SPE 68703

Experience with the Quantification of Subsurface Uncertainties

T. Charles, J.M. Guéméné, B. Corre, G. Vincent, O. Dubrule, TotalFinaElf

Copyright 2001, Society of Petroleum Engineers Inc.

- rather than defining three «mini, median and maxi »
This paper was prepared for presentation at the SPE Asia Pacific Oil and Gas Conference and geological models, we prefer to define these models in relation
Exhibition held in Jakarta, Indonesia, 17–19 April 2001.
with quantitative parameters of interest, such as OIP and
This paper was selected for presentation by an SPE Program Committee following review of
information contained in an abstract submitted by the author(s). Contents of the paper, as
Reserves.
presented, have not been reviewed by the Society of Petroleum Engineers and are subject to We also stress that all our approach merely consists of
correction by the author(s). The material, as presented, does not necessarily reflect any
position of the Society of Petroleum Engineers, its officers, or members. Papers presented at translating uncertainties on input parameters into uncertainties
SPE meetings are subject to publication review by Editorial Committees of the Society of on parameters of economic interest. The quality of the result
Petroleum Engineers. Electronic reproduction, distribution, or storage of any part of this paper
for commercial purposes without the written consent of the Society of Petroleum Engineers is obtained depends on the validity of quantified uncertainties on
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 input parameters.
acknowledgment of where and by whom the paper was presented. Write Librarian, SPE, P.O.
Box 833836, Richardson, TX 75083-3836, U.S.A., fax 01-972-952-9435.
Introduction
Reservoir modelling is a challenging task for two main
Abstract reasons :
Now that tools are available for building 3D earth models, and - the complexity of the physics involved in predicting flow
for quantifying uncertainties on parameters associated with the - the lack of data available for modelling the reservoir
earth model, it is much easier to understand the impact of all This data shortage is mainly due to reservoir complexity and
subsurface uncertainties on field production. In the last years, to the high cost of data acquisition. Therefore, the prediction
we have accumulated a significant experience and know-how of oil-in-place, production profiles, or ultimate recovery, are
addressing this sort of issues in integrated studies. difficult exercice, and any forecast figure is uncertain. A
We briefly recall how we address the problems, what tools and systematic quantification of technical risks is crucial for
methodologies are used for quantifying the joint impact of decision makers to be able to compare or rank possible
subsurface uncertainties. We discuss the key points in a study : projects of a global portfolio, or understand the risk level they
1. clearly stressing which operational decision will be are taking on any new developement1,2. A quantification of the
supported by the uncertainty study. impact of various subsurface uncertainties on economic
2. reformulating the problem in terms of an uncertainty figures may also help justify the acquisition of further data, in
study workflow. order to reduce the uncertainty before major decisions are
3. listing the key subsurface uncertainties impacting the made.
study. A chain of tools has been developped at TotalFinaElf to deal
4. translating these key uncertainties into uncertainties on with the quantification of subsurface uncertainties. This chain
gross-rock volume, oil-in-place or reserves, using the uses as its backbone three main softwares, each dealing with
workflow. specific subsurface uncertainties3 :
Then we illustrate the approach with four field cases. The first ALEA : in-house software developed in the GOCAD4
example highlights how the joint impact of geometrical, environment, which allows the quantification of the joint
geological and dynamic uncertainties on reserves estimates impact of seismic-picking, time-to-depth conversion
can be quantified. The second example focusses on the impact uncertainties and any other uncertainty in seismic parameters
of fault sealing on a development decision for an HP-HT field. on Gross Rock Volume uncertainties5,6. Based on uncertainty
The third example shows how uncertainty figures can be used maps produced by seismic interpreteters, ALEA simulates a
to help optimally locate an appraisal well. The last example large number of structural models of the reservoir, calculates
illustrates how the quantification of structural uncertainties the corresponding Rock Volumes and exports these surfaces
can be used to justify the acquisition or processing of seismic towards JACTATM .
data. We also discuss how new uncertainty quantifications JACTATM : commercial software initially developed for
techniques are impacting some of the standard industry TotalFinaElf in the GOCAD environment. Once geostatistical
workflows. For instance : parameters affecting reservoir properties have been identified,
- scenario approaches can now be run in conjunction with a large number of possible geological reservoir models are
probabilistic approaches.
2 T. CHARLES, J.M. GUÉMÉNÉ, B. CORRE, G. VINCENT, O. DUBRULE SPE 68703

