World Oil Forecasting Program: USER'S GUIDE Introduction We assume that the peak of World oil production will be a watershed in human history. Our goal is to forecast the year that production will reach its all-time peak. To achieve this goal, we employ historic production data, mathematics, heuristics, and special graphics in a unique new 'tool' called the 'World Oil Forecasting Program.' Our strategy is to rerun the Program annually to build up a series of forecasts which, taken together, comprise the 'World Oil Forecasting Method.' The series will inevitably converge on the peak. The website download contains the second in the series of forecasts. You can use the Program in two ways. First, you can study and use-as-is the 'base-line' National and World oil forecasts, as contained in the existing Program. Second, you can make your own oil production forecasts. The Program runs in the STELLA 6.0 environment, the same for both Macintosh and Windows. Some useful nomenclature follows: 'Prod' stands for 'Oil Production' [in billion barrels per year]. 'Q' stands for 'Cumulative Production' [in billion barrels]. 'EUR' means 'Expected Ultimate Recovery.' 'N' is short for 'Numeric model' -- 'G' refers to the 'Guide forecast,' the output of the N-model. 'H' is short for 'Heuristic model' -- 'J' refers to the 'Judgmental forecast,' the output of the H-model. Briefly put: N and H are models; G and J are forecasts. Metaphor: Don't confuse the violin with the concertos it plays. The Program and Method were designed and built by Dr. R. C. Duncan. The base-line forecasts, contained in the file downloads, were made by Dr. Walter L. Youngquist. We begin with the Program base-line study mode. Part I. Study the National Base-Line Oil Forecasts In this part we use US oil production as an example. For instance, you will observe from historic data that US production peaked in 1970 and has declined ever since. Moreover, the base-line forecast depicts a continuing decline in future US production. Here we simply study the base-line US production profile. 1. Launch STELLA 6.0 Run-Time 2. Launch the North America model. The model opens in the 'High-Level Mapping Layer' (explained below). 3. Launch the Graph Pad icon labeled 'Prod NH US' (little red square). The USA production curve appears (thick red line): Years on the horizontal axis and Production on the vertical axis. The Years range from 1960 to 2040. The Production scale ranges from 0.00 to 4.40 billion barrels per year. Production values from 1960 through 1996 are historic data; values from 1997 to 2040 are forecasts. This is the 'base-line' US oil production-and-forecast curve: the so-called 'Judgmental' curve. Note that the curve is tagged '1.' 4. Select Run from the Run menu. The model redraws the base-line J-curve. Note: If you are running STELLA in Windows, a small window will open to show additional options. You can move this window out of the way. We will be using it later. 5. Study Questions: In what year did US production peak? Why was there a temporary increase from 1978 to 1985? Why the overall decline since 1985? Close the Graph Pad. 6. Launch the Graph Pad labeled 'HistData Prod G Prod NH US' (little blue square). A graph appears with three separate curves (same scale as before, but now the curves are shown as thin lines). And grid lines help pinpoint the exact year and production values. 7. Select Run from the Run menu. Watch the model redraw the three curves. Curve 1 (green line, named 'HistData US') shows historic US oil production data from 1960 through 1996; from 1997 to 2040 the green line remains constant (i.e., horizontal) at the 1996 level (very useful to distinguish historic data from future values). Curve 2 (blue line, named 'Prod G US') is a parabolic curve 'best-fit' to the historic data by a special translated coordinate system technique: the so-called 'Guide' curve. Curve 3 (red line, named 'Prod NH US') is the same J-curve discussed in item 3 above, except it's now a thin line, and tagged '3'. 