The Denali fault, a transcurrent fault system that extends from northwestern Canada across Alaska toward the Bering Sea, is partitioned into segments that exhibit variable levels of historical seismicity. A pair of earthquakes (M 6.2 and 6.3) on 1 May 2017, in proximity to the Eastern Denali fault (EDF), exhibited source mechanisms and stress conditions inconsistent with expectations for strike-slip fault activation. Precise relocation of ∼1500 aftershocks revealed distinct fault strands that are oblique to the EDF. Calculated patterns of Coulomb stress show that the first earthquake likely triggered the second one. The EDF parallels the Fairweather transform, which separates the obliquely colliding Yakutat microplate from North America. In our model, inboard transfer of stress is deforming and shortening the mountainous region between the EDF and the Fairweather transform. This is supported by historical seismicity concentrated southwest of the EDF, suggesting that it now represents a structural boundary that controls regional deformation but is no longer an active fault.
The Denali fault (DF) is a 2100-km-long, dextral strike-slip fault in northwestern Canada and Alaska (Fig. 1). In Alaska, the Central Denali fault (CDF) has attracted considerable attention, particularly since the M 7.9 Denali earthquake in 2002 that resulted in surface rupture extending >340 km (Fig. 1; Eberhart-Phillips et al., 2003). This earthquake initiated on a previously unrecognized fault strand and thereupon activated the CDF, but it also splayed southward along the Totschunda fault, thus bypassing the Eastern Denali fault (EDF).
On 1 May 2017, two earthquakes of M 6.2 and 6.3 occurred during a 2 h interval near the border between Yukon and British Columbia, Canada, in proximity to the EDF (Fig. 2). Previous studies of the focal mechanisms of these events, as well as historical seismicity in the vicinity, reveal a mix of reverse and strike-slip faulting styles (Doser and Rodriguez, 2011; He et al., 2018; Feng et al., 2019). This sequence, herein named the St. Elias earthquake sequence, has been interpreted as reactivation of the Duke River fault, a southwest-dipping, terrane-bounding structure that runs subparallel to the EDF (He et al., 2018). Interferometric synthetic aperture radar (InSAR) and GPS data support the interpreted fault geometry (Feng et al., 2019), but the possibility of future activity on the EDF is unresolved, with important implications for earthquake hazards that could impact Whitehorse, Yukon, and Juneau, Alaska.
To address this issue, this study investigated the stress regime and tectonic implications of the St. Elias earthquake sequence. The earthquake doublet and prolonged aftershock sequence were exceptionally well recorded by EarthScope Transportable Array stations. Our results indicate that the current stress regime does not favor slip along the EDF; instead, deformation is distributed across the Insular superterrane between the active Yakutat–North American plate boundary and the Yukon-Tanana terrane.
The western margin of North America is composed of oceanic and island-arc terranes that accreted during the Mesozoic Era (Fig. 1). The DF separates the Insular superterrane from the pericratonic Yukon-Tanana (YT) terrane, a part of the Intermontane belt. The YT terrane is composed of a mid-Paleozoic volcanic-plu-tonic assemblage that was strongly deformed and metamorphosed by Late Triassic time and subsequently intruded by the coast plutonic arc (Mortensen, 1992; Israel et al., 2014). The Wrangellia and Alexander terranes are separated by the Duke River fault (DRF) and together form the Insular superterrane. The Alexander terrane consists of siliciclastic, carbonate, and volcanic rocks, ranging from Cambrian to Late Triassic age, while the Wrangellia terrane is a late Paleozoic island-arc assemblage overlain by Mesozoic flood basalts and sedimentary rocks (Cobbett et al., 2016). During the Jurassic and Cretaceous, the Insular superterrane was intruded by several older arcs, and later by the active Wrangell arc, which has migrated from southeast to northwest since 28 Ma (Beranek et al., 2017).
The Yakutat microplate is presently moving northwest relative to the North American plate and is obliquely colliding with the western continental margin (Fig. 1). The margin-parallel component is largely accommodated by the Fairweather fault (FF) system, which separates the Yakutat microplate from North America (Elliott et al., 2010). As the plate boundary bends to the west, dextral strike-slip displacement is transferred northward to connect with the active Totschunda fault and CDF via the Connector fault (Fig. 1). Historical seismicity reveals diffuse deformation with a mix of reverse and strike-slip faulting in the region between the FF and the St. Elias sequence (Fig. 2A).
