Sei sulla pagina 1di 26

See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/273379862

Uncertainty Analysis in Well Log and


Petrophysical Interpretations
Chapter in AAPG Memoir December 2011
DOI: 10.1306/13301405M963478

CITATIONS

READS

14

571

4 authors, including:
William Moore

Zee Ma

Schlumberger Limited

Schlumberger Limited

15 PUBLICATIONS 15 CITATIONS

120 PUBLICATIONS 258 CITATIONS

SEE PROFILE

SEE PROFILE

Tom Bratton
Tom Bratton LLC
50 PUBLICATIONS 261 CITATIONS
SEE PROFILE

All in-text references underlined in blue are linked to publications on ResearchGate,


letting you access and read them immediately.

Available from: Zee Ma


Retrieved on: 10 October 2016

Uncertainty Analysis in Well Log and Petrophysical Interpretations


William R. Moore, Y. Zee Ma, Jim Urdea and Tom Bratton
Schlumberger, Greenwood Village, CO 80111

ABSTRACT
Petrophysical analysis plays an important role in a reservoir study as it provides primary
input data for characterization of subsurface formations and evaluation of resources.
Common petrophysical input data to a reservoir study include porosity, water saturation,
permeability and mineral or rock volumes. These reservoir variables are typically not
directly measured by well logging tools; instead, they are usually derived through
multiple processes, including acquisition, processing, interpretation and calibration. As
each of these steps involves uncertainty, the resultant petrophysical data will have
uncertainty and limitations. Common uncertainties in interpreting the most important
petrophysical variables from well logs are discussed in this chapter. Uncertainties from
well log and petrophysical interpretations should be analyzed, quantified and explicitly
communicated to the integration team in order to be accounted for in reservoir
characterization and modeling.

INTRODUCTION
Petrophysical analysis is critical in a reservoir study; it provides a primary source
of input data for reservoir characterization and modeling. To characterize subsurface
resources and gain knowledge of the reservoir, it is necessary to acquire data. Wireline
logs provide continuous recordings over the rock formations and give geoscientists
insights on the rock properties. These measurements supply critical information for both
single-well formation evaluations and field-wide subsurface resource evaluations.
Since the first electrical log was recorded in 1927 in a well in a small oil field in
France (Schlumberger 1999), the use of wireline logging has grown dramatically. Today,
a dozen or more types of logging data and measurements are typically acquired for
formation evaluations. These often include gamma ray (GR), spontaneous potential (SP),
density, neutron, sonic, magnetic resonance, and various resistivity logs. Well-log data is
often the main source of information for subsurface evaluation, especially in areas where
core and/or seismic data is sparse or unavailable. It is, however, important to understand
that logging data has uncertainties and limitations that, if not accounted for, can lead to
an incorrect picture of the area around the wellbore and ultimately of the reservoir.
First, all measurements have uncertainties. Several organizations, including the
International Standardization Organization (ISO), the Bureau International des Poids et
Mesures (BIPM) and the National Institute of Standardization and Technology (NIST) all
acknowledge uncertainties in measurements. In fact, the general attitude has changed
from the traditional acknowledgement of random and systematic errors in measurements
to the approach of the best estimate with its associated uncertainties, under the premise
that no measurement is exact and it is not possible to state how well the essentially

