A Process for Evaluating Exploration Prospects1 Robert M. Otis and Nahum Schneidermann Schneidermann 2
ABSTRACT In 1989, Chevron Overseas Petroleum, Inc., developed a process to allow management to compare a wide variety of global exploration opportunities on a uniform and consistent basis. Over the next five years, the process evolved into an effective method to plan exploration programs on a basis of value incorporating prospect ranking, budget allocation, and technology management. The final product is a continuous process and includes, within a single sin gle organiz o rganizationa ationall unit, uni t, the t he integration in tegration of geologic risk assessment, probabilistic distribution of prospect hydrocarbon hydrocarbon volumes, engineering development planning, and prospect economics. The process is based on the concepts of the play and hydrocarbon system. Other steps of the process (geologic risk assessment, volumetric estimation, engineering support, economic evaluation, and postdrill feedback) are considered extensions of fundamental knowledge and understanding of the underlying geological, engineering, and fiscal constraints imposed by these concepts. A foundation is set, describing the geologic framework and the prospect in terms of the play concept—source, reservoir, trap (including seal), and dynamics (timing/migration). The information and data from this description become the basis for subsquent steps in the process. Risk assessment assigns a probability of success to each of these four elements of the play concept, and multiplication of ©Copyright 1997. The American Association of Petroleum Geologists. All rights reserved. 1Manuscript received February 16, 1996; revised manuscript received September 26, 1996; final acceptance February 4, 1997. 2 Chevron Overseas Petroleum, Inc., P.O. Box 5046, San Ramon, California 94583-0946. We acknowledge the champion of this process, M. W. Boyce, without whose continuing, senior-management support this process would not have been possible. We acknowledge the pioneering efforts of C. L. Aguilera, G. A. Demaison, E. J. Durrer, F. R. Johnson, W. E. Perkins, J. L. Reich, and R. A. Seltzer, who established the framework for the process in its early st ages. We also acknowledge the efforts to refine, document, and teach the process during the later stages by S. D. Adams, A. O. Akinpelu, G. A. Ankenbauer, G. L. Bliss, T. J. Humphrey, E. McLean, and D. B. Wallem. Finally, we acknowledge all the people who, over the past several decades, have championed such a process, but fell victim to deaf ears because of high oil prices or dumb luck. These people provided the well-founded basis for the theoretical and practical application of evaluation principles. We also wish to extend special thanks to Gerard Demaison and Erwin Durrer for their continuous support, guidance, and friendship.
AAPG Bulletin, V. 81, No. 7 (July 1997), P. 1087–1109.
these probabilities yields the probability of geologic success. A well is considered a geologic success if a stabilized flow of hydrocarbons is obtained on test. Volumetric estimation expresses uncertainty in a distribution of possible hydrocarbon volumes for the prospect constructed from ranges of parameters obtained from information specific to the prospect, and data described by the parent play concept. With this distribution, engineering support provides development scenarios for three cases—a pessimistic case (10%), the mean, and an optimistic case (90%). Economic evaluation is run for each of the three cases, thus providing a range of economic consequences of the geological, engineering, and fiscal framework. Commercial risk is based on the results of this evaluation, and overall probability of success is the multiplication of the probability of geologic success and probability of commercial success. Postdrill feedback determines whether the individual processes are providing predicted results consistent with actual outcomes.
INTRODUCTION The topic of prospect evaluation has been discussed in the literature for many years and has been recently described in a sequence of reviews by Explorer rer . In recent years, Robert Megill in the AAPG Explo AAPG has encourag encouraged ed discussions discussions on this this subject subject by sponsoring Hedberg research conferences and con vention sessions sessions at which we we presented presented parts of the Chevron system (Otis and Schneidermann, 1994; Otis, 1995). Many of the conference participants requested that we summarize our process in print. This paper is a summary of the exploration evaluation process that has been used to provide estimates of exploration prospect value for the last 7 yr at Chevron Overseas Petroleum, Inc. For obvious reasons, this summary does not include all of the details; however, we hope this paper will stimulate further discussions and encourage the release of similar summaries by other companies. The foundation of the process is knowledge of geology; in particular, the concepts of hydrocarbon systems and the play concept as developed over the years by Dow (1972, 1974), Nederlof (1979), Perrodon (1980, 1983, 1992), Demaison (1984), 1087
1088
Evaluating Prospects
RISK
Testing a Stabilized Flow of Hydrocarbons
PLAY CONCEPT Source Rock, Reservoir, Trap, Timing, and Migration
ECONOMIC ANALYSIS
ENGINEERING
Conceptual Development Plan Facilities Costs Production Profile Recovery Factor
Cash Flow Model and Value Measures
DECISION DECISION
POSTDRILL REVIEW
If Success, Compare Actual Parameters to Predicted; If Failure, Reason Why
OPTIMIZATION VOLUMETRICS
Volumetric Distribution of Hydrocarbons (In-Place and Estimated Recoverable)
Figure 1—The exploration evaluation process incorporates specification of geologic play concept, assessment of geologic risk, estimation of hydrocarbon volumes, conceptual engineering, and a development plan for economic analysis. The process includes a feedback loop for process improvement based on results results of comparisons between predrill and postdrill results.