simulated and the corresponding volumetrics are calculated. - Support to an investment decision during the life of a
Structural uncertainties are incorporated, thanks to the results field (appraisal well location,…).
of the previous ALEA exercise. JACTATM uses a multi-stage The second step consists of defining what will be the crucial
simulation approach, combining depositional environments, parameter to use to support the decision at stake: will it be
rock-types and petrophysical parameters simulations. gross-rock volume, oil-in-place, production profiles, plateau
Uncertainties affecting geostatistical parameters, such as duration…
facies proportions, mean porosity, or variogram range, can The third step is the identification of all the subsurface
also be incorporated. Resulting realizations can be visualised uncertainties which can impact the parameters of interest
in 3D, analysed and exported to the flow simulator. defined in step 2: OWC uncertainty, seismic picks, net/gross,
EST : in-house software for experimental design permeability distribution, fault transmissibility… This
management3,7. A large number of JACTATM geostatistical brainstorming phase is absolutely crucial, since neglecting one
realisations, typically hundreds of them, may be available after major uncertainty may have a severe impact on the economics
the previous stage. Flow simulation can be performed on each of a new development.
realization, but this may prove very costly, and still A frequently-encountered obstacle to an uncertainty study lies
incomplete, since only static (structural and geological) in the habits and culture of engineers who have been trained to
uncertainties are accounted for in ALEA and JACTATM. The defend one single technical choice. The main goal of these
impact of dynamic uncertainties (relative permeability, fault initial discussions is to call the technicians mental scheme into
transmissivity or any flow parameter…) must also be question and to help them to position their basic scenario in
incorporated. The Experimental Design approach provides the global range of possibilities (see Field Case 1). Reviews of
tools and methods for incorporating dynamic uncertainties in project team assumptions may prove to be very instructive, by
the quantification of production profiles or reserves repositioning the base case model with relation to
uncertainty, whilst minimizing the number of flow simulation uncertainties: it is natural for a team to try to promote the
runs. EST functionalities are : choice of the experimental project it has been working on for months but it is essential for
designtype, parametrization and export of the corresponding decision makers to have all the elements to account for this
simulation runs to the flow simulator, post-processing and bias.
interpretation of the results. Once the key subsurface uncertainties have been identified,
Although the workflow is divided in three main steps, related and the workflow has been defined, the ALEA-JACTATM-
to structural, geological and dynamic uncertainties, the key to EST chain can be used to translate these subsurface
the whole approach is multi-disciplinary integration: the three uncertainties into the parameters of interest for the decision-
tools, if used jointly, allow the quantification of the joint making. Hence, the software chain appears to be merely a sort
impact of all the subsurface uncertainties in 3D. Questions of “transfer function” to translate input uncertainties into
such as: “what is the relative impact of structural uncertainties output uncertainties. Obviously, the quality of the results will
versus relative permeability uncertainties on reserves?”, or be highly dependent on the reliability of the input subsurface
“what is the interaction of internal permeability barriers versus uncertainties.
fault sealing on sweep efficiencies” can be addressed. The following field cases will develop examples of these
We will first discuss the value of uncertainty quantification in various applications.
the decision-making process. Then, through four operational
examples, we will discuss our experience with the Field Case 1: 3D integrated uncertainty study on
quantification of subsurface uncertainties in multi- reservoir
disciplinary field studies. To conclude, we will discuss how This first example illustrates the global methodology applied
new uncertainty quantifications techniques are impacting to a reservoir at appraisal stage. The goal of the study was to
some of the standard industry workflows. provide support to the development decision, by addressing
the following questions: what were the key uncertain
A framework to provide support to major investment parameters likely to affect reserves and production profiles
decisions and how robust was the development scheme vis a vis these
As for any deterministic study, the first step of the approach uncertainties. The methodology that was used can be split in 4
consists of a clear definition of the study’s objectives. A main steps :
global discussion of the subject often reveals, very early in
the study, different views among the main actors of the study. Step1 : Quantification of the distribution of Gross Rock
The operationnal impact of the studies we have performed can Volume (GRV) – Geometrical uncertainties
be classified in three main areas : The main uncertainties affecting the GRV were :
- Evaluation of the robustness of a development scheme - uncertainty in seismic time picking
(drilling planning, well architecture,…) with relation - uncertainty in time-to-depth conversion
to the identified uncertainties
- Identification and justification of complementary Picking uncertainty. On the basis of seismic quality, the
measurements for risk decreasing (core measurements, picking uncertainty at top reservoir was estimated at about
static pressure analysis..) 20m (after depth conversion) “far from the wells”. The
SPE 68703 EXPERIENCE WITH THE QUANTIFICATION OF SUBSURFACE UNCERTAINTIES 3

