Professional Documents
Culture Documents
Reserves Estimation - The Challenge For The Industry
Reserves Estimation - The Challenge For The Industry
Introduction
Reserves estimation is one of the most essential tasks in the Production
Discovered in-Place
cally recoverable hydrocarbons in a field, area, or region are Proved
Proved Proved
evaluated quantitatively. Downward revisions of U.S. Security + Probable
+ Probable
and Exchange Commission (SEC)-booked reserves by some Dev. Undev. + Possible
oil companies in 2004 brought the topic under public scruti-
Increasing Maturity
Contingent Resources
ny. Confidence in reserves disclosures became a public issue,
Low Estimate Best Estimate High Estimate
and there were calls from investors and lending institutions
Unrecoverable
for more-reliable reserves estimates. Oil companies have
Undiscov.
responded by revisiting reserves-estimation procedures, and Prospective Resources
SPE, American Association of Petroleum Geologists (AAPG), Low Estimate Best Estimate High Estimate
World Petroleum Council (WPC), and Society of Petroleum Unrecoverable
Evaluation Engineers (SPEE) have launched a joint project Increasing Uncertaint y
to train reserves evaluators. A major goal in this initiative
is preparation of training modules that represent industry’s Fig. 1—Petroleum resource classification, based on
“recommended practices.” Fig. 1 of Petroleum Resources Definitions (2000).
Long before the issue caught the public’s attention, how-
ever, reserves estimation was a challenge for the industry. quantities of petroleum which are anticipated to be com-
The challenge stems from many factors, tangible and intan- mercially recovered from known accumulations from a
gible, that enter the estimation process, and judgment is an given date forward” (Petroleum Reserves Definitions 1997).
integral part of the process. Uncertainty, along with risk, is Commerciality implies commitment or expected commit-
an endemic problem that must be addressed. Consequently, ment to develop reserves within a reasonable time frame.
the industry’s record of properly predicting reserves has been Depending on the degree of uncertainty, three main classes of
mixed. Despite appeals from some quarters, there currently reserves are recognized: proved, probable, and possible, the
is no standardized reserves-estimation procedure. last-named two collectively called unproved. Proved reserves
The purpose of this paper is to discuss various issues are those quantities that have reasonable certainty of being
related to reserves, review reserves-estimation procedures, recovered, indicating a high degree of confidence. Proved
and make suggestions for improvements. Emphasis will be reserves may be developed or undeveloped. Probable reserves
placed on reserves evaluation at the preproduction stage in are more likely than not to be recoverable, while pos-
which estimation errors generally have the highest economic sible reserves are less likely to be recoverable than probable
effect. The discussed procedures pertain to conventional oil reserves. Geological and engineering data form the basis of
and gas reserves. Specific rules relating to booking reserves determination. Proved reserves assume recoverability under
for regulatory purposes are outside the scope of this paper. current economic conditions, operating methods, and gov-
ernment regulations. For unproved reserves, recoverability
Reserves Definition and Classification may be tied to future economic conditions and technology.
Fig. 1 shows how reserves are part of the petroleum resource In the absence of fluid-contact data, the lowest known
base. Under the SPE/WPC definitions, “Reserves are those occurrence of hydrocarbons generally controls the proved
limit. Reserves exclude past production.
Potentially recoverable quantities that do not satisfy the defini-
Ferruh Demirmen, SPE, is an independent petroleum tion of reserves are contingent (discovered but subcommercial)
consultant dealing, in part, with reserves estimation. He and prospective (undiscovered) resources. Subcommerciality
has 40 years of upstream experience, mostly in produc- includes technology limitations (Petroleum Reserves Definitions
tion. Demirmen retired from Shell Intl. Petroleum Mij. as a 1997; Petroleum Resources Definitions 2000).
corporate consultant in production geology responsible for Copyright 2007 Society of Petroleum Engineers
Europe/North Sea, southeast Asia, and South America. He This is paper SPE 103434. Distinguished Author Series articles are general, descriptive
representations that summarize the state of the art in an area of technology by describing recent
holds a BS degree from the U. of Kansas and MS and PhD developments for readers who are not specialists in the topics discussed. Written by individuals
degrees from Stanford U., all in geology. recognized as experts in the area, these articles provide key references to more definitive work
and present specific details only to illustrate the technology. Purpose: to inform the general
readership of recent advances in various areas of petroleum engineering.
4,000
Statfjord
3,500
Discovered 1974
3,000
Production start 1979 Ekofisk
Oil, million bbl
2,500
2,000
Troll West production start
1,500
1,000
500
0
1974 1978 1982 1986 1990 1994 1998 2002
100 3.0
Economic P90
cutoff Untruncated
r Probability, %
Pes 2.5
80 Truncated (30,200)
2.0 Median Reserves Range
Values×10–2
40 1.0 Plausible
P10 maximum
P10 (truncated) Maximum 0.5
20 P90 (truncated)
0
30 25 50 75 100 125 150 175 200
0 Economic
0 25 50 75 100 125 150 175 200 cutoff Reserves, million bbl
UR, million bbl
Fig. 8—Expectation curves for field KK. Fig. 9—Truncated distribution, field KK.
