Uncertainty Correlation in Resource Plays- by Jim DuBois
At the upcoming Hydrocarbon Economics and Evaluation Symposium (HEES) on May 18th in Houston, I will be presenting the paper, “An Investigation into Uncertainty Correlation in Resource Plays.” During this presentation I will discuss the problem of how simply providing uncertainty data around a type curve does little to inform planners about the range of uncertainties they may be exposed to when developing an unconventional play.
It is common for operators to drill a limited number of wells in a play (~6-20), produce them for a relatively short period (6 months to a year), and then create a “type curve” from the resulting data for use in economics and forecasting. This type curve can and should be refined as more data is collected, but the use of this single type curve to represent an area remains common.
Once we are ready to add uncertainty to this analysis, our first step is usually to add a P10 and a P90 around the existing type curve. The validity of the various methodologies used to accomplish this would be an interesting topic for another time, but for our purposes, we need only examine the result: a type-curve that now represents a potentially wide distribution of possibilities.
Problem solved! We have now acknowledged that there is considerable uncertainty in the data that results from our drilling program, and we are prepared to show a range of results that we might expect from that program.
However, unless our period of interest is short – a month, a quarter or a year, depending on the size of our drilling program – our analysis might be quite disappointing. As our program grows, the uncertainty around the program mean shrinks to almost nothing.
This is the law of large numbers at work. When wells are uncorrelated, the mean of a large population will converge to the mean of the population. A case can be made that the forecasted uncertainty for resource plays should be considered to be highly correlated, which will result in quite a different uncertainty profile. My paper and presentation on this topic will consider three examples, let’s look at one here.
Uncertainty in Long Term Decline Parameters
Type curves are often created after gathering a small amount of well history, using data from a small number of wells. While we might start to develop an idea about how IPs are distributed based on a small sample set, the long-term decline characteristics will remain uncertain until we add more data. This uncertainty will tend to be correlated – wells will tend to decline more or less than we assumed as a group, not randomly.
We may feel that all scenarios are equally likely, or we may feel more strongly about some than others. In our analysis, we are free to weight the curves as we see fit. Here, we weight them equally.
Using a method called “Global Dependency,” we can make one factor (the decline future, in this case) correlated, while another (IP, for example) can remain independent. This allows us to represent scenarios or “worlds” where the decline factors are universally higher or lower than our original estimate, but where the natural variability of other well characteristics will remain as we originally estimated.
Figure 5 shows the results of a program of 5 wells simulated in this manner. Figure 6 shows a program of 250 wells. Figures 7 and 8 use the same data, but decline futures are not correlated in these figures.
I include the P0 and P100, which we normally do not show in these graphics, in order to illustrate that uncertainty at the extremes of a data set is much more resistant to being “factored out” by the law of large numbers than is the bulk of the distribution. Still, while the five well programs (correlated and independent declines) are quite similar and quite well dispersed, it is clear by the time we get to the 250 well program that the correlated model shows considerably more portfolio uncertainty than the uncorrelated case. This is particularly apparent if we look at the P10/P90 spread, as opposed to the P0/P100 spread.
Methodologies with credible evaluation parameters are vital for the proper assessment of uncertainty and risk associated with resource plays. It is impossible to estimate the likelihood of an event thoroughly when simply providing uncertainty constraints around a single type curve. I invite you to attend my full presentation, “An Investigation into Uncertainty Correlation in Resource Plays” at 2:30pm on May 18th at HEES to dive into this topic further.
The Hydrocarbon Exonomics and Evaluation Symposium takes place May 17-18, 2016 in Houston, TX. With this year’s theme of “Fundamental Drivers, Commodity Cycles, and the Dynamics of Oil and Gas Valuation,” the event creates an opportunity for E&P professionals to capitalize on the rapidly developing industry and to share technologies and services required to meet the demand. Learn more on the event website.
About Jim DuBois
Jim is a Principal Consultant at 3esi-Enersight. He excels at the critical elements of portfolio management: building customized portfolio models for clients, dealing with issues of data granularity, integrating probabilistic data into the model, and using the model to define, refine and characterize alternate investment options for management to consider. His extensive knowledge of the oil and gas industry, gained while working as an engineer earlier in his career, allows him to bring an insider’s perspective to his client engagements. Learn more about Jim and our other Strategic Consultants here.
3esi-Enersight is the world leading provider of solutions for integrated strategy, planning and execution in upstream oil and gas. From the field, to the boardroom, in operations across 6 continents, 3esi-Enersight is empowering E&P organizations to maximize the value of their upstream portfolios and stay ahead of the competition. Our solutions help customers work more efficiently across teams and functions and make better strategy and planning decisions based on data they can trust.