uncertainty on the flank position was treated through NTG - 3D geological model deformation to match the top
distributions, the one on the channel thickness was negligible map
because of the position of the OWC. - within each AE, simulation of facies and rock types
distributions taking into account the proportions
Time-to-depth conversion uncertainty. Analysis of the match assumptions with Sequential Indicator Simulation8
between seismic and well depths and of the flat spots showed (SIS).
a spatial trend (East-West) with important variations whilst the - simulation of the petrophysics within the Rock Types
evolution along the channel axis was far more continuous. with Sequential Gaussian Simulation8 (SGS)
Thus the vertical depth uncertainty due to the time-to-depth - random draw of the fluid contacts depths within the
conversion was from 0m to 40m. possible distribution
- oil-in-place volume calculation
Combining uncertainties. The uncertainties on depth
conversion and picking were combined by adding the Results.At the end of the JACTATM run, a distribution of
variances (independence hypothesis). Then channels top maps OOIP for each AE was available for each scenario.
simulations, using probability field techniques8, with an - Fig. 1 clearly shows that uncertainties affecting OOIP
anisotropic variogram (correlation length of 4000m for the were highly AE dependent: on AE E, the level of
main axis and of 2000m in the other direction) were relative uncertainty was highest, but for a lower
performed. Two hundred possible top maps were generated absolute value of the volume than on AE A. AE A, D,
and exported to JACTATM . and F appeared to be the most sensitive in terms of
oil-in-place volume.
Step 2 : Quantification of the distribution of OOIP and - Fig. 2 positions the base case accumulation in the
OGIP – Combination of Geometrical and Geological global distribution. Obviously, the picture chosen for
uncertainties the development scheme definition was optimistic.
This happens quite often in operational studies. A
Input data. The simulation of petrophysical properties was systematic quantification of individual uncertainties
guided by the seismic mapping of “Architectural Elements often shows that the so-called base case was
(AE)”. An AE is a volume made of several lithologies whose constructed using optimistic assumptions. Note that
internal organisation is known. This 3D texture is closely the opposite may happen.
related to a sedimentological concept validated on core data
and to a stratigraphical interpretation of the seismic data.. In Step 3 : Selection of dynamic models
the study six different AE were described (Matrix, levee, Step 3 consisted of selecting a subset of the one thousand (five
channel, spill phase, …). These AE were populated with one times two-hundred) models that had been built, in order to :
to two facies (good or bad), each facies being itself filled with - define a relationship between OOIP and recoverable
one to two rock types (nested simulations). On two reserves which characterises the impact of previously
architectural elements, the facies proportions were considered incorporated static uncertainties (geometry, geology
as uncertain parameters. Thus, five different scenarios with and petrophysics) on production.
five different hypothesis of proportion were run. - extract models to be used in the integration of
For each of the facies, net-to-gross, Vshale and porosity uncertainties associated with dynamic parameters.
distributions were defined, based on core data, well logs and Several methods can be applied to sample the models to be
analogs. Generally, triangular probability distributions were extracted. In this study, the criteria used were :
used and an uncertainty on their mean value was introduced to - regular sampling of the OOIP distribution
fully cover the range of uncertainty defined by the geologist. - representative sampling of the OOIP distributions of
Permeabilities and initial water saturations were derived from the main AEs
Vshale values using deterministic relationships. - same number of models for all the scenarios
As far as fluid contacts were concerned, five disconnected The fifty-three models resulting from this selection were flow-
hydrodynamic zones were considered, to account for the simulated using the development scheme considered as the
different flat spots seen on seismic and the OWC encountered most likely at the moment (Fig. 3) The correlation plot clearly
at the wells. shows that, for a fixed oil-in-place value, a scatter may affect
the corresponding reserves even if in this case, the correlation
Simulations. For each of the five scenarios, it was found that is good.
two-hundred realisations were enough to stabilise the resulting Note that this “STOOIP vs Reserves” plot of Fig. 3 can be
distribution of OOIP output. This means that, after two used to define and select a “Minimum, Median and
hundred simulations, the quantiles of the OOIP histogram Maximum” realization out of the fifty-three that were
derived from individual realisations do not vary anymore. For simulated. For instance, if we decide that the “Minimum”
each scenario, the workflow used to generate multiple realization is that corresponding to a q10 on both the oil-in-
realisations was: place and the reserves figures, we clearly provide an objective
- import of a top map realization from ALEA numerical criterion for such a choice. The same principle was
4 T. CHARLES, J.M. GUÉMÉNÉ, B. CORRE, G. VINCENT, O. DUBRULE SPE 68703

