Geophysical analysis of the 30 July 1972 Sitka, Alaska, earthquake sequence

The Queen Charlotte fault (QCF) forms the offshore plate boundary between the North American and Pacific plates from south of Haida Gwaii (Queen Charlotte Islands) to the Icy Bay region of southeastern Alaska (Fig. 1). With slip rates exceeding 50 mm/yr (Brothers et al., 2018; Brothers et al., 2020), it is one of the fastest moving transform faults in the world. This high slip rate led to seven earthquakes of M >7 along the entire offshore fault system since 1927, including the 2012 (M_w 7.7) Haida Gwaii, the 2013 (M_w 7.5) Craig, and the great (M_S 8.1) 1949 Queen Charlotte earthquakes (Fig. 1). In addition, the onshore extension of the system, termed the Fairweather fault, generated the M_w 7.8 (Doser, 2010) Fairweather earthquake in 1958 (Fig. 1). Unlike the San Andreas transform fault system, slip along the Queen Charlotte fault appears localized along a single fault trace with differences in the material properties of the oceanic and continental sides of the fault likely controlling the rupture behavior of large earthquakes (e.g., Aderhold and Abercrombie, 2015; Walton et al., 2015). Thus, study of large earthquakes along the QCF allows an opportunity to determine the material properties that are most influential in controlling rupture and the persistence of these features over time.

Our study focuses on seismicity associated with the segment of the Queen Charlotte fault that ruptured during the M_w 7.6 30 July 1972 earthquake near the town of Sitka, Alaska (Fig. 1). It was felt over an area of ∼130,000 km², causing moderate damage including cracked walls and fallen objects (Stover and Coffman, 1993). It ruptured ∼180 km (Schell and Ruff, 1989) along part of a seismic gap, recognized by Sykes (1971), located between the epicenters of 1949 and 1958 earthquakes (Fig. 1). Considering that the population of Sitka tripled between 1970 and 2010 (State of Alaska, 2018) and that urbanized areas also increased in population in other parts of southeastern Alaska, damage caused by anticipated future earthquakes along the Queen Charlotte fault would be considerably higher than in 1972.

Several previous studies (e.g., Schell and Ruff, 1989; Doser and Lomas, 2000) focused on analysis of the 1972 mainshock, but only a short abstract (Page and Gawthrop, 1973) was published on data collected by a temporary seismograph deployment by U.S. Geological Survey personnel (Fig. 1); the seismograph recorded aftershocks for about two months following the 1972 mainshock. By combining information from the 1972 aftershock survey as well as recent (1973–2015) and historical seismicity (1925–1972), (1) we investigate patterns of seismicity along the Queen Charlotte fault system following the 1972 mainshock, and (2) we assess the relation of this seismicity to structural variations and possible segmentation along the Queen Charlotte fault.

Because much of the Queen Charlotte fault system is located offshore, recent bathymetric studies (e.g., Balster-Gee et al., 2017; Brothers et al., 2018; Brothers et al., 2020) help to define changes in geometry along the fault system that may control its rupture segmentation and moment release. We used results of these studies along with a combination of earthquake seismicity, gravity, and magnetic data to assist in our study of the 1972 rupture zone.

The Queen Charlotte fault system is located in southeastern Alaska and extends offshore from southern Haida Gwaii to the Icy Point area. This mostly dextral strike-slip fault system forms part of the plate boundary between the Pacific and North American plates. Because of these characteristics, it is similar to the San Andreas fault in California. However, unlike the San Andreas fault, slip along the Queen Charlotte fault is localized to a single fault where differences in the material properties between the Pacific and North American plates appear to control rupture behavior (Aderhold and Abercrombie, 2015; Walton et al., 2015). In addition, the Queen Charlotte fault experienced six M>7 earthquakes in the past 100 years, allowing us to better determine the persistence of fault segmentation through the earthquake cycle.

GPS studies (e.g., Elliott et al., 2010) indicate that motion along the Queen Charlotte–Fairweather fault system north of 56°N is translational. South of 56°N, varying degrees of transpressional motion are observed in GPS data as well as earthquake focal mechanisms (e.g., Ristau et al., 2007). At ∼58°N, the Fairweather fault bends and almost parallels the coast. The Fairweather fault is located at the eastern margin of the Yakutat terrane, which is subducting beneath the North American plate at a similar velocity to the Pacific plate (Elliott et al., 2010). The southern boundary of the Yakutat terrane is the Transition fault (Fig. 1). Geophysical studies (Gulick et al., 2007; Gulick et al., 2013) suggest that in this region, plate motion is reorganizing, with the Transition fault now taking up some strike-slip motion from the Queen Charlotte fault system. However, focal mechanisms of earthquakes of the 1973 Cross Sound sequence (Fig. 1) indicate shortening is also occurring along low-angle structures located less than 20 km north of the Transition fault (Doser and Lomas, 2000).

