SPE 102894

Permeability from Production LogsMethod and Application

M.J. Sullivan, D.L. Belanger, M.T. Skalinski, S.D. Jenkins, and P. Dunn, Chevron

Copyright 2006, Society of Petroleum Engineers

This paper was prepared for presentation at the 2006 SPE Annual Technical Conference and
Exhibition held in San Antonio, Texas, U.S.A., 2427 September 2006.
This paper was selected for presentation by an SPE Program Committee following review of
information contained in an abstract submitted by the author(s). Contents of the paper, as
presented, have not been reviewed by the Society of Petroleum Engineers and are subject to
correction by the author(s). The material, as presented, does not necessarily reflect any
position of the Society of Petroleum Engineers, its officers, or members. Papers presented at
SPE meetings are subject to publication review by Editorial Committees of the Society of
Petroleum Engineers. Electronic reproduction, distribution, or storage of any part of this paper
for commercial purposes without the written consent of the Society of Petroleum Engineers is
prohibited. Permission to reproduce in print is restricted to an abstract of not more than
300 words; illustrations may not be copied. The abstract must contain conspicuous
acknowledgment of where and by whom the paper was presented. Write Librarian, SPE, P.O.
Box 833836, Richardson, TX 75083-3836, U.S.A., fax 01-972-952-9435.

Abstract
Estimation of effective permeability at the reservoir scale
has been a long standing challenge in carbonate fields. The
carbonate depositional and diagenetic history can be quite
complex, and this can lead to a permeability field which is
quite difficult to characterize. Permeability in vuggy or
fractured intervals can be dramatically different from the
matrix permeability measured in core plugs. However
realistic estimates of oil recovery, and optimized reservoir
management requires good estimates of the reservoir
permeability.
In the Tengiz field, a giant carbonate reservoir in western
Kazakhstan, a method has recently been developed to
calculate apparent permeability (APERM) based on flow rate
from production (PLT) logs. Incorporation of this flow
calibrated apparent permeability into the static geologic earth
model offers an elegant solution to the long-standing problem
of how to best incorporate dynamic PLT data into a reservoir
model. A reservoir model recently built using APERM
resulted in a step change improvement over previous methods
where only static log based permeability transforms were used
to populate the earth model.
Conventional log based permeability transforms are
designed to characterize matrix permeability but not the
excess permeability due to fractures & vuggy porosity
common in carbonate reservoirs. The APERM method is used
for both accurately characterizing total permeability (matrix +
excess), and for identifying inaccurate permeability
predictions in older wells with poor log quality or limited log
data. The log based permeability predictions are more
accurate in recent wells with modern logs, but hot streak
identification and quantitative permeability estimation from
static well logs is still problematic.
The apparent permeability is calculated by solving Darcys
law on an interval basis, using as input our knowledge of
flowing and static pressures, plus well, reservoir, and fluid
properties.
The method makes several simplifying

assumptions, but the resulting errors are second order in

nature, and the method offers improvements over using
conventional static log based transform permeability.
Application of the method is enhanced by the derivation of
coarse scale zonal layer pressures with multi-rate PLTs.
Accurate zonal layer pressures improve the accuracy of the
permeability derivation. The apparent permeability from PLT
is then used as a benchmark to adjust the transform
permeability derived from static well logs using a variable
multiplier. This technique has the advantage of preserving the
original fine scale heterogeneities of the wireline logs, while
calibrating their magnitudes. It also has the advantage of
identifying higher permeability intervals from rapid changes in
well inflow profile, that may not have been characterized by
conventional (core plug and wireline) estimates.
Recently a full field reservoir model for Tengiz Field has
been constructed using APERM data and petrophysical rock
types. Preliminary history match results from the model show
that use of APERM has increased confidence in the modeled
permeability field before history match. Fewer changes are
required to calibrate the full field reservoir model to well
pressure data (SGS). Permeability-Height (KH) estimates
from the model are a much better match to well test KH, than
models derived from core transforms. The models also have
significantly more heterogeneity, as shown by DykstraParsons calculations.
A similar improvement in history match was recently
observed in a high resolution model constructed using
APERM to monitor gas movements in the Tengiz platforms
area. History match pulse tests have shown that this model
has a much better prediction of the inter-well connectivity than
models without APERM. Models with the APERM data have
provided higher confidence estimates of the future movement
of gas injection in the Tengiz platform.
Future plans are to investigate a correspondence between
APERM and Rock Types as well as statistical transforms with
open-hole logs. The log based transforms can then be used in
wells or intervals without PLT data, improving accuracy of
permeability population into the reservoir model.
Introduction
The fundamental building blocks of a reservoir model are
porosity, permeability and fluid saturation. Porosity and fluid
saturation can be computed from wireline logs with reasonable
confidence, but the same often cannot be said for permeability.
Permeability can be measured directly from core material or
from well tests via pressure transient testing (PTT), pulsetesting between wells, and wireline formation tests. This

paper shows the application of production logging tests (PLT)

to provide improved well scale permeability for use in
reservoir models.
Scale differences in these sources of permeability data
complicate the permeability estimation in a reservoir model3,4.
Core permeability measurements are used to develop wireline
permeability estimates, and pressure transient tests are used to
calibrate these estimates. Determining permeability from well
logs is complicated by differences in scale, measurement
environment, and measurement physics. Well logs have a
vertical resolution of typically 0.5 to 1 meter and provide insitu measurements, compared to 5 cm diameter core plugs that
are measured on the lab bench
In most reservoirs, the amount of core available for direct
permeability measurement is limited, and permeability
estimates are usually made by proxy using wireline log
responses. The science of permeability prediction has evolved
to a high degree of sophistication, with each additional level
resulting in an incremental gain in predictive accuracy1
Even with sophisticated predictive algorithms, accurate
assessment of permeability in carbonate reservoirs is often still
problematic because of significant variations in rock types and
the nature of the pore geometry. When permeability is matrix
dominated and a modern suite of high quality wireline logs is
available, the predictive accuracy can be good. However
carbonate reservoirs can have many different pore structures2
which have widely varying por/perm relationships making
accurate prediction from wireline logs difficult or impossible.
Vuggy porosity and fractures, if they are connected, can result
in extremely high layer flow rates or very low rates if not
connected, for the same value of porosity.
In many carbonate reservoirs the primary source of flow
capacity are fractures. Recognizing the existence of fractures,
and accounting for them in a reservoir model, may be difficult
or impossible in wells with older log suites without image logs
or modern vintage sonic logs. Even with high quality image
logs, recognizing whether fractures are open or closed can be
difficult. Even if open fractures are identified in the borehole,
there may be a great deal of uncertainty whether these
fractures are laterally continuous, and to what degree they are
fed by the rock matrix. These factors increase uncertainty in
productivity prediction and greatly complicate reservoir
characterization efforts.
Several authors have shown how the accuracy of the
reservoir characterization can be improved significantly by
incorporating dynamic data, such as pressure transient tests,
periodic pressure measurements, and production profiles, into
the reservoir characterization process,5,6,7. However, the
industry has lacked a robust work process for flow calibrating
log derived permeability and distributing it between wells in a
multi-layer reservoir model. This obstacle can be overcome
with the use of apparent permeability derived from PLT logs.
This paper describes a new method for deriving apparent
permeability from production log flow profiles. The method
provides a direct integration of static core and wireline log
data with dynamic pressure transient and PLT data. It
calibrates the wireline scale permeability to ground truth
pressure transient results, and retains the character of the log
data. This improves the accuracy of the permeability used in
the model and inherently integrates dynamic data into the