unique true value of the measurement is known, but only how well it is believed to be
known (BIPM, 2009, p. 2-3).
Second, most logging tools do not provide direct measurements of the
petrophysical properties that are used in reservoir modeling and hydrocarbon evaluation.
For example, the two most important petrophysical properties in characterizing the inplace hydrocarbon resource are porosity and fluid saturation. They are not directly
measured but rather are derived from the acquired logs using transforms based on
theoretical and empirical relationships.
Most petrophysical properties used in integrated studies are obtained through a
multiple-step process, including data acquisition (measuring), processing, calibration and
interpretation (Dewan, 1983; Theys, 1997, 1999). Each of these processes has
uncertainties that affect the results of a petrophysical analysis. Most well logs include
basic corrections for the known systematic errors through tool calibrations and
environmental corrections. Due to the statistical nature of most measurements and the
complexity of the borehole environment some uncertainty will remain, especially as the
applied corrections cannot completely eliminate all the errors or cannot be quantified as
doing so.
Acquisition problems in wireline logging, such as total or partial tool failure, bad
borehole conditions, and poor choices of logging suites, either because of the tool
availability or acquisition costs, can affect logging runs and cause uncertainties in the
logging data. In a broad sense, different vintages of well logs can be considered as
acquisition uncertainties because the different vintages imply different tools that
generally have vastly different resolutions and sensitivities, such as sensitivity to
environmental factors. This is especially important in studying a reservoir with older log

suites. Logging suites acquired before the 1960s were much more limited than modern
suites. The differences often cause a dilemma when integrating modern and old logs. On
one hand, old logs may provide valuable information since they represent the reservoir
properties at or close to the initial conditions. On the other hand, quality and sensitivities
of old logging tools may be questionable and require extra processing to extract
appropriate data. Documentation needed to make a correct analysis may be missing or
expertise in analyzing older logs may be lacking. Too often older logs are either
completely ignored or used without thorough analyses. While the analyst should never
ignore the value of information (see Chapter 1 in this volume), careless uses of older
logs can lead to large errors in evaluating formation parameters. Even after careful
analyses, integration of data acquired by diverse tools can still convey significant
uncertainties. The analyst has to put thought and much effort into merging the old log
analysis with the modern log analysis to produce a consistent reservoir picture.
Detailed discussions on acquisition uncertainties are beyond the scope of this
chapter. We focus on discussions of processing, calibration and interpretation related
uncertainties in petrophysical evaluations of formations and hydrocarbon resources.

PETROPHYSICAL INTERPRETATION UNCERTAINTY ANALYSES


Logs acquired using different tool types and different generations of the same tool
type must be adjusted to some standard to produce consistent outputs. The porosity and
resistivity tools have borehole and environmental corrections that vary according to
vendor, tool type and tool generation. Collecting the information for these corrections can
be time consuming. The most recent corrections are often unavailable to the analyst, and
even when they are, there may be no easy way to determine whether they have been

applied correctly. These corrections, however, are critical because serious errors can be
introduced if they are not performed or are inadequately performed.
The logging data must be calibrated to and integrated with other hard data, such
as that from cores, pressures, and flow tests, when these are available and appropriate.
However, the wells with the best logging tool suite coverage may not be the same wells
for which other hard data is available. If detailed data exists for only one or two wells, the
analyst has to make decisions on whether the data is representative of the entire reservoir
interval or must be adjusted over the reservoir areally and vertically for calibration. In
many studies, it may come down to the lowest common denominator, i.e., what you can
do with the most prevalent logging suite. In other words, petrophysical properties can
vary greatly within the reservoir and the parameters chosen in petrophysical analysis are
often a trade-off. This leads to uncertainties in the results of a petrophysical analysis.

Uncertainty Analysis in Interpreting and Deriving Porosity Logs


Porosity can be derived by a number of methods depending on the availability of
raw logs. For example, it can be derived individually from a sonic, neutron or density log,
or it can be derived by combining two or all three of them using various averaging or
crossplot methods. To do this, the analyst has to know the matrix and fluid parameters
used in each method. In particular, the fluid parameters must be considered because they
can change over the life of a reservoir as a result of pressure changes, influx of aquifer
fluids, and water-, steam- or gas-flooding. The choice made for each well may depend on
the data availability as well as on the analyst preference. Typically, single-lithology
intervals can be adequately characterized by a minimum number of measurements, but as
the reservoir becomes more complex, more input data is required for the analysis. Other