953009 95300 9 fre
Ulmishek (1986), White (1988, 1993), Demaison and Huizinga (1991), Magoon (1987, 1988, and 1989, Magoon and Dow (1994). (1994). Ultimately, Ultimately, all estimates of value are based on hydrocarbon volumes, geological risk, and reservoir productivity and performance, which, in turn, are based on the geological characteristics of the hydrocarbons present and the geological nature of the reservoir and trap characteristics. The process, therefore, focuses on estimating the range of resources that may be possible (what nature has provided), the chances of finding a hydrocarbon accumulation, and the requirements for producing the hydrocarbons to add significant value at an acceptable acceptable rate of return. return. The full process, illustrated in Figure 1, begins by establishing the play concept, described by four elements: source rock, reservoir reservoir,, trap (including seal), and dynamics (timing and migration). Based on this description, geological risk is assessed, and the probability of finding producible hydrocarbons hydrocarb ons is assigned a value between 0.01 and 0.99. At the same time, the volume of hydrocarbons present is estimated as a probability distribution of recoverable volumes. The engineering department provides estimates of production profiles and facilities and transportation costs, which are then incorporated with a country economic model and risk to generate economics that correspond to pessimistic, mean, and optimistic estimates from the distribution. If a decision is made to go ahead with the project, results are documented so that predicted and actual outcomes can be compared, added to the knowledge base, and used for process improvement. Methods used in the process are not new. They are based on pioneering publications by Haun (1975), Newendorp (1975), White (1980, 1988,
1993), Megill (1984), and Rose (1987, 1992), as well we ll as in -h -hou ouse se wo work rk by bo th Ch Chev evro ron n (J (Jon ones es,, 1975) and Gulf. The ideas of hydrocarbon system and play concept, as well as descriptive tools, are described fully by Magoon (1987, 1988, 1989), Magoon and Dow (1994), and Demaison and Huizinga (1991). The breakdown of geologic risk into basic risk factors, preparing production production profiles, estimating facilities and transportation costs, and developing economic models are practiced throughout the industr y. Probabilistic techniques are well known from elementary probability and statistics. The three-point method was developed by J. E. Warren of Gulf Oil Corporation in the late 1970s (Warren, 1980–1984, personal communication) and used in the years before the Chevron-Gulf merger. The three-point method is based on an operator for estimating moments of distributions described by Pearson an d Tukey (1965) and Keefer and Bodily (1983). An approach similar to Warren’s was also discussed by Bourdaire et. al. (1985). This process was introduced to Chevron Overseas Petroleum, Inc., in mid-1989 and has since been adopted by the other operating companies upstream in Chevron. Because of its ease of use, transparency, and the built-in mechanism of postdrill feedback, the process has been widely accepted by explorationists and senior management to provide consistent, credible estimates estimates of vaa l u e t h a t c a n b e u s e d t o c o m p a r e a n d r a n k v exploration projects across business unit and operating company boundaries. The use of this process to provide risk, volumetric, and economic input to exploration decision making has all but eliminated the previous gap between predicted and actual results.
Otis and Schneidermann
1089
Figure 2—The timing risk chart (Magoon, 1987) helps to integrate geological knowledge and factual information for risk assessment, volumetric parameter ranges, and engineering considerations.
Recognition of an active petroleum system also serves only as a screening device because it carThe distribution of hydrocarbons in the Earth’s ries no volumetric (and therefore, no economic) crust follows a lognormal distribution typical of value. many other natural resources. Such a distribution implies that hydrocarbons are concentrated in relatively few basins, and that exploration is not an Play equal-chance game. In our assessment process, we In our definition, the play is the elemental part evaluate four different concepts of exploration as a function of the degree of knowledge about the spe- of a petroleum system, and is recognized as havcific project: basin framework, petroleum system ing one or more accumulations of hydrocarbons identified by a common geological character of framework, play, and prospect. reservoir, trap, and seal; timing a nd migration; preservation; a common engineering character of location, environment, and fluid and flow properBasin Framework ties; or a combination of these. Individual plays, Is there a volume of sedimentary rocks capable therefore, have unique geological an d engineering of containing potential ingredients of a working features, and can be used as a basis for economic “hydrocarbon machine”: source, reservoir, trap and characterization. seal, and proper timing and migration? This assessment is a screening device only, and does not Prospect include economic considerations. PLAY CONCEPT
Prospect represents an individual, potential accumulation. Each prospect is perceived as belonging to an individual play, characterized by The petroleum system framework is defined as risk components and a probabilistic range distria volume of sedimentary rocks containing hydro- bution of potential hydrocarbon volumes within carbons and charged by a single source rock. The its trap confines. In frontier areas, geological analogs provide the definition requires manifestations of hydrocarbons (seeps, shows, or a producing well) and is best models for assessing the capability of the evalapplicable in many frontier basins only by analogy. uated basin to yield commercial accumulations of Petroleum System Framework
1090
Evaluating Prospects
hydrocarbons. In more mature areas, the presence of a petroleum system has been proven, and the assessment focuses on play types. Regardless of the maturity of exploration or the amount of existing production, however, each prospect requires a detailed review of the individual risk components. A timing risk chart (Figure 2), modified from the original ideas of Magoon (1987), provides a very useful and user-friendly summary and display of the play concept.
