Professional Documents
Culture Documents
Model Uncertainty. According to Zimmermann4, “certainty” probability distributions according to the known data8-10,
implies that a person [model] has quantitatively and Then, they generate different bodies for the geological model
qualitatively the appropriate information to describe, prescribe by Monte Carlo sampling from the distributions of the
or predict deterministically and numerically a system, its parameters. During the process, geologists can also put
behaviour or other phenomena. The situations which are not specific sedimentary body with certain parameters into the
described by this definition shall be called “uncertain.” The model and match their geological knowledge. The most
causes of uncertainty mainly include lack of information, popular model is the marked point processes. The small-scale
abundance of information (complexity), conflicting evidence, features are simulated mostly by the use of simulated
ambiguity, engineering measurement, and subjective belief. annealing.
From our experience, the major uncertainty in reservoir
modelling is the subjective geological interpretation of the The Major Difficulty. From the geological viewpoints, the
field. This strongly relates to the understanding of the role of reservoir geology in stochastic models is diminishing
reservoir geology. However, the commonly identified problem and is increasingly replaced by two-point statistics (e.g.
inputs in stochastic models are hardly related to reservoir variograms) used in pixel-based models that are far too simple
geology. Hence, there is a great need to re-examine the to parameterise complex geology, and are often
characteristics of the conceptual geological models, which are unrepresentative when the data are sparse, limited and biased.
known to be fully charged with valuable geological The object-based models present a more geologically intuitive
knowledge. concept, but it requires many parameters that are difficult to
interpret geologically and does not consider information
Objective. The objective of this paper is to firstly revisit the derived from sedimentary processes.
current practices of stochastic simulation. This is followed by From our experience, geological knowledge is extremely
a discussion of the conceptual geological models. Lastly, we important and often represents the major uncertainty
will provide a list of future research directions, showing how component in reservoir characterisation. Unfortunately, this is
we can make use of geological knowledge in stochastic not what the current practices of stochastic simulation are
simulation for improved predictions. trying to model. There also exists a general misunderstanding
of what “geological knowledge” really means. The next
Stochastic Simulation section will give a detailed discussion of conceptual
Stochastic simulation is a very fast and flexible approach to geological models and clarify the meaning of geological
generate reservoir models. The use of Monte Carlo methods is knowledge.
extremely convenient to simulate unknowable events. Many
are able to generate multiple conditional realisations for local Conceptual Geological Models
uncertainty analysis. With the increasing interest of stochastic Conceptual models have placed a central role in reservoir
modelling in reservoir characterisation, many stochastic modelling for many decades, even before the birth of modern
models have been developed in the past several years. They computers. Despite the boom of stochastic simulation in recent
can be classified into two broad types: pixel-based models and years, there are still many so-called “conservative geologists”
object-based models. In this section, we will briefly discuss who do not appreciate the value of stochastic simulation,
the characteristics of these models. partly because of the complex mathematics involved, but
mostly because of the lack of geological soundness in the
Pixel-based Models. Most of the pixel-based stochastic simulated models. We forecast that the merging of conceptual
methods are based on kriging or cokriging in a sequential models and stochastic simulation will become an important
manner. Examples are sequential Gaussian simulation, breakthrough in reservoir characterisation.
sequential indicator simulation and truncated Gaussian Construction of conceptual models is in fact a highly non-
simulation5-7. They are all able to obtain estimates of the linear and complex process, which is difficult to be described
necessary conditional distributions by using simple kriging or by precise mathematics. It involves a good understanding of
ordinary kriging. In order to incorporate soft data (e.g. seismic physical and chemical reactions in earth sciences, and depends
impedance), cokriging technique can be used during the heavily on the knowledge of the geologist(s) involved. The
sequential simulation processes. They can directly analyse the next section will expound the meaning of geological
quantitative data and utilise them for prediction purpose via knowledge.
the use of direct-variograms or cross-variograms. Apart from
the sequential simulation algorithms, simulated annealing is Geological Knowledge. The meaning of “knowledge” is
also flexible to incorporate more data types into the model by abstractive and difficult to be quantified11. It is a result of
adding new components to the objective function7. learning and understanding of a certain subject. In the
formalism of knowledge, there is a set of facts contained in the
Object-based Models. Object-based models present a knowledge. When the subject wants to speak about its
promising concept to integrate stochastic geological objects. knowledge, it can only use its concepts12.
They treat each geological body (e.g. a sand body) as an In geology, the subject knowledge can give rise to
object, and parameterise the body by constructing a series of conceptual models and is commonly shown as hand-drawings
SPE 54307 The Role of Geology in Stochastic Reservoir Modelling: The Future Trends 3
for technical communication. The knowledge (conceptual the conceptual model is mostly of a symbolic nature (e.g. sand
model) contains two main groups of information: distribution) rather than a numerical one (e.g. permeability).
