
Vahrenkamp, V. C., Creusen, A., Tull, S., Farmer, A., on Naturally & Hydraulically Induced Fractured
Mookerjee, A. & Al-Bahry, A. 2008. Multi-scale Reservoirs – From NanoDarcies to Darcies, Nafplio,
heterogeneity modelling in a giant carbonate field, Greece, 10–13 April 2011.

You might also like