Potential Link Between 2020 Mentone, West Texas M5 Earthquake and Nearby Wastewater Injection: Implications for Aquifer Mechanical Properties

The M5 Mentone earthquake that occurred on March 26, 2020, was the largest event recorded over the last 2 decades in West Texas within the Delaware Basin, a U.S. major petroleum‐producing area. Also, numerous hydrofracturing and wastewater disposal wells are spread across this region. Within a 30 km distance to mainshock, eight class‐II injection wells for industrial wastewater disposal target the deep porous Ellenburger aquifer at an average rate of 1.36 × 106 barrel (BBL) per month during 2012–2020. Poroelastic models of fluid diffusion show these nearby injectors collectively imparted up to 80.5 kPa of Coulomb stress at the mainshock location, capable of triggering this M5 event. Assuming the Mentone event occurs when pore‐pressure increase is maximum, the time delay between peak injection and the M5 occurrence corresponds with an optimal permeability of 6.76 × 10‐14 m2 for the Ellenburger aquifer layer, in agreement with independent estimates.

• Fluid diffusion from nearby injection wells is likely the primary triggering mechanism for the M5 Mentone earthquake • Two nearby injectors could individually impart adequate Coulomb stress to induce the M5 event • Time delay between injection and seismicity is used to constrain the mechanical properties of the aquifer unit Supporting Information: • Supporting Information S1 • Movie S1 • Table S3 Correspondence to: S. Tung, sui.tung@asu.edu; stung6@wisc.edu

Citation:
Tung, S., Zhai, G., & Shirzaei, M. (2021). Potential link between 2020 Mentone, West Texas M5 earthquake and nearby wastewater injection: Implications for aquifer mechanical properties. Geophysical Research Letters, 48, e2020GL090551. https://doi. org/10.1029/2020GL090551 (RRC, 2020;USGS, 2020) (Figure 1(a)). This earthquake is the largest event recorded within the Delaware Basin since the 1995 M5.7 Alpine earthquake (Frohlich, 2012;Savvaidis et al., 2019;Trabant et al., 2012). There are eight deep disposal wells located within 30 km northwest of the epicenter, injecting fluid at an average rate of 5 × 10 5 BBL/month following 2012 (RRC, 2020) ( Figure 1 and Table S1). Here, we investigate the possible link between the injection operation and the Mentone earthquake by simulating the pore fluid pressure change and poroelastic stresses for 2012-2021. The time delay between the commencement of injection operation and the Mentone earthquake's occurrence is used to constrain the Ellenburger injection layer's bulk permeability.

Local Hydrogeology of Southwestern Delaware Basin
The M5 Mentone earthquake took place in the central Delaware Basin, a subbasin of the larger Permian Basin in West Texas and southeast New Mexico (Figure 1(a)) (USGS, 2020). In this region, the majority of the recent seismicity is linked to wastewater disposal (Deng et al., 2020;Frohlich et al., 2020;Skoumal et al., 2020;Zheng et al., 2019) and hydraulic fracturing (Lei et al., 2019;Savvaidis et al., 2020). The Delaware Basin comprises a thick layer (up to 4 km) of early Permian shale deposits (e.g., Bone Spring formation and Wolfcamp formation) overlain by the midlate Permian sediments (e.g., Delaware Mountain Group) (Matchus & Jones, 1984) (https://www.beg.utexas.edu/resprog/permianbasin/gis. htm). The shale formations typically exhibit a low permeability and seat on a carbonate layer belongs to the Ordovician Ellenburger Group (Hennings et al., 2019). Beneath the shale layer, the porous Ellenburger layer serves as a wastewater disposal reservoir, facilitating fluid diffusion and reducing the pore-pressure buildup during wastewater disposal (Hornbach et al., 2016;Ruppel et al., 2005;H. Wang, 2000) (Table S1).

