SPWLA 56th Annual Logging Symposium, July 18-22, 2015

IMPROVED PERMEABILITY PREDICTION IN HETEROGENEOUS

CARBONATE RESERVOIRS: A NEW APPROACH USING ROCK
TYPING AND WIRELINE FORMATION TESTING
Iulian N Hulea, Daniela Frese and Shyam Ramaswami, Shell International BV, The Netherlands

Copyright 2015, held jointly by the Society of Petrophysicists and Well Log
Analysts (SPWLA) and the submitting authors.
This paper was prepared for presentation at the SPWLA 56th Annual Logging
Symposium held in Long Beach, California, USA, July 18-22, 2015.

ABSTRACT

Carbonate rocks are a synonym for heterogeneity.

Attempts to characterize carbonate reservoirs follow
various rock typing routes in order to predict saturation
and permeability. The rock typing methods focus
mainly on Special Core Analysis output where core
derived permeability and capillary pressures play a
central role. Given the late stage in a typical project
where these results become available, a full integration
with larger volume sampling methods (such as wireline
formation testing, and well testing) is usually not
brought to a closure. For a heterogeneous formation we
attempt to bring this information in agreement and
highlight differences between the results of different
sampling methods. More specifically, we focus on the
agreement between various permeability scales to
accurately model our reservoir.

Also, we will show that by using fluid mobility data

from wireline formation testing pressure measurements
one can improve the accuracy of rock –typing predicted
permeability by comparing mobility against the Fig. 1: Two end members encountered in carbonate
predicted permeability obtained via various averaging formations and their characteristics from thin section
methods. In addition, a new method of QCing and (A,B) to capillary pressure (Pc) models (A2,B2). The
ensuring consistency between static and dynamic thin sections show the porosity in blue and the rock
models will be presented. In populating a reservoir fabric in grey and brown. (A) corresponds to a
model, the permeability and saturation are typically permeable dolomite (see A1) while (B) is an oomoldic
forward modeled in parallel streams without a limestone with poor permeability (see B1) due to the
consistency check between these parameters. Where isolation of the relatively large pore bodies. A1 and B1
core permeability is not available we will show that a show the corresponding porosity-permeability data
permeability curve can be derived based on input data from Routine Core Analysis (RCA) measurements. A2
already available in the model: via capillary pressures. and B2 show the saturation height models (SHM); i.e.
The advantage of this method is that it can be calibrated height above free water level (HAFWL) versus water
directly to well performance. saturation (Sw). They are plotted for the same 6
porosity values ranging from 0.06 to 0.26 v/v for the
same pore fluid to contrast the difference in predicted
saturations between the two rock types shown in A and
B. These two end members can be encountered in a
Permian carbonate where permeability could be
between 1D and 0.01 mD for the same 15% porosity.
1
SPWLA 56th Annual Logging Symposium, July 18-22, 2015

