You are on page 1of 8

ARMA/NARMS 04-551

Uncertainty Shale Pore Pressure from


Borehole Resistivity
Martin D. Matthews, PhD.
Knowledge Systems, Inc. Stafford, Texas, USA
Copyright 2004, ARMA, American Rock Mechanics Association
This paper was prepared for presentation at Gulf Rocks 2004, the 6th North America Rock Mechanics Symposium (NARMS): Rock Mechanics Across Borders and Disciplines, held in
Houston, Texas, June 5 9, 2004.
This paper was selected for presentation by a NARMS Program Committee following review of information contained in an abstract submitted earlier by the author(s). Contents of the paper,
as presented, have not been reviewed by ARMA/NARMS and are subject to correction by the author(s). The material, as presented, does not necessarily reflect any position of NARMS,
ARMA, CARMA, SMMR, their officers, or members. Electronic reproduction, distribution, or storage of any part of this paper for commercial purposes without the written consent of ARMA
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
acknowledgement of where and by whom the paper was presented.

ABSTRACT: The degree of certainty in pore pressure estimates from resistivity data is examined through a case study
interpreted by eight trained analysts. The relevant data included resistivity, gamma ray, density, and mud weight as a function of
depth. A shallow kick pressure and several deeper MDTs were provided for calibration.
While the models and tools are quite simple, the shale pressures are derived from several measurements and decisions. The results
from each analysts interpretation were compiled and combined with some theoretical analysis to estimate the uncertainty attached
to components used to estimate shale pore pressure. Two sources of variation dominate the uncertainty in estimating pore pressure
from resistivity:
1.
2.

trust that sand pressure calibration points are representative of shale pressures, and
choice of the normal compaction trend.

Uncertainties in shale pore pressure ranged as low as plus or minus 0.25 ppge to as much as plus and minus 0.75 ppge depending
on these decisions. Other factors accounted for approximately less than 0.1 ppge in regions of good data. An approach for
minimizing and quantifying the impact they have on an analysis is discussed.

1. INTRODUCTION
Uncertainty is defined by Merriam - Websters
dictionary (2004) as something that is not known
beyond doubt, not clearly identified or defined, and
untrustworthy.
The calculation of pore pressure and other
parameters is commonly done deterministically. It
is assumed, without error that the measurements
respond only to pore pressure as a result of
compaction disequilibrium. Therefore the value
accurately represents the pore pressure at all depths
for which it is calculated. However, when we
examine the output of this calculation we often see
that it has substantial variation. This variation does
not always reasonably represent our interpretation

of pore pressure behavior. In part this variation is


reduced by performing the analysis only on the
best shales. Filtering of the shale response is also
performed to reduce the scatter in the data input into
the calculation, but in many cases scatter remains.
The experienced analyst then reads through this
resultant pore pressure variation to interpret a best
estimate of pore pressure.
The process of reading through variation is
dependent on the analysts vision of how pore
pressure is generated and dissipated in the
sedimentary sequence. The physics of pore pressure
estimation applies only to low permeability
systems, therefore sands are assumed (sometimes
incorrectly) to be in equilibrium with shale pore
pressures.

Although a complete analysis of subsurface


pressure includes three pressure estimates: shales,
sands, and mud weight needed to optimally drill the
well, this discussion will focus on shales. It is
generally assumed that shale pore pressure varies
relatively smoothly with depth, except for possible
rapid shifts to higher pressure due to particularly
effective seals. Shifts to lower pressure are
indicative of major pressure sinks through highly
permeable systems and are generally believed to be
related to the nature of the sharpness of the
transition from shale to sand and back again. As a
result of this paradigm, high frequency swings of
pore pressure are ignored (this is the underlying
assumption for choosing a particular filter function).
It is generally assumed that the shale pore pressure
is represented by the average of these swings, but it
could be anywhere within the envelope of estimates
(high side, low side, median).

systematically included but is assumed to be much


smaller than the other variations. The effects of
salinity due to proximity of evaporites are not
considered in this example and the change in
salinity of pore water between sands and shales is
considered inconsequential and minimized by
sand/shale discrimination techniques and filtering.
Eatons Exponent is held constant for this example.

This variation of pore pressure calculations


represents our uncertainty window of pore pressure.
Knowledge of the origins of the scatter in the input
data allows you to do something intelligent about it,
constraining and evaluating its importance. The
certainty of the estimation can be improved;
otherwise you depend on the fates. Knowledge of
the uncertainty also allows proactive contingency
plans to be prepared. As the well proceeds, and new
data arrives reducing the pre-drill uncertainty, the
driller can make more informed choices such as
appropriate mud weights and casing setting depths.

