Volume 46, Issue 24 p. 14412-14420
Research Letter
Free Access

The August 2018 Kaktovik Earthquakes: Active Tectonics in Northeastern Alaska Revealed With InSAR and Seismology

É. Gaudreau

Corresponding Author

É. Gaudreau

School of Earth and Ocean Sciences, University of Victoria, Victoria, British Columbia, Canada

Correspondence to: É. Gaudreau,

[email protected]

Search for more papers by this author
E. K. Nissen

E. K. Nissen

School of Earth and Ocean Sciences, University of Victoria, Victoria, British Columbia, Canada

Search for more papers by this author
E. A. Bergman

E. A. Bergman

Global Seismological Services, Golden, CO, USA

Search for more papers by this author
H. M. Benz

H. M. Benz

U.S. Geological Survey, Golden, CO, USA

Search for more papers by this author
F. Tan

F. Tan

School of Earth and Ocean Sciences, University of Victoria, Victoria, British Columbia, Canada

Search for more papers by this author
E. Karasözen

E. Karasözen

Department of Geophysics, Colorado School of Mines, Golden, CO, USA

Search for more papers by this author
First published: 05 December 2019
Citations: 19

Abstract

The largest earthquakes recorded in northern Alaska ( urn:x-wiley:grl:media:grl59946:grl59946-math-0001 6.4 and urn:x-wiley:grl:media:grl59946:grl59946-math-0002 6.0) occurred urn:x-wiley:grl:media:grl59946:grl59946-math-00036 hr apart on 12 August 2018, in the northeastern Brooks Range. The earthquakes were captured by Sentinel-1 interferometric synthetic aperture radar (InSAR) satellites and Earthscope Transportable Array seismic data, giving insight into the little-known active tectonic processes of Arctic Alaska, obscured until recently by sparse data availability. In this study, InSAR modeling, teleseismic back projections, calibrated hypocentral relocations, and regional moment tensor solutions resolve two previously unknown, SSW dipping right-lateral fault segments. These are the first active faults identified as conjugate to the NE trending sinistral Canning displacement zone directly to the west, which is therefore a more complex zone of diffuse faulting than previously thought. The northeastern Brooks Range has been characterized as an area of low to moderate seismic hazard, but these earthquakes illustrate the potential for larger, possibly destructive events in a region earmarked for rapid resource development.

Key Points

  • The largest earthquakes recorded in northern Alaska ( urn:x-wiley:grl:media:grl59946:grl59946-math-0005 6.4 and urn:x-wiley:grl:media:grl59946:grl59946-math-0006 6.0) occurred on previously unknown active faults
  • InSAR and calibrated hypocenter relocations indicate that the faults are conjugate to the Canning displacement zone
  • The Canning displacement zone involves a complex fault network and vertical-axis block rotations

Plain Language Summary

The largest earthquakes recorded in northern Alaska (magnitude 6.4 and magnitude 6.0) occurred urn:x-wiley:grl:media:grl59946:grl59946-math-00046 hr apart on 12 August 2018. Few active faults are mapped in this region despite widespread seismicity, and the current tectonic setting remains unclear due to limited available data and the remote location. We use satellite radar images and seismic data to resolve two previously unknown fault segments, along which the magnitude 6.4 earthquake ruptured unilaterally eastward. This fault geometry demonstrates that the Canning displacement zone, the main tectonic feature in the area, is a more complex zone of diffuse faulting than previously thought. These results are also important for reassessing seismic hazard by illustrating the potential for damaging earthquakes on seemingly aseismic faults.

