SPE-196679-MS

An Integrated Workflow for Quantifying the Impact of Geological Uncertainty

and Modelling Decisions On STOIIP Estimates and History Matching - A
Case Study from the Middle East

Mohamed AlBreiki, Saeeda AlAmeri, Sebastian Geiger, and Patrick Corbett, Heriot-Watt University

Copyright 2019, Society of Petroleum Engineers

This paper was prepared for presentation at the SPE Reservoir Characterisation and Simulation Conference and Exhibition held in Abu Dhabi, UAE, 17 - 19 September
2019.

This paper was selected for presentation by an SPE program committee following review of information contained in an abstract submitted by the author(s). Contents
of the paper have not been reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect
any position of the Society of Petroleum Engineers, its officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written
consent of the Society of Petroleum Engineers is prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300 words; illustrations may
not be copied. The abstract must contain conspicuous acknowledgment of SPE copyright.

Abstract
An innovative multi-deterministic scenario workflow was applied to one of the giant and complex
carbonate reservoirs in the Middle East. The application of this workflow had the objective to quantify
how geological uncertainties and different modelling decisions impact the stock tank oil-initially-in-place
(STOIIP) estimates and flow behaviour in this reservoir. In particular, we focused on the uncertainties related
to the presence of fractures, reservoir rock typing, and modelling the initial hydrocarbon distribution.
Based on the available static and dynamic data we considered two key scenarios, the absence of
fractures and the presence of sparse, fault-controlled fractures. In the first scenario, we investigated how
different reservoir rock typing methods impact permeability distributions. We further quantified changes
in hydrocarbon distribution and analysed how a novel approach that combines capillary pressure and log-
derived J-function affects the saturation models. In the second scenario, we used the effective medium
theory to calculate permeability multipliers for the regions where fractures are expected. This enabled us to
effectively represent fractures in a single-porosity reservoir model. The representativeness of the different
models was analysed through blind tests using static data as well as history matching using dynamic data.
The most significant findings of our work are that subtle changes in modelling decisions and reservoir
rock typing have major consequences for the saturation model, leading to up to 20% change in STOIIP
estimates. Such uncertainties must be carried forward in future reservoir management decisions and when
estimating reserves. The blind tests showed that a saturation model based on the combination of core- and
log-derived J-functions gave the most robust STOIIP estimates. These particular saturation models further
led to a much-improved history match, especially for wells located in the transition zone of the reservoir.
The best history matches were obtained once sparse, fault-controlled fractures were included in the reservoir
model using effective medium theory. The presence of fractures specifically improved the history matching
quality for wells located close to the faults; these wells were very difficult to match in the past.
Our work clearly demonstrates that a multi-deterministic scenario workflow is key to explore the
appropriate range of geological uncertainties, and that, equally important, the impact of different modelling
decisions must be accounted for when quantifying uncertainty during reservoir modelling. This is
2 SPE-196679-MS

particularly applicable to giant carbonate reservoirs where relatively minor changes in the workflow and
data interpretation can have major consequences on STOIIP estimates, dynamic behaviours, and reserve
estimates. Multi-stochastic modelling workflows which anchor the reservoir to a single base case are not
capable of achieving this.

Introduction
Carbonate reservoirs contain more than 60% and 40% of the world’s oil and gas reserves, respectively
(Montaron, 2008). These reservoirs are known for their geological heterogeneities, petrophysical
complexity and mixed wettability, which lead to significant uncertainties in the volumetric estimation and
fluid flow behaviour in the reservoir. Therefore, building a geological model that represents and captures a
realistic range of uncertainties (such as structure, reservoir rock typing (RRT), permeability, water saturation
and fracture presence) in these reservoirs is a crucial step for any reservoir development plan.
The hydrocarbon-in-place and fluid flow behaviour uncertainties emerge from the incomplete knowledge
of the subsurface geology which need to be captured consistently in the related modelling workflows,
specifically the reservoir rock typing approaches and generating the initial saturation distributions. In
addition, carbonate reservoirs are often fractured so a critical review of the possible presence of fractures
is important as fractures can be first-order controls on fluid flow in the reservoir even when they do not
impact oil-in-place estimates.
In this paper we focus on four main factors that have been found to significantly influence the initial
hydrocarbon distributions and fluid flow behaviour in a giant carbonate reservoir in the Middle East. These
are (i) the chosen modelling workflow, (ii) the reservoir rock typing approach, (iii) the modelling technique
to estimate initial saturation distributions and (iv) the accounting for the presence of low-intensity fractures.
We provide a brief description of these challenges below.