sources of information, including cores, petrography and outcrop studies, can be used to
help guide the parameter selections.
The uncertainties in the acquisition and processing of logs give rise to more
challenges in the interpretation. A large number of possible choices exist with regards to
what corrections to make, which porosity model to use, how to select the parameters, and
how to handle parameters that may vary over time. Therefore, interpretation can be very
analyst dependent. For example, a number of methodological choices exist for calculating
rock and mineral volumes, and many complexities must be considered before a usable
data set can be produced.
Figure 1 is a crossplot of the core porosity and travel time from a sonic logging
tool, which gave the only measured porosity log. The data was acquired from a
predominantly clastic reservoir in a field in the early development. There are points both
above and below the Wylie time-average porosity line. The analyst must determine if the
error is with the data (inaccurately analyzed core or incorrectly scaled sonic data),
mineralogical effects on the data (clay or cement), light hydrocarbon effects on the data,
or is due to highly laminated intervals. All these possible explanations must be
considered in a petrophysical analysis. A simple linear regression from this crossplot to
generate the porosity values will cause errors and significant uncertainties.
One of the most common uncertainties or inaccuracies in well-log porosity is
caused by the borehole condition. Figure 2 shows an example of the impact of a rugose
borehole in petrophysical analysis. The lithology in the reservoir consists of sand and
shale. The borehole was washed out and extremely rugose. Using data from individual
porosity tools would give overly optimistic answers. Figures 2A through 2C are Pickett
plots of the resistivity log and the porosity logs from various individual tools. The
porosities in these plots are all wildly optimistic, which is mainly due to the effect of the
6

enlarged boreholes. Few porosity values in this reservoir should exceed 15%, certainly
not as high as 40% or higher seen in the plots. In addition, because of the impact of the
bad holes to all the three porosity logs, even the average of these logs could not produce
reasonable porosity values (Figure 2D). It was necessary to perform the borehole- and
shale corrections. Figure 2E is a Pickett plot of the resistivity log and the borehole- and
shale-corrected porosity. While the resultant porosity is perhaps still somewhat affected
by the borehole conditions, and thus has uncertainty, it is much more reasonable.

Figure 1. Crossplot of the sonic travel time and core porosity (in fraction) with the linear
regression line (black) and Wylie time average line (red). The correlation coefficient is
0.75.

(A)

(B)

(C)

(D)

(E)
Figure 2. Crossplots of induction resistivity (ILD) and porosity (in percentage) logs
showing bad hole effects (A) sonic, (B) density, (C) neutron, (D) average of the neutron
8

and density porosities, and (E) borehole and shale corrected effective porosity. Color
represents caliper, blue low, red high.

Clay Volume and Water Saturation (Sw) Uncertainty Analyses


Mineral or rock volumes include clay, sand and various carbonate rock volumes.
Deriving all of these rock volumes can be problematic, though only shale volume is
discussed here, in conjunction with water saturation uncertainty analysis. Figure 3 is a
plot of linear versus non-linear relationships of gamma ray (GR) and shale volume
(VShale). The purpose of the non-linear (Clavier and Stiebner) equations is to correct the
observed deviations from the linear model. The analyst must decide if non-linear
condition between GR and VShale exists in the reservoir and which model best fits the
formation data. Core and other sample analyses can guide the choice of the equations and
verify the calculations. Obviously, a wrong choice can cause erroneous results.

Vclay vs. GR reading for Linear, Clavier, Steibner equations


1.200
1.000

VShale

0.800
Linear
0.600

Clavier
Steibner

0.400
0.200
0.000
10

20

30 40

50

60

70

80

90 100 110 120

GR

Figure 3. Three different relationships between VShale and GR (linear versus non-linear
clay volume calculation)