The geomorphic expression and inferred slip rate of the DF diminish to the east (Fig. 1; Haeussler et al., 2017). Although the DF has accommodated ∼400 km of dextral slip since the Early Cretaceous (Lowey, 1998), i.e., an average rate of >3 mm/yr, the present-day motion of the EDF is <1 mm/yr (Marechal et al., 2018). The EDF is connected to the Queen Charlotte fault by the dextral Chatham Strait fault, which has been quiescent since ca. 13 ka (Brothers et al., 2018). Nevertheless, it remains unclear if the EDF is active, or if deformation has been fully transferred to the DRF or other, as-yet unrecognized, faults in this remote region.
Seismic waveforms for the 2017 earthquake doublet and associated aftershocks were obtained from >60 stations, three of which were located <50 km from the sequence (Fig. 2B, inset). Using initial locations reported by the U.S. Geological Survey (USGS, 2018), we applied a double-difference algorithm (Waldhauser and Ellsworth, 2000) to relocate 1498 events. By significantly improving the precision of relative event locations, this technique aids in elucidating the underlying fault architecture. Hierarchical clustering was then applied to these relocated events to group events on the basis of waveform similarity. Using this approach, events were classified into 136 clusters, of which the three largest clusters included more than 37% of all events. Moment tensors (MTs) were calculated by full-waveform inversion using a grid search and generalized least-squares method. Using the best-fitting double-couple mechanisms for the seven largest events, orientations of the principal stress axes and their associated uncertainties were determined in the epicentral region using linear stress inversion. This stress regime is applicable to the depth range of the mechanisms used in the inversion (4–15 km), but it is not sufficient to determine any depth dependence.
Stress changes associated with an earthquake can increase or retard seismic activity in the surrounding region or trigger other earthquakes by static stress triggering (Freed, 2005). Coulomb3 software (https://www.usgs.gov/software/coulomb-3; Toda et al., 2011) was used to evaluate the static stress triggering effect of the first mainshock to assess whether it triggered the second mainshock. This program calculates stress change on a receiver fault based on the Coulomb failure criterion, in which failure is promoted when the stress change is positive (Toda et al., 2011). Details of the methods used in this study are provided in the Supplemental Material1.
Compared with previous investigations (He et al., 2018; Feng et al., 2019), novel aspects of our analysis included stress inversion, depth dependence of Coulomb stress change, detailed aftershock analysis (clustering), consideration of the regional geological context, and interpretation of changes in plate motion and tectonic history that produced the inferred stress regime. Our MT inversions confirm that the first main-shock was characterized by oblique-reverse slip on a west-dipping fault (based on aftershocks) with a significant non-double-couple (non-DC) component (∼40%). The second event was dominantly DC with left-lateral strike-slip motion on a near-vertical east-west fault. Best-fitting focal mechanisms (Fig. 2B) show that aftershocks exhibited reverse faulting, similar to the initial mainshock. The non-DC components may be indicative of coseismic brittle damage and/or intersecting, overlapping, or nonplanar faults (Zhang et al., 2016), suggesting that new fractures were created within the source region (Julian et al., 1998).
The relocated events revealed distinct clouds of seismicity in a depth range of 4–15 km, illuminating inferred fault planes (Fig. 3). As indicated by interpreted fault segments (bold white lines in Fig. 2B), the orientation of the major axis of each aftershock cloud is aligned with one nodal plane of the corresponding focal mechanism. Waveform-based hierarchical clustering analysis revealed internal details of the sequence as it migrated from the first to the second fault via a diffuse connecting region. A subparallel fault strand west of the first mainshock was also activated, accompanied by episodic pulses of seismicity (Fig. 3; Fig. S3). The overall seismicity pattern is indicative of complex rupture within a relatively immature fault system, consistent with observed non-DC components.
The inverted stress field shows that the maximum principal stress (σ1) is horizontal with an orientation of N50°E, nearly perpendicular to the EDF (Fig. 2B), while the minimum principal stress (σ3) is vertical, indicative of a reverse-faulting stress regime. Although the activated fault planes are well oriented, the EDF is misoriented for slip in the present-day stress regime. If the EDF experienced a large amount of cumulative slip in the past, when it was better aligned with plate motion (see below), it could be characterized by a low effective friction coefficient, as in the case of the San Andreas fault (Townend and Zoback, 2004), potentially enabling slip even with high clamping stress. However, relocated aftershocks of the 2017 earthquake doublet, as well as historical event locations, show a clear concentration southwest of the EDF and a lack of events northeast of the fault (Fig. 2C). This seismicity pattern suggests that the EDF represents a structural backstop, wherein relative plate motion is accommodated by deformation in the region between the EDF and FF.
