SPE 49030 Risk Analysis: Lessons Learned

J.A. Alexander, SPE, and J.R. Lohr, Spirit Energy 76, Unocal
Copyright 1998, Society of Petroleum Engineers, Inc. This paper was prepared for presentation at the 1998 SPE Annual Technical Conference and Exhibition help in New Orleans, Louisiana, 27-30 September 1998. This paper was selected for presentation by an SPE Program Committee following a review of information contained in an abstract by the author(s). Contents of the paper, as presented, have not been reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material, as presented, does not necessarily reflect any position of the Society of Petroleum Engineers, its officers, or members. Papers presented at SPE meetings are subject to publication review by Editorial Committees of the Society of Petroleum Engineers. Electronic reproduction, distribution, or storage of any part of this paper for commercial purposes without the written consent of the Society of Petroleum Engineers is prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300 words; illustrations may not be copied. The abstract must contain conspicuous acknowledgment of where and by whom the paper was presented. Write Librarian, SPE, P.O. Box 833836, Richardson, TX 75083-3836, U.S.A., fax 01-972-952-9435.

6.

7.

Risk methods should be adaptive to allow integration of new techniques, such as seismic direct hydrocarbon indicators (DHCIs), into the process. Results from risked projects should be understood and tracked closely over time to facilitate adjustments and enhancements to the methods.

Abstract During the past 30 years the petroleum industry has generated significant information on risk analysis methods for exploration and development projects. However, some people continue to ask the question Is it worth the effort? or comment It has not worked for me. The authors have learned many lessons during the past few years that resulted in an improved risk analysis process. The following recommendations will ensure a successful risk analysis process. 1. Written guidelines are necessary to avoid misapplication of established methods. These guidelines should be well documented, communicated to everyone involved in the process and regularly updated. Oversight of the risk evaluation process is necessary to maintain consistency, eliminate bias, and check for misinterpretations that occur through lack of understanding. Risk analysis training for both technical professionals and management is essential. Evaluation software should be accessible, standardized, easy to use and adaptable to changing technology and needs. An understanding of dependency between variables is vital.

Introduction C.J. Grayson (1960) is credited with introducing risk analysis to the industry. In 1968, Paul Newendorps Risk Analysis in Drilling Investment Decisions introduced a methodology for assessing and describing the degree of uncertainly involved in evaluating a prospects profitably. Since then, several other papers have been published, books have been written and many training seminars have been offered to assist the industry to understand risk analysis for exploration and development projects. Despite these efforts, why do some continue to doubt the value of risk analysis? We believe it is because they do not have all the essential elements of the processes in place to ensure success. Our experience has helped us identify seven key elements of a successful risk analysis process. We will discuss these elements in this paper while sharing lessons learned and common pitfalls we have noticed in risk analysis. It is not our intent to debate or explain any of the mathematical methods previously introduced in the industry by authors such as Newendorp, Capen, Murtha, Rose, Garb, Smith and Megill. Written Guidelines and Communication Written guidelines are necessary to help avoid misinterpretation of established methods. They should be well-documented, fully communicated and regularly updated. There are many things that can go wrong in risk analysis. Overestimation, underestimation, misidentifying critical risks, overselling projects and underselling projects are some of the problems. To ensure that risk analysis results in better decisions, it must be applied consistently. How can a company choose between two projects if there is no consistency in the

2.

3. 4.

5.

