Today, scientists from Southern Methodist University, the University of Texas, and the U.S. Geological Survey released a new research paper in the journal Nature Communications, entitled “Causal factors for seismicity near Azle, Texas.” Through a process of elimination, the researchers concluded that “brine production combined with wastewater disposal represent the most likely cause of recent seismicity near Azle.” However, several issues in the paper raise questions about its conclusion, including potentially major flaws related to subsurface pressure.
Raising the Bar
While the paper’s methods do raise concerns, it is worth emphasizing that SMU, UT, and USGS deserve credit for developing a model that provides greater understanding of the conditions that can ultimately lead to induced seismicity.
More specifically, the SMU team should be commended for recognizing a basic weakness in much of the existing scientific research on induced seismicity. Many previous assessments that have “linked” oil and gas activities to small earthquakes have done so through mere correlation in time and space. In other words, because an injection well was located in proximity to an epicenter, and the seismic events began after the well went into operation, researchers suggest a “plausible” connection between the two. As this latest paper notes, however, many of these studies “do not evaluate other possible anthropogenic causes of seismicity or do not utilize physical models to quantify stress changes.”
As the SMU team acknowledges, in reference to the fact that simple correlation is insufficient for understanding induced seismicity:
“Although there is an increase in injection volumes in the mid-2013 before the recent events, even higher volumes and pressures are reported in prior years at both injectors, when no felt earthquakes occurred.” (p. 5)
By developing a subsurface model that assesses what actually happens underground – rather than making a simple correlative assessment – the SMU team should be praised for raising the scientific bar.
This study also gives further credence to the need for ongoing site specific assessments. The potential for seismic activity must be addressed based on downhole pressure, injected volumes, and location, including the orientation of certain faults. Peer-reviewed studies have consistently identified these variables as necessary to understand induced seismicity, and not to convey a blanket, one-size-fits-all approach that suggests geological or pressure conditions in any given area are analogous to operations in other parts of the country.
As the study notes, “tens of thousands of currently active injection wells apparently do not induce earthquakes or at least not earthquakes large enough to be felt or recorded by seismic networks.” This corroborates research from Energy In Depth, released last month, that found 99.9 percent of injection wells in the Barnett Shale region of North Texas have not been associated with felt seismicity.
Notably, the SMU paper is an example of ongoing earthquake research made possible in part by active participation by the oil and gas industry. Through partnerships with research institutions like SMU, as well as Stanford University and the University of Oklahoma (among others), the industry has been actively sharing subsurface data with scientists and state geologists to advance public understanding of induced seismicity – and, more importantly, to help researchers arrive at science-based conclusions that can inform solutions.
But as any scientist will likely admit, a model is only as good as its inputs, and it will never provide a perfect explanation. While the SMU report certainly is a great starting point and an excellent example of the collaborative efforts by industry and researchers to find the causes of recent seismic events, there are aspects of the research that deserve a closer look.
Let’s explore a few questions that the report raises, through its conclusions and its methods.
Question 1: Can Football Air Pressure Trigger Earthquakes?
The researchers note that the Azle events originated from a “critically stressed fault,” which the SMU team suggests was triggered to slip by a pressure of about 5 pounds per square inch (psi). That’s roughly half of what a properly inflated NFL football requires – even if you’re the New England Patriots. The depths at which fluids are injected have natural pressures that can exceed several thousand psi, raising questions about whether such a small pressure difference could actually trigger a series of earthquakes large enough to be felt on the surface.
If so, that means even in the absence of wastewater injection activities, the area would not likely have remained earthquake free for very long. The fact that the fault was already “critically stressed” suggests that any number of potential variables could have triggered the event — be they natural causes or man-made.
Question 2: Can Modeled Production Replace Actual Production?
Instead of utilizing all available production data, the study only incorporates some of the actual data, and chose to input a different production scenario into its model. As the report notes:
“We use G-10 production reports combined with gas production reports for the 70 largest brine-producing wells in the region to make first-order estimates of brine production (Fig. 3d).”
The researchers attempt to explain this decision, primarily by stating that G-10 production reporting “provides only a crude estimate of brine production across the region.” But it’s unclear why that data would be inferior to modeling the 70 “largest brine-producing wells in the region” to make “estimates” of production.
Question 3: Why Use Normalized – Instead of Absolute – Pressure Changes?
In Figure 4, the researchers show that they used normalized pressure changes in their model, which help inform their conclusion that natural causes were not responsible for the earthquakes near Azle. But had they used absolute pressure changes, the change over time would not have been as pronounced (i.e. 5 psi in a geologic structure with natural, absolute pressure that can exceed 8,000 psi).
Had absolute pressures been used, the results in the model would not likely have supported the hypothesis ruling out natural causes – or, at least, they would not have provided such a stark suggestion that something other than natural forces were to blame.