The shallow data set is often created from a curve


fitted to shallow core porosity data converted to
density. Figure 1 shows individual measurements
and a fit curve for shallow sand/silt/clay data. Some
of the scatter is undoubtedly due to individual
measurements being dominated by sand, or slit, or
clay.

3. OVERBURDEN
Overburden is ideally determined by the summation
of density data from the surface to total depth (TD).
However, this information is rarely complete from
surface to total depth. Commonly the near surface
data is absent and must be estimated. Deep data
may be available from density logs or can be
estimated from seismic data or checkshots.
Intermediate data often consists of a joining of the
shallow and deep density datasets.

The deterministic estimation of resistivity-based


shale pore pressure depends on several
measurements, empirical reference lines, and
operations. These include Depth, Overburden, Shale
Response/Determination, Shale Line, Porosity
Response (resistivity), Normal Compaction Trend,
Hydrostatic Pressure, and Eatons Exponent
(sensitivity and unloading).
2. FACTORS CONSIDERED TO BE KNOWN
Depth is measured with such low error that its error
is ignored, although cable and pipe stretch can be 50
feet or more in deep wells. That being said, care
should be exercised to assure that all data is brought
to the same reference point.
Hydrostatic pressure is considered to be constant.
The change in hydrostatic pressure with depth due
to temperature and pressure effects can be

Fig. 1. Shallow core density information as a function of depth


for mixed sand/silt/clay sediment. A best fit exponential line
(model) is shown for reference. Note at this shallow depth the
shales are less dense than the sands.

Results of variation in density on the calculation of


overburden have been estimated by Monte Carlo
simulation and shown to be small due to the process
of summing the density information downward,

stacking out the variation. The important part is not


the variation but the suitability of the function. If
you know the shallow sediments are newly
deposited sands or shales, or old compacted
sediments, or carbonates, a different function for the
relationship of overburden to depth is needed .Using
an incorrect function would be a consistent bias and
affect the entire overburden column. The variation
in the deep density information with depth
decreases because of the integration process. Again
what is important is that bias not be introduced,
such is often seen when seismic velocities are
observed to be lower that those observed in sonic
logs. Another sort of bias arises in joining the deep
to the shallow data. Often the two datasets are
joined by a simple linear relationship. In the case
shown in Figure 2 the deep RHOB data does not
begin until well into the overpressured zone. This
linear interpolation results in a kink where the
shallow data is joined, and an unrealistic
compaction profile, as shown by the green line in
the central track of Figure 2. A more realistic
joining is shown in yellow. When all these sources
of variation are accounted for, and used to calculate
the uncertainty in overburden, it is found to vary by
less than plus or minus 0.005 ppge, provided the
appropriate shallow density function is used.

shows densities. The shallow densities, shown in


green, purple, and orange, represent the mean
values and plus and minus two standard deviations
from Monte Carlo simulation of core data. The
intermediate depth RHOBs (bulk density) are
represented by the green straight line interpolation
between the deep and shallow data. The yellow
curve is a hand-drawn smoothed curve through all
data. RHOBs calculated from the checkshot values
are shown in red, and the RHOBs calculated using
the sonic log are represented by deep green. Note
that at these depths the shales are denser than the
sands. In the right track, respective overburden
gradients are shown. The purple curve represents
the overburden gradient calculated using the straight
line interpolation (green) between the RHOB data
and the dark purple curve represents the overburden
gradient calculated using the curved estimate
(yellow). The calculated pore pressures, shown in
light and dark blue, were based on filtered shale
resistivity, with the resulting uncertainty due to
variations in overburden only, all other factors
fixed. Note the two curves are so close that the light
blue one is shown thicker than the dark blue one in
order to see it under the dark blue curve. The effect

Fig. 2. Uncertainty and bias in the estimation of RHOB and


Overburden.

