SPE 115429

Downloaded from http://onepetro.org/SPEATCE/proceedings-pdf/08ATCE/All-08ATCE/SPE-115429-MS/2732820/spe-115429-ms.pdf/1 by Schlumberger Oilfield UK Plc user on 14 February 2023

Advanced Compositional Gradient Analysis
Jesús Cañas, SPE, Julian Pop, SPE, Francois Dubost, SPE, Schlumberger; Hani Elshahawi, SPE, Shell
International E&P

Copyright 2008, Society of Petroleum Engineers

This paper was prepared for presentation at the 2008 SPE Annual Technical Conference and Exhibition held in Denver, Colorado, USA, 21–24 September 2008.

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

Abstract

Steep gradients are common in gas condensate and volatile oil reservoirs, but they are also present in heavy oils reservoirs.
There are numerous publications (Creek, 1985, Lars Høier, 2000, Montel, 2002, Firoozabadi, 1999, Ghorayeb, 2003,
Fujisawa, 2004, Elshahawi, 2005, and Kabir, 2006) that have dealt with complex fluid columns showing compositional
gradients for columns in thermodynamic equilibrium or under steady state conditions. Montel et al. (2002) discuss processes
that arise from recent charging of these reservoirs, which are not in equilibrium but still undergoing for instance a flux of the
light components that diffuse.
Formation testers supply a wealth of information to observe and predict the state of fluids in petroleum reservoirs,
through detailed pressure and fluid analysis measurements. With the correct understanding of fluid characteristics in the
reservoir, reserve calculations and adequate development plans can be prepared. Additionally, flow barriers may then be
revealed, as across such barriers, fluid may appear different. In order to test for the existence of such barriers, pressure or
fluid differences, in their context must be assessed. Formation tester data must therefore be treated by means of a systematic
analysis, so that different sources of information lead to an integrated, preferably, consistent conclusion. Downhole fluid
analysis in conjunction with pressure gradient analysis is simultaneously analyzed to reflect and make deductions concerning
the correct state of fluids in complex fluid columns. As part of the process, different reservoirs need to be well delineated
through appropriate statistical similarity tests of pressure gradients and fluid analysis of the different zones (Kabir, 2006).
In this paper, a novel methodology is presented to account for non-linearity in pressure gradients due to varying fluid
density. We examine two field cases in which compositional gradients were observed with both downhole fluid analysis
measurements and pressure gradients. An equation of state model is used to describe the fluid column according to a simple
fluid equilibrium model. A non-linear pressure gradient regression fit is also examined. The choice and appropriateness of the
pressure gradient model that accounts for the observed fluid density changes observed. The result of the modelled fluid and
pressure analisis are compared to actual downhole measurements of the pressure profile and insitu fluid logs. In particular,
pressures and densities calculated from the multiple sources of information, pressure and fluids, are compared to direct actual
measurements. The reservoir architecture is revealed. For both cases, the fluid column is continuous through the hydrocarbon
intervals down to the water-oil contacts. Flow barrier locations would be found on the basis of anomalies and departures of
the measured data from the modelled pressure and fluid gradients.

Introduction

Compositional variations in reservoir fluids with depth are more common than perhaps normally expected and have been
observed in several reservoirs throughout the world. This phenomenon is not limited to thick reservoirs, but is observed also
over relatively short vertical columns. This is also not only limited to reservoirs containing gas condensates or volatile oils,
but is observed in heavy oil reservoirs as well (Mullins et al, 2005, Mullins et al, 2007).
There are several different mechanisms that may create fluid compositional variations in reservoir fluid columns. Fluid
gradients can originate from gravitation, thermal gradients, biodegradation, water washing, current reservoir charging,
multiple reservoir charges, leaky seals, time and temperature dependent variation of hydrocarbon properties from kerogen
catagenesis, variation in deposited kerogen and production schemes, such as compositionally differential production (e.g.
production below a phase transition), or miscible flooding. Only the first two mechanisms drive the hydrocarbons towards
2 SPE 115429

equilibrium (Lars Høier, 2000).

This wide range of mechanisms and the known limitations inherent in traditional pressure gradient analysis make
compositional grading one of the main challenges to the exploration and production community. Uncertainties in pressure
gradient analysis are associated with the lack of adequate quality and quantity of pressures, poor depth accuracy of the
locations at which pressure are measured and the limited accuracy of the pressure measurement itself (Dubost et al, 2007).
The pressure measurements could moreover be affected by depletion, capillary forces and rock wettability (Elshahawi et al,
1999) and/or supercharging effects.

Downloaded from http://onepetro.org/SPEATCE/proceedings-pdf/08ATCE/All-08ATCE/SPE-115429-MS/2732820/spe-115429-ms.pdf/1 by Schlumberger Oilfield UK Plc user on 14 February 2023

Improper characterization of fluid properties in high-cost production environments such as in deepwater, adversely
affects reservoir evaluation and can lead problems with flow assurance, facility design, reservoir management, production
strategies and reserve estimates, all of which can severely impact the financial performance of a project. Proper fluid
characterization necessitates an accurate understanding of fluid compositional grading. To achieve such an understanding,
modeling of the equilibrium state of the fluid is necessary so that any anomalies can be inferred. This is particularly true
when testing vertical connectivity in a reservoir, a critical aspect for development planning,
In addition to the previously mentioned conventional pressure gradient analysis, other measurements can be performed
using a variety of techniques, at surface or downhole, to enhance the reservoir characterization. Surface measurements
include PVT analysis on formation fluid samples at the wellsite or in a laboratory from samples taken via traditional
Drillstem tests or Formation testers as well as advanced mud gas logging measurements (McKinney et al. 2007). Downhole
measurements, on the other hand, include the already mentioned pretest pressure surveys, in-situ fluid analysis techniques
such as molecular spectroscopy measurements coupled with reflection and fluorescence measurements, and in-situ flowline
measurements with vibrating object densimeter for direct (meaning immersed in the flowline) and precise formation fluid
density and viscosity readings. Downhole electric logs should also support the integrated compositional gradient analysis, or,
at least, aid in helping to delimit the main fluid systems and contacts. Table 1 presents a summary of these various
techniques along with the strengths and limitations of each.
The degree of agreement between predicted fluid properties using an equation of state and laboratory measurements
depends mainly on the accuracy of the experimentally determined composition; however, obtaining a good match implies a
good laboratory analysis but does not mean the sample is representative. The prohibitive difficulty in the past, prior to the
introduction of the in-situ fluid technology has been to justify extensive wireline sample acquisition and laboratory analysis
without a priori evidence of fluid complexities. Today’s in-situ measurements, coupled to novel gradient and fluid models
run in real time (Appendix A), allows an early characterization of fluid gradient complexities. This paper shows how the use
of pressure gradients and in-situ fluid analysis is essential for an accurate representation of the reservoir fluid characteristics
and imaging of flow barriers. The proposed workflow enables optimization of the number of measurements and samples to
be collected downhole. For instance, it allows establishing the case for a continuous homogeneous fluid column in real time,
which helps avoid collecting more samples than necessary.