SPE 102894

static model. The method overcomes the difficulties created

by the differences in scale of various data types mentioned
above. The reservoir characterization process is improved
because calibration to dynamic flow data is built in at the start.
Two reservoir models have been successfully constructed
for the Tengiz reservoir using the APERM technique. These
models include a full-field model used for infill drilling and
reservoir management, and a high resolution model of the
Tengiz platform used to optimize a planned sour gas injection
project. Reservoir models constructed using APERM have
provided a significant improvement in ability to predict well
tests (SGS, pulse tests) with minimal history matching effort.
Improved confidence in the history match of these models has
provided additional confidence in reservoir management
activities such as gas injection and infill drilling of the Tengiz
Field.
The Tengiz Reservoir
Tengiz is a large carbonate reservoir in western
Kazakhstan. Tengiz field (Fig. 1) is located on the south side
of the 500,000 square kilometer (km2) Pri-Caspian basin on
the northeastern edge of the present-day Caspian Sea. Other
giant oil and gas fields located in similar settings include
Karachaganak and Orenberg. The reservoir has a very large
areal extent. Tengiz is over 110 km2 in area at its top and 400
km2 at its base. The oil column is quite thick, as top of the
reservoir is located at approximately 3,850 m subsea and the
approximate oil water contact is at 5,450 m subsea.

Korolev
Tengiz

TENGI

10 Km

Fig. 1 Location of Tengiz Field and the Caspian Sea

The Tengiz reservoir is an isolated carbonate buildup

(platform) with a mesa-like geometry (flat-topped and steepsided). The platform has a slight regional tilt today of less than
one degree to the south. The sides of the platform have much
greater dip (approximately 25 degrees) on the flanks. The
present-day geometry of the Tengiz structure is that of a
topographically high-rim area surrounding a depressed central
platform. It is flanked by a dipping platform-to-basin

SPE 102894

transitional facies along the flank, as shown in Fig. 2. The

Tengiz reservoir formed during the Late Devonian and Lower
to Middle Carboniferous time by the deposition of calcareous
skeletal fragments and lime mud. The cyclicity of these
sediments was in response to the interplay of sea-level rise and
basin subsidence. The platform appears to have drowned in
late Bashkirian or early Moscovian time.

Cell Permeability

Rim

Platform

Fig. 3 Map Showing Increased Permeability in Tengiz Flank.

Flank

Fig. 2 The Tengiz Platform: Rim and Flank regions.

The central platform rock type is dominated by the matrix

porosity and permeability. Previous methods used core
porosity-permeability transforms for converting well log
porosity to permeability. This method works well only with
matrix-dominated permeability rocks and does not work well
with fractured and vug-dominated rocks.
Conversly, the productivity of the rim and flank regions is
dominated by fracture permeability that cannot be easily
characterized from conventional well logs. Permeabililty
estimates from the rim and flank region need to include first
order influence of an effective fracture network. Specialized
software (FRACSIM) is used to simulate the fracture density,
and estimate the effective permeability in the rim and flank
when fractures are accounted for. The permeability height
(kh) is much higher in the flank than the matrix kh when the
effect of fractures is considered (Fig. 3).

Tengiz has a great deal of remaining development

potential, and consequently there is a large focus on reservoir
characterization.
Pressure differences have been noted in
various reservoir intervals, and are related to differences in
reservoir porosity, permeability, and voidage.
Surveillance activities in Tengiz include production
logging surveys, inter-well pulse testing, and pressure buildup
tests.
Most of the wells in the Tengiz reservoir are vertical or
near vertical and all are flowing single phase oil with flowing
bottom-hole pressure above the bubble point. From this
perspective, the conditions for PLT profiling are ideal.
Differential depletion of various reservoir intervals has
occurred in Tengiz, and this is accounted for by the APERM
method described in the next section.
APERM Method
A typical production log in Tengiz includes profiles
acquired under both stabilized shut-in and then stabilized
flowing conditions. After the flowing survey is completed, the
well is shut-in for a pressure buildup test until the infinite
acting radial flow period is reached8. The profiles are then
analyzed with detailed inflow points defining the regions over
which permeability will be calculated. The derivative with
respect to depth of cumulative flow profile is calculated. This
derivative yields the volumetric inflow on a meter by meter
basis (Note that the unit of depth in Tengiz is meters, and so
will form the convention for this discussion - the technique
obviously will work equally as well with depth in feet). This
derivative is equivalent to the traditional interval contribution
commonly seen on PLT analyses divided by the length of the
zonal contribution.
Given the flow contribution per meter from derivative of
cumulative flow, for each meter (h=1), the permeability is
calculated using the familiar Darcys law radial flow equation:

plt

C * Qi * U o Bo
R
[ Ln ( e ) + S ' ]
( Pe Pwf )
Rw

(1)

SPE 102894

where C=141.2 if depth is in feet, or 43.07 if depth is in

meters (Tengiz convention), Qi is flow rate in stbopd, Uo is
viscosity in Cp, pressure in psi, and Re/Rw in consistent units.

results in cross flow into zone 1 when the well is shut-in at

surface (Fig. 6, 7).

PLT permeability (kplt) represents average permeability

over that specific inflow zone. The kplt data is accumulated
over the entire interval to compute khplt. This value is then
normalized to the pressure buildup khbu by multiplying each
kplt value by khbu/khplt.
kplt_norm = kplt (khbu/khplt)
(2)
This normalized PLT permeability (kplt_norm) is used as a
baseline reference to which the wireline transform
permeability from open hole logs is adjusted using a
multiplicative factor (which we call perm_boost). This
process is repeated, inflow by inflow, from the bottom of the
zone upward in a stepwise fashion (Fig. 4). This adjusted
transform permeability curve is called APERM for apparent
permeability. After each boost APERM is accumulated
from the bottom up and compared to the accumulated
kplt_norm. By matching the increases in cumulative KH
between kplt_norm and APERM at each inflow step the final
result is normalized to well test permeability. The technique
of boosting the wireline transform permeability (instead of
simply using kplt_norm) offers the significant advantage of
preserving the vertical heterogeneity visible in the open-hole
wireline logs, while calibrating their magnitudes.
Step 1

Step 2

Fig. 5 Pressure gradient plot showing differential depletion.

The differential depletion causes each layer to flow at a

different percentage drawdown for a given Pwf. These different
individual layer pressures must be accounted for in the
calculation of permeability using the APERM method (Pe in
the denominator of equation 1). We routinely measure
individual layer pressures using the Selective Inflow
Performance (SIP) technique9. This technique involves
acquisition of production profiles at multiple rates and flowing
bottom hole pressures.

Fig. 4 Stepwise calibration to PLT_perm.

A typical Tengiz well with 400m interval and 20 inflow points

can be completed in about 30 minutes using this iterative
technique. We envision this process could be automated with
a purpose built computer code.
Discussion of Details on the Method.
Handling Differential Depletion. The Tengiz reservoir
comprises 3 main flow units that behave as different pressure
compartments (Fig. 5). The upper-most layer (zone 1) is
generally the most prolific so is generally more depleted than
the lower two layers (zones 2 & 3). This differential depletion

Fig. 6- Typical shut-in PLT profile.