used to select two other models : a “Maximum” (q90) and Note the crucial impact of faults transmissivities. Since the
“Median” (q50) realizations. Due to the good correlation development scheme was essentially based on the location of
between OOIP and reserves, 3 models were sufficiant to the main fault network, it is logical to observe that the impact
correctly sample the point cloud. However, the less correlated of secondary faults was more important. Note also that the
OOIP and reserves are, the more numerous models are uncertainty domain associated with the PVT data was narrow
required to obtain a good sampling.The three selected enough to imply that additional measurements would not have
realizations were used for the following quantification of the had much impact on production forecast uncertainty. On the
impact of the dynamic uncertainties. other hand, Fig. 4 provides a direct quantification, in the
response unit, of the likely impact of extra Kr measurements.
Step 4 : Distribution of reserves and statistical production These figures enable economics calculations balancing cost
profiles – Dynamic uncertainties and impact of such a program. Fig. 4 also shows that the
This part aims at the quantification of the impact of the impact of uncertain parameters may change with time (see
dynamic uncertainties on the response. Nine uncertain parameter d). This is important to take into account, since
dynamic parameters had been identified : economic calculations are more sensitive to early production
1. main faults sealing capacity times. The response surfaces allowed the final statistical
2. secondary faults sealing capacity cumulative production profiles calculation and the calculation
3. possible permeability barrier between the channels of the plateau duration pdf based on the five previous listed
within one AE parameters (Fig. 5).
4. absolute permeability The integration of the dynamic uncertainties significantly
5. water/oil and gas/oil relative permeabilities (shapes of decreased the overall reserves on the field.
curves)
6. skin of producers and injectors Conclusion
7. PVT At the end of the study, it appeared that the initial well
8. specific transmissivity reduction for low NTG regions architecture was highly sensitive to the main uncertainties.
9. existence of a possible barrier layer However, this study was totally part of the work of the project
Three different values (minimum, median and maximum) team and happened while the definition of the development
were assigned to each of them. The global integration of these scheme was at its early stage. Thus, the results allowed a
uncertainties was performed using experimental designs. further optimization of the development scheme making it
more robust regarding the subsurface uncertainties.
Principle of the workflow. The purpose of this paper is not to
present the methodology in details3. However, it is important, The initial aim of the study was to quantify the impact of
before focussing on the results, to stress that dynamic subsurface uncertainties on a fixed development scheme. The
uncertainties are addressed through two main stages: results led to a modification of the development wells
1. the screening phase, which leads to the hierachisation architecture and location. On a similar study performed on a
of the impact of the parameters and of their interactions more mature project, the quantification of the possible liquid
on the plateau duration and production profiles. productions led to a redimensioning of the FPSO.
2. a “response surface modeling” (an equation
approaching the response of the simulator) on the Field Case 2 : Impact of dynamic uncertainties on a
domain of uncertainty of the parameters selected in the drilling program
previous stage, allows the statistical estimation of
reserves and production forecasts. Context
Field Y is a satellite of a major field (Field X) currently under
Results. The screening phase showed that five parameters had development. Y and X are HP-HT fields and are separated by
significant impact on both plateau duration and production a main fault initially sealing and by a graben whose fluid
profiles. They are listed below with decreasing order of content is unknown. Field Y can be depleted because of the
influence and banalized for confidential reasons: production of Field X. This potential depletion of Field Y is
1. parameter b important since for technical reason, it is not possible to drill
2. secondary faults transmissivity any well after a depletion of 100 bars. We were asked the
3. Kr question of how long the development of field Y could be
4. main faults transmissivity postponed after the beginning of production of X. Two
5. parameter d alternatives development scenarios could be envisaged for Y:
immediate vertical drilling from a new platform or deviated
Fig. 4 represents the evolution vs. time of the impact of the drilling from the existing plateforms of field X. For this last
domain of uncertainty of each parameter on the production solution, 1.3 years of well certification were necessary for
profiles. The impact is computed as the variation of large deviation qualification.
recoverable reserves induced by the variation of each
parameter from its minimum to its maximum values.
SPE 68703 EXPERIENCE WITH THE QUANTIFICATION OF SUBSURFACE UNCERTAINTIES 5