A. Dependency B. Aggregation
P90 P90
values P10 values
P10
Positively values values
Independent dependent True P90 True P10
Means Mean
Independent Probabilistic sum
Positively dependent Arithmetic sum
prediction of future performance. Although probabilis- Another emerging methodology can be referred to as a mod-
tic approaches have been applied, the common practice eling-based stochastic approach. For calculating in-place
is deterministic. volumes, this approach is similar to the stochastic volumetric
Decline-Trend Analysis. The analysis refers to estimating approach except that seismic-depth uncertainties are handled
reserves on the basis of a reasonably well-defined behavior of through geostatistical analysis performed on seismic-time
a performance characteristic (e.g., production rate or oil cut) and -velocity data. Recovery estimates make a direct linkage
as a function of time or cumulative production. The method between geologic facies and rock and fluid properties, and
usually is used for single-well analysis. The trend established iterative reservoir-simulation runs are used to derive a PDF
from past behavior is extrapolated until the economic limit for recovery. There seems to be no full application of this
is reached. The basic assumption is that the trend estab- method in the literature, but elements of this approach have
lished in the past will govern the future in a uniform man- been proposed (Ovreberg et al. 1992).
ner. Strictly speaking, such estimates are P50 estimates (i.e.,
proved plus probable). Dependencies and Aggregation
Material Balance. This is a conservation-of-matter tech- The issues of dependency and aggregation, if not addressed,
nique whereby the pressure behavior of the reservoir in lead to statistical distortions in reserves estimation. The
response to fluid withdrawal is analyzed in several steps. The distortions can occur whether the estimation method is
fluid properties and pressure history are averaged, treating deterministic or stochastic, although with the former the
the reservoir as a tank. For reliable estimates, there must be distortions are hidden and remedies are impossible, while
sufficient pressure and production data (for all fluids) and stochastic methods enable remedies.
reliable pressure/volume/temperature data, and the reservoir One common way that dependency enters reserves esti-
must have reached semisteady-state conditions. mation is in the calculation of hydrocarbon volumes from
Reservoir Simulation. This procedure represents the reser- multiplication of input data. Statistically, if input variables are
voir with a grid, or a set of interconnected tanks, each con- positively dependent (correlated), but such dependency is
taining rock and fluid properties. A computer model performs ignored, the P90 value will be overestimated and P10 value
a series of material-balance calculations in different cells, and will be underestimated, resulting in reduced dispersion as
migration of fluids between adjoining cells is allowed by use shown in Fig. 10. The mean value also will be underestimated.
of Darcy’s flow equation. A development scheme and operat- With negative dependency, the effect will be the opposite.
ing conditions generally are superimposed on the system. Dependencies (e.g., porosity and hydrocarbon saturation, or
For reliable results, a good match between observed history area and RE) are common in field data; therefore, the problem
and simulated performance is essential. is not trivial.
The aggregation issue arises when reserves are combined
Hybrid Methods (e.g., adding reserves of several reservoirs to obtain the field
In addition to the three reserves estimation methods, there total). Statistically, arithmetic addition will distort the results
also are hybrid methods that attempt to combine the such that P90 values will be underestimated, while P10
strengths of deterministic and probabilistic approaches. values will be overestimated. The mean will be correct. The
These emerging techniques are poorly publicized, but they net effect is too large of a dispersion. The distortion will be
may play a bigger role in future. One such method is the mitigated if the component reserves have positive dependen-
“scenario approach,” a semideterministic method in which cies and aggravated if the dependencies are negative. As the
reservoir uncertainty is depicted by key parameters in a hier- number of summed reserves increases, the difference from
archical manner. Each reservoir attribute is assigned a subjec- the arithmetic sum will become larger, and the probabilistic
tive probability. At the bottom of the hierarchy are reservoir P10 and P90 values will approach the mean value. The dif-
realizations that, linked to suitable development schemes, ference between the arithmetic and probabilistic aggregation
lead to discrete reserves volumes. sometimes is called the portfolio effect.