General Theory on DFA Measurements

Optical measurements: Visible-to-near-infrared (VIS/NIR) spectroscopic measurements are widely used today as the
primary method to investigate reservoir fluids under downhole conditions by a wireline formation tester. In-situ fluid
analysis, which began as a means to assure quality control during fluid sampling, has evolved into a key technique for
characterizing the distribution of reservoir fluid properties and for determining zonal connectivity across a reservoir.
VIS/NIR spectroscopic measurements coupled with reflection and fluorescence measurement have evolved to the degree that
it is possible to measure oil/water/gas fraction, to assess oil-based mud contamination quantification, to measure Gas-to-Oil
Ratio (GOR), hydrocarbon composition (C1,C2,C3-C5,C6+), carbon dioxide (CO2), to provide a fluorescence measurement,
and a pH measurement (Mullins et al 2001, 2004, 2005, Raghuraman, 2007, Schlumberger, 2007).
In-situ density/viscosity measurement with downhole sensor: The in-situ density sensor, or vibrating rod, measures
the density and, under some circumstances, the viscosity of the formation fluid in the formation tester flowline at downhole
conditions (O’Keefe et al. 2007; Godefroy et al, 2008). As with any vibrating object in a fluid media, the resonance
frequency and quality factors are related to the fluid characteristics. The resonance frequency is mainly related to the fluid
density, where added mass around the sensor decreases the resonance frequency. Thus, the heavier the fluid is, the lower the
resonance frequency. On the other hand, the quality factor of the measurement is mainly a function of fluid viscosity; thus the
more viscous the fluid is, the more damping the sensor will experience, and the lower the quality factor. Both resonance
frequency and quality factor are measured from the ring down of the sensing element vibration after an initial phase of
excitation. From both resonance frequency and quality factor, density – and under some conditions, viscosity – are predicted
using a complex physics modeling of the elastic properties of the sensing part and the Navier-Stokes equations describing
fluid displacement around it. As the sensor response is physically modeled, no empirical correlations or internal databases are
required for the computation of density (or viscosity).
SPE 115429 3

Compositional Grading Introduction

The requirements for compositional grading are that the reservoir is continuous or interconnected and that fluid column
properties and composition (including GOR, saturation pressure, API, Saturation/Aromatics ratio, gas mole fraction, etc.)
vary with elevation. The magnitude in grading of a parameter can vary greatly, depending on the geological and geochemical
history of the reservoirs. The time necessary to reach compositional equilibrium is comparable to the geological lifetime of a

Downloaded from http://onepetro.org/SPEATCE/proceedings-pdf/08ATCE/All-08ATCE/SPE-115429-MS/2732820/spe-115429-ms.pdf/1 by Schlumberger Oilfield UK Plc user on 14 February 2023

typical reservoir (Hirschberg, 1988). Therefore, in most cases, if equilibrium conditions are observed across a fluid column,
the conclusion of that fluid column being inter-connected is a valid one.

Fluid compositional grading may be observed in different hydrocarbons that include:

Volatile/ Near critical Oils

For this type of fluid, strong compositional grading generally ocurrs as reservoir conditions approach the critical conditions
(temperature and pressure) for that fluid. Strong variations in fluid properties with depth, such as fluid densities or saturation
pressures are produced even with the fluid column still in equilibrium. That is, the fluid components remain in balance while
subject to gravity and chemical potential forces. In such a cases, the static pressures measured with vertically distributed
pressure measurements will not follow a linear pressure gradient, due to the varying density of the fluid with depth (Wheaton
et al., 1991, Lira-Galeana et al., 1992, Hanafy et al, 2005, Ghorayeb et al, 2000, and Kabir et al, 2006).

Heavy Oils
Similarly, for heavy oils, fluid variations with depth are common. In heavy oils, compositional grading often result from a
loss of lights ends caused by the segregation of asphaltenes or from bio-degradation, resulting in variation of oil coloration,
viscosity and the possibility of tar-mat formation (Tissot, 1978, Hunt, 1979, Mullins, 2007). Such effects can be inferred from
geologic data. These compositional variations also influence field development. The presence of highly viscous oil near the
oil-water contact has forced production from up dip and would be a serious handicap for down dip water injection
(Hirschberg, 1988). A field case is presented in the next section for this type of reservoir.

Proposed Methodology for Compositional Gradient Evaluation

As presented in Table 1 different techniques may be employed to analyze fluids columns. Our proposed methodology
consists of combining pressure data and in-situ fluid logs to predict fluid compositional gradients in real time. This is a
significant step change from prior evaluation methodologies where sampling would be realized rather blindly, and results
discovered weeks later with a possible mismatch with original evaluation objectives; due, for example to lack of experimental
representativity. Fig. 1 and Appendix A show the workflow and detailed methodology that would allow an adequate match
of the objectives. A key element of the workflow is to predict the fluid measurements in real time, so that if actual
measurements deviate from these predictions, further experiments and testing or sampling downhole may be performed.

The detailed suggested workflow for our gradient analysis is presented in Appendix A and following it is a summary:

- Perform a detailed check of each pressure test to eliminate data affected by supercharging, limited stabilization
or any other phenomena which would compromise the validity of the test. In particular, eliminate the tests that
are considered as outliers.
- Run a conventional pressure gradient analysis and check if a single fluid gradient fits the pressure data. This
allows checking if the fluid is in vertical equilibrium. If at all possible identify those curved pressure trends
which arise from variations in the composition of the fluid with depth, eliminating those which are most likely
associated with fluid movement, capillarity, convection, permeability heterogeneity/supercharging, etc.
- Perform an excess pressure analysis (Schlumberger, 1984 and Brown, 2003) to detect complexities in the
vertical pressure profiles and deduce possible variations of density with depth. Subtle pressure trends or
discontinuities in trends which otherwise could be overlooked may be easily identified when the data is
presented in this format. This approach is referred to a possibly hypothetical, constant composition fluid.
Departure from a constant excess pressure trend implies a change in fluid type, compartmentalization or fluid
density - composition variations. These findings must be reconciled with other data sets including geological,
geophysical, petrophysical, and geochemical data.
- Modeling the pressure gradients with straight line and polynomial fits to derive local pressure gradients which
are then converted into a fluid density at the pretests stations, accounting for errors in both depth and pressure;
estimate the error bounds on the density (Kabir et al, 2006 and Collins et al, 2007).
- Formulate an EOS using available PVT equivalent data from in-situ logs. Compositional groups C1, C2, C3-5,
C6+, CO2; density/ viscosity, and GOR are measured today by in-tool sensors. A new scheme is used to delump
4 SPE 115429

these composition groups into length compositional data (Zuo et al, 2007). Then the density and viscosity
measurements are used to tune the model.
- Derive fluid densities from fluid compositional gradient calculations at the pressure pretest stations using
temperatures similar to those from the in-situ density sensor reading. The methodology employed to compute
the density and viscosity EOS derived values is the following: The equation of state model was derived at one
particular station in the reservoir. A compositional gradient is generated, assuming a gravity-chemical potential
equilibrium. Fluid descriptions are then obtained at different depths. Based on these descriptions, perform flash

Downloaded from http://onepetro.org/SPEATCE/proceedings-pdf/08ATCE/All-08ATCE/SPE-115429-MS/2732820/spe-115429-ms.pdf/1 by Schlumberger Oilfield UK Plc user on 14 February 2023

experiments at downhole conditions of pressure and temperature to evaluate density and viscosity at new depths.
- Cross validation of the derived fluid density results from the pressures, EOS and in-situ measurements. In the
absence of focused-probe sampling (O’Keefe et al, 2006) it may be necessary to correct in-situ densities for
OBM contamination (Mullins et al, 2000, Reddy et al, 2007). Optical density measurements give contamination
and linear least squares simultaneous regression gives the density of filtrate and oil. A good match justifies the
fluid connectivity in the interval and the absence of compartments, while differences between pressure derived
densities and EOS-derived densities, possibly, implies the possible following occurrences:
o Inadequate sampling of the interval of interest: a limited number of in-situ stations or pretest surveys
do not allow for unique model matching.
o Non equilibrium conditions, such as a biodegraded column, which does not follow any gravity thermal
equilibrium model
o Existence of flow barriers
- Finally, complementary analysis for fluid “contacts” confirmation and estimation of the error of those contacts.

