WS10 P04
A Chance Calculus for Play-Based Exploration C. Stabell* (Decision Resources & Designs AS)
Summary
This paper presents a rules‐based calculus for prospect chance estimation. A key idea is that estimates are based on a combined categorization of our knowledge level (DATA dimension) and our model of the geological context (MODEL dimension). We call it a chance calculus that combines data and judgement with simple rules. Earlier work (Milkov, 2014) is extended by using two‐level chance rules that distinguish between shared play chance estimates and conditional prospect chance estimates as well as by a second separate step that handles seismic anomalies. The chance calculus provides estimates that are consistent and cover all chance factors and all conventional exploration situations. Explicit MODEL categorization not only provides a transparent, unambiguous basis for the chance estimates, but also eliminates double risking between shared and conditional chance estimates. 79th EAGE Conference & Exhibition 2017 – Workshop Programme
Paris, France, 12-15 June 2017
Introduction
This paper presents a rules-based approach to prospect chance estimation. The key idea is that chance
estimates are based on a combined categorization of our knowledge level and of our model of the
geological context. We call it a chance calculus that combines data and judgement with simple rules.
The approach builds on Risk Tables (Milkov (2014), NPD (Brekke and Kalheim, 1996)). The earlier
work is extended by using two-level chance rules that separate between shared play chance estimates
and conditional prospect chance estimates as well as a second separate step that handles seismic
anomalies.
The Prospect Assessment Workflow
The overall assessment workflow can be described as a problem solving process (Figure 1).
Figure 1. Prospect assessment workflow
A key property of the assessment workflow is that the chance of success (COS) of the prospect is
estimated after the assessment of the success case volumes as the COS is the chance of obtaining the
minimum success case volumes or more.
Another key workflow property is separate assessment of shared play probabilities and conditional
prospect probabilities: the conditional probabilities of a chance factor are conditional on that the
corresponding play-level chance factor is adequate and are assessed accordingly.
Chance play element (CPE) maps are designed to capture spatial variation in the chance of adequacy
(COA). For example the overall “reservoir facies presence” area can be divided into elements
(Figure 2) that reflect that the distal elements are thinner than the proximal elements. The play COA
attached to each CPE is our estimate of the probability of having a reservoir thickness greater than the
prescribed success case minimum.
79th EAGE Conference & Exhibition 2017 – Workshop Programme
Paris, France, 12-15 June 2017
Figure 2. Illustrative CPE Map for P(Reservoir facies presence)
The prospect assessment workflow shown in Figure 1 does not consider seismic anomalies that can be
indicators of the presence of hydrocarbons. The established best practice procedure is to first assess
the COS ignoring the seismic anomaly information. In a second step, a procedure (most often
Bayesian) is used to generate an estimate of a modified COS.
A Two-level Chance Calculus
We propose a chance table (C-Tables) prospect risking calculus. The calculus implies an automated
generation of chance estimates and is designed to provide transparent, unbiased and consistent single
point estimates.
The calculus is based on categorizing the assessment context in terms of how we know about the
context (the DATA dimension) and what we know about the context (the MODEL dimension). The
play MODEL defines the elements of the shared geological context that control the adequacy of the
chance factors. The local MODEL, on the other hand, defines either how the regional processes are
attenuated (absent) in the target location, or how local geological processes might overwrite or
attenuate the regional, play-level geological processes. The conditional COA is the chance that the
regional processes are locally adequate. For example the play level COA of seal is controlled by top
seal being unconformable; locally we assess the COA that the regional processes is not overwritten by
local thinning of the seal.
Key principles for the C-Tables calculus are:
 There are separate C-Tables calculus for each chance factor. The generic format of the tables is
that each chance factor has a play-level component and a target-level component.
 MODEL and DATA categories are chosen to provide requisite discrimination: Not too fine, but
also not too coarse.
 Most of the MODEL dimensions apply to both the play and the target (C in Figure 3).
However, some MODEL dimensions apply only at the target-level (D & E).
79th EAGE Conference & Exhibition 2017 – Workshop Programme
Paris, France, 12-15 June 2017
Fig 3. General structure of C-Table for one chance factor
 There are separate DATA dimensions for the play level chance estimate (A) and for the
conditional target chance level (B). This reflects that play level chance estimates are based on
regional DATA while the conditional chance estimates are based on local, target specific
DATA.
 Where sufficient data is available, the estimates of chance factors reflect quantitative data
attributes.
 With limited data, the calculus follows the “Chance (of) Adequacy Matrix” precept (see Milkov,
2014) and posit a COA value that reflects a baseline estimate of adequacy probabilities
(typically in the 0.4 – 0.6 range).
 Chance estimates are adjusted relative to the difference between target and play minimum
success case volumes.
 Presence of seismic anomaly information is not included in the C-Tables. This data (when
present or when expected, but lacking) is considered in a second Bayesian Risk Modification
(BRM) step (see Stabell et al, 2003).
