New study tracks changes to stability of faults associated with Fort Worth Basin earthquakes
A new study published today in Geophysical Research Letters shows how changes in subsurface fluid pressure due to industrial activities changed the stability of faults in the Fort Worth Basin (Figure 1) over the past decade and a half. The work finds that pressure increased steadily on the faults responsible for most of the region’s recent earthquakes, making them less stable over time. As the figure below shows, the seismicity was associated with quite small pore pressure increases—the largest increase on any fault was less than 1.1 MPa (~160 psi). This is consistent with prior findings of low pressure increases before several potentially triggered earthquakes in Texas and Oklahoma (Lund Snee & Zoback, 2016; Langenbruch et al., 2018; Lund Snee, 2020; Lund Snee & Dvory, 2020).
The Dallas–Fort Worth area has experienced over 30 M3+ earthquakes since 2008, including one M4.0 event, but earthquake frequency has declined in recent years. It is widely thought that most were triggered by fluid pressure changes from oilfield wastewater disposal and other activities associated with oil and gas development (no seismicity was recorded in the area prior to 2008). Over 2 billion barrels of saltwater were injected into the basin between 2003 and 2018.
Figure 2. Map of the Fort Worth Basin showing faults colored by slip potential within the mapped stress field (Figure 2), as well as seismicity. Figure is from Lund Snee (2020), after Hennings et al. (2019).
Figure 3. State of stress in the Fort Worth Basin, northeast Texas. The orientation of maximum horizontal stress (SHmax; shown by the black lines) is consistently ~NE–SW in this area. The style of faulting (colored background) transitions from pure strike-slip in the north (yellow-green) to normal/strike-slip near Fort Worth (teal), to normal faulting further south (blue). Figure is after Lund Snee (2020) and includes data from Lund Snee & Zoback (2016, 2020) and Hennings et al. (2019).
I was a contributor to the new study, part of a team led by Peter Hennings and others at the Bureau of Economic Geology at The University of Texas at Austin. (I’m writing this post in an individual capacity and not on behalf of any organization.) The new work builds upon our prior paper (Hennings et al., 2019), which estimated the slip potential of these faults (Figure 2) within the mapped stress field (Figure 3), in response to a small, fixed (1 MPa = 145 psi) pore pressure increase.
The new study paired a high-resolution 3D fault model by the Texas Bureau of Economic Geology and detailed mapping of the state of stress (by Lund Snee & Zoback, 2016, 2020, and Hennings et al., 2019) with a new model of subsurface fluid pressure changes since 2005 (by Gao et al., 2021). Using these inputs, we estimated the probability that each segment could become critically stressed (reaching the Mohr-Coulomb failure line) over time. The probabilistic analysis employed the Stanford–ExxonMobil Fault Slip Potential (FSP) software developed by Rall Walsh, and it accounted for uncertainties in principal stress magnitudes and orientations, pore pressure, fault orientation, and other geomechanical parameters.
Figure 4. Increase of modeled pore pressure (solid lines), and estimated fault slip potential (dotted lines) over time on faults responsible for two out of ten earthquake sequences studied. Figure courtesy of Peter Hennings, after Hennings et al. (2021).
Figure 4 shows how FSP changed over time on the faults responsible for two of the ten main earthquake sequences in the basin (the other eight are shown in the paper). As the figure shows, FSP increased slightly with pore pressure (ΔPp) over time. In both cases, the onset of earthquakes occurred suddenly, after quite modest increases in FSP. In the ten sequences studied, the FSP at first seismicity ranged from 0.17 to 0.24 (a 17% and 24% chance of being critically stressed, respectively). In some cases, the faults were episodically re-activated as fluid pressure and FSP increased further. The average pressure increase associated with reactivation was only ~0.05 MPa (less than 10 psi).
Interestingly, the seismogenic faults farther from high-volume injection areas failed at lower FSP values than those closest to injection wells. The reason for this is not clear, but one can speculate that it is related in some way to fluid pressure being channelized along conductive fractures away from areas that experienced high injection rates and volumes. These results shows that the triggering thresholds for earthquakes in the FWB are modest and vary to some degree. Although the faults failed at elevated pore pressures, there is not an exact FSP value associated with failure.
Figure 5. Histogram showing the total length of fault segments associated (black) and not associated (gray) with earthquakes. Note that non-seismogenic faults much more commonly experienced little or no fluid pressure increase. (It is not surprising that many fault segments would not produce earthquakes even at increased fluid pressures, since a fault also needs to be oriented properly with respect to the stress field to fail.) In contrast, seismogenic faults were all associated with some pore pressure increase, though the magnitude of the increase was generally modest spanned a significant range (as expected for faults with different but generally favorable orientations and diverse histories). Figure is from Hennings et al. (2021).
Moreover, many other faults (some with similar FSP values to the seismogenic faults) have not yet hosted earthquakes (Figures 2 and 5). There are several possible explanations for this. First, it’s important to note that the FSP values represent probabilities of being critically stressed within the estimated parameter uncertainty ranges. Hence, FSP values potentially associated with earthquakes are often fairly low, usually less than 25% chance of the fault being critically stressed at the elevated pore pressure values. Second, the Earth is heterogeneous, and fault orientations and internal structure are more complicated than can be captured at a basin scale. Finally—and with the greatest implications for our ability to anticipate the location and timing of future human-triggered earthquakes—the failure history of natural faults is nearly always unknown, which makes it nearly impossible using current methods to determine the fault’s status within its earthquake cycle and hence the precise amount of shear stress resolved on it. For example, a well-oriented fault that failed recently in geologic time will not have recovered all of the shear stress that built up on it before the last earthquake. In contrast, a similarly oriented fault nearby might have last failed much longer ago and hence could be closer to failure today. Without a detailed record of the long-term slip history on faults, it is not currently possible to determine when a given fault will fail and at what fluid pressure increase.
Finally, I’d like to draw attention to a similar study (which I was not involved with), also published today, in The Seismic Record. That study, by scientists at UT/BEG and Stanford, and again led by Peter Hennings, applied similar methods to understand the slip potential of faults newly mapped in detail within the Delaware Basin, a sub-region of the prolific Permian Basin of west Texas and southeast New Mexico. That study, which also employed the FSP tool, applied a new, smoothed version of the Permian Basin stress mapping originally published by Lund Snee & Zoback (2016, 2018, 2020) to a detailed new 3D fault map by Elizabeth Horne et al. (2021), identifying the faults of greatest concern that are publicly known to exist in the basin.
The two new studies illustrate the power of pairing detailed stress, fault, and pressure evolution models for identifying the faults of greatest concern for induced seismicity, and they also show the limitations of our current abilities to anticipate specifically which faults will produce earthquakes in response to human-caused changes in subsurface fluid pressure, and at what triggering thresholds. Still, this type of analysis can provide insights into how and why prior earthquakes were triggered, and it provides tools and information that operators and regulators can use to manage future earthquakes.