Brothers et al. (2020) analyzed multi-beam bathymetry data to examine geomorphology along the Queen Charlotte fault. Their analysis suggested a slip rate of 50–57 mm/yr over the past 12–17 k.y. along a straight, narrow fault trace with few local step-overs. They used the trace of the fault to define a small circle path for plate motion and computed its Euler pole. Then they computed along-strike obliquity variations using this new pole. They concluded that obliquity variations appeared to control fault segmentation and development of asperities in M >7 earthquakes along the fault.

Only a few studies focus on the 1972 earthquake sequence. Schell and Ruff (1989) determined the source characteristics of the mainshock and its rupture process using accelerograms of the mainshock and teleseismic data. They calculated a moment magnitude of 7.6, depth extent of 0–10 km, average slip of 6 m, and an average stress drop of 100 bars. They estimated two zones of maximum slip release, one located 0–40 km northwest of the epicenter near Mount Edgecumbe (dashed rectangles, Figs. 2 and 3), which is a Holocene volcano (Brew, 1994), and the second at 60–90 km southeast of the epicenter. In addition, a strong pulse of moment release clearly observed on accelerograms of the earthquake (see Schell and Ruff, 1989) occurred near the northern terminus of rupture (dashed oval, Figs. 2 and 3). Doser and Lomas (2000) observed that a focal mechanism of a larger aftershock of the 1972 sequence (4 August 1972, M_w = 5.8; strike = 167 ± 13, dip = 78 ± 8, rake = 178 ± 8) suggested a small change in fault orientation may occur at the southern end of the 1972 mainshock rupture zone. In a later study, Doser and Rodriguez (2011) relocated 16 historical events (occurring between 1919 and 1971) within the rupture zone of the 1972 earthquake. They also noted a lack of post-1972 seismicity within the regions of highest slip that could be real or could be related to poor station coverage.

The 1927 (M_S 7.1) earthquake, located ∼100 km to the northwest of the Sitka earthquake, is the first large earthquake that was instrumentally recorded in this region. It had a strike-slip focal mechanism, and its bilateral rupture length was ∼35 km (Doser and Lomas, 2000). This rupture length corresponds to most of the gap between the extent of the 1958 and the 1972 ruptures (Doser and Lomas, 2000).

The next large earthquake along the system was the 1949 Queen Charlotte earthquake (M_S = 8.1). Pure strike-slip motion was likely involved, but details of its rupture process are not well known. Seismic radiation studies by Ben-Menahem (1978) and aftershock location studies by Bostwick (1984) estimate a rupture length of ∼495 km that would extend from northern Haida Gwaii to the southern end of the 1972 rupture zone (Fig. 1). On the other hand, a surface-wave directivity study by Bostwick (1984) suggests an ∼265-km-long rupture zone that would extend to only ∼54.5°N.

The 1958 Fairweather earthquake occurred ∼200 km to the north-northwest of the 1972 Sitka event. Doser (2010) estimated its rupture length at 260–370 km using body waveform modeling and the distribution of relocated aftershocks. This event ruptured unilaterally from the mainshock epicenter northwards. Rupture to the south may have been impeded by the 1927 rupture zone (Fig. 1).

The last large event in the northern Queen Charlotte fault region was the M_w 7.5 Craig earthquake on 5 January 2013. Analysis of GPS and regional waveform data for the mainshock by Yue et al. (2013) determined super-shear rupture occurred during the earthquake; they attributed this rupture to the differing mechanical properties of the two sides of the fault. Aderhold and Abercrombie (2015) indicated that rupture in the mainshock was faster along the northern portion of the fault and that super-shear rupture could have occurred. In addition, they suggested that the rheology of the North American plate appears to be more mafic than average continental crust, leading to mainshock rupture behavior more characteristic of an oceanic transform fault. A tomographic study of the rupture zone by Walton et al. (2019) shows that the velocities of the oceanic (Pacific) crust and mantle of the fault zone are 3%–11% slower than those of the continental (North American) side. They suggested that factors other than large differences in the velocities of materials along the fault zone, such as fault zone damage or fault smoothness, could play important roles in super-shear rupture.

Holtkamp and Ruppert (2015) relocated the aftershocks of the Craig sequence, and their results revealed that a complex fault network was activated in response to the mainshock. The distribution of the aftershocks also shows that northern extent of rupture along the Queen Charlotte fault was near the southern limit of the 1972 rupture (Fig. 1).

Walton et al. (2015) suggested that the northern extent of rupture in 2013 was controlled by several factors including the intersection of the Aja fracture zone with the Queen Charlotte fault, causing a 3 m.y. offset in the age of the Pacific plate along the fault and the end of flexure of the Pacific plate as observed to the south. Bathymetry data (Brothers et al., 2018; Brothers et al., 2020) indicate that the Queen Charlotte fault takes a series of three to five steps and bends that define a series of pull-apart basins at the northern end of the 2013 rupture (see Fig. 2), again suggesting structural complexity.