J.A. ALEXANDER, J.R. LOHR

SPE 49030

risking? How can a company successfully manage a portfolio without consistent risking methods? Written guidelines are required to promote consistent risk analysis. Our first experience was with a risk analysis manual introduced in 1992. Unfortunately, users interpreted the manual differently, sometimes taking the guidelines too literally at the expense of sound professional judgment. In the past 1-1/2 years our risk analysis team has reviewed more than 100 projects drilled in 1997 and 1998 and completed a detailed lookback on 75 additional projects drilled between 1994 and 1996. The inconsistencies we observed during these reviews supported the need to update the risk analysis manual and to provide oversight of the risk analysis process. Some of the inconsistencies we found during these reviews include: 1) A variety of nonstandard and occasionally conflicting risking tools being used, 2) Misinterpretation of P10, P50 and P90 definitions when working with distributions, 3) Risking of drilling costs was not commonly done or if it was, the methods were not compatible, 4) Inconsistent methods for risking multiple object zone were practiced, 5) Dependencies between zones and between projects were treated in various ways, 6) Project teams range widely from optimistic to pessimistic 7) Method for truncating or limiting reserve distributions varied. Because of these inconsistencies our guidelines and processes have been updated to improve the consistency of our risk analysis. The results of these changes are positive and are discussed latter. The following are details on a few of the lessons learned. These lessons are being shared to increase the awareness of some of the common problems that surfaced. Lesson 1. Distortions in characterization of reserve distributions have occurred because of the interchangeable use of P90 and P99. In other words, evaluators may have meant the maximum reserve for the project was 10 bcf (absolute maximum), but called it P90 in the calculations of the mean expected reserves. This has a significant impact on the mean expected reserves. Figure 1 is a plot of Percent of Overestimation versus Ratio of P90/P10 Value. For example, if the P90 value is actually P99, P10 is truly P1, and the ratio of P90 to P10 is 10, then the mean expected reserve is overstated by 31%. This might represent a case with a reserve distribution between 1 and 10 bcf as shown in Figure 2. This distortion is the result of only one variable being misused. Since most reserve distributions are a result of more than one variable (i.e. area, pay, or recovery factor), this problem becomes magnified if all individual factors are similarly misrepresented.

Lesson 2. Improper truncation of the lower end of a reserve distribution frequently causes problems. This occurs when there is confusion about whether the reserves should be truncated at the reserve value that made the project profitable or at the reserve value that made the completion profitable. Depending on the ratio of completion cost to dry hole cost, this can significantly impact the economics of the projects. The original reserve distribution should be truncated at the reserve value that represents the possible reserve outcomes if the completion is profitable, not the project. When reserves are truncated, the probability of success (POS) is also adjusted. Figure 3 illustrates this with a decision tree. The four-point method and Monte Carlo simulation are both valid methods for risked economic evaluations. In either method, a reserve distribution associated with the probability of finding hydrocarbons (POSg) is developed. The reserve value at which to truncate the reserve distribution is then determined. The truncated reserve distributions and the revised POS are then used in the economic evaluations. We identify the revised POS as POS(ic) for the probability of incremental commercial success. Because we also need to know the chance the project will be commercial with full cycle economics, we use POSc to define the probability of the project being profitable (commercial). Lesson 3. Use of an Intranet site to maintain and update risk analysis guidelines avoids problems associated with the difficulty of maintaining and distributing hard copy manuals. An Intranet site also can be designed to accommodate changing staff who have a variety of experiences with risk analysis. At the Intranet site, staff also can gain access to examples, lookback results, risk software, references and lessons learned. Risk Evaluation Oversight Process oversight is necessary to maintain consistency, eliminate bias, and check for misinterpretations that occur through lack of full understanding and use of flawed assumptions. One way to achieve this is to assign a crossfunctional team the task of overseeing the risk process for all exploration and development projects. The teams responsibility is to review the risk and reserve estimations for most of the exploration and development projects, update risk guidelines, conduct lookback analysis, and provide training and coaching on a project basis. Lesson 4. A major element of risk analysis is and always will be subjective judgment. It is essential that people with knowledge and critical analysis skills are involved in the analysis process. The peer review process is an effective way to incorporate this critical component. The peer review is a useful tool for enhancing the risk analysis evaluation if these recommendations are followed: First, the review should be done early and not delayed until the final approval stage. Second, peers should be selected that will give