RISK ASSESSMENT Withi n the evaluation process, the risk considered is geologic risk; i.e., the risk that a producible hydrocarbon accumulation exists. We consider a producible accumulation to be one capable of testing a stabilized flow of hydrocarbons. Geologic risk is assessed by considering the probability that the following four independent factors of the play concept exist. (1) Presence of mature source rock (Psource ) (2) Presence of reservoir rock (P reservoir ) (3) Presence of a trap (Ptrap ) (4) Play dynamics (Pdynamics ) or the appropriate timing of trap formation relative to timing of migration, pathways for migration of hydrocarbons from the source to the reservoir, and preservation of hydrocarbons to the present day. The probability of geologic success (P g ) is obtained by multiplying the probabilities of occurrence of each of the four factors of the play concept. Pg = Psource × Preservoir × Ptrap × Pdynamics
If any one of these probability factors is zero, the probability of geologic success is zero. Geological success is defined as having a sustained, stabilized flow of hydrocarbons on test. We do not consider the oil machine to work with only oil and gas shows or flows of hydrocarbons without pressure stabilization. This definition eliminates very low-permeability reservoirs, reservoirs of limited areal extent, biodegraded oils, and other marginal cases that cannot deliver a stabilized flow of hydrocarbons from the success case. In practice, this definition has been easily applied to the range of prospects drilled during the time the process has been used. The probabilities that any of the play (or risk) factors occur are estimated by first analyzing the information available. The risk assessment checklist (Figure 3) was designed to assist the earth scientist in examining as much information as possible. The
checklist has been compiled over several years, with input from person nel insid e and outsid e o f Chevron to ensure all aspects of each play factor are considered. The checklist categorizes the four risk factors with following elements. The risk assessment worksheet (Figure 4) records our assessments of the elements of the risk factors, which are expressed as unfavorable, questionable, neutral, encouraging, and favorable. With little or no data, assessment is based on evaluating the analogs and the likelihood that the model will reflect the analog. As data are acquired, we begin to develop opinions supported by the data. These opinions may be positive (encouraging or favorable) or negative (questionable or unfavorable). Factors with equal probability of positive or negative outcomes are given a probability of occurrence of 0.5. Assessments of encouraging or questionable are based on indirect data that support or do not support the model. Examples of indirect data for an assessment of encouraging include shows, seeps, and presence of direct analogies. Examples of indirect data for an assessment of questionable include lack of shows in nearby wells, thin or poor reser voirs, and evidence of recent faulting. With indirect data, we are more dependent on the model than on the data, and our opinions are supported, but not confirmed, with data. With indirect data supporting the model, probability of occurrence is encouraging, with values between 0.5 and 0.7. When indirect data do not support the model, probability of occurrence is questionable, with values between 0.3 and 0.5. As se ss me nt s of fa vora bl e or un favo ra bl e ar e based on direct data that tend to confirm or disprove the model. Examples of direct data for an assessment of favorable include nearby producing fields or wells with stabilized flows on test, proven hydrocarbon systems with moderate to high source potential index (>5, based on highquality Rock-Eval data) (Demaison and Huizinga, 1991), and maturation models with parameters supported by data from nearby wells. Examples of direct data for an assessment of unfavorable include dry wells testing similar structures defined by good-quality seismic, lack of reservoir in wells, and a hydrocarbon system with very low source potential index (<2, based on high-quality Rock-Eval data). With direct data supporting the model, probability of occurrence is favorable, with value s betwe en 0.7 and 0.99. When direc t data do not support the model, probability of occurrence is unfavorable, with values between 0.01 and 0.3. We record our assessments on the worksheet, and as we complete each factor, we assign a value corresponding to the key at the bottom of the
Otis and Schneidermann
1091
Figure 3—The risk assessment checklist lists the critical aspects of geologic risk assessment to help ensure all aspects have been considered.
1092
Evaluating Prospects
Figure 4—The risk assessment worksheet provides a method for transferring qualitative judgments on geologic risk to quantitative probability of geologic success.
Otis and Schneidermann
Evaluation
Same Play Adjacent Structure
Conventional
Same Play Nearby Structure
New Play - Same Trend Old Play - New Trend
Producing Area Delineation
LOW RISK 1:2
Avg. P g = 0.75
Frontier Area
Play
Hydrocarbon System
HIGH RISK
VERY HIGH RISK
MODERATE RISK 1:4
New Play - New Basin or Play with Negative Data
Emerging Area Prospect
VERY LOW RISK
Frontier
1:8
Figure 5—Risk categorization of “rules of thumb” for geologic risk assessment based on feedback from five years of drilling history.
1:16
Avg. P g = 0.375 Avg. P g = 0.183 Avg. P g = 0.092 Pg= Probability of Geological Success
worksheet (Figure 4). Note that the probability of occurrence for each element depends on the leastfavorable assessment. During the past 5 yr, an understanding of r isk has evolved into five broad categories and general “rules of thumb” that allow characterization of risk and reduce impractical arguments over specific numbers.
1093
Avg. P g = 0.05
VOLUMETRICS
Oil and gas volumes are expressed as a product of a number of individual parameters. Because of uncertainty in the value of each of the individual parameters, oil and gas volumes can be represented as a distribution. The distribution is generally assumed to be lognormal (Capen, 1993). In our process, the distribution represents the range of (1) Very low risk (Pg between 0.5 and 0.99, bet- recoverable hydrocarbons (or reserves, in their ter than 1:2). All risk factors are favorable. This catmost general sense) expected to be found when egory is associated with wells that test proven plays the well is drilled, assuming geologic success (staadjacent to (<5 km) existing production. bilized flow of hydrocarbons on test). It is not the (2) Low risk (Pg between 0.25 and 0.5, between distribution representing the range of commercial 1:4 and 1:2). All risk factors are encouraging to favorreserves, proven reserves, or any other type of able. This category is associated with wells that test reserves tied to economic considerations. Note proven plays near (5–10 km) existing production. (3) Moderate risk (Pg between 0.125 and 0.25, that we use the term reserves as being interbetween 1:8 and 1:4). Two or three risk factors are changeable with recoverable volumes throughout encouraging to favorable—one or two factors are this text based on the general definition of encouraging or neutral. This category is associated reserves being “those quantities of hydrocarbons with wells testing new plays in producing basins that are anticipated to be recovered from a given or proven plays far from (>10 km) existing produc- date forward.” (Journal of Petroleum Technology, 1996, p. 694). We address commerciality during tion. (4) High risk (P g between 0.063 and 0.125, the economics phase of the process. One method that can be used to obtain this disbetween 1:16 and 1:8). One or two risk factors are tribution of reserves is Monte Carlo simulation. The encouraging—two or three factors are neutral or encouraging to neutral. This category is often asso- distribution is obtained by specifying distributions ciated with wells testing new plays in producing for each of the individual parameters and then mulbasins far from (>20 km) existing production or tiplying randomly selected values together many times, thereby creating a highly sampled histogram proven plays in an unproved area. (5) Very high risk (P g between 0.01 and 0.063, that approximates the actual distribution. The worse than 1:16). Two to three risk factors are no number of estimates (iterations) necessary to better than neutral, with one or two factors ques- obtain a satisfactory representation of the distributionable or unfavorable. This category is usually asso- tion ranges from a few hundred to se veral thouciated with wells testing new plays in an unproved sand. Monte Carlo simulation programs are widely available and the calculation can be done in a area far from (>50 km) existing production. few minutes, depending on the number of iterations used. This categorization is summarized in Figure 5.