Two field examples in which compositional gradients have been observed with downhole fluid analysis measurements and
pressure gradients are presented in the following sections.

Compositional Gradient Field Studies

Case 1
In a recent exploration well, downhole fluid measurements were acquired to describe the possible occurrence of a
compositional gradient, understand the reservoir architecture, whether there exists any vertical barriers, and sample to
characterize the fluids.
This job was run using two single-probe modules in the tool string (Fig. 2) to understand the reservoir architecture,
and best characterize the fluid column. A large diameter probe was run for pressure pre-tests (tool zero reference in Fig. 2)
and an extra large diameter probe was run for formation fluid pump out and sampling. In this job for performing the
downhole fluid analysis stations, a combination of optical and in-situ density/viscosity sensors was run upstream and
downstream of the upper pumpout module, plus an optical sensor upstream of the lower pumpout module.
A detailed set of pressures was collected and evaluated in real time as per the workflow in Fig. 1 to perform a
preliminary analysis for apparent compositional gradient in the 150 meters open hole crossing the hydrocarbon bearing
reservoir. Several fluid identification and sampling stations were defined with in-situ logs in real time with the support of the
vibrating rod density sensor to complement the optical spectroscopy sensors. The densimeter recorded density variations
from 0.503 to 0.550 g/cm3 from the top to the bottom of the measured zone. Viscosity varied from 0.118 to 0.160 cp between
four stations spaced 95 meters inside the oil column (Fig. 3). The GOR and compositional results seen with the optical
spectroscopy sensors also detected the same vertical trend. The hydrocarbon fluid filling the reservoir sands were of volatile
oil type, and gradually changing vertically along depth, with lighter fluids at the top and heavier fluids closer to the water oil
contact. All fluid measurements, density, viscosity, compositions and GOR, coming from various distinct sensors
corroborated these finding.
The question arose as to whether the reservoir sand is continuous or if vertical permeability barriers (baffles) exist, and
if these different zones are filled by distinctly different fluids. In order to test this hypothesis, both a pressure gradient and a
fluid gradient model were calculated as per the methodology summarized in Appendix 1. The following are the main results
produced by each step.
• After a detailed check of each pressure pretest, it was possible to define a single fluid gradient, which at first,
appears to fit the pressure data perfectly as shown on Track 1 of Fig. 4. However, the complexities in this 140 m
fluid column are revealed only when the potential or excess pressure plot is examined (center track in Fig. 4). At
least three fluid zones are indicated by this plot: a top zone of about 20 m (form X380 to X400 m) where the fluid
body appears to have a constant density of 0.5012 g/cm3; a second zone of 100 m in extent (from X420 to X520 m)
where the fluid density varies continuously; and a third zone in which only two pressure tests were taken which may
be disconnected from the large zone situated above it.
• Following the proposed methodology for analyzing vertical pressure profiles to deduce variations in density with
depth, a series of low order polynomials was fit through the pressure data. A series of local pressure gradients was
then derived and converted into a series of fluid densities. In this particular case, the best fit to the pressure data was
SPE 115429 5

obtained using a quadratic function (red points and red line in Fig. 4) which implies that the fluid density varies
linearly with depth; the red points in the right-most track of Fig. 4 shows this variation in fluid density (from about
0.5 to 0.54 g/cm3, see also the Table 2). From the statistics of the fit it is possible to derive point bounds for the
density estimates; these are recorded in the fourth column of Table 2.
• The raw in-situ densities were corrected for OBM contamination despite being quite low; under 2% OBM by OCM
analysis (Mullins, 2000, Reddy, 2007, Venkataramanan, 2007). The model to correct the measured in-situ densities
for the level of contamination is described on Appendix A. See results in Table 2 for Case 1.

Downloaded from http://onepetro.org/SPEATCE/proceedings-pdf/08ATCE/All-08ATCE/SPE-115429-MS/2732820/spe-115429-ms.pdf/1 by Schlumberger Oilfield UK Plc user on 14 February 2023

• The fluid was then modeled using an EOS and a gravity-chemical potential equilibrium gradient model was derived
by matching the fluid compositions versus depth generated from the model to optical fluid analyzer compositions
along the fluid column. The equilibrium compositional gradient model was also matched to the observed in-situ
density live fluid density / viscosity. The equation of state model was computed from the fluid descriptions obtained
at depth x421m and extrapolated under an isothermal gravity-chemical potential equilibrium gradient model to the
different other depths. The comparison of the EOS derived fluid parameters and compositions at the different depths
to the in-situ answers (including the corrected in-situ density) are given in Fig. 5. In the Table 2 are given the
temperature references for the flash calculations. The temperatures here follow a gradient of 0.09 oF/ m.
It is worth mentioning that the calculated densities from the EOS used tool temperature, so that the derived fluid
density from pressure, in-situ density or EOS is referred to the same temperature for consistency in the comparisons.
• The cross validation of the derived fluid density results from the pressures (polynomial model), densities obtained
from an equation of state (EOS) and densities obtained by means of a downhole densimeter (in-situ density sensors)
is presented in Table 2 and Fig. 4. Given the subtleties of the variation in properties the agreement between these
independent measurements is quite remarkable. They confirm the existence of a compositional gradient identified in
real time by the optical spectroscopy sensors (C1, C2, C3-5, C6+, CO2 and GOR) and the in-situ density-viscosity
through several downhole fluid stations. PVT lab results, which later became available, also showed excellent
agreement with the pressure- and DFA-derived densities.
• The results generally indicate that the reservoir sands are continuous down to X480m. However, the 20m region
between the top zone and the middle zone (X400 to X420 m) is interesting since the top zone is about 0.5 psi higher
in pressure and appears to show a continuous variation in pressure across it. There are several possibilities that
include a) there are a series of isolated stringers in which the pressure happens to have this distribution. Certainly the
formation appears to be quite heterogeneous; b) some differential depletion is taking place even though this
reservoir has not been produced; or c) there is fluid motion vertically from the top to the middle zone under this
small differential. In the lower reservoir section below X480 m, both the pressures, and the fluid fractions depart
from the compositional gradient model. This also supports the probable occurrence of a subtle flow barrier. The last
possibility (C) has a direct bearing on the computation and validity of the derivation of densities from pressure.
Indeed, if fluid passed from the top zone to the bottom the effect would be to decrease the gradient and therefore
apparent density. A detailed analysis about a possible effect of a vertical fluid flux on apparent density derived from
pressure gradients for Case 1 is presented in Appendix B.