1 Introduction

The urn:x-wiley:grl:media:grl59946:grl59946-math-0007 6.4 earthquake that occurred on 12 August 2018, urn:x-wiley:grl:media:grl59946:grl59946-math-000880 km SW of Kaktovik, Alaska, is the largest ever recorded in the Brooks Range or its foreland basin to the north (Figure 1). The second largest earthquake in this region, a urn:x-wiley:grl:media:grl59946:grl59946-math-0009 6.0 aftershock, occurred urn:x-wiley:grl:media:grl59946:grl59946-math-00106 hr later, urn:x-wiley:grl:media:grl59946:grl59946-math-001135 km to the east (Figure 1b). More than 6,000 aftershocks recorded within 1 year of the mainshock form a WNW-ESE striking trend at the northern margin of the eastern Brooks Range, beneath the Sadlerochit Mountains (Ruppert & West, 2020). The only known faults in this area are mapped as pre-Quaternary and do not align with the 2018 seismicity (Koehler, 2013).

Details are in the caption following the image
(a) Digital elevation model (DEM) of Alaska (USGS 30 ARC-second Global Elevation Data, GTOPO30; https://doi.org/10.5066/F7DF6PQS) with location of epicenters prior to the 2018 Kaktovik sequence in USGS Comcat Catalog (white circles; https://earthquake.usgs.gov/earthquakes/search/) and Quaternary faults are bold black lines (Koehler, 2013). Study area is located in the white rectangle. CDZ: Canning displacement zone. (b) Shaded-relief DEM (Porter et al., 2018) with regional moment tensor solutions for the 2018 Kaktovik mainshock and urn:x-wiley:grl:media:grl59946:grl59946-math-0012 3.5+ aftershocks from the following 5 months plotted at the relocated epicenter locations. White circles represent additional relocated epicenters. Pre-Quaternary faults are thin black lines (Koehler, 2013). SMT: south dipping Sadlerochit Mountains thrust. Location of closest seismic stations used for calibrated relocations are indicated by black triangles.

Despite widespread seismicity in Arctic Alaska, few active faults are mapped, and currently published GPS velocities are sparse (Snay et al., 2016). The current tectonic setting is equivocal (Finzel et al., 2015; Fuis et al., 2008; Haeussler, 2008; Koehler, 2013; Leonard et al., 2008; Mazzotti et al., 2008), but a relatively recent increase in seismic and geodetic data available from this area, with the deployment of the USArray in Alaska since 2014 and launch of the Sentinel-1 satellite pair in 2014 and 2016, permit a more detailed characterization of active tectonics in the Brooks Range.

We use Sentinel-1 interferometric synthetic aperture radar (InSAR) data and elastic dislocation modeling to characterize the fault geometry and slip distribution of the 2018 Kaktovik sequence, which are the most northerly earthquakes ever imaged in this way. We use seismic back projections, calibrated hypocentral relocations and regional moment tensors (RMTs) to map the mainshock rupture and aftershock activity over time and space. These new constraints are used to reassess regional tectonics and provide new information for seismic hazard assessments in an area of keen interest to the petroleum industry for oil drilling.

2 Geologic and Tectonic Setting

The Brooks Range (Figure 1a) is underlain by a thick crustal root where the Moho is mapped to depths of urn:x-wiley:grl:media:grl59946:grl59946-math-001350 km (compared to urn:x-wiley:grl:media:grl59946:grl59946-math-001430–35 km in central Alaska) and the lithosphere is up to urn:x-wiley:grl:media:grl59946:grl59946-math-0015200 km thick (Fuis et al., 1997; Jiang et al., 2018; Miller et al., 2018; O’Driscoll & Miller, 2015; Ward and Lin, 2018). Mantle flow or northward motion of the Yakutat indentor are thought to be driving current tectonic activity (Finzel et al., 2015; Mazzotti & Hyndman, 2002; Mazzotti et al., 2008). According to moment tensor solutions, the northeastern Brooks Range is currently dominated by a transtensional tectonic regime, in contrast to transpression south of the Brooks Range, where the lithosphere is weaker and deforming at a faster rate (Jiang et al., 2018; Leonard et al., 2008; O’Driscoll & Miller, 2015). The main mapped tectonic feature in the northeastern Brooks Range is the Canning displacement zone (CDZ), a NE-SW left-lateral diffuse deformation zone first identified by Grantz et al. (1983) located between the northeastern Brooks Range and the North Slope Deep Magnetic High (Figure 1a), a domain of mechanically strong crust to the west (Saltus et al., 1999).

