We are development geologists working the Permian, and from where we are sitting, it feels as though, as an industry ,we are just beginning to figure out how to develop large-scale unconventional plays and work as unconventional geologists. We are going through a revolution that will change – not just unconventional development – but all subsurface work.
We fully acknowledge that this is not a new insight, but it is worth repeating. Digital transformation and its many polymorphs are the topic of conversation from the C-suite down to the water cooler. However, resumes of entry-level geoscientists, industry conferences and a general finger-on-the-pulse feel of the current state of the industry lead us to believe that we have a great deal of room to grow. The good news is, it has never been easier to join the movement, and young professionals are well positioned to take the lead.
Unconventionals will help to drive this revolution, because unlike in conventional development, it is all but essential for success. Many of the players in this space, even the smaller, non-diversified independents, are recognizing that they have a need for greater mastery of their data. The scale of the data and nature of the play necessitates some degree of expertise in this space and we are seeing more and more companies hiring data scientists or at least data-savvy practitioners. This is driven by the nature of the play, and more fundamentally, the rocks.
The Nature of the Problem Itself
In conventional development, data are often sparse. A few key quantitative products and countless hours of interpretation drive decisions: AVO, high-resolution 3-D seismic inversions, etc. Ultimately, we hope to arrive at a geologic story that convinces us that the critical petroleum system elements are present. These elements, unlike in unconventionals, are essentially binary. There is reservoir or there is not, there is a viable seal or there is not, there is charge or there is not. A popular saying is, “Nobody drills a dry hole in the Permian.” We assume there is charge, we aren’t too worried about seal, and the definition of the word “reservoir” is nebulous. It is a game in which multiple, often subtly changing geologic elements come together with a myriad of engineering decisions to determine whether we end up with a thousands-of-barrels-a-day screamer, or a quickly-declining, marginally-economic hole in the ground. It is a system in which we try to understand the influence of numerous, independent variables on our target variable of choice. Most frequently, this target variable is production, or estimated ultimate recovery, but it could just as well be cost or drilling time.
Those last sentences are critical. We have multiple independent variables, which determine the outcome, which for this conversation, we will say is EUR. Our independent variables include geologic parameters such as clay content, saturation and porosity, as well engineering parameters including proppant and fluid volumes, well spacing and well-bore length. Or stated another way,
EUR = f(x₁,x₂,x₃,…,xn)
and our job is to bring meaningful geologic values for each . And very importantly, we are not saying that this is just a mental model for making sense out of the problem, we are saying that this is the nature of the problem itself. Unconventionals, and the subsurface in general, behave this way fundamentally. This way of thinking and the resulting subsurface workflows are simply a reflection of the system.
In “traditional” interpretation, we often begin with a suite of logs. We create facies cutoffs to build a model in which we correlate packages of genetically-related rock, and while these are frequently based on quantitative cutoffs and quantified rock properties, we typically aim for an end goal that focuses largely on visual representations of data. These visual representations are often siloed, focusing primarily on the geologic domain. Rather than visual representations of geologic interpretations as the end point, we need to work toward generating quantitative attributes that can be consolidated with engineering data for the end goal of quantitative analyses, ideally some form of multivariate regression. Rather than defining facies to describe a landing zone, think of the landing zone as just a portion of our equation, where EUR is a function of landing zone mineralogy, flow properties, saturations and other key drivers. To take it a step further, look for ways to engineer additional features. Perhaps it is not only the absolute value of gamma ray that matters. A highly variable GR curve may suggest an interbedded mixed lithology. By adding in the standard deviation (σ) of the GR over a moving window along with the absolute value of the GR, we have added a new feature that might help to make our model more predictive. Finally, and most importantly, break down the silo and incorporate the engineering data as well.
EUR = f(GR,Sw,σGR,Φ,% silica,% CO₃,% clay, proppant, fluid, distance to other wells, cumulative production in the area, well bore length, …, xn)
Each of these variables is easily quantified. We need to shift our perspective from interpretation to quantification.
Join the Movement
As we said at the beginning, it has never been easier to join the movement. You don’t need to be a professional data scientist or even know a language to be dangerous (though some Python won’t hurt).
So, what can you do?
• Practice thinking quantitatively. Begin by taking your existing maps and interpretations and turning them into quantified features that describe the geology along your well bore or the surrounding rock.
• Create a new spreadsheet: put the API number of a few wells of interest in the first column. Populate the subsequent columns with quantitative descriptions of each well (i.e. clay, quartz, carbonate, saturation, porosity of the landing zone, proppant and fluid volumes, X, Y location, properties of the zone you believe represents SRV, etc.) and in the last column, put in the 12-month cumulative oil production. Then figure out online how to create a correlation matrix, and then how to run a multivariate regression.
• Take advantage of this drop-in activity to begin learning a new skill. If you’re intimidated by programming, start learning power query in Excel or PowerBI and learn what a data model is. If you’re ready to dive in deeper, take one of the countless free or nearly free Python courses online.
• Hit the Internet hard and start reading everything you can find about quantitative geology and subsurface data science applications. There is plenty out there and endless room for growth.