1094
Evaluating Prospects
An alter native met hod to Mon te Carlo simula tion was developed by J. E. Warren of Gulf Oil Corporation (Warren, 1980–1984, personal communication). This method produces distributions that are essentially identical to Monte Carlo simulations, but requires no iterations and no assumptions about the distributions of the reserve parameters. We call the method the three-point method; it is explained in detail in Appendix 1. Briefly, the method uses as input a range for each parameter by specification of values corresponding to the 5, 50, and 95% probability of occurrence. From these ranges, a mean and variance are estimated for each parameter using the Pearson-Tukey operator (Pearson and Tukey, 1965). The means and variances are combined to provide the mean and variance of the resultant reserve distribution. A lognormal distribution is assumed for the reserves distribution and can be calculated from the estimated mean and variance. Advantages of this method are the speed of the calculation, which is essentially instantaneous on any spreadsheet computer program, and that it has no requirement for specifying the parameter distribution. The key to success with this method, therefore, is correctly specifying the ranges. Guidelines include the following: (1) Selecting the 5% value, which is generally near the minimum value expected. For example, for porosity the 5% value would be near the minimum porosity observed in nearby wells; for area, the 5% value would be the area corresponding to the minimum hydrocarbon column expected. The explorationist should keep in mind that the odds of finding a value less than the selection are 1 in 20. (2) Selecting the 95% value, which is generally near the maximum value expected. For example, for porosity the 95% value would be near the maximum porosity observed in nearby wells; for area, the 95% value would be the area corresponding to a maximum hydrocarbon column expected. Likewise, the explorationist should keep in mind that the odds of finding a value greater than the selection are 1 in 20. (3) Selecting the 50% value, which is generally near the middle of the expected range of values. The median is often the most difficult to choose and requires the support of data associated with the play or with an appropriate analog. Analogs should be used with caution. For example, in a purely continental basin, a partial analog with lacustrine source and marine reservoir does not apply. The explorationist should keep in mind that the odds of finding a value less than the selection is equal to the odds of finding a value greater that the selection.
After the ranges for the reserve parameters have been specified, the mean and variance for the reserve distribution are calculated. Figure 6 shows a spreadsheet with an example for a typical small prospect in a deltaic environment, such as the Niger Delta or the Mississippi Delta. The input ranges are as shown, and the output information includes the mean reserves and cases for a pessimistic result (10% or P10) and an optimistic case (90% or P90). In addition to reserves, the spreadsheet calculates values for individual reservoir parameters, including porosity, area, and net pay, that, when multiplied together, will total the pessimistic or optimistic reserve value for use during the engineering and economics phases of the process. These pessimistic and optimistic parameter values are consistent with the variances specified by their corresponding input ranges. Note that the parameter values are not the 10 and 90% values of the input ranges. Figure 6 also shows the cumulative reserve distribution and values for specific percentiles, as well as the mean, median, and mode. In practice, the mean value for the distribution is commonly less than the explorationist’s expectation. At this point it is critical to keep in mind that this result is the consequence of the input parameter ranges. If the input ranges are based on good available data, it may be difficult to alter them significantly, and the explorationist may have to adjust expectations. This dilemma can be resolved by comparing the prospect reserve distribution to field-size distributions of the play or analogs. Questions that arise and responses to them often include the following: (1) Are the predicted values reasonably consistent with reserves found in analogs to date? If so, use the numbers obtained from the input parameter ranges. (2) Are the predicted reserves significantly smaller or larger than those found in analogs to date? If yes, then (3) Are there technical reasons to justify the difference? If so, use the ranges as stated. (4) Are technical reasons for the difference lacking? If so, reconsider values assigned in previous steps and recalculate reserves. When the final reserve distribution is obtained, the information from the process moves to the engineering support and economics stages.
ENGINEERING SUPPORT AND ECONOMICS The amount of time spent making a conceptual development plan for an exploration prospect is minimal. With the small amount of information available concerning the nature and extent of the
Otis and Schneidermann
1095
Figure 6—Three-point-method spreadsheet illustrates volumetric parameter ranges and shows calculations based on Pearson-Tukey estimator and the three-point method. M = million.
Figure 7—An economic summary sheet provides critical economic and geologic information and provides a mechanism for estimation of commercial or economic risk. M = million.
Otis and Schneidermann
Figure 8—A risk histogram of evalution wells, 1989–1990, illustrates predicted and actual results for feedback into the risk assessment process.
10 S L 8 L E W F 6 O R 4 E B M U 2 N
0
1097
1:2
1:4
1:6
1:8
1:10
1:12
1:14
1:16
>1:16
RISK reservoir (or even if there is a reservoir), f luid properties, or amount of resource present, our experience indicates the time and costs of preparing a detailed development plan for a specific case are generally not justified. However, significant attention is given to the credibility of general plans covering a range of cases that rely heavily on analogs or nearby producing examples. This approach is discussed in the following paragraphs. The first step is to take the mean reserve case from the volumetric distribution and construct a “mean” development plan. This plan uses the mean parameters from the volumetrics and mean parameters for reservoir fluid and flow properties to construct a mean production profile. This becomes the mean case (base case) for which facilities, drilling, and transportation costs are estimated. From this information, the revenue profile, based on the production profile and a product price assumption; an investment profile, based on the phasing of drilling, facilities, and transportation costs; an operating cost profile, based on an expected opex/bbl as a function of time; and a miscellaneous expense profile characterize the “mean” development plan and are used as input for the economic model prepared for the prospect. The economic model is then prepared based on the host country contract, if available. If no contract is available, the economic model is based on other known contracts or other published information pertinent to the country. The economic model takes as input the production, investment, operating cost, and miscellaneous profiles and applies the contract terms, resulting in o utput profiles of net income to the company and other tax-related profiles, such as depreciation, royalty, and income tax. The model remains f lexible; if negotiations are not complete, the contract usually becomes a subject of the negotiations and commonly changes.