The Brooks Range is composed of multiple arc and continental margin terranes, which accreted onto the North American margin at the onset of the Brookian orogeny in the latest Jurassic-Cretaceous (Moore, 1992; Moore et al., 1997). The first phase of the Brookian orogeny resulted in thin-skinned deformation in the Brooks Range south of the study area (Moore & Box, 2016) and was followed by middle-Late Cretaceous extension at the southern margin of the Brooks Range (Amato & Miller, 2004; Hannula et al., 1995). A Paleogene episode of thick-skinned deformation formed large-scale duplex structures in the northeastern Brooks Range and is coeval with accretionary deformation in the forearc in southern Alaska and right-lateral strike-slip faulting in the interior (Fuis et al., 2008; Wallace, 1992). This synchronicity suggests that these structures may be linked by a detachment (Moore & Box, 2016).

The 2018 Kaktovik earthquakes occurred beneath the Sadlerochit Mountains, at the northern margin of the northeastern Brooks Range (Figure 1). The geology of these mountains is dominated by Neoproterozoic dolomite overlain by Cambrian to Ordovician limestone (Molenaar et al., 1987, and references therein). Most of the mapped tectonic features in this area formed during the later reactivation of the Brookian orogeny in the Paleogene, expressed on the surface as north vergent listric thrusts and east-west trending folds in the Sadlerochit Mountains and adjacent Kikitat and Shublik Mountains (Moore et al., 1997; O’Sullivan et al., 1993; Wallace & Hanks, 1990).

3 InSAR Analysis

3.1 InSAR Data

The European Space Agency's Sentinel-1 satellites captured the coseismic surface deformation of the 2018 Kaktovik earthquakes. Two 12-day descending track interferograms and one 42-day ascending track interferogram (the shortest repeat time available on ascending orbits) were used in this study, obtained from the automated SARVIEWS program (Meyer et al., 2016) (Figure 2a). Since available interferogram pairs capture the coseismic deformation from both the urn:x-wiley:grl:media:grl59946:grl59946-math-0016 6.4 and urn:x-wiley:grl:media:grl59946:grl59946-math-0017 6.0 events, in this section we discuss the cumulative coseismic deformation without specifying the causative earthquake.

Details are in the caption following the image
(a) Sentinel-1 interferograms, models based on distributed slip with variable rake on two fault planes, and residuals superimposed onto topography from ArcticDEM (Porter et al., 2018). Surface projections of the modeled, buried fault segments are in black. (b) Slip distribution on the fault planes with variable rake. Each fault segment measures 1 km  urn:x-wiley:grl:media:grl59946:grl59946-math-0018 1 km. (c) The model slip area is shaded, with the mainshock and aftershocks superimposed (white circles with 95% confidence ellipses). Colored circles represent back projected 0.2- to 2-Hz energy for the urn:x-wiley:grl:media:grl59946:grl59946-math-0019 6.4 mainshock, colored by time since rupture initiation and scaled by relative energy. The average rupture velocity is 2.6 km/s.

The radar line-of-sight (LOS) coseismic deformation appears as three lobes in the ascending and descending track interferograms, with one northern lobe and two southern ones (Figure 2a). In all interferograms, peak displacements are greatest in the southern lobes, which are LOS range increase in descending interferograms and range decrease in the ascending interferogram. The northern lobes exhibit the opposite sense of motion, consistent with right-lateral faulting along an approximately E-W trend.