Multiple deterministic scenario modelling versus multiple stochastic realisations

In reservoir modelling it is common practice to generate a single base-case to support a decision making
process. Such a base-case model is usually considered as the anchor-point for generating high- and low-
case endmember models, and possibly for adding further models between the endmembers using multiple
stochastic realisations. It is not uncommon to find that such base-case models are built to support decisions
that have already been made, rather than challenging the current understanding of the reservoir geology
(Ringrose & Bentley, 2015). On the other hand, multiple deterministic workflow scenarios aim to challenge
the current reservoir understanding by generating an ensemble of geological models that capture a realistic
range of geological uncertainties and evaluate different modelling decisions. These geological models
evaluate different geological concepts (e.g. horizons, faults, fractures) under different scenario (e.g. fracture
intensity, reservoir rock types, properties distributions, model initialisation).
Figure 1 illustrates the differences between a modelling workflow that is anchored around a base case
and uses multiple stochastic realisations around the base case to account for uncertainty and a multiple
deterministic scenario modelling workflow. The multiple deterministic scenario workflow captures a
broader range of geological uncertainties and hence is able to provide forecasts that capture the real reservoir
performance. It is therefore more suitable to assess the economic risks and opportunities.
SPE-196679-MS 3

Figure 1—Multiple deterministic scenarios workflow versus multiple

stochastic realisations workflow.Modified after Bentley & Ringrose (2015).

Reservoir rock typing (RRT)

RRT is a crucial step in a geomodelling workflow as it controls the distribution of petrophysical parameters
in the geological model (Skalinski & Kenter, 2015). In broad terms, RRT can be defined as a way of
grouping rocks that have similar geological/petrophysical properties. RRTs can be categorised according to
the geological description (e.g. depositional environment, diagenesis, or facies) (Dunham, 1962), or based
on petrophysical properties (e.g. porosity, permeability, capillary pressure), or a combination of the two.
These categories can be established on the small-scale, as micro-pore, to large-scale, from log data or even
seismic data in some cases (Gomes et al. 2008;Hollis et al. 2010;Rebelle 2014).
A considerable amount research has been conducted to identify suitable RRT methods for carbonate
reservoirs. Commonly used RRT methods include Lucia’s rock fabric classes (Lucia, 1995, 1999), which
identify three-rock classes for non-vuggy rock based on the porosity-permeability relationship and grain
size group and suggest an equation to predict permeability for each class. Amaefule et al. (1993) suggested
a procedure to obtain RRT using rock quality indices (RQI) and flow zone indicators (FZI). Similarly,
Corbett and Potter (2004) introduced global hydraulic elements (GHE) by generating the GHE classification
template over the poro-perm cross plot, which divides into ten "petrotype" classes based on the regular
progression ranges of the FZI. The term "petrotype" is now used to refer to a specific petrophysical rock
typeset. On the other hand, Dale Winland (Kolodzie, 1980) showed that using mercury injection capillary
pressure curves can be useful in defining RRT groups, and introduced an empirical relationship between
porosity, permeability and capillary pressure curve when the pore aperture radius corresponds to a mercury
saturation of 35%, the so-called Winland35 method. Recently, Tawfik et al. (2019) introduced the ternary
rock typing plot between porosity, permeability and irreducible water saturation for defining RRT groups.

Initial fluid saturations

Modelling the distribution of the initial fluid saturations is another critical step in a geomodelling workflow
because it can have a significant influence on the oil-in-place estimates and, consequently, the quality of the
history match, reservoir performance prediction, and reserve estimates (AlBreiki et al., 2019). Modelling
the initial oil saturation distribution can be achieved by several techniques. The saturation model can be built
purely statistically by populating the water saturation observed in well logs between the wells. Alternatively,
one can use the method of Cuddy et al. (1993) which correlates log data to relate the water bulk volume
4 SPE-196679-MS

to height. It is also possible to employ capillary pressure curves from special core analysis (SCAL) data
for saturation height modelling using methods such as those of Leverett (1941), Johnson (1987), or Skelt
& Harrison (1995). Bespoke commercial software also exists which allows us to model the saturation
distributions using a combination of the Leverett J-functions from SCAL data and log-derived J-function.