Solution
No fine representation of the heterogeneities or of the structure Field Case 3 : Location of an appraisal well
was necessary to address the question asked, since we were
dealing with depletion modelling only. Thus, only dynamic Objective
uncertainties were taken into account. All the following work This study was performed on a reservoir encountered by two
was based on the flow simulation model already constructed wells, one crossing the whole reservoir in the oil phase, the
by the operator. The following workflow was used: other well encountering water only. A vertical interval of
- definition of the response studied : average depletion about 160m separates the oil-down-to from the water-up-to
in all the layers of field Y depths measured at these two wells. Hence, the OWC depth
- identification, with the reservoir engineer in charge, of was the major uncertainty impacting the Gross Rock Volume,
the key uncertain parameters. Nine were selected: and the question was asked of : “Where to locate an appraisal
- absolute permeability value of the separating well for finding the OWC depth?”.
graben and of its aquifer
- relative permeability (end points) Solution
- West aquifer size The problem was assumed to be only geometrical. It was
- porosity variation within North Aquifer decided to quantify, at each possible location of the appraisal
- main fault transmissivity well, what was the probability to encounter the OWC, and
- main fault offset decide on this basis what the location should be. The
- permeability streaks in the reservoir methodology used to handle geometrical uncertainties was as
- vertical anisotropie follows:
- fluid-fill in graben
- identification of the uncertainty range of the nine Selection of the uncertainties to be treated. All the
selected parameters uncertainties were listed and ranked, in order to select only
- quantification of probability density function (pdf) those with a significant impact on the result. The initial list of
associated with each parameter uncertainty was divided into six subsets:
- use of the experimental design methodology to - data : wells (position, deviation, log quality), seismic
identify the parameters whose impact on the depletion data (positioning, migration..)
were more significant. Only three parameters had an - seismic interpretation : well calibration, picking of the
impact on depletion : absolute permeability, fluid-fill horizons, fault position
of the separating graben and transmissivity of the - velocity model for depth migration
main fault. A response surface was fitted to the - correction at wells
simulation response as a function of the value of these - reservoir thickness
three parameters. - OWC
- Monte-carlo simulation using the parameters pdf’s and An estimation of the uncertainties affecting these parameters
the analytical response surface model, in order to led to keep only those with the most significant impact on the
provide probabilised depletion profiles. reservoir depth maps:
- stacking velocities
Conclusions - seismic picks
Fig. 6 shows two depletion profiles for different hypothesis: - depth-conversion velocities in V0+kZ (especially the
- triangular distribution for fault transmissivity compaction factor k).
- leaking fault (consistent with the inferred fluid - well corrections
migration path). - OWC
Due to the major impact of the fault transmissivity pdf, it was
important to give both results to the operator. In the first Quantification. All the above uncertainties were combined to
scenario, a 100 bar depletion occured after 1.7 years of obtain the global uncertainty (confidence interval) affecting
production of X, whilst in the second case this time decreased the depth of the reservoir. Then the following approach was
to 1.2 years. On the basis of the most pessimistic scenario, it used:
was decided to develop field Y from the existing platform of - simulation of five hundred depth maps of the top
field X and to immediately begin the process of well reservoir around the reference map and within the
certification. A risk of 5% (after 1.3 years of production of confidence interval. The Probability Field technique8
field X, Fig. 6) associated to this decision was accepted. was used with an anisotropic correlation length
honoring the variability of the reference data..
- among all the generated realisations, selection of the
top reservoir maps that were geologicaly plausible
(closed at the water-up-to)
- calculation of the maps of the reservoir base by
addition of the thickness map.
6 T. CHARLES, J.M. GUÉMÉNÉ, B. CORRE, G. VINCENT, O. DUBRULE SPE 68703