SPE 49030

RISK ANAYSIS LESSON LEARNED

constructive feedback and will not introduce their own biases. Third, the peers should have experience with projects similar to the one being reviewed. Fourth, the results of the projects should be communicated back to the peers so the peers can also learn from the results. Training / Understanding Risk analysis training should be required of all employees who use the process and particularly managers who make decisions based on the results. Optimally, this training should be offered in a concentrated fashion, ensuring that the greatest number of employees can be trained within the shortest possible period. Lesson 5. Train everyone including managers in a short time period. This improves communication and will increase the chance of having a successful risk analysis process. Lesson 6. Several papers and books have been written on risk analysis. All of those involved in the risk analysis process should be encouraged to read and review this material to increase their understanding of risk analysis. Lesson 7. External training should be supplemented with focused internal training, mentoring and coaching. This is the responsibility of the risk analysis team. It is achieved during the peer reviews, the project reviews, lunch and learn meetings, lookback meetings, and individual sessions. This has proven an effective technique because it addresses the teams immediate concerns and specific needs. This method of training also keeps the risk analysis team close to the teams and in touch with the current problems. Risk Evaluation Software Evaluation software should be accessible, standardized, easy to use, and adaptable. Over the past years, many versions and varieties of software have been developed to assist those doing risk analysis. A tremendous effort has been invested by individuals developing their own software to meet their specific needs. This software proliferation leads to nonstandard and sometimes conflicting risk software. Teams must reevaluate all risk analysis software to ensure the needs of individual teams are met while providing a consistent evaluation framework. Some of the questions to ask are: 1. Should the software be internally or externally designed? 2. Should it be customized or off-the-shelf? 3. Should the software do risking, reserves, economics, portfolio management and deal screening? 4. Should the programs be PC- or Unix-based? 5. Should it be on the network or stand alone? 6. Does the program correctly calculate multiple zones reserves and POS? 7. Does the software correctly handle dependencies? 8. Should the program do Monte Carlo simulation?

Dependencies / Correlations An understanding of dependency or correlations between variables is also essential for proper risk analysis to prevent overstating or understating the risk and uncertainty. Even though authors such as Newendorp (1975), Garb (1988), Smith (1992), Murtha (1994,1996) and others have written on this subject, people are still struggling with it. Dependency deals with existence or Is it there? while correlation deals with the size or How big is it? Consideration of both provides a seamless fit with the correct and balanced view of risk and uncertainty. Dependency is the alteration in probability of the existence of a factor at location or zone X given the presence or absence of the same factor at location or zone Y. These factors can include reservoir presence, trap (vertical and seal), timing/migration and source. For example, if reservoir is found at location one, the probability of it being present in location two increases by 50%. Correlation on the other hand, is the quantitative or qualitative similarity between existing characteristics in two different locations or zones. For example if the permeability in location one is at P70, then the permeability in location two should be between P60 and P80. Much of this problem can be solved first by recognizing the dependencies or correlations and second by understanding how to handle the effect in the risk analysis. The best tool for recognizing correlations is the cross plot. If actual data for two variables are plotted against each other and a trend is noticed, then a correlation exists between the variables. Most spreadsheet programs can determine the degree of correlation. Some examples of variables with correlations include: Area and Pay Thickness Porosity and Water Saturation Rate and Pay Thickness Number of Wells and Ultimate Recovery Recovery and Well Location in the trap Lesson 8. One of the major correlations not always handled correctly is that of net pay thickness and area. This occurs when a single point value at the wellbore or a single point in the reservoir is used to reflect the distribution of net pay in the reservoir. In reality the net pay thickness may vary within the reservoir, but its thickness is correlative to the area. If this is not recognized, the reserve distribution model will include extreme cases that may not represent any reasonable geological model. This problem causes overestimation of the P90 net pay thickness and therefore the P90 volume and reserves. When

J.A. ALEXANDER, J.R. LOHR

SPE 49030

using distributions for area, pay and recovery factor, the expected reserves can be overestimated by more that 30% if the pay is overstated as shown in Figure 4. A simple check of this problem is to multiply the P90 area and P90 pay together and ask the evaluators if about 5% of the time the reservoir volume would be larger than the resultant product. The answer will be, no, if there is a correlation that has been handled incorrectly. The authors advocate building P90 and P50 reservoir volume maps as a way to have more confidence that the uncertainty of the reservoir volumes is understood. Lesson 9. The existence of dependencies between zones in a multiple-zone prospect is another item that can cause problems. Understanding how the result of one zone impacts the result of other zones is essential. Let us consider a prospect with four zones that have each been assigned a POS of 30%. If the zones are independent of each other, the POS of one of the four zones finding hydrocarbons is 87%. However, if the four zones are fully dependent, the POS of finding hydrocarbons is only 30%. This is a material difference that must be understood. Depending on the degree of dependency between the zones, the correct POS is between 30% and 87%. This concern is also valid for multiple-well prospects or multiple-prospect trends. The dependency between these wells or prospects needs to be understood to conduct proper risk analysis. Lesson 10. When using Monte Carlo simulation as a risk analysis tool, each iteration must be a possible real life scenario. Sometimes this tool is treated as a black box and the evaluator does not understand the effect of the variation in input parameters. If the input has been described correctly and reflects the dependencies correctly, the Monte Carlo output will represent the possible outcomes for the project. Adaptive Risk analysis methods should be adaptive to allow integration of new techniques such as seismic direct hydrocarbon indicators (DHCIs) into the process. Lesson 11. We have learned from reviewing the results of close-in exploration projects between 1994-1996 that the problem of getting on the reserve distribution for projects associated with seismic amplitudes was underestimated by 20% (Figure 6). However, POSg estimates were very close to actual for non-amplitude associated projects during this time. It appears the positive impact of direct hydrocarbon indicators on prospect risk was understated. Many of the wells in the sampling were drilled to Gulf Coast Tertiary sand objectives in and around existing fields. This is a geologic environment where amplitude anomalies are generally reliable indicators of hydrocarbons. Since 1992 the authors have used a geotechnical risk matrix to establish POSg for projects. A portion of the matrix, shown in Figure 10, illustrates the probability of the presence of