Presence of fractures
Carbonate reservoirs often contain fractures. The flow behaviour of a fractured reservoir is generally
more complex than that of an unfractured reservoir because fractures can enhance permeability, leading to
bypassing of oil and possibly super-conductivity channels, all of which cause early water/gas breakthrough
and low recovery factors (Nelson, 2001). Significant research has been conducted to improve the modelling
of naturally fractured reservoirs (Berkowitz, 2002). For brevity, these methods are not reviewed here.
Since the reservoir considered in this study contains low-intensity fractures, we treat it as a single-porosity
system where fractures enhance matrix permeability but do not dominate reservoir performance and apply
workflows to obtain more reliable estimates for such permeability enhancement.

Summary
This paper will address the above issues by applying multiple deterministic workflows to a giant carbonate
reservoir in the Middle East. Specifically, we will:
1. Explore how using different geological modelling scenarios impact STOIIP estimates.
2. Evaluate how RRT and petrophysical modelling approaches change the permeability model and
reservoir connectivity.
3. Investigate how different approaches to model initial saturation distributions, especially in the
transition zone, impact the initial hydrocarbon distribution.
4. Analyse how the enhancement of reservoir permeability in the presence of low-intensity fractures
influences the predicted reservoir performance.
5. Understand how the above four steps impact the history matching process.

Methodology
We employ a geomodelling workflow that combines multiple deterministic scenarios and multiple
stochastic realisations to understand how geological uncertainties and the decision making process during
reservoir modelling impacts STOIIP estimates and predictions of reservoir performance. We apply this
hybrid approach to a giant carbonate reservoir in the Middle East, the so-called "Field X". Field X is
a limestone reservoir deposited in a carbonate ramp during the Upper Cretaceous period. This reservoir
is characterised by six subzones of poor rock properties with an average porosity of 18% and average
permeability of 10mD. Reservoir thickness is around 150ft with a large transition zone which can be up
to 80ft thick.

Structural modelling and reservoir zonation

A 3D geological model of Field X was constructed based on the available data, which includes a 3D seismic
dataset, conventional logs from 100 wells, routine core analysis from 20 wells, 10 image logs, and 2 SCAL
datasets. The static model has 350x350x41 grid cells; each grid block aerial dimensions of 100 × 100 metres
and varying vertical thickness. A structural model was created based on the seismic horizons and formation
tops and includes 80 faults with a main direction of N75W and N45W. The structural model was subjected to
zonation and layering, leading to a division into 16 subzones based on the lithofacies from cores description
and log interpretation, and further 40 layers to capture the vertical variability of the geology.
SPE-196679-MS 5

Property modelling and model validation

Multiple stochastic realisations of the porosity distribution were generated using different seed numbers,
cut-off values, and variogram ranges. The variogram was calculated using the upscaled porosity data Figure
2 All steps were quality checked.

Figure 2—Porosity variogram for Field X.

Figure 3 illustrates the workflow used in this paper. It starts by generating different RRT models,
followed by modelling of the initial fluid saturation. Next, the models are validated based on static and
dynamic tests. At this point, further improvements to the model are made, specifically to enhance reservoir
permeability where needed by accounting for the presence of low-conductivity fractures. We provide a
detailed explanation of each step of the workflow below.
6 SPE-196679-MS

Figure 3—Overview of the property modelling workflow for Field X.

