Can predictive data analytics, a cutting-edge tool for exploration, lead to a future boom in new field discoveries and reserve additions?
If it does, predictive analytics predicted it.
That’s the view of Andy Kemmer, a Kansas independent who has studied the history of exploration cycles and is now applying predictive analytics to midcontinent plays.
A Pattern of Booms
Kemmer said long periods of stasis in the oil and gas industry appear to have been followed by bursts of exploration success. And, oddly, those discovery booms haven’t come when the industry is booming.
“I first became aware of this pattern back in the mid-1980s. Counterintuitively, there appears to be a surge in the discovery rate of large fields when oil prices are low,” he said.
The 1930s brought a surge in discoveries and reserve growth in the United States. Kemmer noted that the period of increased exploration success followed a period of technological advancement.
“What was the trigger? If you look at the spike of the reserves added in the 1930s, it was related to the introduction of seismic reflection/refraction technology,” he said.
“During these periods, new technologies were often the catalysts to these discoveries,” he added.
In a similar way, innovations in the 1990s led to the boom of unconventional resource development as new technologies were applied.
“Obviously, we had significant gains in drilling and completion technologies that made these resource plays viable,” Kemmer observed.
“Only in the 2000s did we realize that people were making big discoveries in the Bakken, in the Eagle Ford, etc.,” he said.
Kemmer identified at least three previous periods of technology development he called “revolutions” for the oil industry:
- 1920s-30s: Seismic reflection/refraction
- 1950-60s: Common depth point seismic
- 1990-2000s: Horizontal drilling and hydraulic fracturing
The Fourth Revolution
“Data analytics is a revolution that we’re just beginning to see in upstream E&P,” he said. “If we’re correct, data analytics is the fourth revolution we’ve experienced in the last 100 years.”
Kemmer was scheduled to discuss data analytics, its application to exploration and the industry’s discovery cycles at AAPG’s 2020 Annual Conference, in the presentation, “Predictive Analytics, Undetected Oil Fields, and Einstein: A Unified Field Theory?”
He said bursts of rapid, intense change have followed long periods of stasis in many diverse industries and areas. Kemmer cites the field of physics as an example, noting that Albert Einstein had four separate, revolutionary papers published in 1905, in a breakthrough known as his “miracle year.”
“It’s not just in our industry we see this. We see it in evolutionary biology, in financial markets, public health, etc.,” he explained.
If the pattern holds, Kemmer said, the 2020s could see the beginning of a surge in discovery rates of large fields.
Predictive analytics draws on statistical techniques from computer learning, data mining, predictive modeling, visualization and other types of data analytics to make assessments of and predictions about the future or other unknowns.
“Using machine-learning algorithms, it (predictive analytics) can find patterns we might not be aware of,” Kemmer noted.
The Data Analytics Breakthrough
Bill Fairhurst, president of Riverford Exploration LLC in Spring, Texas, said he has applied data analytics using small sets from eight wells with a few variables up to large sets with billions of 3-D seismic grid cells with dozens of variables, in both industry and academic settings.
Fairhurst said he has found value “in most all applications,” but he has some caveats about the way analytics is now being used in oil and gas.
“Data analytics – Big Data – cannot solve problems by itself as a stand-alone tool. It should be used with other tools available to experienced petroleum geologists to come up with an optimal solution or answers, including exploration,” Fairhurst observed.
“Predictive analytics in petroleum geology is more reliable in field-reservoir modeling up to basin-wide reservoir characterization modeling than it is in exploration geology, were there are more unknowns or the predictor variables are being used to forecast farther from the data source area or further out in time,” he said.
That doesn’t mean data analytics isn’t useful in exploration, Fairhurst noted, just that the relative reliability and accuracy of the predictions need to be understood and used by data scientists and domain experts properly, then properly communicated to decision makers and investors.
Fairhurst’s biggest caveat is probably a lack of interaction by domain experts – petroleum geologists and engineers – in the data-analytics process and final model results.
“What does that mean? It means that data scientists are taking over the process that formerly used to be done as part of a domain expert’s evaluation,” he said.
The collection and interpretation stages of primary data are 90-95 percent of the evaluation process, Fairhurst noted. For good scientists, analytics is the last 5-10 percent of the process, he said.
The science of data analytics has progressed rapidly during the past decade, and Kemmer said a recent advance in predictive analytics relevant to the oil industry could be traced back to a machine-learning competition around 2010 that became the ImageNet Large Scale Visual Recognition Challenge.
He said a dramatic image-recognition breakthrough made by computer scientist Geoffrey Hinton and his team in 2012 in the ImageNet challenge helped revolutionize the field of computer vision.
“Since then, there are more and more sources for applying machine learning to what we’re doing with seismic” in exploration, he said.
3-D/Analytics Combo
Some explorationists might add 3-D seismic to the list of revolutionary technologies for oil and gas. Kemmer said the effect of 3-D seismic on the industry has been profound, but it’s the combination of data analytics with 3-D seismic in upstream E&P that is truly revolutionary.
“Obviously, 3-D was developed offshore. We had lots of trial and error. Only then did 3-D migrate onshore,” he said. “Machine learning is a different animal. It’s not confined to any one area. And it’s new to everybody.”
The 3-D/analytics combination has typically been applied to very large datasets for offshore exploration. Today, Kemmer’s team and others are applying those techniques onshore.
“Our small group said, ‘Let’s see if that has any applications to some of the onshore areas we’re interested in, like the midcontinent,’” he said.
Kemmer believes multi-attribute analysis and 3-D seismic, even from one survey, can provide sufficient data for application of the predictive-analytics approach.
“As long as you have a single 3-D survey, there’s lots of data you can mine from that,” Kemmer said.
“A lot of information from these signals has not been captured,” he added.
In fact, data analytics might already be influencing play creation and assessment onshore, according to Kemmer.
“It wouldn’t surprise me if people were taking acreage positions based on these tools right now,” he said.
In Kemmer’s view, this is the fourth time in the last 100 years when extraordinary gains are possible for the oil industry due to the confluence of low prices, low activity and a technological breakthrough – in this case, data analytics and machine learning.
Based on that historical perspective, companies might want to consider increasing their upstream budgets, he said.
Does Kemmer have any idea when that next exploration success boom might start?
“I wouldn’t be surprised if it were happening as we speak,” he said.