3.2 InSAR Modeling

To characterize the causative faulting, the LOS displacements were downsampled using a quadtree algorithm (Jónsson et al., 2002) and the reduced data were inverted for slip on rectangular dislocations embedded within a uniform elastic half-space with Lamé parameters urn:x-wiley:grl:media:grl59946:grl59946-math-0020 = urn:x-wiley:grl:media:grl59946:grl59946-math-0021 = 3.2  urn:x-wiley:grl:media:grl59946:grl59946-math-0022 10 urn:x-wiley:grl:media:grl59946:grl59946-math-0023 Pa (Okada, 1985; Wright et al., 1999). The single ascending data set was given equal weighting to the two descending data sets in the inversion. We used Powell's algorithm (Press et al., 1992) to solve for the best-fit fault strike, dip, rake, slip, surface projection center point, length, and top and bottom depths, avoiding local minima by repeating the inversion 500 times with starting parameters sampled randomly from the ranges given in supporting information Table S1 (Wright et al., 1999).

Our preferred slip model was determined using the two-step methodology outlined by Elliott et al. (2012). In the first step, fault geometry was established using a small number of rectangular, uniform slip model faults. A single rectangular fault cannot reproduce the three-lobed fringe pattern (supporting information Table S2 and Figure S1); however, two faults provide a good visual fit (root-mean-square residual displacements of 1.20  urn:x-wiley:grl:media:grl59946:grl59946-math-0024 10 urn:x-wiley:grl:media:grl59946:grl59946-math-0025 m). Both en echelon segments strike ESE, involve right-lateral slip, and dip toward the SSW, explaining the greater deformation observed in the southern lobes of the interferograms (supporting information Figure S2 and Table S3). The urn:x-wiley:grl:media:grl59946:grl59946-math-002615 km-long western segment is steeper (82°) than the urn:x-wiley:grl:media:grl59946:grl59946-math-002712 km-long eastern segment (64°). Uncertainties in these uniform slip parameters were estimated using Monte Carlo inversions of 100 data sets perturbed with realistic noise (Elliott et al., 2012). Strike, rake, length, and center point easting and northing are well constrained for both faults, with relative standard deviations urn:x-wiley:grl:media:grl59946:grl59946-math-00289% (supporting information Figures S3 and S4). Dip is less well constrained for the eastern fault, but a shallower angle than that of the western fault is resolved. The results also show that the rupture did not reach the surface; thus, slip, fault width, minimum depth, maximum depth, and seismic moment are less well constrained (relative standard deviations urn:x-wiley:grl:media:grl59946:grl59946-math-002920%) due to strong trade-offs between these parameters. We also explored the possibility of a listric fault—as the Sadlerochit Mountains thrust is believed to be (O’Sullivan & Wallace, 2002)—which would extend the rupture area farther south. Listric slip models were created using a steeper slip plane near the surface and a deeper segment with a shallower dip; the results were very similar for each model that was created, one such model is described in supporting information Tables S5 and S6. In all cases, slip predominantly occurs on the steeper fault segments near the surface; thus, the slip does not extend farther south in the listric models. Therefore, slip on two steep fault segments remains our preferred model.

In the second step, we solved for the distribution of slip and rake across the two fault planes (Funning et al., 2005). The western and eastern fault lengths were increased to 30 and 20 km, respectively; their bottom depths increased to 10 and 15 km and were subdivided into 1-km  urn:x-wiley:grl:media:grl59946:grl59946-math-0030 1-km patches. We solved for the rake and slip magnitude of each patch using a Laplacian smoothing operator to ensure realistic slip gradients (Figure 2). Allowing for distributed slip and rake reduced the root-mean-square residual displacements to 1.066 urn:x-wiley:grl:media:grl59946:grl59946-math-0031 10 urn:x-wiley:grl:media:grl59946:grl59946-math-0032 m, though notable residuals remain between the two faults and at the eastern termination of the eastern fault (Figure 2a). Rake is consistently right lateral on both faults. Slip is concentrated between depths of 2.7 and 11.0 km on the eastern fault, peaking at 1.8 m at 6-km depth. Slip is concentrated between 1.5 and 9.5 km of slip on the western fault, peaking at 1.3 m at 4-km depth.