The engineering and economic phases generally require refinement and involve a feedback loop to mature the mean case. In other words, t he engineer constructs the conceptual development plan and economics are run. Economic output is examined, and an optimization loop among earth scientist, engineer, and economist generally takes place, resulting in modifications or refinements to th e plan and subsequent economic output. Modifications are generally applied to facilities and drilling plans because of preliminary poor economic indicators. If modifications do not result in economics acceptable for a commercial project, the prospect is generally abandoned at this stage. The construction of this “mean” development plan generally takes from 1 day to 2 weeks, depending on the time available before a decision point and the information available. Once the mean case is completed, pessimistic (P10) and optimistic (P90) cases are run by modifying the mean case input profiles to the economic model. Modifications are based on the pessimistic and optimistic reserve cases from the reserve distribution. Economics are run for these two additional cases, and a range of economic outcomes is established. Volumetrics, development and contract assumptions, and economic results are summarized on a 1-page summary data sheet, as shown in Figure 7. The basic layout of the summary is a synopsis of terms, development assumptions, and a range of vol umetric param ete rs and t heir impact on eco nomic results. Two graphs are displayed that show (1) the volumetric distribution, both cumulative and density, and (2) the resultant ROR (rate of return) for the unrisked case and several risked cases. From these graphs, one can easily see the economic consequences of the expected distribution of reserves, development plans associated with that distribution, and the contract. Additional information, such as NPV (net present value) and NCF (net cash flow), is
1098
Evaluating Prospects
s l l e W f o r e b m u N
2 0 8 6 4 2 1 1
s l l e W f o r e b m u N 6 1 : 1 >
2 0 8 6 4 2 1 1
6 1 : 1
6 1 : 1
4 1 : 1
4 1 : 1
2 k 1 s : i 1 R
2 k s . 1 : i r 1 R y
0 1 : 1 8 : 1
2 9 9 1
6 : 1
4 9 9 1
4 : 1 2 : 1
s l l e W f o r e b m u N
s l l e W f o r e b m u N
2 0 8 6 4 2 1 1
2 0 8 6 4 2 1 1 6 1 : 1 > 6 1 : 1 4 1 : 1 2 1 : 1 k s i 0 R 1 : 1 8 : 1
1 9 9 1
6 1 : 1 >
6 : 1 4 : 1 2 : 1
3 9 9 1
4 f o 0 d 1 : o i 1 r e p 8 a : r 1 e v o 6 k : s 1 i r f o 4 t : n 1 e m s s 2 : e s 1 s a n i t n e m e v o r p m 6 i 1 f : o 1 s > s e r 6 g 1 o : r 1 p w 4 o 1 : h s 1 4 9 2 9 1 : k 1 – 1 s 1 i 9 0 R 9 1 1 r : o 1 f s m 8 : a r 1 g o t s i 6 : h 1 k s i R 4 : — 1 9 e r u 2 : g i 1 F
Otis and Schneidermann
) g % n ( i d n a n h i F T f s o s y e t i l i L b s a e b v r o r e P s e R
Predrill Reserve Distribution 100 80
Actual Reserves, 190 MBO, corresponds to 64th percentile
60 40 20
1099
Figure 10—Predicted distribution of reserves with actual results at the indicated percentile. In this case, the actual reserves of 190 MBO fell on the 64th percentile of the distribution.
0 0
100
200
300
Reserves (MBO)
also plotted at the P10, mean, and P90 cases to illustrate results for those parameters as well. Given the range of possible outcomes for the volumetrics and their economic consequences, an estimate of commercial risk is easily determined. Given the conditions of commerciality, usually a minimum ROR, the probability of a commercial prospect can be read directly from the two graphs. In Figure 7, if a 20% ROR is considered a minimum for a commercial project, from the bottom graph a 20% ROR corresponds to a reserve of 11 MBO (million barrels of oil). From the top graph, 11 MBO corresponds to a 50% probability of finding that reserve or more. Thus, the probability of commercial success is approximately 50%. This will vary from prospect to prospect, but this link is the fundamental driver for this process. In other words, we need to understand what nature has provided, which is the volumetric distribution that describes what we might find when we drill the well. We must also understand the economic consequences; that is, what nature has pro vided may or may not yield satisfactory economics. Analysis of both geologic and commercial risk in this manner allows appropriate decisions regarding risk tolerance and potential reward.
POSTDRILL REVIEW Postdrill information is primarily used as feedback to the risk assessment and volumetric estimation phases of the process. Feedback to the engineering and economics sections generally does not occur within a time frame that can impact the process. In other words, by the time a discovered field is developed and feedback is obtained, the process has already changed because of other, more timely, reasons. Postdrill information is obtained from a postdrill well review conducted within a few months after completing the well. Data analyses are collected and reviewed to (1) determine reasons for failure if
the well is unsuccessful, (2) compare predicted and actual reserves parameters if the well is successful, and (3) review lessons learned regardless of the result. Individual postdrill well reviews are compiled on an annual basis to provide statistical feedback, using simple histograms for both risk assessment and volumetric estimation. The first tool is the risk histogram, a simple plot of well results vs. risk expressed as a fraction of probability of success. Figure 8 shows a risk histogram from an actual 1989–1990 drilling program of wells drilled in producing areas on producing plays (evaluation wells). As is evident from the plot, the bulk of the wells had predrill proba bility of geological success between 1:3 and 1:6 (30–15%). From the histogram, it was immediately obvious that the number of successful wells is inconsistent with the asses sed r isk. Fo r those wells with assessed risk of 1:2, or 50%, 100% of the wells were successful. For those wells with assessed risk of 1:3, or 33%, 87% of the wells were successful, and so on. In fact, the average success rate for all wells drilled was 50% rather than the 20–25% predicted by the mode of the histogram. For this type of well (proven play in a producing area), our first modification to the process was to modify our process of assessing risk to better reflect our actual success rate. Figure 9 shows the risk histogram for each of the subsequent years (1991–1994). Although our efforts to more correctly assess risk were not immediately successful, over the 4-yr period improvement is evident, and by 1994 our predicted success rate is more consistent with that observed. As a side note, examining drilling results prior to 1989 indicated a similar trend. The success rate for wells drilled on proven plays in producing areas is about 50%, or 1:2, whereas the predicted rate was about 0.3–0.2, or 1:3 to 1:5. However, no attempt was made to adjust risk assessment methods until the process was implemented in 1989. Apparently, everyone knew the answer, but without a methodical,
1100
Evaluating Prospects
31 MBO 25%
75 MBO 91%
9 MBO 22%
250 MBO 75%
100
100
100 100
100
50
50
50 50
50
0 0
20 40
60
80
0
0
100 200 300 400
6
0 0
0
10
20
30
40
0
0
25
50
75 100
Percentile Histogram Number of Occurrences
4 2 0 0
20 40 60
80 100
Figure 11—Example of percentile histogram with four predicted distributions and actual results. This histogram is used to calibrate estimation of predrill volumetric parameters with actual results.