INTRODUCTION

Rock typing is a process that attempts to differentiate

volumes of rock that have contrasting reservoir
properties (permeability, saturation, flow) represented
in a 3D model, see (Lucia, 1995), (Hulea et al., 2012)
and (Xu et al., 2012). At the end of the (non-unique)
rock typing process, we usually have 2 main building
blocks with a central role, namely the permeability and
hydrocarbon/water saturation models (see Fig. 1).
Commonly they are modeled in parallel streams.
Although these two properties are closely related, their
consistency is thus not preserved by default in the final
product - the model characterizing a given reservoir
volume. In this work we show how any saturation
height model (SHM) can be converted into a
permeability model, and that additional data can be
used in order to reduce uncertainties resulting from the
rock typing process.
Fig. 2: (a) The RCA porosity permeability data used in
IMPROVING THE QUALITY OF ROCK
the modelling process for the rock type shown in Fig 1
TYPING PREDICTIONS: USING WIRELINE
(A). (b) The permeability obtained by averaging
FORMATION TESTING DATA
Arithmetically, Geometrically and Harmonically for
any given porosity bin. (c) The geometrically averaged
For the permeability prediction from RCA, once a
permeability is fitted using a Lucia (Lucia, 1995) like
certain group of samples has been grouped within a
function: log k= a log Ø +b. (d) the three possible
Rock Type, they undergo an averaging process that
predictions are shown to highlight the variability in the
facilitates the prediction of a permability value for a
predicted permeability, the ratio between the Arithmetic
given porosity, see Fig. 2.
and Harmonic is higher than 4 while the Geometric to
Harmonic is above 2.
Since different averaging methods can be used ranging
from arithmetic (Ar) to harmonic (Ha) a question that is
There are many factors why the mobility inferred
not easy to answer is: which one of those is the most
from Wireline Formation Testing pressure tests
suitable method to be used in the reservoir model? The
could not directly be used for quantitatively
ratio between Ar and Ha is a factor of 4 in this case (for
describing permeability, including uncertainty of the
a well behaved rock type - shown in Fig 1 A1), hence
viscosity of the fluid investigated during the
not insignificant. Based on geological information,
pressure measurement (invasion), the conversion of
preference can be given to arithmetic or geometric
a spherical mobility (which is effectively the flow
based on the structure present in the rock to be
regime during a short pressure test) to radial
modelled.
mobility, the limited volume of fluid tested,
investigation scale, the relative permeability
There is another source of dynamic data that does
between the formation fluid and the mud, etc.
not feature normally in a Rock typing workflow-
However, a qualitative comparison can be carried
mobility data from pressure measurements using
out. In our case, wireline formation testing
wireline formation testers (see Fig. 3). This data can
measurements have been carried out in a well drilled
help answer the above question (Tangyan et al., 2005;
with water based mud in a gas filled reservoir .
Ramaswami et al., 2013). For every formation
pressure measurement, a mobility (Mdd) estimation
can be inferred, with mobility being the permeability
(k) over viscosity µ (Mdd = k/µ) .

2
 SPWLA 56th Annual Logging Symposium, July 18-22, 2015

Fig. 3: Mobility vs log porosity show what we already

know about carbonates: significant scatter in dynamic
properties for any given porosity. For the same porosity
– orders of magnitude of spread in
mobility/permeability are measured as already shown
in Fig 1. Worth noting is that the porosity measurement
scale (~ 1m limited by vertical resolution), is Fig. 4: Permeability vs Mobility plots used to identify
considerably larger as compared to the wireline the best correlation between rock type perm average
formation testing pressure test scale (~ 10 cm, ~ 10 cc- and mobility. Blue lines correspond to a 1:1
depending on formation permeability). relationship. Red lines are fits. The best correlation
is reached for the geometrically average
For the wells where mobility data have been permeability prediction (b) R2=0.49 vs 0.44
acquired, the arithmethic, geometric and harmonic (Arithmetic (a)) and the NMR derived permeability
permeability predictions were made (for a number ((d) R2=0.38).(c) Assuming k ~ Mdd the absolute
of rock types encountered in the well as a function permeability correlation to mobility is Log k(mD)=
of depth). By comparing the various averaged Log Mdd (md/cP) + 0.265 for this example.
permeability values against the measured mobility
we see that although the correlation factor is not too PERMEABILITY PREDICTION IN THE
high (R2=0.5), the best correlation is achieved for RESERVOIR MODEL
the geometrically averaged values. Also in terms of
absolute value the Ge average (see Fig. 4) is the A common issue encountered in measurements
closest to the measured mobility. Worth noting is performed on carbonates is the variability in measured
that the NMR and Arithmetic averaged permeability properties for similar porosity values (Lucia, 1995;
significantly overestimate the measured mobility. Hulea et al., 2012). As described in Fig. 5, this effect
Assuming the same fluid correction has to be can qualitatively be explained by the fact that multiple
applied for all the mobility (Mdd) points within the pore systems combine in different ratios to reach the
same rock type – or in other words k ~ Mdd - the same porosity, but leading to significant scatter in the
absolute perm correlation log k= log Mdd + d with d= dynamic/electric properties responses. See Fig. 1,5.
0.265 is suggested for this example.
If we look into the way we populate our reservoir
models with dynamic properties (i.e. the saturation
height model) we observe that the permeability and the
saturation are modeled in parallel streams (see Fig. 6).
Although these two quantities are closely related, their
consistency is not ensured by default in the final
product. In this work we show how any SHM can be
converted into a permeability model, complementary to
other permeability prediction methods (see Lucia 1995;
Amaefule et al., 1993; Xu et al., 2012).
3
SPWLA 56th Annual Logging Symposium, July 18-22, 2015