SPE 102894

wells. The leftmost point (negative flow rate) on the zone 1

curve in Fig. 8 is the stabilized rate of crossflow into that
zone.
Simplified Approach. If there is no differential depletion
in the reservoir, and the assumption of uniform skin
distribution can be made, a simplified approach may be taken.
With no differential depletion, all flowing bottom hole
pressures will result in the same drawdown being applied to
each layer. The accumulated permeability may simply be
normalized to well test KH, and distributed vertically
according to interval contribution. This simplified approach
has proven successful in other areas10.

Fig. 7 Flowing profile from the same well.

Pressure is plotted against rate for each reservoir flow unit

(Fig. 8). For single phase liquid flow, when the bottom hole
pressures are above the bubble point pressure, the points will
all plot on a straight line. That pressure for which the layer
flow rate would be zero defines the layer pressure. By using
this technique, no zone ever has to be static, i.e. a rate of zero.
Tengiz is an ideal environment for this technique, because all
flow is single phase oil above the bubble point pressure.
When we initially applied this technique in Tengiz we
acquired four PLT profiles; shut-in and 3 different stable
flowing rates. Pressure versus rate data plotted on a straight
line for each zone as predicted by theory for single phase flow.
Modeling and field testing confirmed that the intermediate
flow rates have a minor influence on the computed layer
pressure when a stabilized shut-in profile with crossflow has
been recorded. A practical justification for only conducting
two flow rates (one stable shut-in and one stable flowing) is
that two longer duration and more stable profiles provide
better results than 4 flow rates of shorter duration where each
rate suffers from transient flow effects and is not yet stable.

Significance of Errors
The accuracy of the APERM calculation can be assessed
by examining the certainty factor of the input variables in Eq.
1. As discussed below, uncertainty for many of the input
variables result in an impact on permeability that is
insignificant for our purposes. In the calculation of Apparent
Permeability, we make several simplifying assumptions, but
the resulting errors are second order in nature, and do not have
a significant impact on the result.
To put the errors in perspective, we first must determine
how large of an error is significant. For our purposes, a value
within a factor of two of the true value of permeability is
adequate for reservoir characterization purposes and is a
significant improvement over a permeability field based on
only on log to core transforms. The most important feature of
the APERM method is that it achieves the correct vertical
distribution of permeability, helping to correctly characterize
the flow units in the reservoir.
To illustrate, Fig. 9 is a crossplot of porosity vs.
permeability from 6 cored wells in Unit 1. Rock types are
depicted by color scale. There is approximately 1.5 log cycles
of scatter in the raw data (50x). A difference of a factor of
two is essentially within the noise.

Fig. 8 SIP Plot of Zonal Rates & Pressures.

Tengiz wells are typically shut-in for several days before

logging so a stabilized shut-in profile is available in most

Fig. 9 Core porosity -permeability crossplot.

To further illustrate the impact of a being off by a factor of

two, Fig. 10 is a plot of the transform permeability both
multiplied and divided by two (track 2). The rightmost track
shows the core permeability data (blue dots) plotted on the
transform perm (black curve) that is based on a multivariate fit
between open hole log data and core permeability for specific
rock types. The scatter in core perm data around the transform
permeability is approximately equal to the factor two error
illustrated in track 2 above. The red blocky curve in this track
is the Apparent Perm from PLT. The APERM curve is in
excellent agreement with core data.

SPE 102894

about 2% of the true value. As illustrated in Fig. 11 the layer

pressure is usually well constrained. The leftmost points on
the SIP plot are those operating rates and pressures while the
well is shut in at surface, and crossflowing downhole. The
size of the reservoir compartments are very large in
comparison to the crossflow rate, so the rate of crossflow will
stabilize and remain relatively constant over very long periods
of time (months). The fill up effect in the more depleted
Bashkirian due to crossflow is minor. Because we shut the
wells in for several days prior to commencing a PLT survey,
by the time the shut in PLT passes are made, the rates and
pressures are quite stable.

Fig. 11: Uncertainty in layer pressures is usually low using SIP

technique because the shut-in passes are at stable pressures.
The uncertainty in drawdown pressure is mostly due to
uncertainty in stable flowing bottom hole pressure.

Fig. 10 Permeability shifted to illustrate impact of twofold error,

natural scatter in raw data, and good fit of calculated APERM
data.

Discussion of Potential Uncertainties and their Impact on

Apparent Permeability
1) Fluid Properties: The product of Uo & Bo changes
little over the typical range of pressures we operate in, and is
usually known to within 5%. This does not have a significant
impact on calculated permeability (i.e. less than a line width
on Fig. 10 above).
2) Pressures: An error in drawdown will cause a
proportional error in calculated permeability. We have found
in most cases we have been able to assess stabilized
drawdown within 10% uncertainty.
A potential source of inaccuracy in the SIP technique is
lack of stability (multiphase flow would also be a source of
inaccuracy in other reservoirs). We are normally able to
assess the layer pressure using the SIP technique to within

The crossflow we observe during shut-in conditions helps

the SIP analysis by providing a definitive operating point on
the left side of the SIP plot.
SIP Pressure Confidence. The stabilization time required
for the flowing passes is sometimes an issue. We now open
the well to flow with PLT tools above all perforations, and
monitor rates and pressure for stability before commencing
flowing passes. A real-time log-log pressure derivative plot is
used in the assessment of stability. When analyzing historical
data where this was not done, we may still assess pressure
stability by comparing pass to pass pressure repeatability.
Correction for Transient Flow Instability. On some wells
with lower permeability, the pressure does not stabilize with a
practical amount of time. In practice the flowing passes begin
within 8-12 hours of opening the well to flow even if the flow
has not become perfectly stable by that time. If a well has not
stabilized by that point in time, an additional 12 hours is
unlikely to make a significant difference. After the flowing
passes have been completed, the well is shut-in for pressure
buildup. A log-log derivative pressure diagnostic plot is used
to ensure the Infinite Acting Radial Flow period has been
reached before terminating the buildup. The rate and pressure
data may then be analyzed with pressure transient analysis
software. In doing so, we have constructed an interpretation
model that can be used to extrapolate to what the flowing
bottom hole pressure would have stabilized at had we
continued to flow (Fig. 12). It is this estimate of flowing
bottom hole pressure that we use in the calculation of apparent
permeability. Although this technique is not exact, we believe

SPE 102894

this is a reasonable method of compensating for instability,

and the residual error is reduced to an acceptable level.

Productivity Ratio vs Skin

2.5
Productivity Ratio

2
1.5
1
0.5
0
-5

10

15

20

Skin

Fig. 13 Impact of skin factor, expressed as a proportion of the

productivity the well would have with a skin of zero.

Fig. 12 Well not stable during time allotted for PLT. Modeled
stable pressure is used for calculation of APERM.

It is unlikely to have more than a 20% error in drawdown

pressure, after using the above correction method if necessary.
Because permeability is proportional to drawdown, this will
only produce an error in APERM of 20%. This only affects
the total KH, not the vertical distribution of it, which is the
more important issue.
It is also important to note that the error in drawdown,
expressed as a percentage, remains more constant over a range
of drawdowns. Wells with better permeability stabilize
quicker, have a smaller error bar on pressure, but also have
lower drawdowns. Lower permeability wells have larger error
bars on pressure, but usually have very high drawdowns. The
error bar as a percentage of drawdown may be similar between
the two cases.
4) Drainage Radius & Wellbore Radius. Because these
factors are inside a logarithmic term, the calculated Apparent
Permeability is very insensitive to errors in radius. For typical
Tengiz assumptions, a 20% difference in either radius makes
only a 1.5% difference in calculated permeability.
5) Flow Rate. When possible, we try to flow the well
through a test separator during a PLT but this is not always
possible due to operational constraints. The PLTs that are
conducted with stable flow measured with accurate surface
meters typically match the surface rates with a difference of
less than 10%. Because rate is proportional to calculated
perm, this again has a negligible effect on APERM.
6) Skin Factor: The assumptions made about skin factor
typically receive the greatest amount of attention, for this can
have the impact of twofold difference in permeability, or more
(Fig 13).