Measurements from resistivity tools must be corrected for environmental


conditions. Even though the resistivity logs are usually less affected by changing
reservoir conditions, invasion corrections and bed-boundary and thickness corrections are
sometimes necessary to compensate for significantly changing reservoir conditions.
Once the resistivity and porosity are calibrated or calculated from other logs,
estimates of water saturations can be generated. The Archie equation is the standard for
calculating water saturation in clean, porous rocks. There are several variations to the
basic Archie water saturation equation, including the Indonesian, Simandoux, WaxmanSmits, and Dual Water formulas. All of these models were developed to fit a specific
dataset and have limitations (see, e.g., Fertl and Hammac, 1971). The analyst needs to
determine which of these models best fits the available data and input logs based on the
reservoir characteristics. More specific and detailed discussions are found in SPWLA
(1982) and Crain (1986).
Additionally, uncertainties exist in choosing the model parameters, including
cementation exponent, m, saturation exponent, n and Archie constant, a. First,
Archies equation assumes all porosity is interconnected and that there is a uniform grain
packing. Very few rocks in the real world meet these conditions. All shaly sand equations
(e.g., Waxman-Smits, Dual Water, Simandoux, Indonesian, and Lamimated) try to
account for the discrepancies between the calculated results and the empirical
correlations in Archies equation, but they cannot be perfectly accurate.
Second, laboratory work, although it can guide the analyst in determining the
parameters, has its own range of uncertainty. For example, fractures, vugs, uncommon
minerals or rocks, and other reservoir characteristics can all affect the data analysis.

10

Moreover, these parameters may vary widely over the extent of the reservoir. Core
analysis, Pickett and Magnolia plots are commonly used to derive the average values.
Variables m and n derived from the log data can sometimes be used. For some of the
saturation models, other parameters (shale resistivity, cation exchange capacity, m*) also
have to be determined. All these involve uncertainties that must be mitigated, mainly
through the use of non-log data, such as special core analysis and petrographic studies.
In rocks containing clay or other minerals (e.g., feldspars), the resistivity logs and
any other logs used to derive the porosity are affected and the data must be corrected.
Two types of corrections can generally be distinguished. The first type of corrections
involves subtraction of the effect of the clay or other minerals from the input data and use
of the corrected values in the Archie equation. The questions of what to correct, how
much correction to apply, and which method to use all involve a certain degree of
uncertainty. The second type of corrections involves adding components to the basic
Archie equation to compensate the effect of the clay or other minerals. The analyst must
decide whether to use a clay model which corrects for individual types of clay or a shale
model, which lumps all the non-matrix components together.
Figure 4 shows a log interval with three shaly sand bodies and water saturation
results derived using the equations mentioned above. Similar processing parameters were
used for each equation. The Archie saturation calculation is the most pessimistic; the
various shaly sand equations (Waxman-Smits, Indonesian, Simandoux, Juhasz, and Dual
Water) show greater hydrocarbon saturations. The results for each equation can be
optimized for the interval by varying the parameters. However, the choice of the equation
and related parameters is not obvious and may be dictated by the convention used in the
area. It is also not always possible to optimize all the parameters to obtain consistent

11

results. More discussions on sensitivity of these equations and parameters can be found in
Fertl and Hammac (1971), Worthington (1985), and SPWLA (1982).
Another method is the facies-driven capillary-pressure model, in which the
relationship between capillary-pressure water saturation and height above the free water
level is established for each facies. Uncertainties exist in such a model as well, stemming
from the possible inaccuracies in the core data, non-uniqueness in identifying the facies
(see Chapter 6 in this volume), presence and determination of the free water level and
transition zone, and incomplete imbibition. Similarly, a hybrid method that uses both the
facies-driven capillary-pressure and Sw-porosity relationship (Lucia, 1995) also has
uncertainties in determining the Sw logs for the same reasons.