8. Study Questions: On the US production curve, how can you quickly distinguish the last data point (i.e., 1996) from the first forecasted value (i.e., 1997)? Note that the parabolic curve (blue, curve 2) does not follow the most recent production trend. Why? In that regard, how is it that the 'NH' forecast (Curve 3) does follow the most recent production trend? Hint: Proceed to the items g & h, number 8, Part II; then come back and answer the last question. Close the Graph Pad. 9. The base-line curves for the other North American producers (i.e., Canada and Mexico) can be studied likewise. Close the North America model. Part II. Make Your Own Oil Production Forecasts As previously stated, you can -- and we hope you will -- make your own forecasts for each of the World's top 42 oil-producing nations. Let's start from scratch. 1. Launch (again) the North America model. The file opens in the 'High-Level Mapping Layer.' As was mentioned, in run-time STELLA you can make any changes you wish, run each model, and observe (and print out!) the results. Note: There are three 'layers' in STELLA. For now we'll stay in the 'High-Level Mapping Layer' -- i.e., the layer that automatically opened. (Later we'll visit the lower layers.) 2. Launch the icon named 'HistData&Fcst US' (large blue square, located at the left-hand side of the screen). This is called a 'Graphical Input Device,' or 'GID' for short. On the graph, Years are displayed on the horizontal axis. The vertical axis displays the US historic oil production data (from 1960 thru 1996) followed by the production forecast (from 1997 thru 2040). Use the slider at the bottom of the GID to scroll across the entire US production curve from 1960 thru 2040. Very handy. 3. The GID 'Input' column (left) shows the years from 1960 to 2040. The 'Output' column (right) shows the annual oil production. Scroll down (use the little arrows on the slider bar). Then click on any number in the Output column. The number highlights. Type in a different value. Say 0.0 or 1.8 or 3.9 or 9.3. (Don't be bashful. It's only a demo.) Click OK. The GID closes. 4. Double-click the Graph Pad icon 'Prod NH US' (little red square). Select Run from the Run menu. The curve is redrawn with your change displayed. Close the Graph Pad. 5. Click the little 'Undo' [U] button at the bottom of the GID icon. Again open the little red Graph Pad icon 'Prod NH US.' Select Run from the Run menu. The original curve is redrawn. Very useful, because you can always restore the base-line model. Close the Graph Pad. 6. Repeat Steps 2 thru 5 above, but this time change a lot of production values in the Output column of the GID. 7. Recall that the numbers from 1960 thru 1996 are historic data. 'Water-under-the-bridge.' Leave as is. In contrast, the values from 1997 thru 2040 represent future oil production. Literally, a FORE-cast. These values you can change, or 'play with,' as you see fit. 8. Before making your forecast, click on the little question mark [?] button on the GID icon. Some basic guidelines and requirements for oil forecasting appear. VERY IMPORTANT. Repeated below. Click OK. 9. By definition, "Heuristic denotes a method of solving a problem for which no algorithm exists. It involves trial and error, as in iteration." In this discussion, 'heuristic knowledge' means 'soft,' 'qualitative,' or 'judgmental knowledge.' Although heuristic knowledge is lacking in the Numeric model, it is crucial for oil forecasting. And STELLA makes that easy. Read on. 10. Oil forecasting has its limits, but it isn't arbitrary because there are strict physical constraints and theoretical limits, mostly heuristic. Moreover, a large amount of heuristic knowledge is readily available. Question: 'So what to do?' Answer: 'Put it in the H-model.' In that regard, oil is often called 'black gold.' Compiled from experience, we offer the following 'golden rules' of oil forecasting. The Golden Rules of Oil Forecasting: Oil production must satisfy the following conditions, constraints, and requirements. a. For all practical purposes, no new oil will be formed within the Earth during the human time-frame. b. Oil must be discovered before it can be produced; on average the discovery-to-production delay is 10 to 12 years. c. The largest and most accessible oil fields are usually discovered (and produced) first. 