hydrocarbon system variables of trap, top seal, reservoir rock, source and timing and migration. Descriptors for each element relate certainty to data quality, data type, analogies and trend information. While the matrix was not designed specifically for evaluating the impact of DHCIs, careful considerations of the key risk factors that relate to the DHCIs will help apply them appropriately to risking. Each of the principal hydrocarbon system elements must be evaluated for the real impact of seismic amplitude anomalies. Although the matrix has proved to be an effective tool, it must continually accommodate improvements in technology such as seismic attributes. Trap (lateral seal) is the risk element most affected by DHCIs. Hydrocarbon saturation implies the presence of trapped hydrocarbons and a working hydrocarbon system greatly reducing the chance of failure. Seismic anomalies in general can be either qualitative or quantitative when using the risk matrix. Seismic anomalies, particularly when used as DHCIs, should be tested and calibrated by modeling and verified by analogies. Matching expected versus actual seismic responses is the cornerstone of a thorough quantitative analysis. Qualitative indicators of the presence of hydrocarbons can also be effective risk reducers. They include hydrocarbon related fluid contact flat spots, velocity sags, amplitude dimouts and phase changes. Reservoir risk is not as easily defined by seismic attributes, but some estimates of quality and thickness are possible. Seismic inversion/impedance characteristics that relate velocity to porosity and lithology may define minimum threshold for reservoir quality. Under some circumstances, thickness may be estimated reliably and related to minimum requirements for an active hydrocarbon system. Timing and migration, source and top seal are less directly related to seismic attributes. However, it is possible to distort the risk of an amplitude supported project if non-definitive estimates of these three elements are considered to have the same weight as quantifiable DHCIs related to definition of trap. Caution must be exercised in over- or underestimating the impact of seismic amplitude anomalies on prospect risks. Tracking / Understanding Results Results from risked projects should be tracked closely over time to facilitate adjustments and enhancements to the methods and process. The results tracking (lookbacks) should incorporate both technical review and statistical evaluation of results. A formal process of technical post-project reviews can be designed to promote consistency in risk and reserve estimations, technical best practices, technology transfer and tracking of results. Coordinators from various technical disciplines who serve as facilitators, process champions and mentors can sponsor these reviews. Post-project peer reviews