SPE-196679-MS 7

Reservoir rock typing. Models were generated using three different RRT approaches, namely poro-perm
transforms for each subzone, the GHE method, and Lucia’s pore fabric classes. The permeability model of
the poro-perm transforms was generated using a bivariate distribution technique where pseudo-permeability
clouds were created for the 16 subzones. The GHE method used power regression lines for each of the
petrotype classes. Finally, the Lucia permeability model used the classical equation to estimate reservoir
permeability for each class (Lucia, 1999).
Initial fluid saturations. The distribution of the initial fluid saturations was modelled using two different
approaches, namely the core-derived Leverett J-functions and a hybrid method, which is available in a
dedicated commercial tool and uses both, core-derived J-functions and the log-derived J-functions.
Figure 4 shows the steps for generating the water saturation model using a commercial software
(Geo2Flow). The steps are summarised below and have been discussed in (AlBreiki et al., 2017):
1. Convert capillary pressure data from a porous plate experiment to Leverett J-functions.
2. Define the average, maximum, and minimum curves for the Leverett J-function.
3. Conversion of the log water saturation to log-derived Leverett J-functions.
4. Generate synthetic permeability logs from the 3D permeability model.
5. Define the cumulative distribution functions for the errors in the permeability, irreducible water
saturation, and Archie exponent data.
6. Calculate synthetic saturation logs that honour the saturation from the log- and core-derived Leverett
J-function by changing the values of permeability, irreducible water saturation, and Archie exponents
within the defined error ranges.
7. Construct the 3D saturations model using geostatistical interpolations (kriging).

Figure 4—Workflow to model saturation distributions in Field X using

the combination of the core-derived and log-derived J-functions.

Static validation. We first validated the permeability models by calculating the average permeability in the
drainage area of a given well with the average permeability observed in the corresponding well tests. Then
we compared the core permeability with modelled permeability at each well. Next, we carried out blind tests
for the saturation model by randomly selecting 80% of the available wells as input data for the modelling
process and predicting the properties of the remaining 20% of the wells. This process was repeated several
times to avoid any bias in the validation process.
Validation with dynamic data. A classical history match was performed to validate the resulting models
using dynamic data from 20 years of production. Several wells show mismatch in the bottom hole pressure
(BHP), water cut (WC) and the gas/oil ratio (GOR). Wells with poor match quality were revisited in order to
8 SPE-196679-MS

understand how the model could be updated based on the available geological data (e.g. the understanding of
the regional geology, image logs, and core data). Upon inspection of the data, it was apparent the presence of
fault-related fractures in the reservoir is likely, albeit fracture intensity is low. This observation is consistent
with the understanding that the reservoir has been deformed during two main compressional events (Sirat,
Salman, & Bellah, 2007). Most of the wells where history matching quality is poor are located close to the
faults where fractures appear to be more abundant.
Fracture modelling. In order to estimate how the presence of low-intensity fractures enhances reservoir
permeability and impacts fluid flow, we modelled the fractures implicitly using the effective medium theory
(EMT) approach. EMT allows us to analytically calculate the effective average permeability of both matrix
and fractures (Sævik et al., 2013; Wong et al., 2018). Using an in-house code, we calculated the ratio
between the effective permeability ke for fracture and matrix and the original matrix permeability km, that was
recalculated for different fracture intensity scenarios (Figure 5). The ratio of ke to km provides geologically
consistent permeability multipliers for any given matrix permeability value; that is, these multipliers are not
arbitrary but based on the key fracture properties such as intensity, orientation, aperture, etc. The relevant
fracture input data was acquired from the available image logs and core data, and sensible bounds were
used to compute the ratio of ke to km. In the EMT workflow we assume that the fractures are elliptical,
each fracture set as one orientation only, and a power law relates fracture aperture to fracture size. At the
field scale, we assumed that the fracture intensity values are higher near the fault plane and decrease when
moving away from the fault plane. We assume that the fracture intensity decreases to zero 300, 500 or 1000
m away from the fault plane. Figure 6 illustrates the steps undertaken to update the reservoir permeability
in areas where wells have a poor history match quality and fractures are likely to be present.

Figure 5—Calculation of the ratio of the effective fracture and matrix permeability ke to the original matrix permeability km
for different fracture intensity scenarios (fracture intensity decreases away from the faults). The effective medium theory
(Saevik et al., 2013) is used to compute the ratio of ke to km, which provides a geologically consistent permeability multiplier.
SPE-196679-MS 9

Figure 6—Illustration of the steps undertaken to include fractures in the

reservoir model in areas where individual wells show poor history matches.

Results and Discussion

By combining multiple deterministic scenarios with multiple stochastic realisations, we were able to
evaluate and assess how various geological scenarios and different modelling decisions impact STOIIP and
reservoir performance. We now discuss how our workflow impacts the model outcomes for key reservoir
properties.