- sampling of the OWC from a triangular distribution, large range of possible values for the Gross Rock Volume. In
with a minimum equal to the oil-down-to, a maximum order to better estimate the potential of this field, it was clear
equal to the water-up-to and a mode defined from the that the two major uncertainties had to be reduced.
RFT data.
We built a probability map for the OWC to be inside the Conclusion for Field Case 4
reservoir : for the whole set of valid simulations we produced This work demonstrated that, even for a mature field with
maps of hydrocarbon column height Hu, then at each (x,y) numerous wells, the geometrical uncertainties can still have a
location we calculated the probability for Hu to be between 0 large impact on the Gross Rock Volume. We determined
and the reservoir thickness (when Hu is 0, the OWC is above which parameters had to be improved in order to reduce these
the top reservoir; when it is equal to the reservoir thickness, uncertainties and consequently we recommended the
the OWC is below the base reservoir). This map (Fig. 7) acquisition of a 3D seismic data to be processed using a depth
showed a wide area of high probabilitybut remains below migration rather than a classical time migration.
95%.
Conclusion
Conclusion The industry recognizes the need to attach uncertainty figures
The location of the appraisal well was defined in the highest to production forecasts9. The approach presented in this paper
probability area. The well was actually drilled with a risk of provides a framework to address this issue. The four
failure estimated to about 5%, and the WOC was encountered examples presented here have shown that the proposed
close to the WUT. This result confirmed that the depth of the approach could support operational decisions related to the:
OWC was the crucial uncertain parameter considering that the - robustness of a development scheme with relation to
OWC depth derived from the RFT interpretation was subsurface uncertainties
erroneous. - selection of optimal location for a appraisal well
- decision to acquire more data in order to reduce
Field Case 4 : Justification of complementary data uncertainty
acquisition Our approach also provides a formal framework thanks to
which a number of issues can be addressed:
Objective - ranking of uncertainties
The study focused on the impact of structural uncertainties on - integration of scenario and probabilistic approach
the Gross Rock Volume of a mature field. The depth maps at - selection of Minimum, Median and Maximum models
reservoir level were built from the interpretation of the on objective numerical criteria.
available 2D time migrated seismic data and from the fifty However, we must stress that all this approach merely consists
available wells. The objective of the study was to estimate the of translating uncertainties on input parameters into
possible range of Gross-Rock-Volume, in order to understand uncertainties on parameters of economic interest, such as
which parameters are most uncertain, to justify a possible gross-rock volume, oil-in-place, reserves, plateau duration or
seismic data acquisition and to define the most adapted production profiles. The quality of the result obtained depends
processing. on the validity of quantified uncertainties on input parameters.
There is a great need in the industry to train subsurface
Solution engineers, so that they can provide realistic uncertainty figures
After an exhaustive listing of the uncertain parameters and a associated with their determination of key subsurface
ranking of the most significant ones, we kept the following : parameters.
- velocity field used for seismic time migration, which
drastically impacted the lateral location of the flanks Acknowledgements
of the structure The authors would like to thank TotalFinaElf for permission to
- seismic picks publish this paper. They also wish to thank P. Thore and V. de
- kriging interpolation away from the 2D seismic data Feraudy for their significant contribution to the different field
and well values cases presented in this paper.
- vertical time-to-depth conversion (performed using a
function calculated from the wells)
- OWC
The combination of all these uncertainties resulted in a global
confidence interval around the depth of the top reservoir (Fig.
8). The uncertainty related to the time-to-depth conversion is
not presented because it was constant laterally. We can
observe that the most significant uncertainties are related to
the estimation by kriging and to the seismic migration. The
simulation of 200 possible depth maps around the base case
map and within the calculated confidence interval resulted in a
SPE 68703 EXPERIENCE WITH THE QUANTIFICATION OF SUBSURFACE UNCERTAINTIES 7

