Irving Earthquake Study Relies on Contradictions and Dubious Assumptions
A new study examining seismicity in North Texas relies on questionable assumptions and contradictory claims to link earthquakes to injection operations, according to a review by Energy In Depth. The study (Hornbach et al., 2016), authored by researchers at Southern Methodist University and the University of Texas at Austin, claims it is “plausible” that injection operations triggered earthquakes near Irving, Tex., despite the fact that no injection wells are in the area.
It’s worth noting that this same team of researchers has published multiple reports on North Texas earthquakes. Notably, they collaborated last year for a study on seismic events near Azle, Tex., (Hornbach et al., 2015), as well as a study released earlier this year (Frohlich et al., 2016) that introduced a new “five question test” for assessing earthquake causality. Many of the authors are also part of the Center for Integrated Seismicity Research, including one who serves on the governor’s technical advisory committee.
To connect the Irving-Dallas earthquake sequence to injection, Hornbach et al. (2016) had to make assumptions about the geology of North Texas, many of which were flawed or even contradict what the authors themselves have found in previous studies. Below are a few of the most notable errors in the study, which raise questions about the validity of its conclusions.
From ‘Tremendous Heterogeneity’ to ‘Relative Homogeneity’
In Hornbach et al. (2016), the authors “assume relative homogeneity for the Ellenburger” to arrive at their conclusion. According to the study, that homogeneity allowed fluid pressures to travel up to 40 kilometers to trigger the Irving-Dallas earthquake sequence.
But last year, the same authors (Hornbach et al., 2015) stressed that “tremendous heterogeneity exists in the Ellenburger” across very short intervals. They also noted that “several large karst features also exist in this region,” and that “these features represent zones of significant permeability changes.”
In order for the researchers to link the Irving earthquakes to injection operations, the Ellenburger formation must be homogenous across long distances. But the authors have already acknowledged that is not the case. In fact, permeability in the Ellenburger can vary greatly over very short distances.
Hornbach et al. (2016) also assumes “no significant fluid loss across the 63,000 km2 basin.” But again, the same authors acknowledged last year that “natural fractures could extend through the entire Ellenburger and into the Barnett shale,” and “the change in pressure caused by water extraction [in the Barnett] will still impact other areas of the Ellenburger formation due to the nature of pore pressure diffusion.”
That’s not the only contradiction, though. The homogenous assumption in Hornbach et al. (2016), specifically the claim that “no significant fluid loss” occurs, is contradicted by the researchers’ conclusions within the study itself:
“Currently, there is no significant oil and gas production in the Ellenburger in the central part of the basin. It is therefore perhaps more likely that pressure reductions are caused by natural fluid migration out of the formation.” (emphasis added)
Is there “no significant fluid loss” across the basin, or is there “natural fluid migration out of the formation”? Is the Ellenburger “homogenous,” or are there “pressure reductions” due to geological features in different areas?
The permeability assumptions these researchers used to claim a linkage between injection and the Azle earthquake sequence (Hornbach et al., 2015) are the opposite of what they’re now using to claim a linkage between injection and the Irving-Dallas earthquake sequence.
Hydrostatic Assumption Lacks Basis
Hornbach et al. (2016) asserts:
“Recent spot measurements of pressure in the Ellenburger confirm that elevated fluid pressures ranging from 1.7-4.5 MPa (250-650 psi) above hydrostatic exist in this formation, and this may promote failure on pre-existing faults in the Ellenburger and underlying basement.” (emphasis added)
The implication is that injection activities have increased pressure well above the natural pressures in the Ellenburger formation. The problem is, what the authors consider “hydrostatic” is never defined. The heterogeneity of the Ellenburger suggests the natural reservoir pressures could vary greatly over short distances, as the same research team suggested in its Azle study last year (Hornbach et al.,2015).
To try to justify the hydrostatic assumption, Hornbach et al. (2016) references well data taken from Johnson County in 2015, in which several operators conducted pressure tests during and after injection. After stopping injection, one of the tests showed pressures normalizing at about 340 psi. Another well leveled out around 250 psi after injection ceased, while a third well showed readings of 700 psi.
Even within the same county, there are dramatic differences by well site. We don’t know if those numbers would converge or diverge over a greater time period, but it is that same uncertainty that highlights why assumptions about homogeneity and uniform hydrostatic pressure in the Ellenburger are so problematic. The authors admit the likelihood that “pressure reductions are caused by natural fluid migration out of the formation,” which further underscores why it’s so important to define hydrostatic pressure – and the basis on which such an assumption rests.
At the very least, more analysis is needed to confirm what the equilibrium pressure is in the formation, or if the regional variability of the Ellenburger means any such computation would have to be location specific.
Careless Assumption to Malign Fracking
From the abstract of Hornbach et al. (2016):
“We estimate that a majority of the water being injected into the Ellenburger is flowback water associated with the hydraulic fracturing process. According to the Texas Railroad Commission website, at least 15,000 unconventional wells have been drilled in the Barnett Shale. The average well in the Barnett shale that is hydraulically fractured uses between 11,000 – 19,000 m3 (69,000-119,000 bbls) of water (Nicot et al., 2014). If injected water is ultimately recovered from each well during production, then the total amount of flowback water from Barnett production ranges from 175-285 million m3 (1.1 – 1.8 billion bbls). As we will show, this amount is equivalent to ∼65% to 106% of the total volume of water injected into the Ellenburger since 2005. Thus, the amount of water used to hydraulically fracture the Barnett from 2006-2014 is consistent to first-order with the amount of wastewater injected into the Ellenburger during that same time.” (emphasis added)
This assumption is not accurate.