Porosity model. There is limited variation in the multiple stochastic realisations of the porosity model
(less than 1%), and hence minor influence on the STOIIP estimates. Figure 7 shows example porosity logs
based on different modelling decision and provide a comparison with the real porosity logs during the blind
testing. The lack of variability is not particularly surprising, given the fact that we have access to porosity
logs from more than 100 wells and the logs showed reliable correlation when calibrated to the core porosity.
However, a small impact of the porosity model cannot always be expected because other factors in the
modelling process can impact the calculation of pore volume (e.g. the number of logs, log calibration, well
distribution across the field).
10 SPE-196679-MS

Figure 7—Examples from two wells comparing the porosity model and the actual porosity logs
(black lines) for different stochastic realisations (i.e. different seed numbers as well as the original
porosity model). We also show an example prediction for a well that was used in the blind test.

Permeability model. The results from the three RRT approaches and related permeability predictions
show different permeability distribution and averages. The cloud transformation is applied to each of the
16 subzones while the GHE model was split into five different petrotype groups and two groups were
identified based on Lucia’s classes (Figure 8). As a result, the permeability distribution shows more vertical
heterogeneity for each subzone in the cloud transformation; this method also yields the lowest average
permeability compared to the GHE method and Lucia’s porosity classes (Figure 9).
SPE-196679-MS 11

Figure 8—Comparison poro-perm groups based on the different RRT approaches.

12 SPE-196679-MS

Figure 9—Comparison of the resulting permeability model arising from the

three different RRT approaches. Note the differences in vertical heterogeneity.

Saturation model. The saturation model that uses a combination of log- and core-derived Leverett J-
functions shows a perfect correlation between the log data and the modelled saturation, as the model honours
the logs exactly (Figure 10). One of the significant advantages of this approach is that the saturation profile
in the transition zone area is captured, which is not the case when only core-derived Leverett J-functions
are used. It is worth noting that log saturations are assumed to be correct because they were validated using
a Dean-Stark extraction apparatus from two cores. Figure 11 presents the result of the 3D saturation models
using both saturation modelling approaches for each RRT method.
SPE-196679-MS 13

Figure 10—Comparison of the water saturation models in two wells for the different saturation modelling approaches.

Figure 11—Comparison of six exemplary saturation models generated using different RRT approaches
and saturation methods. Note the significant differences in saturation distribution and movable water.

Figure 12 shows an example of the static validation using blind testing with 80% of the available well
data for the resulting saturation model. In addition, Figure 13 shows a cross-validation of the saturation
model, comparing the modelled water saturations and log saturations for the cases where both, core- and
log-derived Leverett J-functions are used and where only core-derived Leverett J-functions are used. Since
14 SPE-196679-MS

key input parameters are adjusted to fit the saturation model to the log data in the former, the correlation
between saturation model and log data is perfect when using core- and log-derived Leverett J-functions.

Figure 12—Resulting saturation models for an examply blind-tested well.

Figure 13—Comparison of saturation data that use core-derived

and a combination of core- and log-derived Leverett J-function.

STOIIP estimates. The combination of using different RRT techniques and saturation models have a
major impact on STOIIP estimates (Figure 14). STOIIP estimate can increase by up to 17% depending on
the chosen modelling decision, or decrease by 7%. The poro-perm transform using the Leverett J-function
SPE-196679-MS 15

resulted in the lowest STOIIP estimate because they have the lowest average permeability model and the
Reservoir Quality Index (RQI) directly impacts the Leverett J-function and hence saturation distribution.
On the other hand, the saturation model generated from the combination of log- and core-derived Leverett
J-functions yields more consistent STOIIP estimates because we account for error estimates in irreducible
water and Archie exponents, and honour the saturation logs, when modelling the permeability.

Figure 14—Impact on STOIIP estimates as a function of RRT and saturation modelling choices.

History matching. History matching was performed for all models using CMG’s GEM and CMOST
software. The choice of RRT and saturation modelling approach had a direct impact on the quality of the
history match at a well-by-well basis. Figure 15 shows, as an example, the matched WC from two wells
when building the saturation model based on log- and core-derived of Leverett J-functions and the core-
derived Leverett J-function only.