The InSAR moments for the western and eastern faults are 2.798  urn:x-wiley:grl:media:grl59946:grl59946-math-0033 10 urn:x-wiley:grl:media:grl59946:grl59946-math-0034 and 3.259  urn:x-wiley:grl:media:grl59946:grl59946-math-0035 10 urn:x-wiley:grl:media:grl59946:grl59946-math-0036 N m, respectively, and the combined moment magnitude from slip on both faults is 6.5. For comparison, the U.S. Geological Survey W-phase moment magnitude for this event is estimated at 6.4.

4 Seismological Constraints

4.1 Calibrated Earthquake Relocations

We reevaluated hypocenter locations of the Kaktovik mainshock and 109 of the best-recorded aftershocks using the mloc calibrated earthquake relocation technique (Bergman & Solomon, 1990; Walker et al., 2011). This minimizes location bias by assuming that raypaths of events clustered in space and recorded at common stations sample roughly the same portion of Earth, such that travel time differences more likely reflect the relative epicenter locations within the cluster rather than the 3-D velocity structure. The mloc method splits the relocation into two independent steps, each utilizing a specific, tailored set of arrival time data (Jordan & Sverdrup, 1981). First, relative locations of each hypocenter within the cluster are estimated from differences in arrival times picked at common stations at all distances. Second, the absolute location of the hypocentroid—the geometric mean of the cluster—is calculated based on the observed travel times at local distances, since nearby stations will have accumulated less travel time error. This step is a direct calibration—an indirect calibration (i.e., using an independent data set to constrain the absolute location of the hypocentroid) may be used in the event that a cluster of events cannot be reliably calibrated using travel times. The azimuthal coverage is not ideal in this case since only one regional station (C26K) is located north of the cluster.

Using a direct calibration, the relocated mainshock hypocenter lies urn:x-wiley:grl:media:grl59946:grl59946-math-00377 km south of the western model fault plane, as do most aftershocks. Three hypotheses for this discrepancy were explored: (1) a timing error at the only station north of the cluster (C26K), exerting a strong influence on the latitude of the relocated cluster; (2) lateral heterogeneities in crustal seismic velocity, notably those north of the cluster in the Colville basin (Fuis et al., 1997; Moore et al., 1994); and (3) the fault that ruptured was listric; thus, the true slip patch could extend farther south than that shown in Figure 2c. As described in section 3.2, listric fault models were tested; however, the slip in these models is concentrated on the steep parts of the fault, which would not extend the slip patch farther south. On the other hand, the InSAR data are mostly sensitive to shallow slip, thus it is possible that some unresolvable component of slip occurred on deeper parts of the faults that have a shallower dip. As for the possibility of a timing error, clock quality at station C26K was assessed using the IRIS timeseries Web Service (https://service.iris.edu/irisws/timeseries/docs/). From these data, we conclude that time errors are not large enough to affect the location of the calibrated cluster. Moreover, an unreasonably large timing error of 1.2 s at station C26K is required to shift the cluster northward to correct the discrepancy with the InSAR model; thus, a heterogeneous crust is a more likely explanation.

An indirect calibration was performed using the InSAR model faults due to the likelihood of strong lateral heterogeneities in the crust biasing the epicenters urn:x-wiley:grl:media:grl59946:grl59946-math-00387 km to the south (Fuis et al., 1997). The mainshock and urn:x-wiley:grl:media:grl59946:grl59946-math-0039 6.0 aftershock were used as calibration events (Figure 2c and supporting information Table S7). The origin times for calibration were set so that the arrival times at the local station in the Brooks Range, D25K, agreed with a typical crustal model. The urn:x-wiley:grl:media:grl59946:grl59946-math-00406.4 mainshock nucleated close to the western end of the western model fault, implying that it ruptured unilaterally eastward. The urn:x-wiley:grl:media:grl59946:grl59946-math-00416.0 aftershock hypocenter is near the eastern end of the eastern model fault, among a concentration of smaller events.