Seismicity and Injection Before the Mentone Earthquake
The M5 Mentone earthquake sequence occurred on a normal fault dipping ∼37° to the south with an average strike of ∼82° . It was preceded by a swarm of 71 events within 5 km whose moment magnitudes range between 0.6 and 3.8 (enclosed by the red box in Figure 1 and Table S1). The swarm began in late-2019, with 69 located below 5 km depth within the Precambrian basement beneath the Ellenburger reservoir layer, according to the TexNet seismic catalog (Savvaidis et al., 2019). The M5 hypocenter is also located at a depth of 9.5 ± 3.3 km (USGS, 2020) or 6.2 ± 1.9 km (TexNet) within the basement. According to the Texas Railroad Commission (RRC) database, there are eight deep injection wells within 30 km of the Mentone epicenter ( Figure 1). We have obtained the associated locations, depths, and pressurized injection volumes of these deep injection wells. The average injection depth is 4.8 km, and the combined peak injection rate is ∼5 × 10 6 BBL/month, which was recorded between late-2018 and early-2019, roughly 10 times the injection rate at the beginning of 2012 (Figures 1(b) and 2). Two active hydrofracking wells were located near the epicenter, whose activity halted weeks before the Mentone event (see Section 4.2 for more discussion).

Coupled Poroelastic and Seismicity Rate Modeling
The evolution of pore pressure and poroelastic stresses induced by fluid injections is simulated using a oneway-coupled numerical code developed by R. Wang (2012)(2013)(2014)(2015)(2016)(2017)(2018)(2019)(2020) with the inverted triangles designating the individual well-injection start date and the shaded area representing the injected volume of each well (RRC, 2020). The dashed lines represent the cumulative number of farfield (gray) seismicity within the map in (a) and near-field (red) events enclosed by the red box in (a).
for more details). A simplified five-layer Earth model is developed to describe the medium's hydrogeological properties (Table S2) (cf., Deng et al., 2020;Zhai & Shirzaei, 2018). Pore pressure is calculated at the bottom of the injection layer, following earlier studies (Zhai & Shirzaei, 2018;Zhai et al., 2019Zhai et al., , 2020. We applied a directed Monte Carlo search method to estimate the Ellenburger layer's permeability given the Mentone earthquake's delayed occurrence (see supplementary information for more details). Changes in Coulomb Failure stress, ΔCFS, quantify how changing stress and pore pressure brings faults closer to failure Stein, 1999): where Δτ is the shear-stress change parallel to the receiver fault strike/rake, Δσ is the normal-stress change perpendicular to the fault surface, ΔP is the pore-pressure change, and μ = 0.6 is the frictional coefficient. The evolution of relative seismicity rate, R seismicity is thus modeled as a function of background stressing rate and ΔCFS (Dieterich, 1994;Segall & Lu, 2015):˙ṡ where ˙o  is the background stressing rate, which is assumed to be 10 -5 MPa/year (Calais et al., 2006), A = 0.003 is a constitutive parameter in the rate-and-state friction law (Segall & Lu, 2015), and  is the background effective normal stress. Based on a normal faulting regime and depth-dependent vertical tectonic stress, the estimated normal stress associated with the M5 event focal mechanism is around 40 MPa at the seismogenic depth (Fan et al., 2016; Lund Snee & Dvory, 2020) (see supplementary information for more details).