This case illustrates that the combination of downhole fluid analysis (including in-situ density and viscosity) and
pressure measurements is highly effective for the characterization of the reservoir pressures and fluids.

Case 2
In this well, it was required to perform the best possible real-time downhole fluid analysis for a viscous oil analysis and
sampling with low OBM filtrate contamination, with the objective of understanding the reservoir fluid properties to address
issues such as completion and topside designs, production forecasts, and flow assurance predictions. Similar to Case 1, this
job was run using two single-probe modules in the tool string (Fig. 6). A large diameter probe was run for pressure pre-tests
(tool zero reference in Fig. 6) and an extra large diameter probe was run for formation fluid pump out and sampling. In this
job for performing the downhole fluid analysis stations, a combination of optical and in-situ density/viscosity sensors was
run upstream and downstream of the upper pumpout module, plus an optical sensor upstream of the lower pumpout module.
The primary logging objective was to acquire PVT quality reservoir fluid samples with oil base mud contamination
less than 5%. The latest technology optical spectroscopy sensors and density/viscosity vibrating rod sensors were used for
this purpose. Samples with contamination less than 5 % were collected which for the operator was excellent for the type of
oil and formation sampled.
Fig. 7 presents the pressure vs. depth data, together with the in-situ density and viscosity, and in-situ fluid analyzer
GOR evaluated in real time and the comparison with laboratory data. The matches were excellent.
Although multiple pretest stations were taken, only a single fluid sample was collected. A similar workflow as in
preceding example was employed to model the fluid and pressure gradients separately.
In order to infer the existence of barriers or, the alternative, the vertical continuity of fluids and sand, i.e. test the
reservoir architecture, both a pressure gradient and a fluid gradient model were calculated as per the methodology
summarized on Appendix 1. Following is presented the main results of the analysis:
6 SPE 115429

• After detailed check of each of the pretests, it was possible to fit a single fluid gradient to all the pressure data above
X945 m. The line defines a gradient equivalent to a 0.876 g/cm3 density as is shown in Track 1 of Fig. 8.
• The actual complexity of the hydrocarbon column is only revealed when the excess pressure plot is constructed
(center track in Fig. 8) and in more detail in Fig. 9. A gradation is observed along the oil column.
• The integration of pressure data with open-hole logs indicated a shaly zone under X945 m of low permeability,
which is reflected in the lack of effective pretests performed at the bottom of the oil zone. A gradual increase in the
free water saturation is also identified on the openhole logs.

Downloaded from http://onepetro.org/SPEATCE/proceedings-pdf/08ATCE/All-08ATCE/SPE-115429-MS/2732820/spe-115429-ms.pdf/1 by Schlumberger Oilfield UK Plc user on 14 February 2023

• The result of modeling the pressure gradients with polynomial fits to derive the fluid density at the pretests stations
is presented in track 3 of Fig. 8 and Fig. 9. Although the fluid grading is subtle here, the fluid density is still
projected to vary from 0.875 g/cm3 at the top of the oil zone to 0.915 g/cm3 at the bottom, over an interval of about
30 m. See Table 3 for further details, including the error bounds on the density.
• The fluid was then modeled using an EOS with the description provided by the lone fluid sample. For this case,
since the analysis was run after the preliminary laboratory test results, a basic fluid description was available,
including the fluid saturation pressure. This allowed to better tune the equation. Then an equilibrium gradient model
(a temperature gradient of 0.09 oF/ft) was used to derive fluid compositions at the other depths. The equilibrium
compositional gradient model was compared to observed in-situ live fluid density / viscosity (corrected by OBM
contamination).
• Table 3 and track 3 in Fig. 8 and Fig. 9 show the derived fluid density results from the pressures, downhole fluid
measurements (in-situ density sensors), PVT and EOS. These results include the raw in-situ densities and the
corrected value by the OBM contamination (less than 5% OBM by OCM analysis).

There is a slight difference between pressure derived densities (polynomial method) and EOS derived as shown by the trend
deviation in Fig. 9. The pressure data densities (polynomial method) indicates a larger gradient than anticipated if everything
is in gravity- thermal equilibrium (EOS), which looks to be associated to free water starting to show up below X945 m and
biodegradation which tends to show higher densities toward the bottom of the reservoir as per the pressure readings. This
issue could have been resolved by taking additional downhole fluid stations down in the oil zone to settle this uncertainty. A
change of viscosity by a factor close to 2 between top and bottom is identified by the EOS (Table 4), which will impact in
the productivity index of this reservoir. Geochemical data available from this well clearly confirms the presence of heavy
biodegradation in this reservoir.

Conclusions
A novel methodology for compositional gradient analysis is proposed based on rigorous analysis of pressure gradients and
downhole fluid analysis measurements. The choice and appropriateness of the pressure gradient model that accounts for the
observed change in fluid composition through depth is also tested. The resulting gradient values are corroborated with the
distributed insitu fluid density measurements. This approach allows identifying bias in the reservoir characterization in the
early exploration stages and maximizes the use of PVT data analysis for better fluid columns analysis to greatly reduce risks
of large failures in the production of oil associated, on one hand with the lack of fluid reservoir characterization, and on the
other hand with the lack of vertical communication.
A fluid equilibrium model may be fitted to the observed downhole fluid data, justifying the fluid connectivity in the
interval. The reservoir architecture and flow barrier locations may be found on the basis of anomalies encountered while
logging, with respect to modeled fluid compositions.
We have presented two field examples in which compositional gradients were observed with downhole fluid analysis
measurements and pressure gradients. The cases cover a volatile hydrocarbon case and a heavy oil case. Fluid equilibrium
models were fitted to the observed data, and pressure gradients were modelled with polynomial fits to derive fluid densities.
In general, the agreement has been very good, and the results have been corroborated with lab PVT and geochemical data.

Acknowledgements
The authors wish to thank Shell E&P Inc. and Schlumberger for permission to publish results presented in the different case
studies. The authors wish to thank as well all the colleagues who have contributed to this study. We especially gratefully
acknowledge the work of our colleague Sophie Godefroy.

References
Brown, A.: “Improved Interpretation of Wireline Pressure Data”. AAPG Bulletin, 87, no. 2, February 2003: 295-311.

Collins, C., M. Proett, B. Storm, G. Ugueto.: “An Integrated Approach to Reservoir Connectivity and Fluid Contact
Estimates by Applying Statistical Analysis Methods to Pressure Gradients”, paper HH, presented at the SPWLA 48th Annual
Logging Symposium, June 3-6, 2007.
SPE 115429 7
SPE 115429 7

Creek, J.L. and Schrader, M.L.: “East Painter Reservoir: An Example of a Compositional Gradient from a Gravitational
Field”, paper SPE 14411 presented at the ATCE, Las Vegas, USA, September, 1985.

Dubost, F., Carnegie, A., Mullins, O, O’Keefe, M., Betancourt, S., Zuo, J., Eriksen, K.: “Integration of In-Situ Fluid
Measurements for Pressure Gradients Calculations”, paper SPE 108494 presented at the Intl Conference and Exhibition,
Veracruz Mexico, June 2007.