SPE 49030

RISK ANAYSIS LESSON LEARNED

need not be a formal part of the review system for all projects, but the information from pre-drill predictions of risk, reserves, value and cost should be compiled on the riskier wells and wells with more uncertainty. These data, accumulated since 1993, provide the basis for the statistical plots presented in this paper. The information should be disseminate to everyone from the project team to management in a discussion format that maximizes understanding and tailors applications of the risking process. The types of information and depth of detail of the data collection have changed over the past few years to enable more accurate measurement of results, identification of the critical success/failure causes and portfolio management. The risk and reserve team currently manages the project review and results analysis process to enhance the understanding and application of the best and most appropriate risking techniques. Predicted and actual performance data indicate that improvements in the process are having a positive effect. Post-drill information from project reviews conducted shortly after completing wells is collected to compare predicted to actual reserves and to review technical lessons learned regardless of the result. The feedback step is critical to the success of any process because without it, no process could be modified and improved. This is similar to Chevrons method of project risk evaluation and results tracking presented in Otis and Schneidermanns paper (1997). Lesson 12. It is necessary to constantly dig deeper to determine what is driving the results. Figure 5 shows the results of predicted versus actual POS, reserves and NPV (net present value) for 75 close-in exploration projects drilled between 1994-1996. This summary shows the results as acceptable. However, as the details are examined, several observations become evident. It became apparent that one of the 75 projects was driving the results of the summary. This project contributed 37% of the reserves and 52% of the NPV for the 75 well total. When this project is removed from the summary, the predicted versus actual results for the three-year period dropped to 66% for reserves and 54% for NPV. This level of predictability is not sufficient for the evaluation of individual projects or balancing a portfolio. However, the impact of the one project demonstrate the need for a balanced portfolio. When examining the results of lookbacks, one should focus on risk (i.e. POS) and uncertainty (i.e. reserves) separately. The two must be separated to understand the results and impact of each parameter. Some of the various ways the results for POS prediction were analyzed for 1994-1996 are presented in Figure 6. It shows POS prediction results by year, API well category, POS level,

and seismic amplitude response. Such data indicates that predictions for POS were conservative during the three years. The results also show an overall conservative POS driven by conservative estimates for API category 3 & 4 wells, high POS wells and wells with seismic direct hydrocarbon indicators. These results have been shared with the technical professionals and managers to provide the basis for changes to the interpretation of the geotechnical risk matrix and improvements in the results. The POS for the 25 close-in exploration projects drilled in 1997 was also conservative (131% over prediction) but, for a different reason than noticed in the 1994-1996 results. The underestimation in 1997 was driven by six successful wells which were treated as independent wells when in fact such wells were dependent. The strong dependency between the wells was not reflected in the original recommendation. If dependency had been represented correctly in the original recommendation, the actual results would have been closer to 100%. This information should be communicated throughout the organization to improve future results. Figure 7 shows some of the same methods for examining accuracy of predicting the uncertainty of reserves. These data indicate the predictions were understated for API category 3 and 4 wells and overestimated for most of the others. Figure 7 shows that 48% of the completed wells will recover less than P20 reserves. Figure 8 shows that 80% of the completed wells testing seismic amplitude anomalies, found less than P50 reserves. It also shows the wells testing nonamplitude objectives had a more balanced reserve distribution (61% below P50 and 41% above P50). The reason POS is understated and reserves overstated is due to problems described earlier in this paper. Spirit Energy 76 has improved its drilling results in 1997 and 1998 by utilizing the seven-element process. The actual reserves found for the 25 close-in exploration wells drilled in 1997 are 103% of prediction. No single project overly influenced these results as we saw in the 1994-1996 results. Other Reasons and Lame Excuses for Failure in a Risk Analysis Process Seven requirements for a successful risk analysis process have been discussed along with several lessons the authors have learned in their pursuit to improve risking results. Below is a list of other legitimate reasons that will cause the risk analysis process to fail, along with a list of lame excuses that are often given. Reasons Difficulty in quantifying risk and uncertainty or defining probabilities. Inexperienced people doing the analysis. Risk specialists not involved with day-to-day problems or not working as a member of the technical teams.

J.A. ALEXANDER, J.R. LOHR

SPE 49030

Education process is short-circuited. Incorrect analogy data used.

Lame Excuses Poor communication between individuals providing input on the range of uncertainty. Lack of support from management. Lack of professional judgment. Belief that risk analysis is only for the statisticians. Not using the experts. Not willing to change from traditional ways. Misinterpreting data (logs, samples, maps). Personnel turnover. Conclusions As an expansion of the fundamental evaluation techniques used by the industry, risk analysis is well worth the effort, if it is done properly. Seven essential elements of a successful risk analysis program have been presented along with some of the lessons learned by the authors. Successful application of these lessons requires commitment and endorsement from both management and technical professionals. To be successful, the process should include close oversight of risk evaluations including written guidelines, appropriate software, an understanding of dependency, adaptive processes, and results tracking. In the end, all risk evaluations need to be tempered by sound technical and professional judgment. Risk analysis does not replace professional judgment. It can supplement judgment, and if properly used, can improve project evaluations and selection. We hope this paper will add value to the industry by helping others avoid many of the common problems noted. Nomenclature DHCI Direct Hydrocarbon Indicator (seismic) P10: 10% probability the occurrence is less than this level P50: 50% probability the occurrence is less than this level P90: 90% probability the occurrence is less than this level POS: Probability of success. POSg: Probability of geological success. POSic: Probability of incremental commercial success. POSc: Probability of commercial success. Reserves: Proved and unproved reserves Acknowledgements We thank Bill Haskett, John Collins and our many colleagues and supporters for the insight they provided for this paper. References
1. Grayson, C.J., Decisions Under Uncertainty: Drilling Decisions by Oil and Gas Operations, Harvard Univ., Div. Of Research, Graduate School of Business Administration (1960), 402.