12

Figure 4. Results from various water saturation equations for a shaly sand interval

13

Uncertainty Analysis in Deriving Permeability


Permeability is a critical variable for reservoir simulations and performance
forecasts. Several methods can be used to generate a 3D permeability model. Quite
accurate permeability values can be derived in-situ from nuclear magnetic resonance
logs. The 3D permeability model can be generated using the geostatistical collocated cosimulation in relation to the porosity (Ma et al., 2008) or cloud transform or p-field
method (Bashore et al., 1994). In many studies where modern logs are not available, the
permeability is commonly derived using the empirical relationship between the core
porosity and permeability. The validity of the core measurements and the core to well-log
adjustments, including scale effect due to different supports (Delfiner, 2007; Jensen and
Lake, 1988), are some of the main areas that introduce uncertainties in the permeability.
In general, such a derived permeability may be quite accurate in clean isotropic rocks, but
is usually less so in other rock formations. In some cases, uncertainties in the
permeability prediction can be reduced by incorporating facies data. When available,
permeability values from long interval tests or point measurements can also be used to
validate or adjust the equation in the core porosity and permeability calibration.
Figure 5 shows some of the problems in generating a permeability model using
the core porosity and permeability calibration. As commonly observed, a cloud
relationship (i.e., with a considerable spread in the plot) between the core porosity and
permeability exists. The spread in the relationship may be partly due to their true
correlation, and partly due to measurement inaccuracy. Several workflows for generating
the permeability model can be chosen from:
(1) A linear regression, as shown in Figure 5B, is directly used in generating a 3D
permeability model based on the 3D porosity model that is built first;

14

(2) The cloud relation is modeled to generate a 3D permeability model using


either collocated co-simulation in relation to the 3D porosity model (Ma et al., 2008) or
cloud transform or p-field method (Bashore et al., 1994);
(3) The linear regression is used to generate permeability values at the wells that
are subsequently used to condition the 3D permeability model while honoring the cloud
relationship between the porosity and permeability, by using collocated co-simulation or
p-field method.
The permeability regression model can also be generated for separate facies
and/or using a different regression method, such as the major axis regression (Figure 5C).
All these choices of the parameters and workflows are a direct result of the complexity in
permeability and limited data, and the resultant permeability model has uncertainties.
One of the issues in the core to well log calibration is the so called support effect
or scale effect (Delfiner, 2007). Because a linear regression is a smoothing transform, the
frequency distribution of the transformed property can be very different from the original
distribution (Ma, 2010). In the above case, if the linear regression in Figure 5B is used,
the generated permeability in the well log has a histogram that is quite different from the
histogram of the core permeability (Figure 5D). In general, the histogram of a sogenerated permeability is skewed toward the low-permeability values and reduces the
overall permeability. Nonetheless, in some cases, a few extreme permeability values are
generated by such a transform so that the arithmetic average of the generated
permeability is greater. Take the example shown in Figure 5D, the arithmetic average of
the transformed permeability is 59 mD while that of the core permeability is only 22 mD.
But, notice that the overall transformed permeability has much more low permeability
values.

15

(A)

(B)

(C)

(D)

Figure 5. Illustration of uncertainties in calibrating core porosity (in fraction) and


permeability (in millidarcy or mD). (A) Crossplot of core permeability (logarithm) and
porosity. (B) Same as (A), but overlaid by the linear regression and facies (red designates
dolomite and blue, limestone). (C) Same as (B), but overlaid by two linear regressions
based on the two separate facies using the major axis method. (D) Comparison of the
histograms for the core permeability (green) and the predicted permeability (orange)
using the linear regression in (B).

16

A SIMPLE EXAMPLE OF INTEGRATED GEOLOGIC AND PETROPHYSICAL


UNCERTAINTY ANALYSIS
Reservoir characterization and resource evaluation, almost by definition, require a
multidisciplinary