Downloaded from http://onepetro.org/SPEATCE/proceedings-pdf/08ATCE/All-08ATCE/SPE-115429-MS/2732820/spe-115429-ms.pdf/1 by Schlumberger Oilfield UK Plc user on 14 February 2023

Elshahawi, H., Fathy, K., and Hiekal, S., "Capillary Pressure and Rock Wettability Effects on Wireline Formation Tester
Measurements", paper SPE 56712, presented at the SPE Annual Technical Conference and Exhibition, Houston, Texas, 3-6
October 1999.

Elshahawi, H., Hashem, M., McKinney, D., Ardila, M., Ayan, C.,”The Power of Real-Time Monitoring and Interpretation in
Wireline Formation Testing Case Studies”, SPE 94708, presented at the ATCE, Dallas, USA, October 9-12, 2005.

Elshahawi H., Venkataramanan, L., McKinney, D., Flannery, Mullins, O., Hashem, M.: “Combining Continuous Fluid
Typing, Wireline Formation Testers, and Geochemical Measurements for an Improved Understanding of Reservoir
Architecture”, SPE 10040, 2006.

Firoozabadi A.: “Thermodynamics of Hydrocarbon Reservoirs”, Mc Graw Hill, 1999.

Fujisawa, Go., Betancourt, S., Mullins, O., Torgersen, T., O’Keefe, M., Terabayashi, T., Dong, C., Eriksen K.: “Large
Hydrocarbon Compositional Gradient Revealed by In-Situ Optical Spectroscopy”, paper SPE 89704 presented at the ATCE,
Houston, USA, September, 2004.

Ghorayeb, K., Firoozabadi, A., Anraku, T.: “Interpretation of the Unusual Fluid Distribution in the Yufutsu Gas-Condensate
Field”, paper SPE 84953 presented at the Asia Pacific Conference of Integrated Modeling for Asset Management,
Yokohama, Japan, April 25-26, 2000.

Godefroy, S., Zuo, J., Fujisawa, G., O’keefe, M., Ardila, M., Cañas, J., De Santo, L., Koksal, C.:”Discussion on Formation
Fluid Density Measurements and Their Applications”, SPE 114648 presented at the 2008 SPE ATCE in Denver, USA,
September 2008.

Hanafy, H.H., Mahgoub, I.S.: “Methodology of Investigating the Compositional Gradient within the Hydrocarbon Column”,
SPE 95760 presented at the 2005 SPE ATCE in Dallas, USA, September 2005.

Hirschberg, A.: “Role of Asphaltenes in Compositional Grading of a Reservoir’s Fluid Column”, JPT, pg. 88-94, Jan.1988
Hunt, J.M.: “Petroleum Geochemistry and Geology”, W.H.Freeman and Co., San Francisco, 1979.

Kabir, C. and Pop, J.:” How Reliable is Fluid Gradient in Gas/Condensate Reservoirs?” SPE 99386, 2006.

Lars Høier, SPE, Statoil and Curtis H. Whitson.: “Compositional Grading – Theory and Practice”, SPE 63085, NTNU, 2000
Lira-Galeana, C.: “Discussion of Treatment of Variations of Composition with Depth in Gas-Condensate Reservoirs”.
SPERE 7(1): 158, 1992.

McKinney, D., Flannery, M., Elshahawi, H., Stankiewicz, A, Clarke, E., and Breviere, J., “Advanced Mud Gas Logging in
Combination with Wireline Formation Testing and Geochemical Fingerprinting for an Improved Understanding of Reservoir
Architecture”, Paper SPE 109861 presented at the SPE Annual Technical Conference and Exhibition, Anaheim, California,
11–14 November 2007.

Montel, F., Bickert, J., Caillet, G., Le Goff. C.: “Modeling the effect of external gas flux on reservoir fluid description”,
paper SPE 77383 presented at the ATCE, San Antonio, USA, 2002.

Mullins, O.C., Schroer, J.: “Real-time determination of filtrate contamination during openhole wireline sampling by optical
spectroscopy,” paper SPE 63071 presented at the ATCE, Dallas, Texas, October 1–4, 2000.

Mullins, O., Daigle, T., Crowell, C., Groenzin, H., Joshi, N.: “Gas-Oil Ratio of Live Crude Oils Determined by Near-Infrared
Spectroscopy”, Applied Spectroscopy, 55, 197, 2001.

Mullins, O., Hashem, M., Elshahawi, H, Fujisawa, G., Dong, C., Betancourt, S. and Terabayashi, T.: “Hydrocarbon
Compositional Analysis In-Situ in OpenHole Wireline Logging”, SPWLA 45th Annual Logging Symposium, Jun. 6-9, 2004.
8 SPE 115429

Mullins, O.C., Elshahawi, H., Hashem, M., Fujisawa, Go.: “Identification of Vertical Compartmentalization and
Compositional Variation by Downhole Fluid Analysis; Towards a Continuous Downhole Fluid Log”, paper presented at the
SPWLA 46th Annual Logging Symposium, June 26-29,2005.
Mullins, O.C., S. S. Betancourt, M. E, Cribbs, F.X. Dubost, J.L. Creek, A. B. Andrews, and L.Venkataramanan: “The
Colloidal Structure of Crude Oil and the Structure of Oil Reservoirs. Energy & Fuels, 2007.

Downloaded from http://onepetro.org/SPEATCE/proceedings-pdf/08ATCE/All-08ATCE/SPE-115429-MS/2732820/spe-115429-ms.pdf/1 by Schlumberger Oilfield UK Plc user on 14 February 2023

Mullins, O., Betancourt, S., Dubost, F., Cribbs. M.: “Asphaltene Gradient in a Deepwater Oil Reservoir as Determined by
Downhole Fluid Analysis”, SPE 106375, 2007.

O’Keefe, M., Eriksen, K. O., Williams, S., Stensland, D., and Vasques, R., “Focused Sampling of Reservoir Fluids Achieves
Undetectable Levels of Contamination”, paper SPE 101084 presented at the SPE Asia Pacific Oil & Gas Conference and
Exhibition, Adelaide, Australia, 11-13 September 2006.

O'Keefe, M., Godefroy, S., Vasques, R., Agenes, A., Weinheber, P., Jackson, R., Ardila, M., Wichers, W., Daungkaew, S.,
De Santo, I.: "In-situ Density and Viscosity Measured by Wireline Formation Testers", paper SPE 110364, presented at the
SPE Asia Pacific Oil & Gas Conference and Exhibition held in Jakarta, Indonesia, 30 Oct-1Nov 2007.

Raghuraman, B., O’Keefe, M., Eriksen, K.O., Tau, L.A., Vikane, O., Gustavson, G., Indo, K.: “Real-Time Downhole pH
Measurement Using Optical Spectroscopy”, SPE Reservoir Evaluation & Engineering, Volume 10, Number 3, June 2007.

Reddy, M.C., Nelson, R.K., Sylva, S.P.,Xu, Li.,Peacock, E.A., Raghuraman, B., Mullins, O.: “Identification and
quantification of alkene-based drilling fluids in crude oils by comprehensive two-dimensional gas chromatography with
flame ionization detection”, Journal of Chromatography A, Elsevier, 2007.

Schlumberger: “Application Document – Repeat Formation Tester”, 1984.

Schlumberger: "Fundamental of Formation Testing" 06-FE-014, 2007.

Stewart, G and L Ayestaran: “The Interpretation of Vertical Pressure Gradients Measured at Observation Wells”, paper SPE
11132 prepared for presentation at the SPE Fall Technical Conference and Exhibition, New Orleans, LA, Sept. 26–29, 1982.