References

1. Tyler, K., Sansdalen, L. , Maeland, J.O., Aasen, J.O.,

Siring, F., and Barbieri, M., “Integrated Stochastic
Modelling in Reservoir Evaluation for Project Evaluation
and Risk Assessment,” paper SPE 36706 (1996).
2. Lia, O., Omre, H., Tjelmeland, H., Holden, L. and
Egeland, T., “Uncertainties in Reservoir Production
Forecasts,” AAPG Bulletin, V.81, No. 5, May 1997, p.
775-802 (1997).
3. Corre, B., Thore, P., de Feraudy, V. and Vincent,
“Integrated Uncertainty Assessment for Project
Evaluation and Risk Analysis,” paper SPE 65205 (2000).
4. Dubrule, O., Bombarde, S., Samson, Ph., Segonds, D. and
Wonham, J., “Reservoir Geology using 3-D Modelling
Tools,” paper SPE 38659 (1997).
5. Samson, P., Dubrule, O. and Euler, N., “Quantifying the
Impact of Structural Uncertainties on Gross-Rock Volume
Estimates,” paper SPE 35535 (1996).
6. Vincent, G., Corre, B., and Thore, P., “Managing
Structural Uncertainty in a Mature Field for Optimal Well
Placement,” SPE Res. Eval. & Eng. 2 (4), Aug. 99, p.
377-384.
7. Damsleth, E., Hage, A., and Volden, R.: “Maximum
Information at Minimum Cost: A North Sea Field
Development Study Using Experimental Design,” Journal
of Petroleum Technology, Dec. 92, p. 1350-135
8. Deutsch, C.V., and A.G. Journel, 1992, GSLIB:
Geostatistical Software Library and User's Guide, New
York, Oxford University press, 340 p.
9. Jonkman, R.M., Bos, C.F.M., Morgan, D.T.K., Spencer,
J.A., and Sondena, E., “Best Practices and Methods in
Hydrocarbon Resource estimation, Production and
Emissions Forecasting, Uncertainty Evaluation and
Decision Making,” paper SPE 65144 (2000).
8 T. CHARLES, J.M. GUÉMÉNÉ, B. CORRE, G. VINCENT, O. DUBRULE SPE 68703

OOIP DISTRIBUTIONS PER ARCHITECTURAL ELEMENTS

1.0 PARETO PLOT ON TOTAL FIELD PRODUCTION

1
0.9 MINIMUM
0.9
MEAN
0.8 0.8
MAXIMUM
0.7 Kr

IMPACT (NORMALIZED)
0.7 parameter a
0.6
parameter b
0.6 0.5
PVT
OOIP .