periodic performance review, little was done to modify the process. Thus, the feedback step is considered critical to the success of any process; wi tho ut i t, no proce ss wil l be mod ifi ed a nd improved. Volumetr ic esti mation fee dback is somewhat more complicated because it requires a method to determine whether distributions are being accurately estimated. Our volumetric feedback process consists of two steps. The first step is to determine whether reser ve distributions are accurate. The second step is to determine whether the individual reserve parameters are accurate. The method is the same for both steps and uses a second tool, the percentile histogram. The percentile histogram is constructed in the following way. Given a set of successful wells, each with a predicted distribution of reserves, calculate the probability of occurrence for the actual reserves on the predicted parameter distribution. For example, in Figure 10 a predicted distribution of reserves is shown where the actual reserves of 190 MBO correspond to the 64% probability of occurrence. Extending this to the set of four wells, as shown in Figure 11, the percentiles of the actual reserves on the predicted reserve distributions 1–4 are 25, 75, 21, and 91%, respectively. If these probabilities of occurrence for the four distributions are plotted as a histogram of occurrences in the ten dectiles (ten 10% intervals), the result is a percentile histogram, also shown in Figure 11.
The percentile histogram can be used to diagnose a variety of problems, as shown in Figure 12. The desired response is “flat.” In other words, if we are estimating distributions correctly there is an equal probability that the actual reserves will fall within any one of the ten dectiles (ten 10% intervals). It is analogous to rolling a ten-sided die, because each side (a 10% interval) has an equal probability of occurrence. Diagnostics are relatively simple. If the histogram is heavy to the low, or downside, we are tending to overestimate potential. In other words, most of the actual results are on the downside of the distribution. If the histogram is heavy to the high, or upside, the opposite is true; most of the actual results are on the upside of the distribution, indicating a tendency to underestimate reserves. If the histogram is heavy on the ends and light in the middle, prospect reserve ranges are too narrow and need to be broadened. If the histogram is heavy in the middle, ranges need to be reduced. Figure 13 shows the percentile histogram for reserves for Chevron Overseas Petroleum, Inc., in 1989–1990. The histogram is heavy to the downside; thus, we had overestimated potential in the majority of cases and needed to account for the large number of small discoveries we had made. We knew we had to correct this problem, but the primary cause required additional analysis. To determine what was causing the overestimation of reserves, we applied the same method
Otis and Schneidermann
d e r t e h t g n i e e C w •
to individual parameters. The percentile histograms for the individual parameters are shown in Figure 14. The following observations were made:
s i n o i t e u d b i i r t w s o i o D t •
n o l d s e a n d i d a o - s h m w g i i o B l h •
s i n o w i t o u r r b a i r n t s o i o D t •
e o d t i s w e h g k i S h •
o o n t o n c i o t i t i s u e b m i i d r t s i s s s i e p D p u •
n o y r o e t c d a i s f n s i t w a o S d •
o e t d i w e s k w o S l •
o o t n n o o c e i t i t d u s i i b s i n r m t i t s w i p o D o d •
n o y r o t c a e f s d i i t s a p S u •
n o i t d m u e r b r o i i r s f i t s e n i D u d •
y r o s t n c o a i t f s u i t b i r a s t s i e r D a •
1101
. s n o i t a t e r p r e t n i c i t s o n g a i d h t i w s m a r g o t s i h e l i t n e c r e p f o s e l p m a x E — 2 1 e r u g i F
(1) Estimates for gross pay and area were consistently overestimated. (2) Estimates of net-to-gross ratio (N:G), porosity, hydrocarbon saturation, and formation volume factor (FVF) were too narrow. (3) The geometry factor was not being estimated correctly. Modifications were made to tie ranges of gross pay and area to the expected hydrocarbon column. Research indicated columns associated with previous ranges of gross pay and areal extent were grossly overestimated, so considerable attention was given to hydrocarbon columns expected for different seals, especially fault seals. Other modifications included widening ranges for N:G, porosity, hydrocarbon saturation, and formation volume factor, as well as introducing a different approach to estimating geometry factor. Figure 15 shows the reserve histogram and Figure 16 shows the parameter histograms for 1993–1994. The reserves and all parameters have percentile histograms that are within the statistical tolerance of being acceptable for the number of samples, and it is obvious they are being estimated with improved accuracy. The histograms are much closer to the desired “flat” response. Based upon this feedback for both risk assessment and volumetric estimation, we observed a discrepancy between predicted and actual results, analyzed the data to determine where improvements could be made, implemented those changes, and observed a favorable response when predicted and actual results were in better agreement. The feedback was absolutely necessary to establish credibility and build support for the continued use of the process.
CONCLUSION Since its inception in 1989, application of this process has resulted in a consistent method of assessing risk, estimating volumes of hydrocarbons, and, thus, calculating economic indicators that can be used to judge the potential of exploration prospects. Through yearly feedback and modifications, credibility has improved, and the process has been accepted by Chevron upstream operating companies as a basis to assess the potential of opportunities in Chevron’s worldwide exploration prospect inventory. The process is used routinely in international exploration activities and has been the subject of numerous training sessions with partners and host countries.