A single skin factor equal to that value calculated from the

pressure buildup test is used as a starting point for the
calculation of APERM. Under ideal circumstances, this can
be a valid assumption. The ideal case is one where the PLT
was run after the well was stimulated with several
reciprocations of a coil tubing acid wash resulting in an even
stimulation and a calculated skin factor of -4 or less (highly
stimulated) based on the post stimulation pressure build up
analysis. Approximately half of our APERM interpretations
fall into this category. The rest are complicated by a variety of
factors. Any time a well has had a workover on it, where the
well was killed with drilling mud, and the well may now have,
for example, a +2 skin factor, there is reason to question the
assumption of the same skin factor for all intervals. A well
with +2 skin may have a 0 skin over a higher pressure zone
with shallow invasion and a +5 skin over lower pressure, more
deeply invaded zone.
Our certainty factor on computed APERM is influenced by
how the APERM value compares to that predicted by
permeability transform. If a highly positive boost must be
applied to the transform permeability to match the PLT perm
there is more confidence that the high perm streak is real and
not the product of over-stimulation, especially if we assumed a
negative skin in the calculation of APERM. With matrix
acidizing, the skin factor cannot be much more negative than
about -5. If a negative skin is assumed in the APERM
calculation and a boost factor of 10 is required to match the
PLT_perm, the only possible explanation can be that the
permeability is indeed as high as it appears to be.
Low Permeability vs. Damage. If the productivity of a
zone is much less than predicted, then we must assess whether
the permeability really is low, or whether this layer is
damaged or unstimulated. To make this assessment, we must
take into account factors such as; drilling history (whether
fractures were encountered, amount of mud losses into those
fractures indicating high permeability), workover history,
prior flow tests (especially pre-workover tests), stimulation
history (type and volumes), as well as vintage and quality of
open hole logs.
For example, when a well that has been drilled and logged
with modern logs of high quality on the reservoir platform
(where the porosity is primarily intercrystaline and the
transform permeability is usually fairly accurate), shows much

less inflow than was predicted by the transform perm, and the
measured skin factor was -2, we would be inclined to conclude
that the reason for the low productivity was a high skin factor
over this interval, and leave the transform permeability unadjusted.
Other wells may have poor open hole log quality, limited
logging suites, or logs made with by tools that have been
poorly characterized by western standards. In these cases, we
often have much less confidence in the resulting predicted
permeability, and so would be less hesitant to adjust the
permeability downward. Fig. 14 is an example of this. Track
1 shows porosity from open hole logs (red line), porosity from
PNC ratio normalized to PHIT (black line), with difference
shaded in grey. Acid effect (explained later in this text), is
shaded in magenta. Track 2 is a wellbore diagram, showing
perforations and two strings of casing. Track 3 is transform
permeability from open hole logs (green), apparent
permeability from PLT (black bars), and the resulting adjusted
APERM (red). Track 4 shows the multiplicative permeability
adjustment factor used to calibrate the transform perm to PLT.
Positive boosts are shaded in magenta; negative boosts are
shaded in grey. Also posted in this track is the difference
between the open hole and PNC porosity (linear scale). Track
5 is the PLT results.

Fig. 14 Example of reduction of transform permeability after

calibration to PLT perm.

In this case we felt confident in adjusting the permeability

downward because we believed the original open-hole
porosity (from Soviet era uncalibrated neutron log with poor
excentralization) was too optimistic over the lower section.
The porosity values were higher than offsetting wells, whereas
the PNC porosity was consistent with offsets. The difference
between the two tracks well with the required downward
adjustment factor. A reciprocating coil tubing acid stimulation
was conducted over the entire interval, and the PNC acid
effect shows the entire interval was stimulated, so we do not

SPE 102894

suspect damage is responsible for the lower than predicted

flow.
Flank Wells fractured reservoir. In the rim and flank
areas of the Tengiz reservoir, fractures are frequently the
primary source of productivity. Matrix porosity is very low,
in the range of 0-5% versus 10-15% for the reef platform.
KH is often under predicted by 2 or 3 orders of magnitude
by the permeability transform. The permeability transform is
designed to predict matrix permeability only. The excess
permeability due to fractures is difficult or impossible to
predict. At best, we can recognize the presence of fractures
with FMI image logs which are surprisingly good in the oilbased mud. However, both healed and open fractures are
shown as high resistivity and it is difficult to separate them.
To classify open fractures, we are using other fracture
indicators such as: Pef (barite invasion in the fracture),
Stoneley wave reflection and mud losses.
Matrix Permeability. Permeability transforms were
based on the statistical relationships between log porosity and
core permeability. Klinkenberg core permeability was
corrected for overburden, assuming an average net overburden
pressure of 3000 psi.
A total of 8056 plug data from 16 wells (5 platform and 11
Flank) constituted an initial database. The plugs with
indication of fractures (mostly drilling induced) were rejected
to obtain an accurate representation of matrix permeability.
Two transform types were created: porosity-based (linear
or non-linear) and multivariate.
Porosity based transforms were used to populate
permeability in the geologic model. Multivariate transforms
(with porosity generally as a main predictor) were used for the
definition of APERM.
A total of 22 transforms were defined for combinations of
stratigraphic horizons and reservoir gross facies (inner
platform, outer platform, upper slope, middle slope and toe of
slope).
Fig. 15 shows an example of matrix permeability
transform for zone 2 in the platform. Application of
multivariate transform improves permeability prediction error
from 0.9 log cycles (porosity based) to 0. 5 log cycles
(multivariate) for matrix permeability on wells with modern
logs.

SPE 102894

100

2
876
909
31

100

10

Permeability (mD)

0.270

0.300

0.240

0.210

0.180

0.150

0.120

0.090

0.060

0.030

0.000

Matrix Permeability Transform for Serp-Unit 1

10

0.1

0.1

0.270

0.001
0.300

Porosity (frac)

0.240

0.210

0.180

0.150

0.120

0.090

0.000

0.001

0.060

0.01

0.030

0.01

0.5

10.5
Petrophysical Rock Types

2
3
K _ sm4 = 10( 3.34802 + 50.7596 229.738 + 869.252 )

Perm = 10 (2.503 + 17.412 CMFF 0.239 log(Cxo) 0.0692SGR )

Fig. 15 Example of permeability transform for zone 2. Porosity

rd
transform is defined by 3 order polynomial and multivariate
transform by NMR Free Fluid Volume from NMR log, spectral GR,
and logarithm of conductivity.