In Figure 2, the left track shows gamma ray (blue)


with shale reference lines (red). The central track

of not appropriately modeling the shallow densities


is significant, as is done when a constant

overburden is assumed. This affects not only the


shallow pressure estimates but propagates throughout the section. That gross blunder is not considered
here.
4. SHALE DISCRIMINATION
Because pore pressure estimation from electric logs
and seismic interval velocities is typically based on
effective vertical stress shale compaction models,
pore pressures are only estimated in purer shales.
Marls, sands, evaporites, silt rich shales, etc. should
not be utilized. This is accomplished through the
use of a shale discriminator data set such as gamma
ray or SP, and one or more thresholds. The selection
of shales/not shales does not directly enter into the
calculation of pore pressure but rather influences
which resistivity values will be used to calculate
shale pressure. By selecting the purer shales we
often wind up with those deposited in deeper water
which minimizes lateral and stratigraphic variation.
However, there is still variation in these shales. It is
felt that a different threshold is justified for each
casing interval because of hole size effects.
The use of multiple shale lines or a single shale line
within a casing interval does not significantly affect
the trend of pore pressure when a filter is used to
smooth out the variation in resistivity. It does,
however, affect the number of resistivity points and
therefore the filter characteristics and the frequency
of variation (Figure 3).

Fig. 3. Effect of multiple vs. one shale threshold casing


interval.

The use of multiple shales lines does, however,


allow the analyst to selectively control both the
magnitude and variation of the pore pressure
estimates. It is recommended that one shale line per
casing interval be utilized. Exceptions justifying
multiple shale lines with a casing interval include:

change in mud properties (density and/or


potassium content: mica LCM or swelling
inhibitors),

unusually high gamma ray responses


interpreted as petroleum source rocks with
high concentration of radioactive minerals;

logging hardware considerations, e.g. failure


of part of a Geiger-Mueller tube array.

Formation density also affects GR response but is


negligible.
In Figure 3 the left track shows gamma ray (dark
blue) with both one (light blue) and multiple
(yellow) shale thresholds per casing interval. The
central track shows resistivity (red) with filtered
trends and slightly different normal compaction
trend lines in corresponding colors. The right track
shows pore pressure in corresponding colors with
RFTs in red squares and overburden in purple.

Note the similarity of trends in both pore pressure


curves. Note also the different frequency content.
The light blue pore pressure curve is based on more
shale points than the yellow curve. To counter this
effect a smoothing filter length 10 times that used in
construction of the yellow curve was used. These
two pore pressure estimates, from the same data,
illustrates the interaction of number of shale points
and filter size. It also demonstrates the difficulty in
deciding what the true value of the shale pressure is.
The pressure regression in the blue pore pressure
curve at 17,000 feet is a case in point. This
regression is coincident to the sand rich section at
that depth. Do these sands bleed the pressure down
from the shales in that section and adjoining shales?
Or is the regression, which does not show in the
yellow pore pressure curve because of the more
conservative selection of shale points, caused by a
facies change within the shales?
The effect of selecting more and more restrictive
definitions of shale is shown in Figure 4. A thick,
uniform, relative pure, shale was selected for this
analysis. The magnitude of pore pressure is
generally not affected by this decision. In this study,
surprisingly, the variation of resistivity values is
also not significantly effected by the more
aggressive thresholds. What is affected, however, is
the number of resistivity points available for
filtering, and therefore the characteristics of the
filter. The choice of a conservative threshold
(orange line) results in few points and requires a
long filter for smoothing. The choice of a less
selective threshold (while still eliminating obvious
non-shales) results in more points and therefore a
shorter filter can be utilized, if desired.

Fig. 4. Effect of shale threshold value.

In Figure 4 the left track shows gamma ray (dark


blue) with light blue, yellow and orange shale
threshold lines. The right track shows resistivity
(red) with light blue, yellow and orange symbols
showing the resistivity values selected based on
their matching shale threshold colors. Numbers
refer to the number of resistivity values
corresponding to the conservative orange shale
threshold.
5. SELECTION OF RESISTIVITY DATA
Resistivity data has already been selected based on
a shale threshold. The short term, high amplitude
swings in resistivity, such as shown in Figure 4 near
7320, 7370, 7390, and 7470 depths suggest LWD
(logging while drilling) real-time telemetry errors
and can be edited out in the uncertainty analysis.
Note that this type of swing in the data does not
introduce bias in the filtered result but only affects
certainty of interpretation.
Additionally, shale streaks may be dominated by
lithologic effects, such as marly zones and changes
in shale material (insulating organics and/or super
conductive pyrite, chlorite etc.). These can not
always be recognized by their gamma ray response
and can be incorrectly interpreted as a response to
pore pressure. These shales may need to be
eliminated as outliers. This process does not affect
the magnitude of the pore pressure estimate to a
significant extent because of the filtering process,
but can significantly affect the uncertainty. Thirty
thick shales, with relatively uniform gamma ray
response, were selected to determine the effect of
selective removal of resistivity outliers (spikes).
Analysis showed the majority of shale resistivity to
have a standard deviation under 0.1 ohms,
regardless of depth or magnitude. One of only two
exceptions to this is shown in Figure 5. The removal
of two abnormally high resistivity measurements
reduced the uncertainty in the resistivity estimate
while having minimal effect on the average value.
Selective spike reduction to remove outliers that are
most probably not representative of the compaction
state of the shale is desirable.