integration

using

geological,

petrophysical,

geophysical

and

engineering analyses. An integrated analysis can often help discern inconsistencies in the
data, narrow the uncertainty range and make a better decision. The process does not
always have to be long. An example in which a simple integration with uncertainty
analysis would have resulted in the right business decision demonstrates this.
Figure 6A shows a Pickett plot of well-log porosity versus the resistivity, RT,
from 21 wells in a tight gas field. Each point represents separate sand body averages.
Nineteen of the wells were initially drilled and completed over a 2-year period using the
same drilling, logging and completion techniques. Two additional wells, plotted as red
dots, were drilled. Only basic wireline logs, including induction, density, neutron, GR
and SP, were acquired because of the marginal economics. The geologic interpretation of
the depositional environment was also available for evaluating these wells. The objective
of the analysis was to determine if the two new wells should be completed. The initial
analysis (Figure 6A) indicates that the two additional wells are apparently productive and
worth completing, even though their boreholes likely intersect both good and mediocre
sands.
Figure 6B shows a plot with the same data, but with an overlay of the production
results from the 19 completed wells. Blue represents non-productive or non-commercial;
green, marginally productive; black, commercially productive; and red, the two
uncompleted wells. The scatter is large with commercial and non-commercial wells
overlapping. The two uncompleted wells are apparently within the range of commercial
17

production, albeit with no definitive trend. Using only this information, the
recommendation would still be to complete the wells.
However, production results by facies indicate that commercial production
requires a significant presence of the fluvial or bay-estuarine sands. Marine and nonchannel fluvial sands are generally non-commercial or poor producers. Figure 6C shows
the same plot, but with overlay of the depositional facies: fluvial channel, bay-estuarine,
marine, and non-channel facies. Figure 6D shows a plot of the depositional facies for the
uncompleted wells only. There are mostly non-channel fluvial facies with some marine
sands and no fluvial or bay-estuarine sands. This suggests that the two wells would be
non-commercial poor producers. Because this analysis was not actually incorporated in
the decision, the two wells were completed and produced non-commercial quantities of
gas. The integrated facies and petrophysical analysis, had it been incorporated, would
have helped better understand the reservoir quality of the two new wells and a better
economic decision would have been made.

(A)

(B)

18

(C)

(D)

Figure 6. Pickett plots of the resistivity, RT, and porosity. (A) Data from the 19 tested
wells (black) and the two non-tested wells (red). (B) Good production (black), mediocre
production (green) and poor production (blue). Data from the two uncompleted wells
(red). (C) Same data including overlay of the depositional environments:
fluvial channel - orange, bay-estuarine - green, marine - blue, and overbank - black. (D)
Subset of (C) showing that the two uncompleted wells are in poor facies.

DISCUSSION: UNCERTAINTY ANALYSIS OF WELL-LOG DATA FOR


RESERVOIR MODELING
Data from well-log and petrophysical analysis are usually the main input for
reservoir modeling. Typically the data is given to the reservoir modeler without
communicating the uncertainty in acquisition, processing, and interpretation. Reservoir
modelers often assume the well-log data as hard data without uncertainty. To account for
the uncertainty, both petrophysicists and modelers need to question the data and
communicate the uncertainty from the various processes. Some simple questions can
greatly help understand the uncertainty in the well-log data, including

19

Are the data simply raw data dumped into a quick model or have they undergone
a set of rigorous checks and environmental corrections? Time spent on checking
and correcting the input data should increase the level of confidence in the results.

Have the data been normalized and if so, to what standard? Normalization should
remove systematic errors and noises from the data. However, it is very dependent
on the standard chosen to normalize to. If a certain facies is chosen, for example,
a thick marine shale or a low porosity zone, all the wells need to have that interval
for the normalization to be representative. Otherwise, true variability of the rock
in the wells will be masked or eliminated. The reservoir modeler should be
informed of the normalization process and related issues.

Is there a description of the correction and analysis process and a justification for
the analysts parameter selection and methodology? If so, the modeler can use the
data more confidently. If the data is of undocumented curves from unknown
sources, then uncertainty is obviously higher.

Was all the available data used and incorporated in the analysis? Everything being
equal, the more data is available and appropriately used in the processing
methodology, the less uncertainty is in the final petrophysical results and reservoir
model.