Newendorp, P.D.: Risk Analysis in Drilling Investment Decisions, JPT (June 1968) 579-85. 3. Newendorp, P. D.: Decision Analysis for Petroleum Exploration, PennWell Publishing Co., Tulsa, Okla. (1975). 4. Newendorp, P.D.: A Method for Treating Dependencies Between Variables in Simulation Risk Analysis Models, JPT (Oct. 1976) 1145-50. 5. Garb, G.A.: Assessing Risk in Estimating Hydrocarbon Reserves and in Evaluating Hydrocarbon-Producing Properties, JPT (June 1988) 765-776. 6. Smith, M.D. and Jones, D.R.: Trend Analysis, The Business of Petroleum Exploration, The American Association of Petroleum Geologist, Tulsa, Okla. (1992) p.215-229. 7. Murtha, J.A.: Incorporating Historical Data into Monte Carlo Simulation, SPECA (April 1994). 8. Murtha, J.A.: Estimating Reserves and Success for a Prospect with Geological Dependent Layers, SPERE (Feb. 1996). 9. Megill, R.E.: An Introduction to Risk Analysis, Petroleum Publishing Co., Tulsa, Okla. (1977). 10. Rose, P.R.: Exploration Economics, Risk Analysis and Prospect Evaluation (Telegraph Exploration, Inc. 1996 edition). 11. Capen, E.C.: The Difficulty of Assessing Uncertainty, JPT (August 1976) 843-850 12. Otis, R.M. and Schneidermann, N.: A Process for Evaluating Exploration Prospects AAPG (July 1997) 1087-1109 AAPG paper

2.

100 90 Two variables (I.e. pay and area)

Overestimate of Mean Value (%)

80 70 60 50 40 30 20 10 0 0 2 4 6 8 10 12 14 16 One Variable (I.e. reserves)

Ratio of Original P90 to P10 Value

Fig.1. Percent reserves are overstated if the estimator's P90 & P10 estimates are actually P99 & P1 values. The effect for various P90/P10 ratios are shown for one and two variables. Note the compounding effect of two variables being misrepresented.

SPE 49030

RISK ANAYSIS LESSON LEARNED

100
Reserves (BCF)
Interpretation 1

10
Interpretation 2

1 0.1 P1 P10 P50

Probability

P90

P99

Fig. 2. Impact of referring to P99 as P90 and P01 as P10 is illustrated. The expected mean reserves for Interpretation 1 is 4.7 bcf and 3.6 bcf for Interpretation 2. Expected mean for Interpretation 1 is 31% higher than Interpretation 2.

Fig. 4. - Example of area and pay correlation. As drainage area increases, the average pay thickness decreases for location X, however increases for location Y.

P10 Yes 70%

30%

8%

Actual / Predicted (%)

Completed Well Reserve Distribution. P50 P90 40% 30% 11% 8%

Yes

40%

Chance to complete well No 30% 12%

Drill

Chance to find HC No 60% 60%

140 120 100 80 60 40 20 0 POS Reserves NPV

Drill Decision Dont drill

Fig. 3. - Decision tree illustrating why reserve distributions should be truncated at the reserve level that makes the completion profitable.

Fig. 5 - Prediction results (actual / predicted) for 75 close-in exploration projects drilled between 1994 and 1996. Summary results are acceptable. However, one project contributed 37% of the reserves and 52% of the NPV.