porosity with depth is an adequate model [e.g. Athy,


1930].
Shale pressures can not be measured directly
because their low permeability usually does not
allow measurements in a reasonable time frame.
Measured pressures and flow events are
representative of the more permeable units such as
sands. Uncertainty in the value of sand pressures for
calibration falls into two categories:
Fig. 5. The affect of spike removal on resistivity.

Applying the 0.1 ohm standard deviation estimated


above to the entire filtered resistivity trend results
and combining it with the overburden uncertainty
results in a pore pressure uncertainty of about plus
or minus 0.07 ppge.
6. NORMAL COMPACTION TREND
The selection of a normal compaction trend, from
which to estimate pore pressure using resistivity
data, accounts for the majority of the uncertainty in
the estimates. This uncertainty arises from six main
sources:
(i) Uncertainty in the shape of the normal
compaction trend (logarithmic, power law,
complex, etc.);
(ii) Uncertainty in the relationship of calibration
data (sand pressures)
(iii)Operator variability in locating the normal
compaction trend, relative to the calibration
data;
(iv) Salinity variation
(v) Temperature effects
(vi) Unconformities and faults.
In this study salinity changes with depth were not
noted. Unconformities and faults were also not
recognized. Temperature increase with depth
undoubtedly occurs, but is assumed to be
approximately linear. This linear drift is normally
handled by modifying the slope of the normal
compaction trend, rather than numerical
normalization of resistivity for temperature.
Uncertainty due to the shape of the normal
compaction trend is not evaluated. All shales are not
equal, due to their original depositional conditions
and changes in their ability of their grain structure
to support load as burial and diagenesis progress.
Despite this potential complexity many geologic
studies have shown that a logarithmic decay of

confidence in the measure itself


the degree to which the analyst believes it
represents the pressures in the shale.

Table 1 shows the calibration data available in a


well in order of confidence.
Table 1. Data sources of sand pressure in well and their
reliability in order of confidence.

Data

Lithology

Bias

Problems

RFT/MDT
RFT/MDT
RFT/MDT

Sand
Sand
Sand

Low
High
High or low

Shut in
Drill Pipe
Pressure

Sand

Slightly low

Kill mud
weight

Sand

Slightly
high/low

Mud
weight in
general

Various

Previous
experience
in the area

Bad seal
Centroid
Uncertainty due to
extrapolation
towards a long time
frame due to low
permeability
Uncertainty in
cuttings volume,
even after
circulating bottoms
up, and friction in
lines.
Adjusted until they
can live with it.
May be high to kill
it for sure or low
and live with a
little fluid/gas
influx
Too uncertain
without additonal
information from
the well such as
connection gas,
mud loss/gain,
cutting
morphology, etc.

The centroid effect [Finkbiener et. al. 1998],


coupled with buoyancy, is responsible for the
analysts doubt of the applicability of the sand

pressure to accurately represent shale pressure. It is


critical to understand and remove this effect. Wells
are most always drilled on high points of structure.
This is in the region where the sand pressure can be
higher than the surrounding shales due to pressure
transference in the more permeable sand (just like
pressure transference up a well bore). Thus, most
often, the far field shale pressure adjacent to the
sand is below this pressure by an amount equal to
the depth to the centroid and the density of the
fluids in that sand. The pressure in the shale in
contact with the sand may be almost in equilibrium
with the sand but the fall off of pressure between
the two is comparatively short because of the low
permeability in the shale. This region is of no
interest in this study.
Operator variability in locating the normal
compaction trend relative to the calibration data is
intimately tied to the trust the analyst has in the
calibration data. This uncertainty was analyzed by
giving the same overburden, filtered shale
resistivity and calibration data to eight geopressure
analysts with over 140 total years of experience.
Each analyst had received similar training in strict
company best practices, and had been working
together in peer groups for several years. These
analysts used the same techniques, only their
selection of normal compaction trend positions
varied. The results are shown in Figure 6.