Did the log analyst perform sensitivity studies of porosity and saturation
parameters? Sensitivity analysis can help understand and identify the ranges of
these properties, and therefore help quantify the uncertainty in the data.
Figure 7 shows a simple example of the sensitivity analysis in deriving an

effective porosity. The pessimistic estimate of the effective porosity has an average of

20

2.3%, the base case estimate, an average of 4.0%, and the optimistic estimate, an average
of 5.1%. This data is from older well logs in a sparsely tested area. Porosity calculations
and shale corrections therefore have a high degree of uncertainty. The sensitivity was
done using optimistic matrix values and a minimal shale correction for the optimistic
estimate, normal porosity matrix values and average shale correction for the base case
estimate, and pessimistic matrix values and a maximum shale correction for the
pessimistic estimate. The range from this sensitivity analysis gives the reservoir modeler
some idea of the uncertainty in the well-log porosity.
Raw log data or un-rigorously processed log data can cause either over- or underestimation of subsurface resources in the reservoir model as a result of data error (Ma,
2010). One recent study of a clastic reservoir was to compare the oil in place (OIP) using
raw input curves and rigorously processed data. The raw data gave a 25% higher OIP
estimate. A rigorous analysis has helped identify the bad-hole effects on the log data that
caused the initial overestimation of OIP. The data correction process helped discern not
only the problem of the overall overestimation, but also identification of stratigraphic
zonations and areal locations of the overestimations. Obviously, there are also cases
where logs may be not sensitive enough to read the true values of porosity or resistivity,
and careless uses of such logs may underestimate the subsurface resources.
Propagation of the uncertainty from well-log and petrophysical analysis into
reservoir model can be quite complex, especially for local uncertainties in the distribution
of the reservoir properties within the reservoir. Nevertheless, global uncertainties of
composite reservoir properties often can be quantified based on the sensitivity analyses of
the individual contributing variables. For example, hydrocarbon pore volume can be

21

expressed as a function of the porosity, hydrocarbon saturation and net-to-gross ratio over
the reservoir. Once the reservoir dimension is defined, what impacts the global
hydrocarbon volumetrics the most is the average values of these properties. For example,
the three different well-log effective porosities in Figure 7 will imply large differences in
the pore volume of the reservoir model. Everything else being equal, the pessimistic
porosity will imply just more than half of the pore volume (consequently hydrocarbon
pore volume) of the base-case porosity in the model. On the other hand, the optimistic
porosity will imply 27.5% more pore volume in the reservoir model than the base-case
porosity (i.e., (0.051 0.04)/0.04 = 27.5%).

Figure 7. Histograms of the effective porosity (in fraction) derived with


three different sets of parameters. Red is the pessimistic estimate of
effective porosity derived with shale corrections and has an average of
2.2%; Dark blue is the base case estimate of effective porosity derived with
most-likely shale-correction parameters and has an average of 4.0%; and
light-blue is the optimistic estimate of effective porosity derived using the
most optimistic shale corrections and has an average of 5.1%;

22

CONCLUSIONS
Petrophysical analysis includes well log acquisition, calibration, processing and
interpretation. Each of these steps has uncertainties that can affect the results. The analyst
has many techniques, equations and processes available to choose from. Geologic and
petrophysical interpretations are non-unique-solution problems, and there is no absolutely
correct recipe to follow.
In this chapter we have briefly discussed the uncertainties that are due to
acquisition and processing in the primary logs. Reconciliations of logs acquired by tools
of different generations with varying qualities and sensitivities can help reduce
uncertainties in reservoir characterization. Many petrophysical properties derived from
logs used in reservoir characterization and modeling are interpreted instead of directly
measured. While a good interpretation can help narrow the range of the uncertainties
inherited from the acquisition and processing, inaccurate interpretations can cause added
uncertainties.
The key objectives for petrophysical analysis in a reservoir study is to reduce the
systematic errors caused by borehole effects, tool and vendor types, resolution
differences, depth shifts and other acquisition factors, and to perform a consistent
analysis between wells. While there may still be uncertainty in the results caused by
random error, these should be minimized for the ongoing interpretation. The output can
then be used more confidently in the reservoir modeling.
Integration is the key to any reservoir study that can validate petrophysical
analysis. Too often, each discipline - petrophysics, geology, geophysics, or reservoir
engineering, focuses on its own interpretation and analysis with little input from the
others. Uncertainties inherent in the individual disciplines are often not communicated to

23

the reservoir modeler, who uses the data as the accurate hard data, rather than as the
result of a number of choices and assumptions in the interpretations. Integration enables
team members to be involved in defining the uncertainty in the data and an acceptable
uncertainty range for the project. Only when the uncertainty range of the data from the
well log and petrophysical analysis is communicated to the other team members and
management, can the uncertainty be accounted for in reservoir characterization and
modeling.