Only a fraction of the water injected during the hydraulic fracturing process is recovered. The numbers vary, but they range from 16 percent to 30 percent. The rest of the water injected during fracking remains in the formation.
While testifying before the Texas House Committee on Energy Resources in 2015, seismologists from Southern Methodist University said, “we’re not talking at all about fracking.” They went on to say that it’s “driving us crazy” that the media kept misinterpreting their research, which focused on wastewater injection.
In a 2013 interview, Dr. Stephen Holditch, a professor emeritus of petroleum engineering at Texas A&M University, said only between 10 and 30 percent of the water injected during fracking will be returned back to the surface. The rest of the water, according to Holditch, stays in the formation.
As the Railroad Commission of Texas also notes on its website:
“The overwhelming majority of injected fluid is oilfield brine, which is also sometimes referred to as produced water. Oilfield brine is the water, with varying levels of salinity that is found in the same geologic formations that produce oil and gas. This produced water comes up simultaneously with the production of oil and gas. However, small quantities of substances used in the drilling, completion and production operations of a well may be mixed in this waste stream.” (emphasis added)
Hornbach et al. (2016) assumes that all injected water during fracking in the Barnett is eventually recovered, which is not true. It then tries to support that assumption by comparing the total amount of water injected during fracking to the total amount of water injected into the Ellenburger. This reveals an irrelevant similarity, since the fluids injected into the Ellenburger largely consist of brine, not flowback.
That a Barnett Shale well might produce an amount of brine that is similar to the quantity of water used during fracking does not prove that the water used during fracking is being recovered.
Does Subsurface Pressure Matter – or Not?
According to Hornbach et al. (2016):
“Fully addressing the induced seismicity hazard requires understanding not only subsurface pressure changes but also the local stress regime.”
But the same research team published a study in May of this year (Frohlich et al., 2016) that claimed subsurface pressures were not important in assessing causality of induced earthquakes. According to that earlier study:
“We no longer include questions related to subsurface pressures and modeling; this information is available for few events and, when reported, often relies on somewhat arbitrary (and arguable) assumptions about subsurface structure and flow properties.”
In a manner of two months, the authors have gone from saying subsurface pressure data are based on “arbitrary” assumptions and thus should not be relied on, to saying that “fully addressing” the hazard from induced seismicity “requires” understanding subsurface pressures.
Earthquakes and Injection Well Proximity: A Moving Target?
In Frohlich et al. (2016), the authors used a question-based system to assess whether a particular earthquake was natural (tectonic) or induced. Two of the five questions examined spatial features, asking if any injection or production operations were within “5 km for well-determined epicenters or within 15 km otherwise.”
But Hornbach et al. (2016) expands the allowable proximity nearly three-fold (i.e. up to 40 km) in order to claim that injection is a “plausible” explanation for the Irving-Dallas earthquake sequence.
Either the question-based system in Frohlich et al. (2016) is flawed, or Hornbach et al. (2016) is drawing a “plausible” connection between earthquakes and injection that is inappropriate.
Hornbach et al. (2016) does explore the possibility of an injection well that is approximately 15 km from the epicenters of the Irving-Dallas sequence. But according to Frohlich et al. (2016), that distance was only appropriate for assessing earthquakes with epicenters that are not well-determined. It is inappropriate, however, to categorize the Irving-Dallas epicenters in that fashion.
How do we know that the Irving-Dallas epicenters are well-determined? Because the authors themselves have said so. Dr. Brian Stump, a seismologist at SMU and a co-author of Hornbach et al. (2016), said last year with regard to his team’s research of the Irving-Dallas earthquakes:
“Now that we know the fault’s location and depth, we can begin studying how this fault moves – both the amount and direction of motion.” (emphasis added)
According to the question-based system in Frohlich et al. (2016), the proximity of injection operations in the case of the Irving-Dallas sequence would force researchers to answer at least two of the five questions with scores indicating a natural origin. But the same research team is now using the same proximity to claim instead that it’s “plausible” the earthquakes were manmade.
Curiously, two of the studies cited in Hornbach et al. (2016) to justify these larger distances appear to undercut the authors’ case. One study, Zoback and Hickman (1982), suggested that well pressures would have expanded beyond 10 km over just a matter of months in significantly lower permeability (i.e. 1 mdarcy) than what Hornbach et al. (2016) estimated for the Ellenburger (10-300 mD). Hornbach et al. (2016) also assumed that it would take six years for pressure to travel through higher permeability. Another study, Hsieh and Bredehoeft (1981), explored Colorado earthquakes that occurred nearly 15 years after initial injection operations, but which were still less than 10 km away from the injection site.
It is not unheard of for research to evolve as new analyses take place. In fact, that’s one of the goals of publishing scientific papers: to advance public understanding, while also sparking new questions and possibly new assumptions about how to explain certain phenomena.
However, it is difficult if not impossible to justify the flaws in Hornbach et al. (2016) with that explanation, because the authors’ assumptions are such a dramatic departure from their prior work. That the Ellenburger is not homogenous is dismissed with a simple assumption, even though heterogeneity was a key assumption in the model that allowed the same researchers to link the Azle, Tex., earthquakes to injection. Relevant proximities are also ignored, without any discussion as to why a significantly larger distance between injections and epicenters can now provide a “plausible” explanation for induced seismicity.
Studying the Irving-Dallas earthquake sequence is an important scientific pursuit, particularly given its rather anomalous characteristics among other seismic events in North Texas. But using bad assumptions to blame the earthquakes on injection does not advance public understanding; if anything, it only hinders our ability to arrive at the truth.