Open-Hole Log Quality, and confidence in the

Petrophysics.
Before the TengizChevroil joint-venture was formed in
1994, a total of 67 wells were drilled and logged using Soviet
technology. Half of the wells were logged with full logging
suites including density, neutron and sonic logs. However,
calibration of the Soviet logs was not up to the western
standard which triggered a need for the normalization effort.
Since the joint venture was formed, a total of 41 wells were
drilled and logged using a modern suite of logs. These logs
were used to normalize Soviet log data, first in 1999 then in
2005.
Uncertainty analysis indicates that maximum error in
porosity determination for modern logs in 1.5 PU while the
wells with Soviet era logs can have porosity uncertainty up to
4 PU. A major source of uncertainty was inability to predict
bitumen content from Soviet logs. In modern wells bitumen is
estimated using Nuclear Magnetic Resonance logging and
advanced multimineral modeling. Fig. 17 shows comparisons
of log versus core porosity for 2 key platform wells.
Core Porosity
0.045

1.0

0.040
0.8

0.035

Phi_cor- phi_log

0.030
0.6
0.025
0.020
0.4
0.015
0.010

0.2

0.06

0.0

0.05

1.0

0.005

Statistics

0.8

0.300

0.270

0.240

0.210

0.180

0.150

0.120

0.090

0.060

0.030

0.000
0.000

0.04

Wells:

0.6

Phi_av= 8.63 %

0.03
0.4
0.02

Log Porosity

0.2
0.01

0.8

0.035
0.030

0.6

0.100

0.080

0.060

0.040

0.020

0.000

-0.020

0.0

St ti ti

-0.040

0.00
-0.060

1.0

0.040

-0.080

0.045

-0.100

Excess Permeability
Permeability derived from PLT is in agreement with
matrix permeability in most of the reservoir zones. However,
several inflow zones yield a boost factor >1 which results in
APERM higher than log derived matrix permeability. The
excess permeability named EPERM is related to secondary
porosity which corresponds most likely to the vuggy/moldic
porosity in the platform and fractures in the flank (Fig. 16).

W ll

Delta Phi_av = - 0.0017

0.025
0.020
0.4
0.015
0.010

0.2

0.005

Matrix Perm
Prediction Area

Eperm events

St ti ti

0.300

0.270

0.240

0.210

0.180

0.150

0.120

0.090

0.060

0.030

0.0
0.000

0.000

W ll

Phi_av= 8.97 %

Permeability

Fig. 17 Comparison of Core porosity with Log-derived porosity

for 2 key wells. Multimin porosity yields unbiased values with low
uncertainty.

Matrix
transform
from Core
APERM values from PLT
Porosity

Fig. 16 Excess permeability is defined as a PLT permeability

which falls above matrix permeability prediction range.

Confidence Factor
A confidence factor is assigned to each interpretation to
communicate the level of certainty in the final APERM value.
In the example illustrated in Fig. 4 the permeability of the
upper zone has a positive boost. A key point with this method
is that when a much higher permeability than predicted by
transform is encountered there is higher confidence that this is
real and significant.

Using Fig. 9, a porosity uncertainty of 1.5 PU translates to

a matrix permeability uncertainty of 0.3 logarithmic cycles.
The 4 PU porosity uncertainty of Soviet logs would translate
to almost one order of magnitude uncertainty in matrix
permeability.
EPERM Prediction from OH Logs
Matrix permeability was predicted using core permeability
vs porosity transforms (Fig 15, 16). An attempt was made to
predict EPERM (secondary permeability) from logs in order to
predict this dynamic property in wells without PLT. Since,
secondary perm is usually driven by properties other than
porosity (amount of connected vugs, fractures, rock types etc),
predictive variables were selected to consider all logs which
might carry this information.
Use of Stepwise Multiple Regression allows to evaluate
the predictive significance of input variables. The dependent
variable used was EPERM defined as :
EPERM = APERM - Kmatrix

(3)

Where: K_matrix matrix permeability predicted from logs

using Core PHI-k transforms

10

SPE 102894

Input independent logs included:

Porosity corresponding to different T2 cutoffs:
o FFI(1-8)for:8;16;32;64;128;256;512;1024
(>cutoff) and BVI below the same cutoffs.
o K_sdr,K_tim permeabilities from CMR
logs.
Ct,Cxo and Delta_con (short and deep
conductivities and their difference )
PHIE, PHIE-vug (porosity from Multimin and vug
porosity derived from sonic using Lucia method)
Raw logs such as: RHOB,T2LM, CGR,SGR, U,
Pef, CMFF,CMRP
For variables exhibiting lognormal distribution a
logarithmic transform was applied. Multiple regression was
run separately for Bashkirian and Serpukhovian formation and
resulted in predictive regression equations.
The most
significant message from this statistical exercise was:
1. EPERM can be predicted from logs in statistically
reliable way ( R= 0.8-0.85 standard prediction error
approx 1 log cycle)
2. Top predictive variables include consistently
variables such as FFI7,FFI8, (porosity corresponding
to high T2 cutoffs), Phi_vug which indirectly
indicate that the EPERM in platform is caused mostly
by connected vugs.
Those are encouraging results showing possible link
between static log data and dynamic PLTderived
permeability. Regression analysis is based on data from three
wells, and we plan to obtain more PLT data in wells with
modern logs to validate the prediction and create more robust
transforms.
Predicted EPERM was than used to obtain predicted
APERM using:
APERMpred = EPERMpred + kmatrix

water saturation is approximately 10% of pore volume. The

salinity of the connate water is low, only about 60 kppm.
When an interval is acidized, the acid reaction products are
essentially saturated with calcium chloride. The elevated
chlorine concentration imbibes into the irreducible water
increasing its chlorine concentration in the near wellbore area.
The spent acid water flows back during clean up but the
elevated chlorine levels in the irreducible water remain, and
may be detected with a PNC sigma log. We believe this effect
is permanent, or nearly so, because we have observed elevated
sigma readings due to acid several years after the stimulation.
With only clean oil production, there is no mechanism to
dilute the chlorine concentration with time in the near
wellbore area.
Most of our wells only have sigma data acquired after the
acid stimulation has been completed so a comparison of
measured sigma before and after stimulation was not possible.
We have been able to overcome this by comparing the
measured sigma to a synthetic baseline sigma generated with a
predictive transform. A few wells did have pre-stimulation
sigma logs, and a multivariate transform was developed to
predict sigma from commonly available open-hole wireline
data (GR, NPHI, DT, and Rt). The predictive algorithm works
well, with a standard deviation of error of 0.5 cu. Fig. 18
illustrates the good match between predicted and measured
sigma on one well logged with a PNC tool prior to
stimulation.

(4)

Acid Effect with PNC Sigma

A key factor in assessing whether low productivity is due
to low permeability or high skin factor is the effectiveness of
the acid stimulation. Most of the wells in Tengiz receive acid
stimulations, either reciprocating coil tubing acid washes or
bullhead diverted treatments. We have run several Pulsed
Neutron Capture (PNC) logs in the central area of the Tengiz
field as baseline data in preparation for a future hydrocarbon
miscible flood. Although not the primary reason for running
the PNC logs, we have found the sigma curve useful in
assessing placement of acid stimulations.
The PNC log records the thermal neutron capture cross
section of the formation by measuring the thermal neutrons
rate of decay. Chlorine is one of the strongest neutron
absorbers of the common earth elements so the response of the
PNC log is determined primarily by the chlorine present in the
formation water (usually as NaCl). The sigma reading also
has a pronounced effect due to chlorine in the acid reaction
products in the form of calcium chloride11. We have found a
significant increase in sigma due to the HCL acid stimulations
that are conducted on most wells in Tengiz (some wells are
prolific enough that stimulation is not needed). Nearly all
wells in Tengiz produce water free. The irreducible connate

Fig. 18 Comparison of Measured vs. Predicted Sigma on

Unstimulated Well.