There was also an increase in seismic activity southwest of the study area starting in late July 2018 and peaking in October 2018 (Ruppert & West, 2020). These events will be referred to as the Niviak cluster, after the nearby Niviak Pass. Some of these events cluster in multiple NW-SE trends similar to the Kaktovik cluster, while others form more diffuse groups. We investigate whether both clusters are related to the larger-scale left-lateral motion of the CDZ by relocating the 155 best-recorded earthquakes in this area using a direct calibration (supporting information Figure S5).

4.2 RMTs

RMT solutions were calculated for the mainshock and 87 best-recorded aftershocks. The RMTs were computed using the same Green's functions, fit function, and filtering strategy found in Herrmann et al. (2011). Unlike the grid search approach of Herrmann et al. (2011), we used a linear inversion to solve for the moment tensor components and assumed a pure double-couple source. For events urn:x-wiley:grl:media:grl59946:grl59946-math-0042 4.0 or smaller, waveform filtering is typically done in the passband 16–50 s, while for larger events waveform filtering is typically done in the passband 20–50 s. For all events, we used the central U.S. velocity model of Herrmann et al. (2011), an observational distance range 0–500 km, and three-component waveforms that included body waves and surface waves. RMT solutions were computed in the depth range 2–24 km in increments of 1 km. The RMT solution for each event is the one with the best fit as a function of depth. All RMT solutions were computed using the single event locations found in the U.S. Geological Survey earthquake catalog (https://earthquake.usgs.gov/earthquakes/map/). Given the good azimuthal distribution of stations surrounding the source region and longer period signals used in the inversion, we expect little differences in computed mechanisms whether using the single or multiple locations in this study.

The large majority of RMTs are consistent with E-W right-lateral strike-slip motion. The southward dips (mostly urn:x-wiley:grl:media:grl59946:grl59946-math-004360°) of the E-W nodal planes are in agreement with the dips of the modeled fault segments (Figures 1b and S6 and Table S8). Centroid depths range from 0–22.7 km, with most urn:x-wiley:grl:media:grl59946:grl59946-math-004410 km. The mainshock has a centroid depth of 2.2 km.

4.3 Teleseismic Back Projection

We applied a phase-weighted relative back projection method (Tan et al., 2019) to track the rupture energy in time and space. This provides a measure of relative energy release, rather than true moment release, since the amplitudes are normalized and the phases are used as a weighting factor. The phase-weighted stacking emphasizes both large amplitudes and the coherency of the signal, reducing biases introduced by incoherent signals with large amplitudes.

We performed back projections for both the urn:x-wiley:grl:media:grl59946:grl59946-math-0045 6.4 mainshock and urn:x-wiley:grl:media:grl59946:grl59946-math-0046 6.0 aftershock using teleseismic stations from the contiguous United States, and the teleseismic travel times are estimated using the IASP91 reference model (Kennett & Engdahl, 1991). The 10-s window that was used gave rise to artifacts in the form of multiple energy radiators at the same latitude and longitude, but with slightly different source times. Of the energy radiators at the same coordinates, the result with the median source time was chosen and the rest discarded. Other array configurations were considered; however, they had less suitable distance and azimuth ranges (Tan et al., 2019). Relocated epicenters from section 4.1 were used to constrain the location of each rupture. For the mainshock, the back projection indicates linear rupture propagation from WNW to ESE at an average velocity of 2.6 km/s (Figures 2c and S7a and Table S9). Coherent back projected energy encompasses the full length of the InSAR model faults, with peak energy release occurring after urn:x-wiley:grl:media:grl59946:grl59946-math-004710 s on the eastern fault. Though close to the limit of what a back projection can resolve, the urn:x-wiley:grl:media:grl59946:grl59946-math-0048 6.0 aftershock also seems to have propagated from WNW to ESE and thus likely occurred on the same eastern fault segment rather than on a conjugate fault (supporting information Figure S7b and Table S9).