0.4
Main Faults
0.5
0.3 parameter c
0.2 parameter d
0.4
0.1 parameter e
0.3 0 Second. Faults

-0.1
0.2
-0.2
0.1 -0.3
0 2 4 6 8 10 12 14 16 18 TIM E (YEARS)
0.0
A B C D E F Fig. 4 : Field Case 1: Impact of dynamic parameters
N° AE uncertainty on reserves.
Fig. 1 : Field case 1: Uncertainty on individual Architectural
Elements OOIP.
STOOIP CUMULATIVE DENSITY FUNCTION PLATEAU DURATIONDISTRIBUTION
1.0

0.9 2500 1.0

0.8 0.9
CUMULATIVE PROBABILITY

0.7
2000 0.8
Base Case
0.6
0.7

CUMULATIVE PROBABILITY
NUMBER OF EVENTS

0.5
1500 0.6
0.4
0.5
0.3
1000 0.4
0.2
0.3
0.1
500 0.2
0.0
0.40 0.45 0.50 0.55 0.60 0.65 0.70 0.75 0.80 0.85 0.90 0.95 1.00
0.1
NORMALIZED STOOIP
0 0.0
Fig. 2 : Field Case 1: STOOIP distribution – The Base Case
2.00
2.06
2.11
2.17
2.22
2.28
2.33
2.39
2.44
2.50
2.56
2.61
2.67
2.72
2.78
2.83
2.89
2.94
3.00
volume appears to be on the high end of the distribution.
NORMALIZED TIME

STOOIP / RECOVERABLE RESERVES CORRELATION Fig. 5 : Field Case 1: Plateau duration distribution
1

0.9
NORMALIZED PRODUCTION

0.8
q90
0.7

0.6
q50
0.5

0.4 q10

0.3

0.2 q10 q50 q90

0.1

0
0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00
NORMALIZED STOOIP

Fig. 3 : Field Case 1: STOOIP vs Recoverable Reserves cross-

plot, based on the flow simulation of 53 JACTATM
realisations.
SPE 68703 EXPERIENCE WITH THE QUANTIFICATION OF SUBSURFACE UNCERTAINTIES 9

P ro p o s e d w e ll lo c a tio n
Triangular Law for Fault transmissivity

100 Bars
200 100%
95%
180
90%
160 80%
70%
DEPLETION (Bars)

140 60%

120 50%
40%
100 30%
20%
80
10%
60 0%

40
20
0
0 0.5 1 1.5 2 2.5 3 3.5 4
TIME (Years)

Fig. 6: Field Case 2: Depletion profiles for the two different

scenarios Fig 7 : Field Case 3: Probability map of encountering the
WOC within the reservoir
Leaking Fault

200
100 bars
180 100%
95%
160 90%
DEPLETION (Bars)

140 80%
70%
120 60%
50%
100
40%
80 30%
20%
60 10%
0%
40
20
0
0 0.5 1 1.5 2 2.5 3 3.5 4
TIME (Years)
10 T. CHARLES, J.M. GUÉMÉNÉ, B. CORRE, G. VINCENT, O. DUBRULE SPE 68703

PICKING + SPARCE DATA (2D) + TIME MIGRATION = GLOBAL UNCERTAINTY

α meter α - 12α
8α α
αm αm
< 4α
1.25α α - 20α
12α α
αm
1.5α α - 8α
4α α αm
> 20α
Fig. 8 : Field Case 4: Top reservoir global confidence interval