Tissot, B.P. and Welte, D.H.: “Petroleum Formation and Occurrence”, Springer-Verlag, Berlin, 1978.

Venkataramanan, L., Elshahawi, H., McKinney, D., Flannery, M., Hashem, M., Mullins, O.: “Downhole Fluid Analysis and
Fluid-Comparison Algorithm as Aid to Reservoir Characterization”, SPE 100937, 2006.

Wheaton, R.: “Treatment of Variations of Composition with Depth in Gas-Condensate Reservoirs’. SPERE 6 (2): 239–244,
1991.

Zar. J. H.: Biostatistiacal Analysis, 4th Ed., Prentice Hall, 1999

Zuo, J. Zhang, D., Dubost, F., Dong, C., Mullins, O. O’Keefe, M., Betancourt, S.: “EOS-Based Downhole Fluid
Characterization”, SPE 114702, 2007.
SPE 115429 9

Technology Strength Weakness

Pressure gradients Ø Can be performed with simple and quick
formation tester (for pressure only)
Ø Limits the risks of the tool getting stuck
Ø Not affected by sample or flowline
contamination
Ø Depends on pressure and depth accuracy
Ø Accuracy depends of number of pressure
points and bed thickness, with respect to
reservoir fluid
Ø Affected by borehole conditions

Downloaded from http://onepetro.org/SPEATCE/proceedings-pdf/08ATCE/All-08ATCE/SPE-115429-MS/2732820/spe-115429-ms.pdf/1 by Schlumberger Oilfield UK Plc user on 14 February 2023

Ø Larger survey than sampling, higher vertical Ø Can fail to positively identify
resolution compositional gradients
In-Situ fluid analysis Ø Downhole conditions Ø Affected by flowline conditions
Ø Does not especially require sample Ø Affected by contamination, however
Ø Accurate for both gas and oil correction methods are available
Ø Measurements that not necessarily depend
on correlations
Ø Allows to be able to separate phases
downhole
Ø Larger survey than sampling
Ø Real-time answers
Sampling for PVT Ø Potentially excellent measurement accuracy Ø Laboratory conditions
measurements Ø Sample quality and transfer
Ø Sample representativity
Ø Survey limited to the number of sample
Drillstem tests Ø Big volumes samples Ø Limited to a very few set of perforations
Ø Good chance to have low contaminated Ø Limited to thick reservoir to be able to
samples validate density or compositional
Ø High investigation radius variations vs. depth
Ø Surface samples are not always valid
Openhole logs Ø High frequency data Ø Influenced not only by fluids but the
Ø Common available reading early in the formation itself
reservoir life and at time-lapse sequence Ø Lack of vertical resolution for fluid
Ø Support GOC and WOC delineation identification
Ø Support gas column definition (neutron- Ø Limited investigation radius which make
density log separation) then prone to invasion influence

Table 1— Summary of the different measurement techniques for reservoir fluids

 Downloaded from http://onepetro.org/SPEATCE/proceedings-pdf/08ATCE/All-08ATCE/SPE-115429-MS/2732820/spe-115429-ms.pdf/1 by Schlumberger Oilfield UK Plc user on 14 February 2023
SPE 115429

Table 2—Field Case 1. Derived fluid density results from the pressures, downhole fluid analysis measurements (in-
situ density) and EOS
10
SPE 115429 11

Downloaded from http://onepetro.org/SPEATCE/proceedings-pdf/08ATCE/All-08ATCE/SPE-115429-MS/2732820/spe-115429-ms.pdf/1 by Schlumberger Oilfield UK Plc user on 14 February 2023

Table 3 - Field Case 2. Derived fluid density results from the pressures, downhole fluid analysis measurements (in-
situ density sensors), PVT and EOS

EOS
Depth EOS Pressure Temperature Density EOS Viscosity InSitu like Fractions (wt%)
(m) (Psia) (F) (g/cc) (cp) C1 C2-C5 C6+

X927.3 X332.1 137.4 0.877 24.318 3.262 0.324 96.325

X928 X333.0 137.5 0.878 24.555 3.256 0.324 96.332
X938.5 X346.3 138.4 0.886 28.356 3.155 0.317 96.441
X941.5 X350.1 138.7 0.888 29.535 3.127 0.315 96.471
X943 X352.0 138.8 0.889 30.14 3.113 0.314 96.486
X944.3 X353.6 138.9 0.890 30.674 3.102 0.313 96.499
X945 X354.5 139 0.891 30.965 3.095 0.313 96.506
X959 X372.3 140.3 0.901 37.318 2.972 0.305 96.640
X965 X380.0 140.8 0.905 40.374 2.921 0.301 96.695

Table 4 - Field Case 2. EOS derived viscosity versus depth

 Downloaded from http://onepetro.org/SPEATCE/proceedings-pdf/08ATCE/All-08ATCE/SPE-115429-MS/2732820/spe-115429-ms.pdf/1 by Schlumberger Oilfield UK Plc user on 14 February 2023
SPE 115429

Fig. 1 - Proposed Methodology for Compositional Gradient Evaluation

12
 Downloaded from http://onepetro.org/SPEATCE/proceedings-pdf/08ATCE/All-08ATCE/SPE-115429-MS/2732820/spe-115429-ms.pdf/1 by Schlumberger Oilfield UK Plc user on 14 February 2023
13

Fig. 2 - Field Case 1. Formation tester configuration with downhole fluid analysis capabilities
SPE 115429
 Downloaded from http://onepetro.org/SPEATCE/proceedings-pdf/08ATCE/All-08ATCE/SPE-115429-MS/2732820/spe-115429-ms.pdf/1 by Schlumberger Oilfield UK Plc user on 14 February 2023
SPE 115429

Fig. 3 - Field Case 1. Light fluid compositional gradient validated by pressures data and downhole fluid analysis
14
 SPE 115429 15

X360
HGR Praw -Pmodel Oil Gradient
Oil Gradient ~ 0.5012 gm/cc
Omitted Data
~ 0.5012 gm/cc Poly Model EOS Model
Omitted Data Selected Data Poly Model
X380 Selected Data in-situ Density
PVT Lab

Downloaded from http://onepetro.org/SPEATCE/proceedings-pdf/08ATCE/All-08ATCE/SPE-115429-MS/2732820/spe-115429-ms.pdf/1 by Schlumberger Oilfield UK Plc user on 14 February 2023

I

X400

II

X420

X440
Depth (m)

III

X460

X480

X500

X520

IV

X540
X800 X850 X900 X950 -1 0 1 2 3 0.5 0.52 0.54

Pressure (psi) Excess Pressure (psi) 3

Density (gm/cm )

Fig. 4 - Field Case 1. Derived fluid density results from the pressures, downhole fluid analysis measurements (in-situ
density) and EOS
 Downloaded from http://onepetro.org/SPEATCE/proceedings-pdf/08ATCE/All-08ATCE/SPE-115429-MS/2732820/spe-115429-ms.pdf/1 by Schlumberger Oilfield UK Plc user on 14 February 2023
SPE 115429