5 Discussion

5.1 Faulting in the 2018 Kaktovik Sequence

The InSAR modeling suggests rupture of two en echelon, ESE striking right-lateral faults within the Sadlerochit Mountains (Figure 2a). This is supported by distributions of back projected energy and aftershocks, which are both oriented WNW-ESE, and their RMT solutions (Figures 1b and 2c). The lack of decorrelation between the northern and southern lobes in the 6–18 August 2018 interferogram (Figure 2a) suggests that neither segment ruptured to the surface, in agreement with our InSAR slip model. The mainshock centroid depth of 2.2 km is much shallower than both the relocated focal depth of 11 km and most of the slip in the InSAR model (Figure 2b). The hypocenter depth is located near the bottom of the rupture, as is often the case (Karasözen et al., 2016). We interpret that this discrepancy reflects uncertainties of urn:x-wiley:grl:media:grl59946:grl59946-math-00495 km in RMT centroid depths (Herman et al., 2014).

Back projection results indicate that the urn:x-wiley:grl:media:grl59946:grl59946-math-0050 6.4 mainshock ruptured both faults, with greater relative energy release on the eastern segment, consistent with the high slip patch in the InSAR slip model (Figure 2). Relative energy mapped at 18 s is located urn:x-wiley:grl:media:grl59946:grl59946-math-005115 km SE of the modeled rupture area (Figure 2c), and this most likely constitutes a “swimming artifact” since this is approximately the same azimuth as the seismic stations used in the contiguous United States (Koper et al., 2012), though we cannot rule out an off-fault, near-instantaneous aftershock (Fan & Shearer, 2016). The urn:x-wiley:grl:media:grl59946:grl59946-math-0052 6.0 aftershock probably enlarged the eastern slip patch or ruptured a shallower part of the fault (supporting information Figure S7). Smaller aftershocks concentrate in this same area, with a number located farther east and south, a spatial bias in the direction of mainshock rupture propagation that may indicate a component of dynamic triggering (Gomberg et al., 2003).

The western fault cuts obliquely across the Paleogene Sadlerochit Mountains thrust, mapped as a south dipping listric thrust in this area (O’Sullivan & Wallace, 2002). However, the eastern fault (strike 98°) is parallel to the eastern segment of the Sadlerochit Mountains thrust; thus, it is possible that the steep, shallower part of this listric fault was reactivated. The 2018 Kaktovik earthquakes may have ruptured a combination of unknown and known faults. The steep, southward nodal plane dip angles in nearly all of the Kaktovik focal mechanisms also hint that the structural fabric of the Sadlerochit Mountains may have influenced the geometry of the strike-slip rupture plane.

5.2 Regional Tectonics

The Kaktovik earthquakes exposed the first known active faults in the northeastern Brooks Range that are conjugate to the NE-SW left-lateral CDZ directly to the west (Figures 1a and 2). Previous moment tensor solutions have been interpreted as representing NNE-SSW left-lateral faulting based on the known slip sense of the CDZ. However, diffuse zones of shearing are often associated with conjugate strike-slip faults (Cunningham, 2005; Ghods et al., 2015; Soumaya et al., 2018), and the faults in the Sadlerochit Mountains are optimally oriented with respect to the local maximum horizontal principal stress for right-lateral slip (Hanks et al., 2000).

To explore whether faulting of this orientation may be widespread within the eastern Brooks Range, we performed an additional relocation of the Niviak cluster (Figure 3). The results highlight several discrete concentrations of events along similar NW-SE trends, which presumably represent similar faults conjugate to the CDZ. These right-lateral faults must rotate counterclockwise around vertical axes in order to accommodate overall left-lateral motion along the main CDZ trend (Kim et al., 2004).

Details are in the caption following the image
Relocated epicenters for the Kaktovik mainshock and best-recorded aftershocks from the following 12 months. The Niviak cluster includes the seismic events southwest of the Kaktovik sequence. Some of these events potentially highlight similar conjugate structures to the CDZ (highlighted in red). The arrows demonstrate a simplified version of the block rotation model for northeastern Alaska where the primary faults are NE-SW oriented left lateral, and the secondary faults are NW-SE-oriented right lateral.