Fig. 5: Simplified qualitative picture of a carbonate

sample under test. While the porosity is composed of
multiple pore systems ranging from vugs to
microporosity, it is the relative ratio of those
components that dictate the variability in dynamic and
electric flow properties.

Although in some instances permeability is used as Fig. 6: Overview of the workflow used to populate the
input into the SHM, the SHM is commonly not used to model relying on core analysis for permeability (RCA)
derive a permeability that can be compared against the and special core analysis (SCAL) as well as logs for the
input (or independent) permeability data. Occasionally, saturation height model (SHM). In the model,
that comparison can be performed at the well level if consistency is commonly not preserved between the
the SHM has an explicit permeability dependency (for SHM and permeability.
example by resolving the permeability for a predicted
water saturation, and checking the agreement to the log Our permeability prediction model is based on our
derived water saturation in the petrophysical software earlier work connecting the capillary pressure data
tool, see (Techlog, 2014)) and by making additional (originating from MICP) to permeability with the
assumptions on fluid properties, and the free water level following mathematical expression:
(FWL).
log [ ]= log Ø [μ ] + [1]
In this work, in order for the comparison to be
∑ ∆ ∗ !"
meaningful, wells were chosen where core data were = [2]
available in order to compare to welltest permeability
values in formations where no significant fracture Where A=1.55 and B=2.35. ∆Swi is the water
influence has been identified. The comparison relies on saturation change at the i-th pressure step behind pore
the comparison between measured core and the well throats with a radius #$. Our method has an advantage
test derived permeability, suggesting a factor of 5 over others (e.g., Swanson or Pittman (Pittman, 1992)
difference could be observed (see Fig. 7). That in that it uses the whole pore system for
comparison is necessary in order to avoid possible characterization rather than using a single point in the
disagreement between our method, which predicts the whole pore distribution as a representative value. This
matrix permeability and the permeability inferred from provides confidence that mixtures of pore systems
pressure transient analysis (K_DST)of the well test due (occurring so often in carbonates) are properly
to fractures. Just like the saturation, this workflow will represented in our method. This allows all pore systems
have as output a matrix permeability. The fracture contributing to Sw as counting quantitatively towards
influence on the MICPs ( Mercury Injection Capillary the permeability. The permeability coming out of this
Pressure) is insignificant, see (Hulea , 2013). workflow is the equivalent of a pseudo stressed air
permeability but without being corrected for brine
4
 SPWLA 56th Annual Logging Symposium, July 18-22, 2015

(which from correlations requires a reduction by a up to 3. On the basis of the rock type and Rav (see Fig.
a factor of 2, see Fig 13c). The same workflow could be 8) a permeability value is predicted for every
followed using other permeability predictors connecting pair (Rock type, Porosity)
capillary pressure to permeability (Thomeer, 1983; 4. Carry out first pass checks – compare against
Pittman, 1992; Clerke, 2009 ). RCA
5. Compare against well test values