Fig. 5 - Field Case 1. Fluid properties compared to downhole fluid measurements

16
 Downloaded from http://onepetro.org/SPEATCE/proceedings-pdf/08ATCE/All-08ATCE/SPE-115429-MS/2732820/spe-115429-ms.pdf/1 by Schlumberger Oilfield UK Plc user on 14 February 2023
17

Fig. 6 - Field Case 2. Formation tester configuration with downhole fluid analysis capabilities
SPE 115429
 Downloaded from http://onepetro.org/SPEATCE/proceedings-pdf/08ATCE/All-08ATCE/SPE-115429-MS/2732820/spe-115429-ms.pdf/1 by Schlumberger Oilfield UK Plc user on 14 February 2023
SPE 115429

Fig. 7 - Field Case 2. Dense fluid compositional gradient validated by pressures data and downhole fluid analysis
18
SPE 115429 19

X920
Scaled GR Residual
Ref Gradient EOS Model
~ 0.8756 gm/cc Poly Model
Data Data
X925

Downloaded from http://onepetro.org/SPEATCE/proceedings-pdf/08ATCE/All-08ATCE/SPE-115429-MS/2732820/spe-115429-ms.pdf/1 by Schlumberger Oilfield UK Plc user on 14 February 2023

X930

X935

X940
Depth (m)

X945

X950

X955

X960

X965
Oil Dens
~ 0.876 gm/cc
Water Dens
~ 1.111 gm/cc
X970
EOS Model
Poly Model
in-situ Density
PVT Lab
X920
X300 X350 X400 X450 X500 -5 0 5 10 15 20 0.8 0.9 1 1.1 1.2

Pressure (psi) Excess Pressure (psi) 3

Density (gm/cm )

Fig. 8 - Field Case 2. Derived fluid density results from the pressures, downhole fluid analysis measurements (in-situ
density sensors), PVT and EOS. It shows the oil zone above X945 m and the water zone under X955m. Although the
pressure trends are subtle the fluid density is projected to vary from 0.875 gm/cc at the top of the oil zone to 0.90
gm/cc at the bottom, over a depth span of about 30 m. See the Table 3 for further details.
20 SPE 115429

X920
Scaled GR Residual Oil Dens
Ref Gradient EOS Model ~ 0.876 gm/cc
~ 0.8756 gm/cc Poly Model Water Dens
Data Data ~ 1.111 gm/cc
X925 EOS Model
Poly Model

Downloaded from http://onepetro.org/SPEATCE/proceedings-pdf/08ATCE/All-08ATCE/SPE-115429-MS/2732820/spe-115429-ms.pdf/1 by Schlumberger Oilfield UK Plc user on 14 February 2023

in-situ Density
PVT Lab

X930

X935

X940
Depth (m)

X945

X950

X955

X960

X965

X970
X300 X350 X400 X450 -0.5 0 0.5 1 0.88 0.9 0.92

Pressure (psi) Excess Pressure (psi) 3

Density (gm/cm )

Fig. 9 - Field Case 2. Details for the derived fluid density results from the pressures, downhole fluid analysis
measurements (in-situ density sensors), PVT and EOS in the oil zone
SPE 115429 21

Appendix A— Pressure Interpretation Methodology

Recently there has been renewed interest in using formation testers to determine the existence of and to characterize the
variation in fluid properties along a vertical column, primarily as a result of the ability of modern formation testers to acquire
multiple, substantially uncontaminated samples; it is not uncommon for upwards of a dozen samples to be acquired in a
single descent of the tool (Elshahawi, 2006, Venkataramanan, 2006). Moreover, with the greater confidence being displayed
in the ability to perform downhole analysis of fluid properties there is not always the need to acquire a sample at every station

Downloaded from http://onepetro.org/SPEATCE/proceedings-pdf/08ATCE/All-08ATCE/SPE-115429-MS/2732820/spe-115429-ms.pdf/1 by Schlumberger Oilfield UK Plc user on 14 February 2023

of interest; it may be sufficient to merely characterize the in-situ fluid by downhole measurements, sometimes called fluid
“scanning”, with the acquisition of physical samples being left to key stations. Such operations are becoming increasingly
more common and vertical fluid profiling is a becoming a reality (Dubost, 2007).
As a result of such profiling operations discontinuities in fluid properties, such as GOR and, more subtly, composition, have
led to conclusions concerning the vertical continuity, or lack of continuity, of fluid columns (Fujisawa, 2004, Mullins, 2007).
In some cases discontinuities in fluid properties have been accompanied by discontinuities in the pressure profile, in other
cases a clear discontinuity in the sample profile has been accompanied by an apparent continuous pressure profile and vice
versa (Mullins, 2005). But this is not the primary interest of this work.
Following the fundamental work performed by Ayestaran on accuracy in pressure gradient measurements, only a small
portion of which was presented in (Stewart, 1982), more recent work has extended this work and has provided a convenient
formula for the prediction of accuracy in fluid densities derived from pressure gradients when there are errors in both the
pressure and depth measurements and these errors are known and most importantly, the fluid density is constant with depth
(Kabir, 2007, Collins, 2007). Estimates are available for the error in the fluid contact under similar conditions when the
formation thickness is not too small and the gradients are truly linear.
Attention in this work, however, is focused on cases where there is evidence that the formation fluid is compositionally
graded, i.e. there is a continuous variation in the composition of the fluid which leads to a pressure gradient which no longer
is constant but increases with depth. Often this effect is very subtle and not at all obvious; fitting a straight line to the data in
these cases may appear entirely appropriate. As one might expect curved gradients are most often encountered when the
formations are thick and the fluid is close to being critical (Wheaton, 1991, Lira-Galeana, 1992). Other factors that may result
in a curved gradient, for example capillarity and convection, are not treated here.
We begin by illustrating how the potential plot (Schlumberger, 1984) or the “excess pressure” plot as it is now called (Brown,
2003), can be used to detect gradient curvature. Assuming that curved pressure gradients may be adequately represented by a
polynomial relation between pressure and depth (in particular, a polynomial of order 2 or 3) we illustrate a statistical test for
the inadequacy of the linear model as a fit to the data. In those cases where the polynomial proves to be the better model the
fluid density as a function of depth is derived and displayed at the test stations together with estimates of the error. It should
be noted that the approach of fitting a low-order polynomial to pretest pressure data in order to obtain point-wise estimates of
a changing fluid density with depth is no substitute for acquiring adequate data that would allow the construction of a
properly tuned EOS-based model.
Wherever possible, densities derived from pressure gradient should be compared to direct measurements. Recently a direct
in-tool measurement of the density of the sampled stream has been made available [in-situ Density (O’Keefe, 2007,
Godefroy, 2008)]. The comparison of densities derived from pressure gradients, an in tool in-situ Density and densities
derived from samples either through laboratory measurement or the construction of an EOS-based model based on sample
composition is particularly powerful since the approaches may be considered as essentially independent and affected by
different factors; for example, fluid densities derived from pressure gradients tend not to be influenced by mud filtrate
contamination but they do directly reflect the influence of temperature gradients and in-situ fluid movement. When all --
essentially three methods – are in agreement there should be a high degree of confidence in the results.
With the above as background the methodology adopted here can be summarized as follows:
a. If at all possible identify those curved pressure trends which arise from variations in the composition of the fluid with
depth, eliminating those which are most likely associated with fluid movement, capillarity, convection, permeability
heterogeneity/supercharging, etc.
b. If necessary identify zones within the hydrocarbon column, which may be associated with fluids of “one type”, typically,
gas, volatile oil, heavy oil etc. The variation in fluid properties with depth may be very subtle and not at all obvious on
the scale that pressure data is usually viewed. It is helpful to plot the data in the potential or excess pressure format
where the difference between the actual pressure data and the pressure obtained by a hypothetical fluid with a specified
constant density is presented. Subtle pressure trends or discontinuities in trends which otherwise could be overlooked
may be easily identified when the data is presented in this format.
c. Fit both a straight line to the data and a low-order polynomial, accounting for errors in both depth and pressure. It was
found convenient in this work to take the form of the polynomial used to represent the variation in pressure with depth
as:
22 SPE 115429