The northeastern Brooks Range exhibits low seismic deformation rates and is underlain by a thick Moho ( urn:x-wiley:grl:media:grl59946:grl59946-math-005350-km depth) and thick and strong lithosphere (Fuis et al., 2008; Jiang et al., 2018; Leonard et al., 2008; O’Driscoll & Miller, 2015). We interpret that faults involved in the Kaktovik sequence and the CDZ compose a diffuse zone of active faulting that accommodates slow strain between two mechanically strong lithospheric domains: the NE Brooks Range to the east and the North Slope Deep Magnetic High to the west (Jiang et al., 2018; O’Driscoll & Miller, 2015; Saltus et al., 1999). This has parallels with other diffuse deformation zones such as in northwestern Iran (Ghods et al., 2015), the Mongolian Altai (Cunningham, 2005), and the Alboran-Rif domain in northwestern Africa (Soumaya et al., 2018), many of which exhibit low levels of instrumental seismicity, and low strain rates, but are capable of hosting large earthquakes. The Kaktovik earthquakes thus highlight the importance of reassessing the seismic hazard of areas of low internal deformation.

6 Conclusion

The 12 August 2018 urn:x-wiley:grl:media:grl59946:grl59946-math-0054 6.4 and urn:x-wiley:grl:media:grl59946:grl59946-math-0055 6.0 Kaktovik earthquakes occurred on previously unknown active right-lateral faults that are conjugate to the CDZ, striking ESE. The urn:x-wiley:grl:media:grl59946:grl59946-math-0056 6.4 mainshock nucleated on the western fault and propagated unilaterally eastward onto the eastern fault, where most of the slip and energy release occurred. The urn:x-wiley:grl:media:grl59946:grl59946-math-0057 6.0 aftershock likely further extended the slip area of the mainshock. A direct calibration results in hypocenters systematically biased to the south by urn:x-wiley:grl:media:grl59946:grl59946-math-00587 km, possibly due to the 1-D velocity model and less-than-ideal azimuthal coverage used in the relocations. Using an indirect calibration, relocated mainshock and aftershock hypocenters lie on the rupture area of the InSAR-derived model faults. These earthquakes are the largest-magnitude events recorded in northern Alaska and the first determined as conjugate to the CDZ. Other areas with potential NW-SE right-lateral faulting have been identified from relocated earthquake trends south of the study area. Together, these right-lateral faults may accommodate the overall left-lateral motion of the CDZ by rotating about vertical axes. The relatively low seismicity and deformation rates in the northeastern Brooks Range reflect its thickened crust and lithosphere, but the Kaktovik earthquakes nevertheless highlight the potential for damaging earthquakes on seemingly aseismic faults.

Acknowledgments

É. G. was supported through a Graduate Scholarship from the University of Victoria and an Alexander Graham Bell Canada Graduate Scholarship from the Natural Sciences and Engineering Research Council of Canada (NSERC). E. N. was supported through NSERC Discovery Grant 201704029, a Canada Research Chair, and grants from the Canada Foundation for Innovation and B. C. SAR interferograms were made using European Space Agency Copernicus Sentinel data [2018] and the University of Alaska Fairbanks SARVIEWS web portal (Meyer et al. 2017; http://sarviews-hazards.alaska.edu/Earthquakes/), which is funded by the National Aeronautics and Space Administration (NASA). Seismic back projection data were obtained from Incorporated Research Institutions for Seismology (https://ds.iris.edu/wilber3/find_event) and data used in the relocations and regional moment tensors were gathered from the U.S. Geological Survey (USGS) National Earthquake Information Center (NEIC; https://earthquake.usgs.gov/earthquakes/map/). Computer programs used in the paper are available from the references cited in the text. We also thank two anonymous reviewers and Morgan Moschetti (USGS) for helpful comments.