SPE 102894

The post stimulation sigma is compared with synthetic

baseline sigma to determine acid effect. The difference
between the two curves is computed and used as an indicator
of acid stimulation. Fig. 19 shows the sigma increase
(magenta) rescaled to overlay on porosity (green).

11

This example, and other similar ones, have increased our

confidence in this diagnostic technique, which helps in
computing apparent permeability and the frequent decision
whether lower than anticipated flow rate is due to lack of
stimulation or permeability that is lower than predicted.
The very distinct acid signature still visible 4 years after
the treatment supports our belief that the acid effect is
permanent in a reservoir with low salinity connate water at
irreducible saturation.
Fig. 20 is an example of a well with uneven acid
penetration that was caused by crossflow from the higher
pressure lower zone into the lower pressure upper zone during
the coil tubing acid stimulation. The crossflow out of the
lower zone during the stimulation did not allow matrix
stimulation over that interval, and all the acid was swept up
into the upper interval.
In this example, for the calculation of APERM, rather than
assuming the skin of -2.8 from the pressure transient test was
evenly distributed, we assumed a skin of -4 over the upper
interval, and a skin of 0 over the lower interval. As shown in
Fig. 13, this will increase the apparent permeability of the
lower zone by a factor of about 2.

Fig. 19 Example of PNC acid effect showing poor stimulation

effectiveness, confirming prior interpretation of high skin with
analysis of time lapse PLTs

The post-stimulation sigma logs are a great help in

validating our assumptions about distribution of acid
penetration. Sometimes, these sigma comparisons show us the
reservoir has been unevenly stimulated, causing us to use a
different skin factor by zone in our computation of APERM.
Fig. 19 shows the utility of the sigma acid effect in assessing
stimulation effectiveness. Track 4 on the right shows the
results of a PLT conducted in 2001. A pressure transient test
at the time yielded a skin factor of +1 (blue curve). After a
scale problem was identified, a bullhead acid treatment with
no diverter was conducted the following year to restore
production. In 2006, the PLT illustrated in track 3 was
conducted. A noticeable reduction in flow contribution was
observed below 4287m, and a variable skin of up to +20 (blue
curve) had to be used to reconcile the 2006 results with the
flow observed in the 2001 PLT. A PNC log was then run to
assess the effectiveness of the prior acid treatment. The
measured sigma was compared to a synthetic baseline sigma,
and a difference curve was calculated. This difference curve
(sigma increase) is over posted on porosity in track 1. The
PNC acid effect in track 1 clearly shows that little or no acid
went below 4287m, as one might expect from an undiverted
bullhead acid treatment. This lack of visible acid effect
confirmed our prior interpretation of damage in this zone.
Although the reduced flow clearly indicates the region below
4287m has become damaged, the mechanism for this damage
is not certain. We speculate it may be due to wellbore scale
deposition.

Fig. 20 Example of uneven acid penetration of

reservoir due to crossflow during stimulation.
Constraining APERM with MLT Analysis
Another technique used for determining individual layer
permeability and skin is with Multi-layer transient pressure
testing12. These procedures are operationally more complex,
but further constrain the APERM solution. We have either

12

Zone 1

Tool Position Drawdown 1

Zone 2

Tool Position Drawdown 2

and final build up

Tool Position Drawdown 3

Zone 3

two or three layers which each form separate pressure

compartments. Multi-layer transient tests are integrated into a
multi-rate Production Logging procedure13.
1. Shut the well in two days before PLT operations
begin to allow stabilized crossflow to develop. Record up and
down passes to measure the stabilized shut-in flow profile
with crossflow.
2
Place the tool above the top layer and record the
drawdown with the low flow rate until IARF is observed on
the log-log plot (typically 4-6 hours). At the end of this
drawdown, record up and down flowing passes with the PLT
to measure the flow profile and determine the zonal
contributions.
3. Station the PLT above the middle layer. Increase the
production rate to the medium flow rate. Record the
drawdown until IARF is observed on the log-log plot (4-6
hours). At the end of the drawdown, record up and down
passes with the PLT to measure the flow profile and determine
zonal contributions.
4. Place the PLT above the bottom layer. Increase the
production rate to a high flow rate. Record the drawdown
until IARF is observed on the log-log plot (4-6 hours). Record
up and down passes with the PLT at the end of the transient.
5. Place the PLT between at the point of maximum
crossflow that was observed in step 1 (this is usually the same
location drawdown number 2. Close the well for shut in and
record the pressure buildup. At the end of the transient record
one shut in up and down pass with the PLT across the
reservoir. Remove PLT from well and rig down. It is
assumed that the transients produced by changing the surface
flow rates from low to medium will induce a sufficient
transient to be measured by the PLT pressure gauges and
flowmeter and that they are within their respective resolution
and tolerance. If the transients are not of sufficient magnitude,
then the drawdowns should be recorded with two different
flow rates: low and high. In this scenario the drawdowns
should be recorded with the first drawdown at the maximum
rate and the second one with the low rate and the third one
with the maximum rate. Due to operating time restrictions the
drawdown transient between zone 1 and 2 is omitted in
Tengiz. This coincides with the point of maximum crossflow
so the build-up transient is used to constrain layer 2 properties.

SPE 102894

Fig. 21 Depiction of PLT tool position during Multi-layer

transient testing.

Analysis of the transient data proceeds from the bottom

zone to the top zone. Analysis of the transient data from the
drawdown 3 is a single layer analysis of the lowest layer. The
reservoir properties of the lowest layer is thereby constrained,
and used in a multilayer analysis of the second drawdown,
which is a combination of layers 2 and 3. The properties of
layer 2 can then be determined. The properties of layers 2 & 3
are input as constraints into the analysis of the first drawdown,
which is the combined response of layers 1, 2 & 3, to solve for
the properties of the top layer. Layer pressures that are
required for the multilayer analysis are taken from the SIP
analysis that was previously described. When the second
drawdown is not recorded we use the pressure and crossflow
recorded during the final build-up to estimate layer 2
properties. In this fashion, the permeability and skin of each
layer may be determined, and used to constrain the APERM
analysis.
Fig. 22 is an example of a well where the results of a
multi-layer transient test match well with the APERM
interpretation. Track 1 shows the porosity trace (green
shading) with the acid effect from calculated sigma increase
over-posted in magenta. The porosity on this well was also
computed from a Soviet era uncalibrated neutron log,
increasing uncertainty on the resulting predicted permeability.
Track 2 is APERM (red), transform perm (green), and
apparent perm from PLT in black.
Track 3 is the
multiplicative factor used to adjust the transform perm to the
PLT perm. Note the adjustment factors are all highly positive.
This is likely because the open hole porosity is reading too
low, resulting in a low predicted perm. Track 4 is the PLT
results. Also posted in track 4 is a log of the calculated skin
factor from the multi-layer transient test, which was a value of
-3 in the upper zone and +1 in the lower zone. This is
consistent with the acid effect observed in the upper zone but
not in the lower.

SPE 102894

13

to the APERM geostatistical population process. Both SGS

with collocated-cokriging and SGS with cloud transform
methods have been used to populate APERM. An example of
the cloud of APERM data for a shallow reservoir interval at
Tengiz is shown in Figure 22.

Fig. 23. Cross plot of porosity and APERM for a shallow interval
in the Tengiz platform.

Fig. 22 Example of good match between APERM, Acid Effect,

and Multi-Layer Pressure Transient Interpretations.