Results
Using poroelastic models and injection records obtained from the RRC, we simulate the stress impacted by each injection well to the Mentone earthquake fault. Pore pressure and poroelastic stresses are calculated at the bottom of Ellenburger, and we assume that the basement and Ellenburger group are hydraulically connected. Snapshots of the injection rate, ΔP, ΔCFS, and seismicity rate (R seismicity ) are shown in Figure 2, assuming a reference reservoir diffusivity (D Ellenburger ) of 1 m 2 s -1 (cf., Barbour et al., 2017). The magnitude of µΔP, the pore-pressure component of ΔCFS (Equation 1), is roughly an order of magnitude larger than and opposite in sign to that of the poroelastic stress component, Δτ + µΔσ (Equation 3). Thus, the pattern and amplitude of ΔCFS are mainly controlled by pore-pressure change (Figures 2, 3(a), and S1). We investigate the stress field's temporal evolution (Figure 2) due to injection at wells that have become active since 2012, approximately one new well per year (Figures 1(b) and 2 and Table S1). As the fluid diffuses within the permeable Ellenburger layer (Figures 2(c), 2(g), 2(k), 2(o), 2(s), 2(w), and 2(a1)), the potential of induced seismicity northwest of the M5 epicenter increases due to the elevation of the positive ΔCFS. See the supplementary Movie S1 for the graphical animation of ΔCFS evolution and Figure S1 for models of different reservoir permeability.
In 2012, pore pressure was high in the vicinity to well 1 launched early that year (Figure 2(a)). In the following, the pressure front of 100 kPa propagated radially and merged with that originated from well 2 that became active in late-2013 to the north, resulting in a broader zone of positive ΔP (Figure 2(e)) and ΔCFS (Figure 2(g)). In mid-2015, the zone of positive pore pressure expanded southward to join that of well 3 that is only ∼19 km away from the M5 epicenter and injected at a rate of 5.24 × 10 5 BBL/month (Figures 2(i) and 2(k)). The M5 fault has not received a significant ΔCFS, although a combined volume of 2 × 10 8 BBL has been injected since 2012 (Figures 1(b), 2, and 3(a)). Well 5 (∼12 km away) from the M5 epicenter with an injection rate of 8.73 × 10 5 BBL/month appeared to play a pivotal role in the M5 event's occurrence. In 2018, well 6, 7, and 8 began their operation (Figure 2(q)), amounting the local injection rate to a maximum of 5.3 × 10 6 BBL/month in 2019 and resulting in ΔCFS M5 increase at a rate of ∼40 kPa/year (Figures 1(b), 2(u), 2(w), and 3(a)). In early 2020, positive ΔP and ΔCFS appeared to affect the majority of the study region (Figures 2(y), 2(a1), and 3(c)), and at the same time, the tip of ΔP = 100 kPa and ΔCFS = 50 kPa reached the location of M5 event and most (>97%) swarms (Figures 2(y), 2(a1), and 3(c)), bringing the local faults closer to rupture (Figures 2(y), 2(a1), and 3(c)). The ΔCFS incidence at the mainshock's location and timing and the TUNG ET AL.  Temporal evolution of ΔCFS experienced by the M5 event (red line) and near-field seismicity (black lines) whose occurrence is denoted by the orange dot and red dots, respectively. All near-field swarms and the M5 event happened with ΔCFS incidence > 10 kPa. (d) ΔCFS M5 evolution is compared for the scenarios of D Ellenburger = 0.24, 0.5, 1, 2, and 2.5 m 2 s -1 and the preferred value is 1. (e) Positive relation between the mean/maximum injection rate and ΔCFS M5_incidence received from each well (with exponential fit). (f) Linear regression between ΔCFS M5_incidence gained from each well and its distance from the epicenter. The ΔCFS is averaged over the receiver fault orientations of O 1 and O 2 (USGS, 2020) (also see Figure 2). nearby 71 swarms is, on average, 86 kPa (Figure 3(c)). Some of the swarms could be related to the postseismic effect of the Mentone mainshock. However, we focus on studying the potential link between the nearby injection and the mainshock. The detailed investigation of individual aftershocks and their relationship with the mainshock is beyond this study's scope and the subject of future work.
Next, the seismicity rate, R seismicity was estimated using the ΔCFS time series (see supplementary information for details), showing a spatiotemporal pattern similar to ΔCFS (Figure 2). We obtain a seismicity rate ∼3 times higher than the background rate during the Mentone event (Figure 2(b1)).