ACKNOWLEDGEMENTS
The authors thank Schlumberger Ltd. for the permission to publish this work and
colleagues for the discussions on uncertainties in petrophysical analysis.
REFERENCES CITED
Bashore, W. M., U. G. Araktingi, M. Levy and W. J. Schweller, 1994, Importance of a
geological framework and seismic data integration for reservoir modeling and
subsequent fluid-flow predictions, In Stochastic Modeling and Geostatistics, Yarus
and Chambers (Eds), p. 159-175.
BIPM, 2009, Evaluation of measurement data An introduction to the Guide to the
expression of uncertainty in measurement and related documents, Joint Cmte for
Guides in Metrology, JCGM 104:2009, 28p.
Crain, E. R., 1986, The log analyst handbook, Pennwell Books, Tulsa, 700p.
Delfiner, P., 2007, Three pitfalls of Phi-K transforms: SPE Formation Evaluation &
Engineering, Dec. 2007, p. 609-617.
Dewan, J. T., 1983, Essentials of modern open-hole log interpretation, PennWell Books,
Tulsa, 361p.
Fertl, W. H. and Hammac, G. W, 1971, A comparative look at water saturation
computations in shaly pay sands, SPWLA 12th Annual Logging Symposium

24

Jennings, J. W., 1999, How much core-sample variance should a well-log model
reproduce? SPE Reservoir Eval. & Eng., vol. 2, no. 5, p. 442-450.
Jensen, J. and Lake L. W., 1988, The influence of sample size and permeability
distribution on heterogeneity measures: SPE Formation Evaluation & Engineering,
vol. 3, No. 2, p. 629-637.
Larsen, J. K. and Fabricius, I, 2004, Interpretation of water saturation above the
transitional zone in Chalk reservoirs, SPE Reservoir Evaluation & Engineering, Vol.
7, No. 2, April 2004.
Lucia, J. F., 1995, Rock-fabric/petrophysical classification of carbonate pore space for
reservoir characterization, AAPG Bulletin, vol. 79, no. 9, p. 1275-1300.
Ma, Y. Z., 2010, Error types in reservoir characterization and management, J. Petrol. Sci.
& Eng, doi: 10.1016/j.petrol.2010.03.030.
Ma, Y. Z., Seto A. and Gomez, E., 2008, Frequentist meets spatialist: A marriage made in
reservoir characterization and modeling: SPE 115836, SPE ATCE, Denver, CO.
Ma, Y. Z., E. Gomez, T. L. Young, D. L. Cox, B. Luneau and F. Iwere, 2010, Integrated
reservoir modeling of a Pinedale tight-gas reservoir in the Greater Green River Basin,
Wyoming, In Ma and La Pointe (Eds), Uncertainty Analysis and Reservoir
Characterization and Modeling, AAPG.
Schlumberger, 1999, Log interpretation principles/applications, 8th print, Schlumberger
Educational Services, Sugarland, Texas.
SPWLA, 1982, Shaly Sand Reprint Volume, July.
Theys, P., 1997, Accuracy Essential information for a log measurement, SPWLA 38th
Annual Logging Symposium, Paper V.
Theys, P., 1999, Log data acquisition and quality control, 2nd edition, Technip, 453p.
Worthington, P. E. 1985, The evolution of shaly-sand concepts in reservoir evaluation,
The Log Analyst, Jan-Feb.

25

Potrebbero piacerti anche