Construction of the Static Model using APERM

The successful use of reservoir simulation is critical to the
proper management of Tengiz reservoir.
The efforts
associated with delineating, developing, and producing Tengiz
reservoir are substantial. Optimizing these efforts with a
robust reservoir simulation can have large positive effects on
improved oil recovery, and optimizing expenditures. For the
reservoir simulation model to be successful, it must be
populated with accurate geologic model data, rock property
descriptions, fluid properties, and well information.
Additionally, the detail captured in the simulation model must
match the scope of the studies for which the modeling effort
will be applied.
Several generations of reservoir models have been
constructed for Tengiz Field. These models have accurately
predicted the primary depletion of the field, and have
successfully been used to guide infill drilling. Tengiz
reservoir models have conventionally been built using porosity
and permeability transform, which is based on core plugs.
However, to model secondary recovery processes (such as
sour gas injection) even more accuracy is required.
The acquisition of modern wireline logs and PLT logs in
the Tengiz platform has provided a rich source of data to
properly calibrate the reservoir model. APERM traces are
input into the reservoir simulation model to improve the
reliability of the permeability field, and improve the accuracy
of secondary recovery estimates.
Geostatistical methodogies have been used to populate
APERM data into Tengiz reservoir models. In the area of
investigation at Tengiz, most of the wells have reasonable
quality suite of wireline logs. Construction of the model
follows several discrete steps in a sequential workflow.
Wireline porosity was first distributed in a 3D earth model
using Sequential Gaussian Simulation (SGS). Porosity is a
fair predictor of permeability in Tengiz, and provides a guide

These well established geostatistical techniques honor the

spatial variability of the APERM data, while following the
trend of the background porosity as "soft" data. These
processes provide a way to condition the model at the well
location with the input APERM traces while honoring the
spatial variability of the APERM data in each stratigrahic
interval of the reservoir. They provide a set of equiprobable
images of the permeability between the measured APERM
traces. The resulting permeability images are calibrated with
all the known geologic and engineering data, and provide a
reasonable (albeit non-unique) realization for predicting
reservoir performance.
Restoration of Natural Heterogeneity:
As illustrated in figure 9, carbonate reservoirs are naturally
heterogeneous, with a permeability variiance often exceeding
1 order of magnitude for a given porosity value. The very
nature of regression based transforms reduces this
heterogeneity to a single value for a given facies/porosity pair,
leading to models that are inherently too homogeneous. This
is compounded by the tendency to either under-recover core in
fractured high permeability intervals, or undersample in nonpermeable intervals. The excess homogeneity results in
predicted flood responses that are too piston-like, and actual
breakthrough times earlier than predicted. As illustrated in
figure 23 above, the APERM method restores that
heterogeneity. This should lead to more realistic predictions,
and improved reservoir management planning.
Results of Incorporating APERM into the Static Model
Introduction of the APERM data into the reservoir model
has substantially improved the reservoir model. Compared to
model built with a traditional matrix permeability transform,
the new APERM based model has measurably higher
permeability. Fig. 24 compares two reservoir models, one
constructed with a conventional transform, and another with
APERM traces. The hotter colors in the right figure indicate
the presence of higher K values after the introduction of
APERM.

14

SPE 102894

Conventional Matrix Perm

APERM

This is clearly shown by the scatter in Fig. 26 A. Note that

before the use of APERM, there is considerably more scatter,
and there is tendency for many wells to have inadequate KH in
the model compared to well test KH. However, after the
permeability field was constructed using the APERM traces
there is a much better match between the well test Kh and
model derived Kh, as shown in Figure 25 B.

A
.001

permeability (md)
.01
.1
1
10

Kh Model
without
APERM

100

Fig. 24 Conventional Reservoir model for Tengiz (left) and

model constructed using APERM (right)

The introduction of APERM data has also helped calibrate

the permeability levels in different stratigrahic intervals of the
reservoir. Some specific stratigrahic intervals are observed to
have more statistical difference from the conventional matrix
permeability transform. A cross section through the central
platform area of the Tengiz Field (Fig. 25) shows the tendency
of zone 1 to have a positive permeability correction (blue
color) relative to the conventional matrix transform.
The introduction of APERM data has also substantially
increased the heterogeneity of the reservoir model relative to a
conventional permeability transform in the platform. DykstraParsons coefficient calculations in some areas of the Tengiz
platform have mean values as low as 0.50 for a conventional
permeability transform. This level of heterogeneity implies a
fairly homogeneous permeability distribution. However, after
the introduction of APERM data, the Dykstra-Parson
coefficients are much larger, with some calculations having a
mean value as high as 0.95, suggesting a very significant level
of heterogeneity.

B
Kh Model
With
APERM

Kh well
Fig 26 Crossplot of KH from welltest in the Tengiz platform
versus KH from a full field reservoir model. A shows the
comparison from a model without APERM calibration, and B
shows the same wells after APERM calibration

Validation of APERM Model Improvement with Pulse

Tests
The Tengiz injection wells have permanent downhole
pressure gauges and have been shut-in while preparing for gas
injection (Fig. 27). Surrounding producers are pulsed and the
pressure response is measured at the injector. The results have
shown both higher levels of connectivity than previously
modeled and different levels of connectivity between various
wells pairs (colored arrows).

Fig. 25 Difference between APERM and

permeability transform for one area of Tengiz field.

conventional

Match to well test KH

One benefit from the integration of APERM data into the
reservoir models is quite clear when comparing cumulative
permeability (KH) from the reservoir model to well test (PTT)
derived cumulative permeability. After the introduction of
APERM, the Kh of the reservoir model in the platform area
now has a much better match to the Kh observed in well tests.
Accumulated Kh from the reservoir model has a poor match to
well-tests, when the permeability is not properly calibrated.

MOA

*
SGI

*
220
5442

Diffusivity (kh/phih)
*

Tag Fill above

Serp

< 50
50 to 100

4431, 1351m

> 100
Fig. 27 Sour Gas Injection pilot area with results of pulse testing

SPE 102894

showing relative connectivity between wells. Injectors are in the

center of the patterns (red arrows).

The pressure response to the pulse test flow sequences

were simulated using the flow-calibrated APERM based
reservoir model and compared to the simulated results from
the prior model, where the permeability was based solely on
transforms from wireline logs.
A significant improvement in the predictive accuracy was
observed with the new model after the introduction of
APERM. This important history match calibration confirms
that the simulation model now has approximately the correct
amount of connected pore space and permeability. Fig. 28
shows a comparison of actual pulse test results (black lines)
and simulated pressure response with both the APERM based
earth model (red lines) and the previous transform perm based
model (blue). The bottom panel shows the flow sequence.
The top panel is the pressure response, and the middle panel is
the derivative of pressure. This analysis shows that the model
after the introduction of APERM has a much better match to
the pulse tests than the conventional reservoir model. In
general, the model with the introduction of APERM provided
a significantly improved match of the inter-well connectivity
as indicated by the pulse tests.

15

wish to thank Peter Hillock of Exxon Mobil for his thoughtful

collaboration and insight. Dennis Fischer with TCO provided
critical geologic support and assisted constructed the earth
model. Dennis has been an enthusiastic supporter of the
APERM process. Rob Kimmel provided a synthesis of Tengiz
geological setting. Steve Johnson with Chevron conducted the
reservoir simulation and pulse test matching for gas injection
monitoring and provided helpful feedback. Travis Billiter
with TCO demonstrated the ability of the APERM data to
reach a rapid history match for the Tengiz full field model.
We also wish to thank our colleagues Ed Neubauer and John
Clarke for modeling and analysis of the multilayer transient
data and many fruitful discussions that led to improved
estimates of layer properties.
References
1.