Fig. 7: A direct comparison of the average

permeability from core (RCA) and nuclear magnetic
resonance (NMR) vs average welltest permeability
(over the tested interval). A 5x factor can be
expected in the difference between the RCA and
welltest permeability. This is the level of accuracy
Fig. 8: Average pore throat (Rav- see Equation [2])
our method aims at. Also the average NMR perm
vs porosity for a number of rock types. The Rav (µm)
derived using the Schlumberger Doll Research
appears to have a different dependency on porosity
Formula (SDR) is not suited for permeability
for various rock types ((a)-(f) represent 6 rock types
estimation in carbonates in its raw form and
ranging from permeable dolomite to dolomitic
requires proper core calibration (Volokitin et al.,
limestone), each MICP curve is condensed to one
2001).
point in the above cross plot.
For a given volume of rock in the subsurface we can
As an intermediary check, once the MICP based
apply the same method to predict a permeability if an
permeability is predicted along the well it can be
SHM is available since it should carry the same pore
compared against the RCA measured permeability. In
throat information. In practice we have designed the
Fig. 9 two such instances are shown for the rock types
following workflow for some cases where reliable data
at the opposite sides of the permeability spectrum. In
are available- MICP:
Fig 9 (a) the same permeable rock type that was
discussed in Fig 1 is shown. In Fig 9 (b) a much poorer
1. At the well location detect from logs the Rock
rock type (from a permeability point of view) is shown
Type distribution vs depth
corresponding to the well tests shown in Red in Fig. 10.
2. For every rock type on the basis of sampled
MICP extract an average pore throat vs
porosity (see Fig. 8) trend (porosity is a
reliable well property)

5
SPWLA 56th Annual Logging Symposium, July 18-22, 2015

method underestimates the absolute permeability (see

Hulea et al., 2012).

Fig. 10: Predicted permeability based on the capillary

pressure model vs welltest permeability. The average
permeability range is quite wide – welltest permeability
varying between 0.1 mD and 1 D. With one exception
all the points are falling within the 5x range already
seen in the core (see Fig 7). The point falling out of the
range is at the limit in the core domain as well (see Fig
7), suggesting the core might not capture all the
Fig. 9: Blind check- comparing MICP derived formation heterogeneity. The colors correspond to
permeability (magenta) vs core data (black) and rocks originating from different geological formations,
geometrical averaged permeability (green). (a) for ranging from Lower Cretaceus, Jurassic to Permian.
one of the most permeable rock types also shown in The blue dashed line is a guide to the eye
Fig 1. (b) one of the lowest permeability examples corresponding to a 1:1 relationship. The red dashed
with a poorer agreement between core averaging lines show a 5x higher and lower predicted
and MICP perm prediction. In obtaining the permeability for any DST derived average permeability.
magenta predicting correlation, no direct use has
been made of the measured k-phi relationship. The resulting agreement between the average Pc
modeled permeability and the average well test
The final check of our prediction is to compare against permeability is exceptional. With a single exception, all
well test (DST) derived permeability as shown in Fig. the points are falling within a 5x range that has been
10. Once a permeability profile is calculated based on observed in the DST to core comparison, see Fig 10.
log derived porosity and rock type (where necessary),
an average permeability can be calculated. Over the The deviations from the 1:1 relationship observed for
same interval where the welltest (DST) K_DST has the low permeability rock (Fig 9b) is most probably
been derived, average DST permeability and Pc (SHM) caused by the fact that the corrections (brine, stress) are
derived permeability can be compared, see Fig 10. relatively more important for low permeability. Hence
by performing a brine correction the red points would
A number of geological formations are included in the shift to lower predicted average permeability, see Fig
study with average DST permeability ranging from 0.1 13.
mD to 1D, hence spanning 5 orders of magnitude. We
are aware that at permeability values around 1 D our

6
 SPWLA 56th Annual Logging Symposium, July 18-22, 2015

The workflow can be simply adjusted for a case when a

reliable SHM is present in the absence of capillary
pressure (MICP) data by simply calculating the
corresponding pore throats for any SHM (that
information is contained in the model just like in a
capillary pressure measurement). We have followed the
workflow discussed above for a well that did have a
reliable SHM (log derived) and a RCA dataset in
addition, but without using the measured permeability
in the prediction process, see Fig. 11.