'Creaming' it's called. Small fields are left for last, if at all. d. Oil production starts at zero, increases to one or more peaks, and sooner-or-later ends at zero. The 'oil life-cycle.' e. If the life-cycle profile is symmetric, then the peak year coincides exactly with the midpoint of cumulative production. If asymmetric, then virtually never. In this study, for example, the World midpoint occurs in 2003, the peak in 2007. i.e., The midpoint precedes the peak by 4 years. f. A large portion of the primal oil in the Earth (called 'oil in place') will remain underground -- forever unused by Humankind. The unused portion ranges from 60% to 100%. g. The first forecasted value must connect back to the most recent historic data point. (e.g., The 1997 value must connect back to the 1996 value.) h. The most recent production trend (e.g., up, down, or flat) usually continues for several years into the future. i. The closer the peak, the more important the recent trend becomes in predicting the peak. j. If a historic maximum occurred many years ago and since then production has stubbornly declined, then the historic maximum is likely to prove the all-time peak. Peru, Romania, and the US are examples. k. Although we don't know what the Ultimate Recovery (UR) will be, we do know the EUR forecasts published by sundry petroleum 'experts.' Knowledge of the distribution, values, and range of these EUR estimates is very useful in making a new (and better!) forecast. l. 'Net-Energy Limit:' When it takes more energy to extract, transport, and process a barrel of crude oil than can later be obtained from that same barrel, then production will stop no matter what the price. m. 'Depletion Dominance:' Obdurate to 'technology advance,' the oil life-cycle inexorably terminates with low-grade ores. n. The World oil production peak will be determined by oil production data only. The myriad of 'reserve' estimates will remain an irrelevant, historic enigma. o. The oil life-cycle curve (profile) determines the Expected Ultimate Recovery (EUR). Not vise versa! See the 'Curve-Area Axiom' below. p. Strictly speaking, the World oil peak will never be 'forecasted,' because several years of post-peak decline will be necessary to verify the peak. Thus, the World peak will only be 'BACK-casted.' No cigar Charlie. 11. In your forecasts you should use all the heuristic knowledge available -- such as the declining worldwide trend of 'giant' field discoveries; OPEC 'quotas;' the values, range and reliability of EUR estimates; 'political reserves;' API degree of the oil; regional geology; hostile environments; reservoir depth, porosity, pressure; etc. 12. Importance of the Oil Production Profile: The combination of historic data and the Judgmental forecast, displayed in the GID named 'HistData&Fcst US' (big blue box), depicts the US oil production profile from 1960 thru 2040. Bottom Line: This production profile is the MAJOR OUTPUT of the US model: in brief the J-curve. Likewise for all other national models. 13. With items 9-12 (above) in mind, again launch the US GID. Scroll right until the year 1992 appears on the left. This time use the cross-hairs and mouse to click-and-drag in your J-forecast (i.e., sketch in future production from 1997 thru 2040). Much faster. Click OK. 14. Launch the Graph Pad labeled 'HistData Prod G Prod NH US.' Select Run from the Run menu. Your J-forecast is drawn (red line, tagged '3'). Close the graph. 15. Be creative. Try several other forecasts. Run each and observe the different EUR values (automatically calculated after each run). Then select one of the curves as the US life-cycle profile (i.e., the historic data plus by your J-forecast). Next some clarification about the 'oily' EUR. Part III. Calculation of the Expected Ultimate Recovery There are two main approaches to oil forecasting. One approach (not ours) begins by assuming an EUR value; then (with the EUR area in mind) a production curve is drawn, i.e., a life-cycle profile. The second approach (ours!) begins by predicting the life-cycle profile and then, from the profile, the Program calculates the exact EUR value. This gives a specific one-to-one correspondence. The advantage is encapsulated in the following Axiom. Curve-Area Axiom: A given EUR value (i.e., an area) maps into an infinite number of arbitrary life-cycle curves (i.e., all different shapes). In contrast, a given life-cycle curve (i.e., a zero-peak-zero profile) maps into one, and only one, EUR value. It follows from the Axiom that, for one-to-one correspondence, the EUR can be calculated only after the life-cycle production curve has been forecasted. The difference is crucial because EUR estimates are 'guesstimates' at best. They vary wildly (e.g., by a factor of 7 or more). In fact, in the oil business the only hard numbers available are the historic production data. Bottom Line: The production profile comes first. The EUR follows. To calculate the EUR, drafters use a planimeter. Mathematicians use integration. STELLA uses 'Stocks' and 'Flows.' The following hands-on example is worth 10,000 words. 