Geological rules. Geological rules are the result of the Secondly, it represents only the global trends of
basic theories in petroleum geology. They include the heterogeneities and is not able to provide any fine-scale
scientific rules of structural geology, stratigraphy, features. The problems are worse if a full 3D model is being
sedimentology, paleontology, diagenesis, geochemistry and sought, as the traditional approach is based on 2D techniques
others13. We can call them “objective knowledge,” as they are (e.g. cross-section method) with limited interpolation choices.
generally agreeable among geologists. The spatial This is why stochastic models are useful as they have the
distributions of lithofacies and reservoir properties are ability to handle such problems efficiently. However, the
basically controlled by the geological rules for a given simple statistical measures and distribution functions should
sedimentary environment. never replace the contribution of the conceptual model, which
Geological experience. Geological experience from long- provides a geological framework of the reservoir. More
term geological practice is extremely valuable for reservoir research works are required to develop methodologies for
prediction, especially when there is insufficient reservoir data combining the usefulness of the conceptual models and the
to support the geological rules. As opposed to the geological efficient stochastic models.
rules, geological experience is generally considered as
“subjective knowledge,” as they are almost disagreeable Future Trends
among geologists, especially those with different research This final section outlines five major research directions
experience. It is therefore important for geologists to work in showing how we can make use of geological knowledge and
various types of environments and subsequently improve the conceptual geological models to bridge the gap between
confidence of their own intuition. Because of the existence of reservoir geology and stochastic simulation practices.
subjectivity in geological intuition, it becomes the major cause
of uncertainty in reservoir modelling. Formalisation of Geological Knowledge. A strong emphasis
on how the geological knowledge is evolved is required. The
Model Construction. The construction of conceptual model formalisation of the knowledge allows the development of a
involves a number of steps. The first step is to construct a flexible computer system for the generation of conceptual
structural model based on seismic, regional geology and well models with different scenarios. As the major uncertainty lies
data. The next step is to identify a depositional model. This on the subjective interpretation of the reservoir data, it is
identification is based on the regional geological information necessary to consider multiple or alternative geological
(tectonics and stratigraphy), global sea level eustatic interpretations15,16. This will be a dramatic improvement on
interpretation, seismic interpretation (structural and the current methodologies (e.g. the use of different random
stratigraphic), sequence stratigraphy, well logs, core seed) for simulating multiple, equally-probable realisations,
description, paleontological, geochemical and analog data. which are only indirectly related to reservoir geology.
Based on the depositional model, sedimentary facies model The potential research tool to understand and model
comprising reservoir and non-reservoir facies along with geological knowledge is artificial intelligence (AI). AI is a
reservoir geometry (external form) can be inferred. High specialised field of computer science concerned with concepts
resolution 3D seismic and statistical database regarding and methods of symbolic inference by computer and symbolic
reservoir geometry in different environments, if available, can knowledge representation for use in making inferences. AI can
also be used to derive the reservoir geometry. Orientation of be seen as an attempt to model aspects of human thought on
reservoirs can be inferred from 3D seismic (amplitude slices) computers. The particular AI technologies are rule-based
and dip meter log interpretations. expert systems (or knowledge-based systems) with a strong
As the original reservoir characteristics are greatly affected focus on the incorporation of fuzzy-based information. These
by diagenesis, the next important step is to envisage diagenetic are extremely useful in modelling uncertain, qualitative and
facies model. This model is constructed based on post- linguistic information, which contribute significantly to the
depositional, physio-chemical, biological and mechanical derivation of the conceptual geological models17.
processes, detailed core description, logging, petrophysical
and geochemical data. Based on the diagenetic facies model, Use of Possibility Theory in Reservoir Modelling. So far,
reservoir flow units and barriers can be identified. The final probability theories dominate the fundamentals in stochastic
model is a 3D conceptual geological model depicting simulation, especially in facies simulation. From the
structural picture, reservoir geometry, spatial distribution of geological viewpoints, however, some geological facies
flow units and barriers along with the qualitative knowledge cannot simply be described by probability. For example, the
regarding reservoir heterogeneities. It represents the diagenetic facies is generally identified in core plugs. There
geological knowledge and remains as an important tool for are a number of transitional types between the end facies
reservoir modelling. types, and the definition of target diagenetic facies types is
often imprecise, or fuzzy. Unlike simple depositional facies,
Model Limitations. There are a number of limitations in the which can be expressed by facies proportions in clastic
use of conceptual models for reservoir simulation14. Firstly, sedimentary sequences, the definition of diagenetic facies is
4 D. Tamhane, L. Wang and P. M. Wong SPE 54307
not as straightforward. Possibility theory is a prominent some criteria. There is a potential research area for
methodology to handle such fuzzy information16. constraining the upscaled model to the conceptual model. It is
There is a fundamental difference between probability and because the conceptual model represents the large-scale
possibility. Probability expresses a binary decision (yes or no), geological trends of the reservoir features. When larger grid
but possibility is for multi-valued decision. Consider an size is used, the upscaled model should reflect the same
attempt to describe a core plug using two minerals: type “A” geological trends. Again, we may develop rule-based systems
and type “B.” Very often, we describe that the plug looks to evaluate the performance of various upscaling techniques.