At the end of the process we compared the measured

permeability to the predicted one based on the
saturation height model. The quality of the agreement is
impressive, less than a factor of 2 for porosity higher
than 10 %- see Fig. 12. For this particular formation the
relevant porosity range is between 10 and 20%. For the
highest porosity and permeability the agreement is the
best between the averaged permeability and SHM
predicted.

Fig. 12: Blind test comparing core permeability to

predicted permeability from the SHM obtained
following the workflow as shown in Fig. 11.Magenta is
the permeability resulting from the SHM while the
green is the geometrical average of the RCA
Fig. 11: An example of the workflow to be followed measurements (black). The log derived SHM is
when no MICP data are available (a) check the SHM obtained on a porosity range from 0.1 to 0.2 (porosity
describes the log saturations (in Equivalent histogram shown in the insert) hence outside that range
hydrocarbon columns- EHC). (b) the saturation the predicted values are less reliable, as seen in the
dependency on HAFWL- height above free water level – difference between green and magenta
derived based in the log domain. (c) Average pore
throat Rav (um) vs porosity that is used to calculate the
permeability. The dots correspond to calculated Rav
while the red line is the fit.

7
SPWLA 56th Annual Logging Symposium, July 18-22, 2015

DISCUSSION

Although feared to be hindered by various non-

reservoir related aspects, the mobility was shown to be
a valuable source of information for permeability (k)
prediction within a rock typing scheme. Despite the fact
that pretest derived mobilities from wireline formation
tester measurement assume the entire pressure drop is
related to the system mobility and does not resolve for
skin, relative permeability and fluid viscosity, it is
worth noting that on average they appear to agree well
with other permeability sources. The observed
correlation (that is most probably not universal) of log
k= log Mdd + 0.265 would imply for Mdd = k*Kr/ µ
(Kr/ µ the ratio relative permeability to viscosity)
Kr/ µ =1.85. For water based mud assuming a
viscosity of 0.8 cp this results in a Kr=0.5, within
the range of possible values.

The level of agreement observed between the welltest

permeability and capillary pressure model (Pc)
predicted permeability is remarkable, we are only a
factor of 5 away. The same level of accuracy is
achieved by direct RCA measurements, as shown in Fig
7. The Pc-predicted permeability can be a powerful tool
in predicting the permeability when a robust SHM is
available (derived from logs, core or a combination) or
as a QC tool to evaluate our upscaling process as well.
This work also suggests that by ensuring consistency
between the capillary pressure derived saturation height
model and the permeability model, it brings us closer to
understand the dynamic reservoir performance. If we
ensure properties consistency, we might be one step
closer in understanding near well permeability and its
relationship to the welltest observed permeability (in
case the flow is governed by matrix). The difference
between Pc predicted permeability and well test
average can be related to corrections that haven’t been
fully implemented. In cases where the predicted Pc
Fig. 13: (a,b) Permeability histograms for the 2 points permeability exceeds the well tests (red dots in Fig 10)
with an average welltest permeability between 1 and 2 the maximum permeability encountered is the lowest
mD. The upper histogram corresponds to the blue dot (see Fig. 13)- suggesting that the difference between the
(better agreement between prediction and prediction and measurement could be related to an air-
measurement) while the lower corresponds to the red brine correction (that will be less than a factor of 2, see
dot in Fig 10. As the maximum observed permeability also Fig 13c).
(over the tested intervals) is lower, the corrections
become more important. In red we highlight the high In case no MICP data are available, a robust log derived
permeability component that makes the difference saturation height model could provide the same
between the two points. (c) The liquid to air information (see Fig 11 &12). Special care should be
permeability correction observed in a field example. taken in building a SHM in such a case since the entry
For permeability around 100 mD the correction is of pressure (likely coupled to permeability) is relative to
the order of 0.6 while the correction is more significant the Free Water Level (FWL) position (Chen et al,
when the permeability decreases. 2014), hence susceptible to errors if FWL is an
8
 SPWLA 56th Annual Logging Symposium, July 18-22, 2015