1102
Evaluating Prospects
Figure 13—Actual percentile histogram for years 1989–1990. Diagnostics indicate distribution estimates were too optimistic on downside uncertainty (downside and median estimates were too large).
Figure 14—Actual percentile histograms for parameters of reserve distribution for years 1989–1990. Note problems with area, gross pay, geometry factor, porosity, and hydrocarbon saturation.
Otis and Schneidermann
1103
Figure 15—Actual percentile histogram for years 1993–1994 after modifications to process. Note distributions are more consistent with desirable uniform distribution.
Figure 16—Actual percentile histograms for parameters for years 1993–1994 after modification to process. Note problems have essentially been eliminated and distributions are consistent with desirable uniform distribution.
1104
Evaluating Prospects
Figure 17—Step 1 of three-point method for calculating reserve distributions: specify parameter ranges. M = million.
APPENDIX 1: THREE-POINT METHOD The three-point method, as developed by J. E. Warren (1980–1984, personal communications) for reserve estimation, uses the general equation shown below, which combines individual parameters in calculating recoverable reserves, R. R(oil) = 7758 × A × h × φ × S h × (1 B oi ) × R fo R(gas) = 43560 × A × h × φ × S h × (1 B gi ) × R fg R(condensate) = 43560 × A × h × φ × S h × (1 B gi ) × R fg × CR
where A = areal extent of prospect in acres, h = average net pay in feet, f = average porosity, S h = hydrocarbon saturation (1 – S w , where S w = water saturation), Boi = initial oil formation volume fac-
tor in reservoir barrels/stock tank barrels (STB), B gi = initial gas formation volume factor in reservoir cubic feet/surface cubic feet, R fo = recovery factor for oil, R fg = recovery factor for gas, CR = condensate recovery factor in STB/ft3, 7758 = conversion factor from acre-feet to barrels, and 43560 = conversion factor from acre-feet to cubic feet. The parameters are combined by multiplication; therefore, if the parameters are assumed to be probabilistically independent, the reserve distribution, R, will be lognormal in the limit as provided by the central limit theorem. Likewise, the first and second moments of R [m(R) and m 2(R)], respectively, will be the product of the first and second moments of the parameter distributions, respectively, as shown. Note that the first moment of the distribution is the mean. m[R ( oil )] = 7758 × m( A ) × m( h) × m(φ) ×
( h ) × m(1 B
m S
oi
) × m(R fo )
(1)
Otis and Schneidermann
1105
Figure 18—Step 2 of three-point method for calculating reserve distributions: calculate parameter means and variances. M = million. m 2 [R ( oil)] = 7758 × m2 ( A ) × m2 ( h) × m2 (φ) ×
( h ) × m (1 B
m2 S
2
oi
) × m2 (R fo )
(2)
Wi th th e fi rs t an d se co nd mo me nt s of R, th e lo gn or ma l reserve distribution is completely specified. Even if probabilistic independence is not strictly valid, the results are a useful approximation, given the level of information generally available to an exploration project. In practice, the uncertainty in specifying the ranges of input parameters is far greater than the amount of uncertainty introduced by assuming parameter independence. The first and second moments of R are calculated using equations 1 and 2 and estimates of the first and second moments of the
input parameter distributions. These estimates are obtained using the Pearson-Tukey estimator (Pearson and Tukey, 1965; Keefer and Bodily, 1983). An example for the area, A, is m ( A ) = 0.185 × P5( A ) + 0.63 × P50( A ) + 0.185 × P95( A) 2
2
2
m 2 ( A ) = 0.185 × P5( A ) + 0.63 × P50( A ) + 0.185 × P95( A)
where P5 = the 5% probability of occurrence of the area distribution, P50 = the median of the area distribution, and P95 = the 95% probability of occurrence of the area distribution.
1106
Evaluating Prospects
Figure 19—Step 3 of three-point method for calculating reserve distributions: calculate mean and variance of reserve distribution. M = million.
The Pearson-Tukey estimator is used because of its robustness in estimating mean values from a wide variety of nonsymmetric distributions, including the popularly used triangular distribution. Thus, the estimated mean values estimated are not restricted to any assumptions of distribution, such as those necessary for a Monte Carlo simulation, and allow the Earth scientist a reasonable amount of freedom in choosing the input values for the P5, P50, and P95 estimates. At this point it is useful to introduce a more convenient parameterization, ∂2, the variance of the natural logarithm of R. ∂2 is calculated using the following formula.
[
2
∂2 = ln m 2 (R ) m(R )
]
It is easy to show that the variance of the natural logarithm of R is the sum of the ∂2 of the individual parameters. Thus,
∂2 [R ( oil )] = ∂2 ( A ) + ∂2 ( h) + ∂2 ( φ) + ∂2 ( S h ) + ∂2 (1 B oi ) + ∂2 (R fo )
Otis and Schneidermann
1107
Figure 20—Step 4 of three-point method for calculating reserve distributions: calculate values for different probabilities of occurrence. M = million.
and any percentile value of the lognormal distribution can be calculated using the formula ∂z x R x = P50(R ) × e ( )
where P50( R) = m( R) * e-0.5 ∂2 (the median of the distribution), x = the probability of occurrence desired, z(x) = the value or zfactor corresponding to the x-percentile of the standard normal
distribution (obtained from tables given in most probability textbooks). Figures 17–20 show a spreadsheet with the example from the text and illustrate the calculation process. Step 1: Specify the parameter ranges. Step 2: Calculate a mean and ∂ (variance) for each parameter. Step 3: Multiply the parameter means and sum the ∂ to obtain the mean and ∂ of the reserve distribution. Step 4: Calculate values for different probabilities of occurrence as listed in the table and plotted on the cumulative distribution.