Critical Stress Change
To investigate the stress criterion needed for triggering the Mentone event, we invoked the Coulomb failure criterion, in which fault shear strength ( s  ) is related to effective normal stress ( n  ), rocks cohesion ( 0  ), and the coefficient of friction () through where n   is related to principal stress ( ) and pore fluid pressure (P) via The failure may occur when shear stress exceeds the s  for a given n  . Shear stress can alter due to nonzero differential stress changes caused by the imparted poroelastic stress. Also, n   depends on the magnitude of the stresses and the orientation of the fault concerning the tectonic stress field. The principal stresses' magnitude can reduce due to increased pore fluid pressure (Equation 4). Although we suggested that pore-pressure change due to injection might have triggered the Mentone earthquake, we note that establishing a threshold for pressure change to trigger an earthquake is not trivial (e.g., Talwani & Acree, 1985). Following earlier studies (e.g., Townend & Zoback, 2000), we assumed that faults are critically stressed if they have not ruptured recently. Therefore, a small perturbation of the stress field due to fluid diffusion can likely trigger earthquakes. Some examples include seasonal modulation of the seismicity due to hydrological unloading (Carlson et al., 2020;Christiansen et al., 2007;Johnson et al., 2017), triggering earthquakes due to tides (Tanaka et al., 2002;Wilcock, 2001) and induced seismicity due to pore-pressure change by seasonal snowmelt (Montgomery-Brown et al., 2019;Saar & Manga, 2003). Several studies suggested that a pore-pressure increase of 0.01-0.1 MPa can trigger seismicity (Harris, 1998;Roeloffs, 1996;Stein et al., 1992Stein et al., , 1994.
The fault nodal planes of the Mentone earthquake ( Figure 1) are aligned well with the orientation of S hmax (Lund Snee & Dvory,2020;Lund Snee & Zoback,2016, which indicates that the normal fault of the mainshock is near-optimally orientated concerning the stress field. A first-order calculation using the Mohr-circle suggests that a 10° deviation of the fault strike from the optimal orientation requires an extra 0.9 MPa/km stress change to initiate the failure.

Possible Impact of Hydrofracking and Other Triggering Mechanisms
Apart from deep wastewater injection, hydrofracking is also suggested to be a driver of seismicity within the Delaware Basin Schultz et al., 2020;Skoumal et al., 2020). According to the FracFocus database, two nearby hydrofracking wells were active during some of these swarm occurrences, while no fracking activity was documented at the time of the M5 Mentone earthquake. These sites are respectively located 2.4 and 2.9 km away from the epicenter averagely at a true vertical depth (TVD) of ∼3.3 km (Figure 1). This TVD depth represents the distance from the surface to the deepest point of penetration. However, their operations respectively ended 48 and 102 days before the M5 event though HF-induced aseismic slip might be able to trigger swarms over a long period of months (e.g., Eyre et al., 2020).
Furthermore, hydrofracking operations often occur within the shallow shale layer. Thus, without a preexisting pathway, such as faults conduits, hydrofracking fluid's vertical diffusion to the deep basement is not feasible. The presence of fault conduits in this area is yet to be confirmed in independent studies. In short, we cannot rule out the possibility of triggering the Mentone event by hydrofracking operations. However, we argue that this effect is of second-order importance compared with that of deep-seated wastewater injection. More work is needed to investigate the possible link between the nearby hydrofracking and the Mentone event, which requires additional data and new models beyond this study's scope.
The other triggering mechanisms include injected fluid reaching the Mentone hypocentral area from a different direction, such as the southeast, through unmapped faults. As suggested in the literature (e.g., Bhattacharya & Viesca, 2019;Eyre et al., 2019;Guglielmi et al., 2015), injection-induced aseismic fault slip can also trigger earthquakes at farther distances and explain the delay between peak injection and seismicity, which are subjects of future studies. It is also worthy of future investigation if the preceding swarm could cause the Mentone earthquake in late 2019.
Furthermore, the ΔCFS M5_incidence exerted by each well correlates (R 2 = 0.91) with the inverse square of its distance from the M5 epicenter. Thus, closer wells play a more critical role in triggering the Mentone event (Figure 3(f)), as suggested in earlier studies (Peterie et al., 2018;Walsh & Zoback, 2015;Yeck et al., 2016;Zhai et al., 2020). For instance, substituting the injection rate of well 6 (17.7 km from epicenter) with that of well 5 (12 km from epicenter) reduces the imparted ΔCFS M5_incidence from 45 kPa to less than 10 kPa (Figures 3(b) and 3(f)). This implies that a ∼50% increase in the well-fault distance can result in a ∼80% drop in received ΔCFS (Figures 3(b) and 3(f)). This first-order knowledge is useful when planning injection within the vicinity of an active fault. For instance, the M5 fault is influenced by a ΔCFS M5_incidence < 10 kPa when the injectors are located at >25 km distance (i.e., injection occurs only at well 1, 2, 4, and 7) (Figures 3(b) and 3(f)). This information can be used to regulate and plan injection operations to minimize the likelihood of triggering major earthquakes (e.g., Delgado et al., 2001;Marufuzzaman et al., 2015).