N
p( z ) = ∑ cα gα ( z )
α =0
N
= ∑ cα ( ( z − z ) / sz )
α

α =0

Downloaded from http://onepetro.org/SPEATCE/proceedings-pdf/08ATCE/All-08ATCE/SPE-115429-MS/2732820/spe-115429-ms.pdf/1 by Schlumberger Oilfield UK Plc user on 14 February 2023

where z denotes the station depths at which the pressure measurements were taken, z is the mean of those depths and
sz represents the standard deviation of the station depths, cα are the parameters of the model; the coefficients in the case
of polynomials, and N is the order of the polynomial.
d. Test which model is better justified by the data, a linear model or polynomial model. There are several statistical tests
that one might apply in order to aid in this decision. The straightest forward is one associated with a t-statistic (Zar,
1999). After having fit a polynomial of given order n, starting with a straight line, the significance of the coefficient of
the highest order term may be assessed by constructing the statistic
cM
t=
se(cM )
Where cM is the coefficient of the M-th (typically, the highest power) of z and se(cM ) is the standard error of that
coefficient. The null hypothesis, that the coefficient cM = 0 , is rejected if t > t β (2),( n − M −1) , where tβ (2),ν represents the
(two-sided) Student t-distribution at the (1- β ) significance level with ν degrees of freedom. Polynomials of
successively higher order are tested until, if the data warrants the procedure, two consecutive orders lead to acceptance of
the null hypothesis. It is often helpful to plot the difference between the predicted density from the polynomial against
the constant density obtained from the best-fit straight line in the form of a “residual” plot.
e. Once the appropriate order of the fitting polynomial has been established, the variation of density/ pressure gradient with
depth may be computed: in the case of the linear trend this is a constant, in the case of the polynomial it is computed as
the tangent to the curve defining the pressure trend with depth, that is,
1 N
α cα ( ( z − z ) / sz )
α −1
γ = dp / dz = ∑
s z α =1

f. If possible compute error bounds for the gradient/density. It may not be easy to compute the error bounds on density
directly so some form of numerical scheme may be required.
g. If available, compare the estimated fluid densities with those predicted by a PVT (EOS) simulator or point measurements
derived from optical spectrometry or direct measurement (in-situ Density). If a direct measurement is made on the
sampled stream a correction needs to be applied to the density for the contamination by mud filtrate of the sampled fluid
(Mullins, 2000, Reddy, 2007). If the mud filtrate density is known, such corrections may be applied during data
acquisition using known techniques (Venkataramanan, 2006) or after the sample has been analyzed in a laboratory. For
the first field case discussed in the main text the mud filtrate density was known and the sample contamination was
measured for each of the recovered samples. Assuming the mud filtrate to be completely miscible with the oil being
sampled the following formula was used to correct the downhole density measurement for the residual contamination –
which was small:

( )
ρo = ρ DVR −νρ f / (1 − ν )

Where ρo represents the uncontaminated density of the sampled oil, ρ DVR is the fluid stream density measured
downhole, ρ f is the mud filtrate density and ν is the volume fraction of mud filtrate,

h. Investigate alternate models and assess their relevance. Let

n
e( N ) = ∑ ( pk − p ( zk , N ) )
2

k =1
where e( N ) is the residual sum of squares for the N-th model, ie the sum of squared differences with the model
parameters having values which yield a minimum squared error, n represents the number of pressure data and the pk
are the pressure data values. Suppose that it is required to compare two models, the “full” model with N F + 1 parameters
and the “reduced” model with N R + 1 parameters, where N F > N R , for example two polynomials of order N F and N R .
SPE 115429 23

This may be done through constructing a statistic F:

e( N F ) − e( N R ) n − ( N F + 1)
F= ≥ F ( β (1), N F − N R , n − ( N F + 1))
e( N F ) NF − NR

In the above the expression to the right of the inequality represents the (single-sided) F-distribution with

Downloaded from http://onepetro.org/SPEATCE/proceedings-pdf/08ATCE/All-08ATCE/SPE-115429-MS/2732820/spe-115429-ms.pdf/1 by Schlumberger Oilfield UK Plc user on 14 February 2023

( N F − N R , n − ( N F + 1)) degrees of freedom at the 1 − β significance level. The more complex model is rejected if the F-
statistic is greater than expected value of the F-distribution.
i. Estimate the position of the gas-oil and water-oil contacts, if any and estimate the error in locating the contact.

APPENDIX B: The Effect of a Vertical Fluid Flux on Apparent Density from Pressure Gradients
The issue was previously raised of whether an active flux of fluid from zone I in Field Case 1 (Fig. 4) through the tightest
zone II might influence the pressure gradient and therefore the fluid density estimated from the pressure distribution at the
top of zone III. An order-of-magnitude analysis of the effect is given below.
Assuming linear vertical flow the fluid speed is given by

 K  ∂p ∂p
νz =  z  ≡ λII
 µ  II ∂z ∂z
Where λII is the mobility in zone II. By continuity the apparent change in density due to this linear flow is given by

1 µ  1 1 1 λII ∂p
∆ρ =   νz = νz =
g  K z  III g λIII g λIII ∂z

From the excess pressure plot, Fig 5, the pressure difference across zone II is about 0.5 psi.
Pretest
Depth Pressure
Mobility
(m) (Psia)
(mD/cP)
X428.0 X872.22 101.1
X424.0 X869.42 123.8 Top of main oil zone
X421.0 X868.53 230.2 (Zone III)
X420.0 X866.54 930.9
X418.0 X865.03 152.5
X416.0 X863.73 163.6
X414.0 X862.35 449.9 Transition zone
X409.5 X859.24 84.4 (Zone II)
X407.5 X857.80 7.6
X406.1 X856.87 0.77
X404.0 X855.41 1971 Top zone (Zone I)

Using a) the available high resolution logs (in particular, GR and NMR) to define individual layers; b) using the NMR-
derived permeabilities and scaling them with the pretest derived mobilities to obtain a continuous mobility distribution; and
c) assuming the layers to be isotropic with respect to mobility, the effective mobility (obtained through an harmonic average)
of Zone II in the above table is estimated to be about 2 mD/cP. Taking the thickness of the zone to be 11 m the pressure
gradient across the zone II is about 0.014 psi/ft. The effective mobility of the zone being fed by this fluid flux, Zone III in the
above table, is estimated to be approximately 200 mD/cP. The density deficit due to the flux of fluid is then calculated from
the last equality in the second equation to be less than 0.0005 gm/cc, which is less than the confidence interval on the density,
derived from fitting the pressure data.