Fig. 6. Uncertainty due to selection of the normal compaction


reference line.

In Figure 6 the left track shows gamma ray (dark


blue) with multiple (yellow) shale thresholds per
casing interval. The centeral track shows resistivity
(dark red) with filtered trend in red and eight
normal compaction trend reference lines (one for
each analyst. The right track shows pore pressure in
corresponding colors with RFTs/MDTs (repeat
formation tester, multiple depth tester) in red
squares and overburden in purple.
Three groups of compaction trends and their
resultant pressure profiles are immediately
noticeable. The central group of five analysts fit the
calibration data and represents a reasonably tight
cluster, approximately plus or minus 0.1 ppge.
There is one analyst that fit the calibration
reasonably well but used a negatively sloped normal
compaction trend (indicative of porosity increasing
with depth). The justification for this surprising
trend is that no temperature correction was made to
the resistivity data and thus the negative slope
adjusted for this effect. This analyst felt that such a
correction was justified based on the high
sand/shale ratio in the lower portion of the well and
the slight decline in RFTs with depth in that portion
of the hole. Had the resistivity data been corrected
for temperature all slopes would have been positive.
There were two analysts that had prior knowledge
of this well and that colored their choice of a normal
compaction line.

They suspected the kick at 8500 feet was a shallow


flow due to centroid effects and therefore made sure
their pressure estimates were below the kick
pressure. Instead they assumed the upper portion of
the hole was essentially normally pressured.

The degree of trust and estimation of bias in the


calibration data is the greatest source of uncertainty
in shale pore pressure estimation. It can only be
minimized by an independent analysis of the
relationship of each measured sand pressure to the
existence of shale pressure compartments and
pressure transference/buoyancy in each individual
sand.

7. SUMMARY AND CONCLUSIONS


It is, perhaps, not surprising that human decisions
are responsible for the greatest part of uncertainty in
pore pressure estimation and the measured data
responsible for the least part.
Uncertainty in shale pore pressure due to selection
of a reference normal compaction line, given that
the calibration data (measured in sands) is
considered representative of shale pressure, is
approximately plus or minus 0.1 ppge. Improving
the reproducibility of the fit to the calibration data
represents the greatest potential gain in certainty in
the pore pressure estimate. Corrections of measured
sand pressures for centroid and buoyancy effect
represent a potential gain in certainty if correctly
done.
Total uncertainty in shale pore pressure due to
uncertainty in log data is expected to be under plus
or minus 0.08 ppge. In regions where the resistivity
data has abnormally high frequency variation that is
not suppressed by filtering it can be more.
Filtering changes the frequency and magnitude of
pore pressure variation. This uncertainty was not
evaluated. The decision on what portion of the
resistivity response is due to pore pressure and what
to lithologic and pore fluid conductivity is complex
and represents the most misunderstood factor in
estimating pore pressure at a given depth.
The following recommendations are made in order
to minimize uncertainty.

Care should be taken in the selection of


density information used in construction of
the overburden curve. The choice of an

appropriate shallow density curve and the


joining of this curve to the deeper density
data is particularly important.
Use a single shale line in each casing
interval.
Do not be overly aggressive in determining
shale non shale, thereby unduly limiting
the number of shale values.
Selectively eliminate resistivity spikes prior
to filtering (minimal effect on pore pressure
estimates).
Corrections for bias in the relationship of
calibration data to shale pore pressure needs
to be considered. This is the major source of
uncertainty.
A more systematic, less subjective, method
of fitting the normal compaction trend to the
calibration data needs to be utilized.
Although not discussed in this paper
normalization of resistivity measurements
for salinity and temperature is
recommended. Normalization for salinity
often removes the necessity of multiple
normal trend lines.

REFERENCES
1.

Athy, L. F., 1930. Density, porosity and compaction of


sedimentary rocks. AAPG Bull. Vol 14, pp 1-24.

2.

Finkbeiner, T., M. Zoback, P. Flemmings and B.


Stump, 1998. Stress, pore pressure and dynamically
constrained hydrocarbon columns in the South Eugene
Island 330 field, Northern Gulf of Mexico. AAPG Bull.,
Vol 85, pp 1007- 1031.

3.

Merriam-Webster OnLine Dictionary,


http://www.m-w.com/cgibin/dictionary?book=Dictionary&va=uncertainty&x=1
2&y=14

You might also like