uncertain property. Also, deriving a SHM based on 14. A rather large difference is seen in the porosity
wells above the transition zone could potentially lead to dependency of Winland’s R35 approach (see also( Xu
large errors in predicted permeability since close to Swi et al., 2012)).
the SHM sensitivity to depth is significantly reduced.
Our work shows that there is a porosity dependence in
One route we still have not fully explored is to follow the average pore throat for a given Rock type. This
the described workflow in opposite direction – using as implies a Rock type is not constrained by a constant
starting point permeability and having as output a R35. This fact is not intuitive but still directly measured
SHM. In case only matrix permeability values are (see Fig 8). The average pore throat dependency on
available (routine core analysis), the workflow needs porosity is easily explained by visualizing a Rock type
further adjustments, in essence connecting the that has an entry pressure dependency with porosity
equivalent of Rock Fabric Number , see (Lucia, 1995) (see Fig 1). This can be explained by a changing Rav
to a SHM. More work is currently being performed to with porosity (or permeability). Following the same
understand this relationship better and as such, we do reasoning, R35 would depend on porosity in a similar
not have a proof of universality yet that would connect fashion, hence assuming a Rock Type is defined by a
the permeability (unimodal rocks) to a SHM. In case constant R35 with porosity is inconsistent with the
NMR data is available, it would be interesting to look at SHM and porosity- permeability data.
the consistency between an NMR-SHM and
permeability that are derived in the same process CONCLUSIONS
(Volokitin et al., 2001).
By using wireline formation testing mobility data it was
shown that the geometric averaged permeability should
be used for a given rock type.

The saturation height model was shown to carry

valuable permeability information that compares well
with well test derived average permeability. A
workflow has been designed to convert any saturation
height model (capillary pressure data- MICP or log
derived) into a permeability model that was shown to
be less than a factor of 5 away from the measured
(RCA or DST) permeability. This is a consistency
check bridging the gap between static (saturation) and
dynamic (well test) data. By ensuring consistency
between the SHM and the permeability we expect the
need for using multipliers in the dynamic domain to be
reduced. A new check can be performed between the
permeability and the SHM at the fine scale and contrast
Fig. 14: A rock type example shown in Fig 1 is plotted to the upscaled model paving the way for checking the
against the geometrical averaged perm (green), Lucia upscaling method.
(red) and Winland (blue) lines highlighting Winland
constant R35 unsuitability in characterizing in general REFERENCES
a rock type.
Amaefule, J.O., Altunbay, M., Tiab, D. , 1993,
A more fundamental aspect of this work refers to the Enhanced Reservoir Description: Using Core
porosity dependency of permeability for one rock type. and Log Data to Identify Hydraulic (Flow) Units
When compared to default permeability dependency for and Predict Permeability in Uncored
a given Rock type as given by Winland (Pittman, 1992) Intervals/Wells, Presented at the SPE Annual
or Lucia (Lucia, 1995) one can see quite a difference, Technical Conference and Exhibition, Houston,
see Fig. 14. Our rock types display a slightly weaker 3–6 October, SPE 26436.
permeability dependency as compared to Lucia’s but Chen A., 2014, The Practice of Graphical Fluid-
not too far off. Most of the permeability range is bound Gradient Interpretations of Formation Tester
by a RFN 1 to 2 range in this particular case, see Fig.
9
SPWLA 56th Annual Logging Symposium, July 18-22, 2015