Question 4: What are the Limitations of Not Utilizing Multi-Phase Flow?
As the study notes on page 5:
“Owing to uncertainties in gas production and gas volumes in the Ellenburger, the model currently does not account for multiphase flow.”
In a nutshell, the SMU model’s pressure assessment could not handle the interplay between gas and water in the formation, so the model assumes the Ellenburger formation is essentially a box of water. The presence of natural gas, however, can change pressure considerably, as gas can absorb pressure in water. Considering the fact that the researchers identified a pressure change of 5 psi as being the trigger for the earthquakes, a lack of attention to how the presence of gas in the formation studied can impact pressure is a shortcoming that is not fully addressed in the paper.
Question 5: How Did Permeability Assumptions Impact Modeled Pressure?
The SMU model establishes a permeability barrier at the fault, which means it assumes the fault forms an effective seal between one “side” where brine is being produced and another “side” where fluids are being injected. This creates a major pressure difference in the SMU model, which provides the explanation for induced seismicity: high pressure on one side, lower pressure on the other, and a fault that was “critically stressed.”
The problem is, as a formation that’s being used for underground wastewater injection, the Ellenburger is actually quite permeable. Otherwise, companies would not be using it for disposal.
Although the analogy is quite crude, imagine a sponge that is ripped in half, and then the two sides are placed against each other once again – tightly. If water is applied to one side, the other side will also become damp; the tear won’t seal one side from the other.
Curiously, although the researchers added a permeability barrier at the fault, they did not create one vertically. This serves to increase pressure at the fault in the model, without adjusting for vertical pressures. In short, this is a very unlikely actual scenario.
Even small assumptions on permeability can yield major differences in results. Considering, again, that just 5 psi was the triggering pressure, if the model had allowed even moderately more permeability, the pressure difference across the fault would have been far smaller.
A recent paper found that “the U.S. Geological Survey in collaboration with the University of Colorado, Oklahoma Geological Survey and Lawrence Berkeley National Laboratory, suggests that it is possible to reduce the hazard of induced seismicity through management of injection activities.” And, as the SMU team acknowledges early in its paper, seismic activity and potential correlations need to be assessed carefully, since most wastewater injection activities have no connection to seismicity.
In 2012, the National Research Council (NRC) estimated that one out of every 4,000 disposal wells nationwide are associated with triggered seismicity. The NRC said wastewater disposal was only suspected as a likely cause of induced seismicity in eight areas in the past several decades, and much of the research since then has focused on those same regions. The U.S. Geological Survey says “only a small fraction of these disposal wells have induced earthquakes that are large enough to be of concern to the public.”
The U.S. Environmental Protection Agency – which manages the Underground Injection Control program – says “very few” of the nation’s disposal wells have been linked to felt seismic events, and has applauded states for how they have managed their disposal well programs.
Most studies have called for increased collaboration between industry, scientists, and regulators – and that’s exactly what has been happening in recent years. The State Oil & Gas Regulatory Exchange recently created an induced seismicity working group, which brings together state regulatory agencies and geological surveys, along with the Ground Water Protection Council, to collaborate and share science and research.
This kind of collaboration has happened alongside major regulatory and operational changes in Texas and Oklahoma, which is where much of the induced earthquake discussion is focused. In Oklahoma, the Corporation Commission has increased its scrutiny for new injection wells in seismically active areas. According to EnergyWire, there are at least 25 injection wells that did not get drilled in Oklahoma where the companies had requested, and roughly 20 applications were never filed after regulators informed them that the proposed site was a “red light” zone. More than 80 permits are currently held up for additional review.
Here in Texas, the Railroad Commission just adopted new rules for disposal wells to further address seismicity concerns, which the industry supported. The rules require, among other things, that permit applicants provide historical earthquake data from the U.S. Geological Survey within 100 square miles of a newly proposed well, and more frequent disclosure of injected fluid pressures and volumes. The rule also gives the Commission the authority to modify or even terminate a permit if a disposal well is determined to be contributing to seismic activity.
This latest study from SMU builds upon past analyses, and is bolstered by industry’s data and statewide subsurface knowledge. This kind of active collaboration bodes well for all interested stakeholders – not just scientists and regulators, but also the general public, which wants to know what is being done to address their concerns about earthquakes.
While the SMU study released today appears to suffer from some modeling flaws, those issues should not overshadow the fact that the research team has provided all of us with a valuable tool for better assessing induced seismicity. Being the “first mover” in science is not easy, as science advances knowledge through trial and error. Models are proposed and subsequently improved, ultimately enhancing our understanding of certain processes. The SMU paper was released in that spirit, and hopefully the issues and concerns identified here will provide constructive feedback to improve future analyses.