Attention to Pitfalls. Some of the pitfalls in reserves estima- take should provide improvements. When implementing
tion are of a geoscientific or engineering nature and usually these measures, the proactive role of management cannot
will be avoided by the experienced professional. Still, errors be overstated. Depending on circumstances, a probabilistic
happen and caution is needed. Harrell et al. (2004) describe approach and emerging methodologies generally should be
recurring mistakes in reserves estimates related to mapping preferred over a deterministic approach.
and other procedures. One pitfall to avoid in decline-curve
analysis, for example, is to aggregate well-decline trends to Acknowledgments
represent a composite decline for the whole field. Errors in The author is indebted to Chapman Cronquist, Ron Harrell,
seismic mapping caused for example by polarity reversals and and Ed Capen for reading the draft manuscript and making
tuning effects should be avoided by seismic modeling, and valuable suggestions. Thanks also are due to Ed Capen for
stringent guidelines should be followed for seismic evidence extended discussions on reserves issues.
of hydrocarbons presence. A probabilistic approach should
not be considered complete until dependency and aggregation Nomenclature
issues are addressed. A= Area (oil zone)
boi= Initial oil shrinkage factor (=1/Boi)
Long-Term Outlook. A forward-looking outlook that Boi = Initial oil formation volume factor
attempts to foresee and plan for future incremental-recovery h= Thickness (gross, oil zone)
projects and technology developments in a life-cycle context hn= Net thickness (oil zone)
could reduce the incidence of reserves-estimate surprises n/g= Net-to-gross thickness ratio (oil zone)
during production. Recoveries from such projects will quali- Pes= Probability of economic success
fy as unproved reserves and contingent resources. Soi= Initial oil saturation
fi= Initial porosity (oil zone)
Look-Back Analysis. Comparison of reserves estimates with
actual results on a post-mortem basis will provide valuable References
learning points. The comparison should include quantitative Capen, E.C. 1992. Dealing with Exploration Uncertainties. In
assessment of bias and accuracy of estimates. Companies can The Business of Petroleum Exploration, ed. R. Steinmetz, Tulsa:
significantly improve estimating abilities by keeping records American Assn. Petroleum Geologists, 29–61.
and tracking estimation performance. Castle, G.R. 1985. Assessment of North Sea Field Performance.
Petroleum Economist, October: 358–359.
Training. Reserves estimators should be trained toward Cronquist, C. 2001. Estimation and Classification of Reserves of Crude
proficiency in volumetrics, risk assessment, avoidance of Oil, Natural Gas, and Condensate. Richardson, Texas: Society of
bias, and an open-minded attitude with respect to alterna- Petroleum Engineers. 416.
tive interpretations. Ethical conduct should be cultivated, Demirmen, F. 2005. Reliability and Uncertainty in Reserves: How
and the interdisciplinary character of reserves estimation and Why the Industry Fails, and a Vision for Improvement.
involving geoscientists, engineers, and economists should Paper SPE 94680 presented at the SPE Hydrocarbon Economics
be emphasized. & Evaluation Symposium, Dallas, 3–5 April. DOI: 10.2118/
94680-MS.
Reserves Governance. The management should issue EAGE 1996. Workshop Report: Uncertainty in Reserve Estimates.
reserves-evaluation guidelines, set peer reviews and audits, Petroleum Geoscience 2 (4): 351–352.
and establish a reserves-governance structure. Harrell, D.R., Hodgin, J.E., and Wagenhofer, T. 2004. Oil and
Gas Reserves Estimates: Recurring Mistakes and Errors. Paper
Conclusion SPE 91069 presented at the SPE Annual Technical Conference
Reserves estimation is a complex process affected by many & Exhibition, Houston, 28–29 September. DOI: 10.2118/
factors, not all of them transparent. Uncertainty and sub- 91069-MS.
jectivity are inherent in the process. The process, however, Hefner, J.M., and Thompson, R.S. 1996. A Comparison of
must be underpinned by sound geoscientific and engineer- Probabilistic and Deterministic Reserve Estimates: A Case
ing practices. Study. SPERE 11 (2): 43–47. Paper SPE 26388. DOI: 10.2118/
Field histories reveal a mixed record for the industry as far 26388-PA.
as consistency in reserves estimates. Many fields show wide Oil and Gas Reserves Committee Mapping Subcommittee Final Report—
fluctuations in reserves over time. Fluctuating reserves esti- December 2005: Comparison of Selected Reserves and Resource
mates entail cost penalties even in cases in which estimated Classifications and Associated Definitions. 2005. Richardson, Texas:
recoveries are better than anticipated. Improving reserves Society of Petroleum Engineers. http://www.spe.org/web/ogr/
reliability is a challenge for the industry and requires a OGR_Mapping_Final_Report_15Dec05.pdf. (Accessed 2 April
multipoint approach. Efforts to improve reserves-estimation 2007.)
reliability should parallel risk-reduction efforts through use Ovreberg, O., Damsleth, E., and Haldorsen, H.H. 1992. Putting Error
of opportunity portfolios. Bars on Reservoir Engineering Forecasts. JPT 44 (6): 732–738.
There are no quick-fix remedies for reserves-estimation Paper SPE 20512. DOI: 10.2118/20512-PA.
problems. However, certain measures that oil companies can