Figure 15—Comparison of history match for the water cut at two wells using different saturation modelling approaches
(green colour represent the core-derived Leverett J-functions, blue colour the core- and log-derived Leverett J-functions).
16 SPE-196679-MS

Although the quality of the history match improved on a well-by-well basis if saturation model based
on log- and core-derived of Leverett J-functions is used because the fluid mobilities are estimated more
accurately, a small number of wells that are located in the vicinity of the faults still show unsatisfactory
history matches. When using EMT to estimate the enhancement of the effective matrix permeability due
to the presence of low-intensity fractures in the vicinity of the faults, a clear increase in permeability
can be observed around these wells (Figure 16). Figure 17 shows an example of the improvement in the
history match quality on a well-by-well basis after applying EMT to estimate the enhancement in effective
permeability due to the presence of low-intensity fractures. The increased permeability improves reservoir
connectivity and, hence, the quality of history matching for WC, GOR and BHP. It is important to note that
these new history matched models are likely to lead to new reservoir performance forecasts which could
have a major impact on the field development strategy.

Figure 16—Comparison of permeability enhancement in the vicinity of the faults when using the Effective Medium
Theory to estimate the effective permeability for the rock matrix in the presence of low-intensity fractures.
SPE-196679-MS 17

Figure 17—Comparison of history matches (top) at three wells located in the vicinity of faults (bottom,
colours represent the depth of the top surface) after increasing the reservoir permeability using the
Effective Medium Theory to account for the presence of low-intensity fractures in the vicinity of the faults.

Conclusions
This study has shown that a multi-deterministic scenarios workflow combined with stochastic realisations
enabled us to explore how a range of geological uncertainties (e.g. presence of fractures) and uncertainties
related to different modelling decisions (i.e. reservoir rock typing and saturation modelling) impact STOIIP
estimates and reservoir performance in a giant carbonate reservoir. Using the same dataset, different
reservoir modelling approaches can lead up to 17% change in STOIIP estimates. Based on extensive
validation, we suggest that a saturation model that uses a combination of core- and log-derived Leverett
J-function method is likely to provide the most robust STOIIP estimates. Such a model also improves
the history match quality on a well-by-well basis. Where individual wells cannot be history matched, the
application of the effective medium theory to calculate the permeability enhancement of the matrix due to the
presence of low-intensity fractures in the vicinity of the fault has shown further, significant improvement in
the quality of the history match. The updated reservoir models are likely to lead to new reservoir performance
forecasts which could have a major impact on the field development strategy and reserve estimates. Our
work re-emphasises the needed to carefully consider both, reservoir modelling decisions and geological
uncertainties, through the appropriate multi-deterministic geomodelling workflows.

Acknowledgements
The authors would like to thank Abu Dhabi National Oil Company (ADNOC) for providing access to data
and the scholarship for Mohamed. Sebastian acknowledges support from Energi Simulation for his Chair in
Carbonate Reservoir Simulation. We are also grateful to Vicky and Dan O’Meara for access to the Geo2Flow
software and CMG for access to GEM and CMOST.

AlBreiki, M., Geiger, S., & Corbett, P. (2017). A Consistent Workflow to Re-evaluate the Hydrocarbon Distribution
in a Giant Carbonate Reservoir in the Middle East. In 79th EAGE Conference and Exhibition 2017. http://
dx.doi.org/10.3997/2214-4609.201701539
AlBreiki, M., Geiger, S., Corbett, P., & AlAmeri, S. (2019). Assessing The Uncertainty In Hydrocarbon
Distributions In The Transition Zone Of A Giant Carbonate Reservoir. In First EAGE Reservoir
Characterization and Modelling Workflows for Giant Carbonate Field Developments of the Middle East. http://
dx.doi.org/10.3997/2214-4609.201900196
18 SPE-196679-MS