1108
Evaluating Prospects
REFERENCES CITED
and methods, in L. B. Magoon, ed., The petroleum system— status of research and methods, 1990: U. S. Geological Survey Bourdaire, J. M., R. J. Byramjee, and R. Pattinson, 1985, Reserve Bulletin 1912, p. 1–9. assessment under uncertainty—a new approach: Oil & Gas Magoon, L. B., and W. G. Dow, eds., 1994, The petroleum sys Journal, June 10, v. 83, no. 23, p. 135–140. tem—from source to trap: AAPG Memoir 60, 655 p. Capen, E. C., 1993, A consistent probabilistic approach to reserves Megill, R. E., 1984, An introduction to risk analysis: Tulsa, estimates: Society of Petroleum Engineers Hydrocarbon Oklahoma, PennWell Books, 274 p. Economics and Evaluation Symposium, SPE Paper 25830, Nederlof, M. H., 1979, The use of habitat of oil models in explop. 117–122. ration prospect appraisal: Proceedings of the 10th World Demaison, G., 1984, The generative basin concept, in G. Demaison Petroleum Congress, p. 13–21. and R. J. Murris, eds., Petroleum geochemistry and basin evaluaNewendorp, P. D., 1975, Decision analysis for petroleum explotion: AAPG Memoir 35, p. 1–14. ration: Tulsa, Oklahoma, PennWell, 668 p. Demaison, G., and B. J. Huizinga, 1991, Genetic classification of Otis, R. M., 1995, Five year look back at risk assessment and estipetroleum systems: AAPG Bulletin, v. 75, p. 1626–1643. mation of hydrocarbon volumes (abs.): AAPG 1995 Annual Dow, W. G., 1972, Application of oil correlation and source rock Convention Program, p. 73A. data to exploration in Williston basin (abs.): AAPG Bulletin, Otis, R. M. and N. Schneidermann, 1994, A process for valuation of v. 56, p. 615. exploration prospects (abs.): AAPG 1994 Annual Convention Dow, W. G., 1974, Application of oil correlation and source rock Program, p. 228. data to exploration in Williston basin: AAPG Bulletin, v. 58, Pearson, E. S., and J. W. Tukey, 1965, Approximate means and no. 7, p. 1253–1262. standard deviations based on distances between percentage Haun, J. D., ed., 1975, Methods of estimating the volume of undispoints of frequency curves: Biometrika, v. 52, no. 3–4, covered oil and gas resources: AAPG Studies in Geology 1, 206 p. p. 533–546. Jones, R. W., 1975, A quantitative geologic approach to prediction Perrodon, A., 1980, Géodynamique pétrolière. Genèse et répartition of petroleum resources, in J. D. Haun, ed., Methods of estimatdes gisements d’hydrocarbures: Paris, Masson-Elf Aquitaine, 381 p. ing the volume of undiscovered oil and gas resources: AAPG Perrodon, A., 1983, Dynamics of oil and gas accumulations: Pau, Studies in Geology 1, p. 186–195. Elf Aquitaine, p. 187–210. Journal of Petroleum Te chnology, 1996, SPE/ WPC draft reserves Perrodon, A., 1992, Petroleum systems: models and applications: definitions: Journal of Petroleum Technology, v. 48, no. 8, Journal of Petroleum Geology, v. 15, no. 3, p. 319–326. p. 694–695. Rose, P. R., 1987, Dealing with risk and uncertainty in exploration: Keefer, D. L., and S. E. Bodily, 1983, Three-point approximations how can we improve?: AAPG Bulletin, v. 77, no. 3, p. 485–490. for continuous random variables: Management Science, v. 29, Rose, P. R., 1992, Chance of success and its use in petroleum no. 5, p. 595–609. exploration, in R. Steinmetz, ed., The business of petroleum Magoon, L. B., 1987, The petroleum system—a classification exploration: AAPG Treatise of Petroleum Geology, Handbook scheme for research, resource assessment, and exploration of Petroleum Geology, p. 71–86. (abs.): AAPG Bulletin, v. 71, p. 587. Whi te, D. A., 1980 , Asse ssi ng o il a nd g as pla ys i n faci es- cyc le Magoon, L. B., 1988, The petroleum system—a classification wedges: AAPG Bulletin, v. 64, no. 8, p. 1158–1178. scheme for research, exploration, and resource assessment, in Whi te, D . A., 1988, Oil and gas pla y maps in exp lor ation and L. B. Magoon, ed., Petroleum systems of the United States: U.S. assessment: AAPG Bulletin, v. 72, no. 8, p. 944–949. Geological Survey Bulletin 1870, p. 2–15. White, D. A., 1993, Geologic risking guide for prospects and plays: Magoon, L. B., 1989, The petroleum system—status of resea rch AAPG Bulletin, v. 77, p. 2048–2061.
Otis and Schneidermann
1109
ABOUT THE AUTHORS Robert M. Otis Bob Otis is supervisor for Cabinda B/C Exploration, Chevron Overseas Petroleum, Inc. Previous Chevron experience includes manager, exploration evaluation division, coordinator Argentina exploration, and coordinator Middle East exploration. Before joining Chevron, Bob worked one year for the Western Division of Sohio (California and Alaska) and eight years for Mobil in Gulf Coast and Alaska exploration. He received a B.S. degree in 1969 and a Ph.D. in 1975, both from the University of Utah.
Nahum Schneidermann Nahum Schneidermann is director of international technical relations, executive staff, Chevron Overseas Petroleum, Inc., San Ramon, California. A native of Zayadin, former Soviet Union (now Uzbekistan), Schneidermann received his bachelor’s and master’s degrees from the Hebrew University of Jerusalem, Israel, in 1967 and 1969, respectively, and his Ph.D. from the University of Illinois, Urbana, Illinois, in 1972. His career in the industry started in 1974 with Gulf Oil, where he held various positions at the Houston Technical Services Center. In 1985 he started his tenure with Chevron Overseas Petroleum in San Ramon, serving as manager, basin studies and geochemistry, for the exploration department prior to being named to his present position.