Multiple basic MODEL dimensions are often combined into composite categories. Categories are
lumped in order to produce 5 - 7 composite MODEL categories. Composite categories are ranked in
terms of adequacy (see Figure 4). Ranks are defined on a scale from 1 to 20 that translate into a
multiple of one or more 5% COA intervals.
Figure 4. From dimensions to estimates
Figure 5 shows an illustrative C-Table for the COA of “Reservoir facies presence”.
79th EAGE Conference & Exhibition 2017 – Workshop Programme
Paris, France, 12-15 June 2017
Chance Play Element (CPE) Maps
Discovery in CPE
Well in CPE w Reservoir > R Min
Discovery in Play Fairway
Well in Fairway w Reservoir > R Min
Wells in play element (CPE)
Wells in play fairway
Fraction wells wIth reservoir
3D Seismic over target (R Visible)
2D Seismic over target (R Visible)
M: Shallow marine blanket deposits,
carbonates
YES
YES
YES
YES
YES
NO
YES
YES
YES
NO
NO
YES
YES
YES
NO
NO
NO
YES
NO
NO
NO
YES
YES
YES
NO
ASSUME
YES
YES
NO
YES
NO
NO
ASSUME
YES
Conditioning
processes:
> 2 wells
YES
ALL
FRACTION ALL LACK
YES
ALL
> 2 wells
FRACTION
<3 wells
> 0 wells
YES (R Visible)
<3 wells
> 0 wells
NO
YES (R Visible)
<3 wells
> 0 wells
NO
YES (R NOT Visible)
<3 wells
0 wells
NO
YES (R NOT Visible)
1.00
FRACTION
0.60
0.95
FRACTION
1.00
0.95
0.85
0.50
1.00
FRACTION
0.55
0.90
FRACTION
1.00
0.95
0.90
0.50
1.00
1.00
CPE COA CPE COA
1.00
1.00
0.50
1.00
1.00
CPE COA CPE COA
1.00
1.00
0.50
M: Deep-water turbidites (channels, fans)
1.00
1.00
CPE COA CPE COA
1.00
1.00
0.50
●NONE
0.75
FRACTION
0.20
0.65
FRACTION
1.00
0.95
0.50
0.40
C: Eolian
1.00
1.00
CPE COA CPE COA
1.00
1.00
0.50
◌ Non-deposit
0.95
FRACTION
0.45
0.85
FRACTION
1.00
0.95
0.70
0.50
C: Lacustrine
1.00
1.00
CPE COA CPE COA
1.00
1.00
0.50
◌ Erosion
0.90
FRACTION
0.45
0.80
FRACTION
1.00
0.95
0.70
0.50
C: Alluvial fan, braided streams
1.00
1.00
CPE COA CPE COA
1.00
1.00
0.50
◌ Faulted
0.85
FRACTION
0.35
0.75
FRACTION
1.00
0.95
0.60
0.50
C: Meandering channels
1.00
1.00
CPE COA CPE COA
1.00
1.00
0.50
◌ Other:
0.80
FRACTION
0.25
0.70
FRACTION
1.00
0.95
0.50
0.50
Fractured basement or porous lava
1.00
1.00
CPE COA CPE COA
1.00
1.00
0.50
0.55
FRACTION
0.05
0.45
FRACTION
1.00
0.95
0.50
0.40
M: Coastal, fluvio-deltaic, deltaic, tidal
Figure 5. Illustrative C-Table for “Reservoir facies presence” COA
Conclusion
The C-Tables chance calculus provides estimates for a complete set of chance factors that are
consistent and cover all conventional exploration situations. Explicit MODEL categorization not only
provides a transparent, unambiguous basis for the chance estimates, but also eliminates double
risking. The two-level separation between estimates of shared play chances and conditional target
chances provides a unique basis for play based exploration.
The chance calculus is clearly much more than a simple table lookup: The calculus is being
implemented as an APP over multiple platforms: A map-based environment (ArcGIS) as well for
desktop and the web with direct interaction with play and prospect applications via a common
relational database.
References
Brekke, H., J-E. Kalheim (1996) “The Norwegian Petroleum Directorate’s assessment of
undiscovered resources of the Norwegian Continental shelf – background and methods” in Dore,
A.G., Sinding-Larsen, R. (Eds.) Quantification and Prediction of Petroleum Resources, NPF Special
Publication 6, Elsevier, Amsterdam, 91 – 104.
Milkov, A.V. (2015) “Risk tables for less biased and more consistent estimation of probability of
geological success (PoS) for segments with conventional oil and gas prospective resources” EarthScience Reviews, 150, 453–476.
Stabell, C.B., Lunn, S. Breirem, K. (2003) “Making Effective Use of a DFI: A Practical Bayesian
Approach to Risking Prospects for Seismic Anomaly Information”, Proceedings of HEES
(Hydrocarbon Economics and Evaluation Symposium), SPE 82020, Dallas, April 5-8, 2003.
79th EAGE Conference & Exhibition 2017 – Workshop Programme
Paris, France, 12-15 June 2017