1. Click the Undo [U] button on the US GID. Select Run from the Run menu. Watch the run indicator (i.e., the advancing horizontal bar with speeding numbers). For Windows users, this will appear in a small window. For Macintosh users it will appear at the top of the screen. The base-line models are restored. 2. Click the 'Down-Triangle' at the right of the 'USA' label (blue). (This Triangle is just one of many 'Navigation Arrows' used to display STELLA's 3-layer operating environment.) The 'Model Construction Layer' appears: USA in view. Locate the 'Numeric Display' that shows the number '271.2' (i.e., in a little horizontal rectangle under the label 'EUR NH US'). The base-line US Expected Ultimate Recovery (EUR) was calculated to be 271.2 billion barrels. Keep that number in mind. 3. Click the 'Up-Triangle' (there are two; either will do). The High-Level Mapping Layer appears. 4. Double-click the US GID icon (big blue square). Scroll the graph to the far right and, for example, increase all the values in the Output column to 3.0 for the years 2030 through 2040. That increases the production forecast and, of course, the area under the curve. Click OK. 5. Click the Down-Triangle at the right of the 'USA' label. The Model Construction Layer appears: USA in view. Select Run in the Run menu and watch the 'EUR NH US' numbers (they shudder a bit as the run progresses). All zeros (boring!) until the final number 322.4 pops up. Question: How'd I know that? Bottom Line: The EUR is an output of the World Oil Forecasting Program. It is calculated from the oil production profile (i.e., historic-data-plus-forecast) at the completion of each run. In other words, every time you change the production profile, you change the EUR. Again, the EUR is calculated from the production profile. Not vice versa. (Duncan & Youngquist, 1998a; forthcoming). 6. Of interest -- This is our last chance to look at the lowest layer of STELLA's operating environment. Click the Down-Triangle in the upper left-hand corner of the window. The 'Equations View Layer' appears (i.e., Layer 3). Scroll up and down; have-a-look. Select 'Equation Prefs' from the Equation menu. A sub-menu appears. Select 'Sector.' Click OK. The North America model equations appear grouped conveniently by nation: i.e., Canada, Mexico, and USA. Scroll down the list. Have-another-look. Suggestion: Try the various options in the 'Equation Prefs' sub-menu. Observe the different arrangements and study the equations. Note: Most of the equations in Layer 3 are used by the Numeric model to generate the Guide curve (i.e., the blue line, Curve 2, in many graphs). 7. Click the Up-Triangle. The Model Construction Layer re-appears (Layer 2). Close the North America model. Part IV. World Base-Line Oil Forecast: Study Mode This section studies the base-line forecasts of the second running of the World Oil Forecasting Program. The forecasts appear as graphs and tables. You can make new tables and graphs, as you like, and print them out. It is worth noting, however, that the commercial version of STELLA 6.0 greatly facilitates the inclusion of your own national forecasts into the World Summary. 1. Launch the World file. The module opens in the 'Model Construction Layer' (i.e., STELLA's Layer 2). In this part we'll stay in Layer 2. 2. Open the Graph Pad labeled 'Prod World' (little blue square). A graph showing World oil production appears (thick red line), scaled on the vertical-axis from 0.0 to 32.0 billion barrels per year. World oil production, you'll recall, is the sum of the production curves of 42 nations. Note that the World peak appears somewhere between the years 2000 and 2020. But in exactly what year? (Hint: See item 3 below.) Close the Graph Pad. 3. Open the Table Pad labeled 'Prod Q EUR World' (little green square). Column 2, headed 'Prod World,' displays World oil production. Scroll down the table and note that maximum production occurs at the rate of 30.31 billion barrels per year in the year 2007. Voila! The World peak! (continued) 4. Also observe that Column 3, headed 'Q World,' displays the World cumulative oil production in billion barrels. Scroll down to the bottom and observe that the World cumulative oil production in 2040 (STELLA calls it 'Final') is forecast to be 1,861 billion barrels. Likewise in Column 4, headed 'EUR World,' the Expected Ultimate Recovery (EUR) is forecast to be 2,014 billion barrels. In other words, by 2040 we will have produced 92% (i.e., 1,861/2,014) of all the oil that will ever be extracted from the Earth. 