Pressure Data, AAPG Bulletin, v. 98, no.7, p. ABOUT THE AUTHORS

1431-1448.
Clerke E. A., 2009, Permeability, Relative Iulian N. Hulea is a Senior Petrophysicist working for
Permeability, Microscopic Displacement Shell Global Solutions BV, Projects and Technology in
Efficiency and Pore Geometry of Bimodal Pore the Netherlands. Currently working on carbonates field
System in Arab-D Limestone: Society of development planning. Before this position he held a
Petroleum Engineers Journal, v. 14, no. 3, p. 524 Research Petrophysicist position, also in Shell. He
– 531. holds a Master (Bucharest University, Romania) and a
Hulea I. N. and Nicholls C., 2012, Carbonate rock Phd (Leiden University, The Netherlands) in
Characterization and Modeling– Capillary experimental physics. After completing the PhD (2004)
Pressure and Permeability in Multimodal Rocks he held a Postdoctoral position at the Delft University
– A Look Beyond Sample Specific of Technology, Kavli Institute for Nanoscience also in
Heterogeneity, AAPG Bulletin, v. 96, no.9, p. The Netherlands.
1627-1642. Daniela Frese is a Senior Petrophysicist working for
Hulea I. N., 2013, Capillary Pressure and Permeability Shell Global Solutions BV, Projects and Technology in
Prediction in Carbonates Rocks –New Methods the Netherlands. Currently working on Fluid Evaluation
For Fractures Detection and Accurate Matrix and Sampling Technologies. She holds a Masters in
Properties Prediction, SPE 164251. Geology from the University of Aachen, Germany.
Lucia, F. J, 1995, Rock fabric/ Petrophysical Shyam Ramaswami is a Senior Petrophysicist working
Classification of Carbonate Pore Space for for Shell Global Solutions BV, Projects and
Reservoir Characterization: AAPG Bulletin, v Technology in the Netherlands. Currently working on
76, p 1275-1300. Fluid Evaluation and Sampling Technologies. He holds
Pittman, E. D., 1992, Relationship of Porosity and a Masters in Aeronautical Engineering from the
Permeability to Various Parameters Derived University of Bristol.
from Mercury Injection – Capillary Pressure
Curves for Sandstone: AAPG Bulletin, v. 79,
Acknowledgements
no.2, p 191-198.
The authors would like to thank Mirano Spalburg for
Ramaswami S., Hows M., Frese D., Dong C., and
Elshahawi H., 2013, Pressure Transient Data his input in the technical work, the stimulating
from Wireline Formation Testers: When and discussions and reviewing the manuscript. Discussions
How to use it?, SPE 164733. with Chris Nicholls around the rock typing ideas are
Tangyan L., Zaitian M., Junxiao W. and Hongzhi L., acknowledged as well as discussions with Jerry Lucia
2005, Integrating MDT, NMR Log and on the relation between saturation and permeability.
Conventional Logs for One-Well evaluation, The authors would also like to thank Harm Dijk,
Journal of Petroleum Science and Engineering, v Anneke Kleinpenning, Mart van Stiphout, Maria Boya
46, p 73-80. Ferrero, Hajo van Hasselt and Wim Looyestijn for
Thomeer, J.H.M., 1983, Air Permeability as a Function reviewing the manuscript and Shell for supporting the
of Three Pore Network Parameters: Journal of work and allowing this work to be presented.
Petroleum Technology, v 33, no 12, p 2498-
2505.
(Techlog, 2014)
http://www.software.slb.com/Lists/SalesandMar
ketingDocuments/Attachments/207/techlog_for_
reservoir_engineers_ps.pdf
Volokitin Y., Looyestijn W. L., Slijkerman W. F.J.
and Hofman J. P., 2001, A Practical Approach
To Obtain Primary Drainage Capillary Pressure
Curves From NMR Core And Log Data,
Petrophysics, v 42, p 334-343.
Xu C. and Torres-Verdín C., 2012, Saturation-Height
and Invasion Consistent Hydraulic Rock Typing
using Multi-Well Conventional Logs, 53rd
Annual Logging Symposium.

10