J.A. ALEXANDER, J.R. LOHR

SPE 49030

POS Prediction Results

180 160 140 120 100 80 60 40 20 0

1994 3 year Total

Annual

API Well Category

Actual / Predicted (%)

Seismic DHCI (Amplitude)

POS Estimate

1995

1996

No

25-50

50-75

0-25

Analysis Detail

Fig. 6. - Various ways to look at POS prediction results for 75 close-in exploration wells drilled between 1994-1996. Results show the POS was understated for API Well Category 3 and 4 wells, the wells with direct hydrocarbon indicators and the high POS wells. The results for the other categories were closer to 100%. The actual/predicted results for POS for 25 close-in exploration wells drilled in 1997 was 131%. The 1997 results were driven by a strong dependency between six wells that was not originally recognized.

Reserves Prediction Results

350

Actual / Predicted (%)

Annual

API Well Category

300 250 200 150 100 50

1994 1995 1996

Seismic DHCI (Amplitude)

POS Estimate

No

0
1 2 3 4 5 3 year Total

Analysis Detail

Fig. 7. - Various ways to look at reserve prediction results for 75 close-in exploration wells drilled between 1994-1996. Although the results for the three year total were acceptable, the detail analysis suggest significant room for improvement. The actual/predicted results for 1997 was 103% with no single project overly influenced the results as in 1994-1996.

75-100

25-50

50-75

0-25

Yes

75-100

Yes

SPE 49030

RISK ANALYSIS LESSONS LEARNED

Reserve Prediction Results

60

Reserve Prediction Results for Seismic Hydrocabon Indicators 100 Occurance (%) 80 60 40 20 0 0-50
P Value

Occurance (%)

50 40 30 20 10 0 0-20 20-40 40-60 60-80 P Value (%) 80-100

DHCI No DHCI

50-100

Fig. 8. - Actual reserve P values (probability from original reserve prediction) for completed close-in exploration wells drilled in 1994-1996. Reserve P value results for the 1997 wells are more evenly distributed.

Fig. 9 - Results from 75 close-in exploration wells drilled in 1994-1996 indicate more low reserve P values occur when seismic hydrocarbon indicators were present.

PROBABILITY (Principle Descriptors (=what am I risking) 0.95 1.00 Condition is virtual to absolutely certain and data quality/control is excellent. 0.65-0.95 Condition is most probable and data quality/control is good. Most likely interpretation. 0.35-0.65 Condition is probable or data quality/control is fair. Less favorable interpretations possible. 0.05-0.35 Condition is possible or data quality/control is poor. Less favorable interpretations more likely. 0.00-0.05 Condition is virtually to absolutely impossible and data quality/control is excellent.

TRAP Closure of any trap(includes: structure, stratigraphic and faulting) Presence of trap hydrocarbon retention capability (lateral sealing) Identical trap in immediate vicinity successfully tested and defined by unambiguous data (such as seismic, well control, outcrop, and engineering) clearly verified closure and lateral seal. Analogous trap within trend successfully tested or defined by convincing data (such as well control, seismic, outcrop, engineering), indicating a probable closure and lateral seal. Similar trap within other trend/trends successfully tested and/or limited data suggests a probable closure and lateral sealing capacity. Trap is poorly defined and/or structurally complex, or on geological concepts only, unconvincing seismic and/or well data hints at closure and lateral sealing capacity. Identical trap proven unsuccessful in trend, and unambiguous data (such as seismic, well control, outcrop, and engineering) clearly establishes the absence of both closure and lateral seal.

Fig. 10. - Part of Geotechnical Risk Matrix used to risk trap potential. The remaining elements, top seal, reservoir rock, source rock and timing/migration are handled similarly. Wording and descriptions are used to facilitate communication and consistency. The basic assumption of the matrix is that risking is done to the probability of a P1 resource outcome. For purpose of this exercise, assume that P1 resource is the resultant of P10 area, recovery and net pay. Trap = probability of closure covering P10 area. Source = probability of presence of adequate source rock within the fetch area to produce P1 resource. Vertical Seal = probability of continuous interval with sealing capacity over an area of P10 acres. Reservoir Rock = probability of P10 thickness in average net pay over a P10 area with a P10 recovery factor. Timing/Migration = probability of timely migration route available to P1 resource into P10 area.

10

J.A. ALEXANDER, J.R. LOHR

SPE 49030

Fig. 11. - API well categorie