'Running on empty!' Close the Table Pad. 5. Open the Graph Pad labeled 'Graph 1' (little blue square). Double-click the grid. A dialog box opens. Scroll down the list of 'Allowable' variables, and double-click 'OPEC_Prod_Gap.' The new variable jumps (right) into the 'Selected' list. Click OK. 'Graph 1' reappears. Choose Run from the Run menu. Observe that new variable, 'OPEC_Prod_Gap,' plots out along with the former variables, each on its own scale. Neat huh! Close the Graph Pad. 6. Single-click on the icon labeled 'Graph 1' (i.e., the same little blue square as above). The label highlights. Rename it 'Apple Strudel.' Click once outside the Graph Pad. The 'Apple Strudel' digests (pun intended). 7. As you wish, continue exploring the latest World oil forecasts. Happy discovering! Part V. World Oil Forecasting Method: Summary The peak of World oil production is assumed to be a watershed in human history. The main goal of the World Oil Forecasting Method is to systematically predict the year that world oil will peak. Historic data of the world's top 42 oil-producing nations is our basis. The World Oil Forecasting Program is our tool. Each nation is modeled separately. World production is the sum. The ongoing series of annual updates, taken together, is our Method. The strategy of the Method is twofold. First, 'encircle the World oil peak' (i.e., 'bracket the target,' as Navy gunners say). Second, zero-in on the peak with a series of annual updated forecasts. In US Navy-speak, 'Hit the target with a series of barrages.' Perhaps the first will score. Perhaps the tenth. The first running of the Program forecasted that the World peak will occur in 2005 and by the year 2040 production will decline by 61.7% -- i.e., US production falls from 28.5 to 10.9 billion barrels per year in just 35 years (ref. Duncan, 1997). The second running of the Program (i.e., the current models at this website) forecasted that World production will peak in 2007 and by 2040 decline by 62.0% -- i.e., from 30.3 to 11.5 billion barrels per year in only 33 years (ref. Duncan & Youngquist, 1998a). Thus after two forecasts, we bracket the peak by years 2005 and 2007, a spread of 3 years. (Note: Italy was excluded from the first forecast, but included in the second.) As we continue the series of forecasts, sooner-or-later the Method will converge on the World peak. However recalling the USA experience where the pessimists finally prevailed (i.e., the pessimistic prediction of a 1970 peak proved to be correct), I feel that the first two forecasts are a bit optimistic, and the World peak is likely to occur between 2002 and 2004. Barely 4 to 6 years hence. With no alternatives in sight. Production data, and production data alone, will be the final arbiter. ------------------------------------------------- References: 1. Duncan, R. C. (1996) The Mexican petroleum 'play' in two 'acts': Taking hold of oil production data. System Dynamics Conference Proceedings, pp. 141-144, Vol. 1. System Dynamics Society. Cambridge, MA. This paper introduces the translated coordinate system method for analyzing oil production data. 2. Duncan, R. C. (1997) The world petroleum life-cycle: Encircling the production peak. Proceedings of the Thirteenth SSI/Princeton Conference on Space Manufacturing, pp. 267-274. Space Studies Institute. Princeton, NJ. This paper presents the first application of the World Oil Forecasting Program. 3. Duncan, R. C. & Youngquist, W. L. (1998a, forthcoming) Encircling the peak of world oil production. This paper presents the second application of the World Oil Forecasting Program. 4. Duncan, R. C. & Youngquist, W. L. (1998b, forthcoming) World oil forecasting: A numeric-heuristic approach. This paper describes the technical details of the World Oil Forecasting Program. Hard copies of the papers are available ASAP from . R. C. Duncan Rev. 8-28-00