Amaefule, J.O., Altunbay, M., Tiab, D., Kersey, D. G., & Keelan, D. K.(1993). Enhanced Reservoir Description: Using
Core and Log Data to Identify Hydraulic (Flow) Units and Predict Permeability in Uncored Intervals/Wells. In SPE
Annual Technical Conference and Exhibition. Society of Petroleum Engineers. http://dx.doi.org/10.2118/26436-MS
Bentley, M., & Ringrose, P. (2015). Modelling for comfort?Petroleum Geoscience, 2014-2089. http://dx.doi.org/10.1144/
petgeo2014-089
Berkowitz, B. (2002). Characterizing flow and transport in fractured geological media: A review. Advances in Water
Resources, 25(8-12), 861-884. http://dx.doi.org/10.1016/S0309-1708(02)00042-8
Corbett, P., & Potter, D. (2004). Petrotyping: A base map and atlas for navigating through permeability and porosity data
for reservoir comparison and permeability prediction. International Symposium of the Society of Core Analysts Held
inAbu Dhabi, UAE, 5–9.
Dunham, R. J. (1962). Classification of Carbonate Rocks According to Depositional Textures. In Classification of
Carbonate Rocks-A Symposium.
Gomes, J. S., Ribeiro, M. T., Strohmenger, C. J., Negahban, S., & Kalam, M. Z. (2008). Carbonate Reservoir Rock Typing
— The Link between Geology and SCAL. Abu Dhabi International Petroleum Exhibition and Conference 2008, 1-24.
http://dx.doi.org/10.2118/118284-MS
Hollis, C., Vahrenkamp, V., Tull, S., Mookerjee, A., Taberner, C., & Huang, Y. (2010). Pore system characterisation in
heterogeneous carbonates: An alternative approach to widely-used rock-typing methodologies. Marine and Petroleum
Geology, 27(4), 772-793. http://dx.doi.org/10.1016/j.marpetgeo.2009.12.002
Johnson, A. (1987). Permeability Averaged Capillary Data: A Supplement To Log Analysis In Field Studies. In
Proceedings of SPWLA 28th Annual Logging Symposium.
Kolodzie, S. J. (1980). Analysis of pore throat size and use of the Waxman—Smits equation to determine 00IP
in spindle field. Annual Fall Technical Conference of Society of Petroleum Engineers. http://dx.doi.org/10.1017/
CB09781107415324.004
Lucia, F. J. (1995). Rock-fabric/petrophysical classification of carbonate pore space for reservoir
characterization. American Association of Petroleum Geologists Bulletin, 79(9), 1275-1300. http://
dx.doi.org/10.1306/7834D4A4-1721-11D7-8645000102C1865D
Montaron, B. (2008). Carbonate Evolution. Oil and Gas Middle East, (August), 26-32.
Nelson, R. (2001). Geologic Analysis of Naturally Fractured Reservoirs. Geologic Analysis of Naturally Fractured
Reservoirs. http://dx.doi.org/10.1016/b978-0-88415-317-7.x5000-3
Rebelle, M. (2014). Rock-typing In Carbonates: A Critical Review Of Clustering Methods. Abu Dhabi International
Petroleum Exhibition and Conference, (November), 10-13. http://dx.doi.org/10.2118/171759-MS
Ringrose, P., & Bentley, M. (2015). Reservoir Model Design. http://dx.doi.org/10.1007/978-94-007-5497-3
Sævik, P. N., Berre, I., Jakobsen, M., & Lien, M. (2013). A 3D Computational Study of Effective Medium Methods Applied
to Fractured Media. Transport in Porous Media, 100(1), 115-142. http://dx.doi.org/10.1007/s11242-013-0208-0
Sirat, M., Salman, S., & Bellah, S. (2007). Characterisation and Modeling of Fractures in Large Onshore Abu Dhabi
Carbonate Field. In Abu Dhabi International Petroleum Conference and Exhibition (Vol. 81, p. 9). Society of Petroleum
Engineers. http://dx.doi.org/10.2118/162342-MS
Skalinski, M., & Kenter, J. A. (2015). Carbonate petrophysical rock typing: integrating geological attributes and
petrophysical properties while linking with dynamic. Geological Society, London, Special Publications, 406(1),
229-259. http://dx.doi.org/10.1144/5P406.6
Tawfik, B., Corbett, P., & Pickup, G. (2019). Ternary Rock Typing: A Novel Solution to Multi-Scale Multi-Discipline Rock
Typing for Carbonate Rocks. http://dx.doi.org/10.3997/2214-4609.201901312