Implications for Aquifer Hydrological Properties
Delayed induced seismicity and fault destabilization are observed in Texas, Oklahoma, and Kansas (e.g., Deng et al., 2020;Langenbruch & Zoback, 2016;Zhai et al., 2019), which are associated with the distance between injection and seismicity and rock permeability (e.g., Barbour et al., 2017;Keranen et al., 2013;Walsh & Zoback, 2015). Assuming that the Mentone event occurred at the peak of the ΔP M5 due to wastewater injection, the observed seismic location and timing could be used to constrain the aquifer's hydrological properties, such as permeability. This parameter determines the rate and amplitude of pore-pressure evolution, which in turn controls the seismicity rate changes derived from physics-based induced earthquake models (Chang & Segall, 2016;Langenbruch et al., 2018;H. Wang, 2000;R. Wang & Küumpel, 2003;Zhai et al., 2019). Here, we used the delay between the occurrence of the M5 Mentone earthquake and peak ΔP M5 to search for the most favorable permeability (see supplementary information for details of the search algorithm). An aquifer permeability of Ellenburger k  = 10 -13.17±0.4 m 2 allows the M5 event to occur when the largest ΔP M5_incidence of 166 kPa (or ΔCFS M5_incidence of 87 kPa) (Figure 4(b)) is imparted to the area of Mentone event in early 2020. This is consistent with the permeability range of 10 -13.7 to 10 -13 m 2 suggested from prior studies (Domenico & Schwartz, 1998;Hornbach et al., 2016;RRC, 2020) (Figure 4(a)).
To test the sensitivity of optimized k* to the M5 location accuracy, we repeated our search, but this time considered an average horizontal location error of ∼2 km for the relocated M5 event (e.g., Savvaidis et al., 2019). The optimum k* is defined so that the associated pore-pressure change is maximum within a square box of 4 km × 4 km centered at the Mentone earthquake epicenter. We find that a k* between 10 -13.04 m 2 and 10 -13.17 m 2 fulfills this criterion, which is in the same range as our initial estimation.

Data Availability Statement
The Railroad Commission of Texas provides the injection data at http://gis.rrc.texas.gov/GISViewer/ (last access in 2020 November). The FORTRAN code implementing R. Wang and Küumpel (2003) approach is available at https:// www.gfz-potsdam.de/en/section/physics-of-earthquakes-and-volcanoes/infrastructure/tool-development-lab/ (last access in 2020 November). Information on the hydrofracking wells is TUNG ET AL.  k  = 10 -13.17 m 2 (or optimal aquifer diffusivity, Ellenburger D  = 1.64 m 2 s -1 ) maximizing ΔP M5_incidence and such permeability falls within the ranges of existing studies (Domenico & Schwartz, 1998;Hornbach et al., 2016;RRC, 2020). (b) ΔCFS M5_incidence , ΔCFS _M5_max , and ΔP M5_incidence are plotted for each MC-sampled permeability. (c) Arrival time of ΔCFS M5_min , ΔCFS M5-ve>+ve , ΔCFS M5 = 10 kPa, and ΔCFS M5_max . (d) Temporal evolution of ΔCFS M5 of each MC-sampled permeability, k Ellenburger between 10 -14 and 10 -12 m 2 . The red line shows the evolution of maximizing the pore-pressure changes at the M5 location, as modeled with the optimal aquifer permeability, k*. Due to the substantial focal-depth uncertainty and ambiguous permeability structures within the basement (Savvaidis et al., 2019;USGS, 2020), the modeling results are investigated at the basement top of the epicentral location, assuming that permeable faults connect the Ellenburger aquifer to the in-depth M5 hypocenter for rapid fluid migration. available at the FracFocus database (http://fracfocusdata.org/digitaldownload/FracFocusCSV.zip, last access in 2020 November).