Figure 3 shows the distribution of aftershocks prior to the second event and the calculated Coulomb stress change from the first mainshock. At the hypocenter of the second event, Coulomb stress increased by ∼3 bars, similar to static stress changes associated with earthquake triggering elsewhere (King et al., 1994). This suggests that static stress changes from the first event were sufficient to trigger the second mainshock.
Source mechanisms for the 1 May 2017 earthquake doublet are incompatible with right-lateral strike-slip activation of the EDF, although historical earthquakes with similar source mechanisms (Fig. 2A) have occurred within an ∼100-km-wide corridor between the FF and the EDF (Doser and Rodriguez, 2011). Our inference that strike-slip activation of the EDF has ceased is supported by its lack of clear geomorphic expression. Taken together with the GPS records, stress vectors, and strain-rate data (Doser and Rodriguez, 2011; Marechal et al., 2015), we interpret that the Insular superterrane is presently undergoing distributed deformation and shortening; the Yukon-Tanana terrane thus appears to be a rigid buttress that serves as a regional backstop for deformation.
The Yakutat microplate is an oceanic plateau that has been transported northward along the western margin of North America since >30 Ma (e.g., Plafker et al., 1994). Today, the leading edge of the subducted Yakutat slab extends as far northward as the CDF in south-central Alaska (Fig. 4; Eberhart-Phillips et al., 2006). The direction and speed of Yakutat microplate motion have varied through time (Fig. 1; Müller et al., 2019). The most recent change at ca. 6 Ma to a more northern motion of the Pacific plate resulted in collisional uplift of the St. Elias Mountains and transpression along the FF (Eberhart-Phillips et al., 2006; Hyndman, 2015). Geomorphic studies along the Queen-Charlotte fault revised the Euler pole position, suggesting an even higher degree of transpression along the FF (Brothers et al., 2020). Additionally, crystalline basement of the Yakutat microplate thickens toward the southeast (Worthington et al., 2012), producing stronger interplate coupling. Oblique convergence of progressively thicker crust has accelerated reverse faulting, surface uplift, and rapid rock exhumation parallel to the plate boundary, forming the St. Elias and Fairweather ranges (Pavlis et al., 2012; Enkelmann et al., 2015). Deglaciation of the coastal mountains has resulted in more rapid surface uplift (Elliott et al., 2010), likely generating additional gravitationally driven stresses. Taken together, these processes have led to deceleration, and ultimately termination, of transcurrent motion on the EDF as its role transitioned to a tectonic backstop (Fig. 4). This model is consistent with suggested outboard transfer of plate-boundary deformation from the Coast shear zone to the Queen Charlotte fault (Fig. 1; ten Brink et al., 2018). Our interpretation is similar to the inferred cessation of activity on the contiguous Chatham Strait fault (Brothers et al., 2018), which occurred due to localization of displacement on the northern Queen Charlotte fault as a result of changes in relative plate motion (Brothers et al., 2020). This shift in the plate boundary to the continent-ocean boundary from former arc lithosphere differs from other transform plate boundaries along the Pacific–North American margin (e.g., California, Baja California, Dead Sea) and is contrary to basic rheological arguments. It has been postulated to arise from reduction of the lithospheric strength of the Yakutat terrane due to its proximity to young and hot Pacific lithosphere (ten Brink et al., 2018).
Analysis of a doublet earthquake sequence near the EDF provides new insights into regional tectonic evolution. Strain partitioning along the Fairweather transform fault, coupled with gravitational potential caused by high topography (>4000 m peaks), appears to be driving regional compressional deformation inboard of the plate margin in northwestern Canada and Alaska. Our analysis of the earthquake sequence showed that these driving forces have imposed increasing levels of compressional stress toward the northeast, orthogonal to the strike of the EDF. The present-day stress regime was likely established during the past ∼6 m.y., leading to a stress state that is geometrically unfavorable for strike-slip activation of the EDF. With the YT terrane acting as a structural backstop, the locus of deformation has shifted to the southeast into the mountainous region between the EDF and the FF. These inferred long-term changes in seismic activity may be impactful for regional earthquake hazard assessments for population centers near the EDF.
This work was supported by Natural Sciences and Engineering Research Council of Canada Discovery Grant RGPIN/03823–2017. We thank Diane Doser, Uri ten Brink, Nathan Miller, and an anonymous reviewer for their constructive suggestions that improved this manuscript.