2.

3.
4.
5.

6.
7.
8.
Fig. 28 Simulated vs. Actual Pulse Test Responses with APERM
Based Model and Prior Transform Perm Based Model.

Ongoing Tengiz Modeling Efforts

A full field reservoir modeling effort is underway at
Tengiz, and the APERM data are providing an important data
source to calibrate the permeability. For this effort, and for all
future modeling efforts, it is anticipated that the APERM
method will continue to be used and the methodology further
refined.
An aggressive program of acid stimulation and PLT is
underway which will contribute to the current database of
APERM wells. This database of high quality permeability
measurements will further improve the confidence of the
reservoir model, and be used to help optimize reservoir
management activities at Tengiz.
Acknowledgements
We wish to thank Tengizchevroil (TCO) and its JointVenture Partners for permission to publish this paper. We also

9.

10.
11.

12.
13.

Skalinski, M., and Sullivan, M.: Application of Improved

Method for Permeability Estimation in Complex Lithology
Reservoirs. Paper C, presented at the 2001 SPWLA
Annual Logging Symposium, 17-20 June 2001, Houston,
Texas.
Lucia,
F.
J.,
1995b,
Rock-fabric/Petrophysical
Classification of carbonate pore space for reservoir
characterization; American Association of Petroleum
Geologists Bulletin, V. 79, P. 1275-1300
Al-Henshiri, M., Arisaka, K., Al-Hassani, H., Al-Yasi, T.,
and Fujita, N.: SPE 93475 Integration of Dynamic &
Geostatic Data Improves Reservoir Characterization.
Ahmed, U., Crary, S., and Coates, G.; SPE 19604
Permeability Estimation: The Various Sources and Their
Interrelationships.
Bahar, A., Ates, H., Al-Deeb, M. Salem, S., Badaam, H.,
Linthorst, S., and Kelkar, M.: SPE 84876 An Innovative
Approach To Integrate Fracture, Well-Test, and Production
Data Into Reservoir Models.
Sibley, M., Bent, J., and Davis, D.: Reservoir Modeling
and Simulation of a Middle Eastern Carbonate Reservoir.
SPE Reservoir Engineering, May 1997.
Mezghani, M., Van Lingen, P., Cosentino, L., and Sengul,
M.: SPE 65122 Conditioning Geostatistical Models to
Flow Meter Logs
Chambers, K.T., Hallager, W.S., Kabir, C.S., and Garber,
R.A.; SPE 72598 Characterization of a Carbonate
Reservoir with Pressure Transient Tests and Production
Logs: Tengiz Field, Kazakhstan.
Stewart, G., Wittmam, M.J., and Lefevre, D.: Well
Performance Analysis: A Synergetic Approach to Dynamic
Reservoir Description, paper SPE 10209 presented at the
1981 SPE Annual Technical Conference and Exhibition,
San Antonio, Oct. 5-7.
Harrison, C. Personal correspondence: Technique for
production log analysis using simplified assumptions.
Al-Saif, A.S., Cochrane, J.E., Edmundson, H.N., and
Youngblood, W.E.; SPE 5443: Analysis of Pulsed
Neutron Decay Time Logs in Acidized Carbonate
Formations
Ehlig-Economides, C.A., and Joseph, J.A.: "A New Test
for Determination of Individual Layer Properties in a
Multilayered Reservoir," SPEFE, Sept., 1987
R. R. Jackson, and R. Banerjee, Advances in Multilayer
Reservoir Testing and Analysis using Numerical Well
Testing and Reservoir Simulation. SPE Annual Technical
Conference and Exhibition, Dallas, Texas, 14 October
2000.

16

SPE 102894

14. Kuchuk, F.J., Karakas, M., and Ayestaran, L.: Well

Testing and Analysis Techniques for Layered Reservoirs,
SPEFE, Aug. 1986

ABOUT THE AUTHORS

Michael J. Sullivan is a Petroleum Engineer working on
the Tengiz reservoir surveillance in Kazakhstan. E-mail:
[email protected]. Prior to moving to Tengiz, Mike worked
as a Petrophysicist at ChevronTexaco Exploration and
Production Technology Company in Houston, Texas. He had
previous assignments as Open Hole Formation Evaluation
Specialist and Cased Hole Formation Evaluation Specialist for
Chevron in Cabinda Angola, and Cased Hole Formation
Evaluation Specialist for Chevron Canada Resources in
Calgary. He has also held a variety of positions in Petroleum
and Production Engineering starting in 1979. His interests
include integration of reservoir surveillance into reservoir
characterization studies. Mike has a BS in Petroleum
Engineering from Montana College of Mineral Science and
Technology. He is a member of the SPWLA, SPE, Petroleum
Society of CIM, and CWLS
David L. Belanger is a Petroleum Engineer working on
Tengiz reservoir surveillance in Kazakhstan.
E-mail:
[email protected]. His main interests include reservoir
surveillance and the integration of dynamic data in reservoir
characterization.
David was previously a Senior Staff
Formation Evaluation Specialist at ChevronTexaco
Exploration and Production Technology Company in Houston,
Texas, where he provided formation evaluation support to
ChevronTexaco operations worldwide. David is a member of
the SPWLA and SPE and he is a former Review Chairman for
the SPE Formation Evaluation Journal. He holds a BS degree
in Mechanical Engineering from the University of California
at Berkeley
Mark Skalinski is currently a Formation Evaluation
Supervisor with Tengizchevroil in Atyrau, Kazakhstan.
Email: [email protected]. He has M.Sc. (1971) and
Ph.D. (1979) degree in Geophysics, from Mining University in
Cracow. His previous assignments include: Chevron Canada
Resources in Calgary, CABGOC in Cabinda, Angola, Husky
Oil in Calgary, ONAREP in Morocco (as a Petrophysicist) and
University of Mining & Metallurgy in Cracow Poland (as
assistant professor). Marks interest includes: petrophysical
multimineral modeling, application of statistical methods for
facies, permeability and rock type prediction and integrated
petrophysical field studies. Mark is a member of SPE and
SPWLA
Steven D. Jenkins is an Earth Scientist working on the
exploration and reservoir geology of Tengiz Field, Atyrau, and
Republic of Kazakhstan. Prior to moving to Atyrau, Steve
worked as a geologist with Chevron Nigeria Limited in Lagos
Nigeria, with Caltex Pacific Indonesia in Sumatra, Indonesia
and with Chevron USA in Denver Colorado. Steves interest
includes: reservoir characterization using seismic methods, the
application of statistical methods for facies, permeability and
rock type prediction and integrated petrophysical field studies.
He has held a variety of positions in Exploration and
Development Geology, and in Reservoir Geophysics starting
in 1981. Steve has a BA in Geology from the University of

Tennessee and MA in Geophysics from Indiana University.

He is a member of the SEG and the AAPG.
Peter C. Dunn is currently the Sour Gas Injection Team
Leader for Tengizchevroil located in Atyrau, Kazakhstan. Email: [email protected]. Prior to moving to
Kazakhstan, Peter worked for ChevronTexaco Exploration and
Production Technology Company in San Ramon assessing and
evaluating acquisition opportunities for Chevrons worldwide
operations and in Western Canada as a petroleum and
reservoir engineer for Chevron Canada Resouces. His interests
include field operations support, reservoir surveillance and the
art of reservoir simulation. He holds a BASc degree in
Chemical Engineering from the University of Waterloo.