more like “A” than “B.” If we use probability to implement On the other hand, we may also use the conceptual model as
such a statement, the best we can do is to use a X% chance the starting point and then downscale it using stochastic
belonging to “A” and a (100-X)% chance belonging to “B,” simulation. This is similar to the ideas presented in Wang et al.
where X>50%. This in fact does not reflect the true meaning (1998)19, but further investigation is required.
of the statement. In the statement, the plug actually looks like
both types of minerals, but is certainly not composed of X% of Acknowledgments
“A” and (100-X)% of “B.” The authors would like to thank the Australian Petroleum
In possibility theory, we are able to assume the co- Cooperative Research Centre for funding the work published
existence of both types, and use X% to express the degree of in this paper.
membership belonging to “A” and Y% to express the degree
of membership belonging to “B,” where X>Y. The major References
characteristic of possibility theory is that (X+Y)% is not 1. Simlote, V.N. and Nikolayenko, S.A.: “Dealing with
100%. From this example, we can clearly see the failure of Uncertainty: Geostatistics and the New Era in Reservoir
probability in describing a core plug, which is visible to us. Studies,” paper SPE 39546 presented at the SPE Iindia Oil
The problem is worse if we use probability concepts to and Gas Conference and Exhibition, New Delhi, India
describe the underground reservoirs. Therefore, research in the (1998) 383-397.
use of possibility theory in stochastic models will offer a 2. Oreskes, N., Shrader-Frechette, K. and Belitz, K.:
completely new insight to reservoir modelling practices. “Verification, Validation, and Confirmation of Numerical
Models in the Earth Sciences,” Science (1994) 263, 641-
Quantification of Reservoir Geology. The most commonly 646.
available quantitative geological model is the iso-porosity 3. Foley, L., Ball, L., Hurst, A., Davis, J. and Blockley, D.:
contours which are often hand-sketched by geologists. As “Fuzziness, incompleteness and randomness: classification
discussed previously, the incorporation of as many inputs as of uncertainty in reservoir appraisal,” Petroleum
possible into the modelling algorithm can reduce the Geoscience (1997) 3, 203-209.
uncertainty of the model. There is no reason why geological 4. Zimmermann, H.-J.: “A fresh perspective on uncertainty
hand drawings cannot be used. Although they represent only modeling: Uncertainty vs. Uncertainty modeling,”
global or large-scale features, they do contain valuable Uncertainty Analysis in Engineering and Sciences: Fuzzy
geological knowledge18. If stochastic simulation is used, the Logic, Statistics, and Neural Network Approach, B.M.
small-scale features can be simulated, and the final model can Ayyub and M.M. Gupta (eds), Kluwer Academic
be constrained by the large-scale features. Techniques such as Publishers, Massachusetts, USA (1998) 353-363.
neural network residual simulation are particular useful19. As 5. Goovarts, P.: “Comparison of CoIK, IK, mIK
the quantitative information is so useful, it is a great challenge performances for modelling conditional probabilities of
to provide more quantitative maps. Methods to convert the categorical variables,” Geostatistics for the next century,
conceptual models into quantitative hand drawings (e.g. M. Armstrong and P.A. Dowd (eds), Kluwer Academic
expressing the sedimentary facies model and the diagenetic Publishers, Netherlands (1994) 18-29.
model) will contribute to many stochastic models. 6. Gotway, C. A. and Rutherford, B. M.: “Stochastic
simulation for imaging spatial uncxertainty: Comparison
Rule-based Ranking of Multiple Realisations. Ranking and evaluation of available algorithms,” Geostatistics for
multiple realisations from stochastic models is important. the next century, M. Armstrong and P.A. Dowd (eds),
Many criteria can be used to choose the “best” image20. If Kluwer Academic Publishers, Netherlands (1994) 1-21.
geological knowledge could be converted into rules, we may 7. Deutsch, C. V. and Journel, A. G.: GSLIB: Geostatistical
develop rule-based systems to rank the multiple realisations software library and user’s guide, Oxford University
and examine if any of the realisations are geologically Press, New York (1992).
interpretable or worth further investigation. This is strongly 8. Lia, O., Tjielmeland, H. and Kjellesvik, L. E.: “Modelling
dependent on the success of the formalisation of geological of facies architecture by marked point processes,”
knowledge discussed previously. Geostatistics Wollongong ’96 Volume 1, E.Y. Baafi and
N.A. Schofield (eds), Kluwer Academic Publishers,
Rule-based Reservoir Upscaling. Reservoir upscaling is a Netherlands, (1997) 386-397.
challenging field. It aims to convert fine-scale reservoir model 9. Syversveen, A. R. and Omre, H.: “Marked point models
to a coarser one, but to retain some characteristics according to for facies units conditioned on well data,” Geostatistics
SPE 54307 The Role of Geology in Stochastic Reservoir Modelling: The Future Trends 5