To the south of our study area, transpression across the Queen Charlotte fault increases. In 2012, this change in plate motion led to a M_w 7.8 thrust earthquake (e.g., Lay et al., 2013) that occurred two months before the Craig mainshock near the west coast of the Haida Gwaii (Fig. 1). It generated a tsunami that affected at least 170 km of the coast line with maximum run-ups of over 7 m (Leonard and Bednarski, 2014).

About a week after the 1972 Sitka mainshock, the U.S. Geological Survey deployed a temporary seismic network of eight portable, smoked-paper seismographs on the landward side of the rupture zone (R. Page, 2015, written commun.). A week later, three additional semi-permanent seismographs were added to the network with signals telemetered to Palmer, Alaska, for recording on microfilm (R. Page, 2015, written commun.). Phase data for the aftershocks were handpicked from paper seismograms and microfiche, entered into digital form, and used to determine locations. Picking accuracy was estimated at 0.2 s for smoked-paper seismograms and 0.05 s for telemetered data recorded on microfiche (R. Page, 2015, written commun.). Over 250 events were located during the nearly two months of network operation. A summary of this preliminary analysis is found in an abstract by Page and Gawthrop (1973). The original seismograms and digital phase data were then misplaced, and further analysis remained unpublished.

In 2015, we obtained paper copies of the original computer printouts of the phase data, including P and S arrival times, phase amplitudes, and durations, as well as information on original locations (W. Ellsworth, 2015, written commun.). We converted the arrival times to digital form by scanning paper copies, saving them as images, and transforming the images to text files using optical character recognition software. We manually corrected the text files for errors by comparing them to the original computer printout and then reformatted the files for use in relocation algorithms with the help of a MATLAB script.

We then relocated the earthquakes using a double-difference algorithm (HYPODD, Waldhauser and Ellsworth, 2000) with the parameters and relocations given in Supplemental Material 2¹ and Supplemental Material 1, respectively. Because we could not access the original seismograms, we could only use P and S arrival times in the algorithm. To obtain more reliable relocations, we used phases with an initial residual smaller than 0.5 s. We used the one-dimensional (1D) velocity model of Matumoto and Page (1969) (Table 1), the standard model used by the Alaska Earthquake Center for southeastern Alaska, in our relocation process. Due to the reduced number of stations, the geometry of the station array with respect to the aftershocks and the fact that we only had the tabulated first motion information, we were not able to determine focal mechanisms or reliable focal depths.

Figure 3 shows the events before and after the relocation process. From the data set of over 285 earthquakes, we only successfully relocated 87 events (see Fig. 3 and Supplemental Material Item 1 [^{footnote 1}]). This small number of relocations is mainly due to the fact that many phase residuals were higher than 0.5 s and consequently were not used.

Before interpreting our relocations, we performed a number of tests to estimate our location errors (Fig. 4). We did this by perturbing the top four layers of the 1D velocity model by ± 0.1 km/s and then comparing the relocations derived from these eight perturbed models to the relocations from the original model. For latitude, 98% of the epicenters fall within 5 km (∼0.05°); and for longitude, 94% fall within 5 km (0.08°).

After the first five days of network operation (Fig. 5), few aftershocks were recorded south of 57°N. This effect does not appear to be related to the magnitude of the events or the number of stations used in the locations. Written notes from R. Page (2015) indicate the weather grew progressively worse over the time of the deployment, and access to seismographs in the southern portion of the network grew difficult. Figure 5 illustrates how the number of events decreased with time, as is expected, over the first ten days of network operation. After ten days, the number of relocated events remained almost constant.

Before relocation (Fig. 3A), aftershocks were mostly distributed in two regions, with the remainder scattered across the area. The first group of events was located to the north of the mainshock epicenter and east of the Queen Charlotte fault near its intersection with the Transition fault. The second group was distributed along the Queen Charlotte fault south of the mainshock epicenter. After the relocation process, the general patterns of epicenters remain (Fig. 3B). The group of events north of the mainshock epicenter still appears to be primarily located east of the Queen Charlotte fault and to form several linear bands (arrows, Fig. 6) with a strike similar to the Transition fault.

Relocated aftershocks appear to primarily occur outside of the regions of maximum moment release (Schell and Ruff, 1989) observed during the 1972 mainshock (dashed rectangles, Figs. 2 and 3). Note that the strongest pulse of energy release in 1972 (Schell and Ruff, 1989) (dashed oval in Figs. 2 and 3) occurred at the southern end of the northern cluster of aftershocks. The northern cluster extends to the intersection of the Queen Charlotte fault with the Transition fault, suggesting that this intersection may serve as a barrier to rupture. The northwest-southeast lineations of aftershocks within the north cluster (arrows, Fig. 6) also suggest activation of faults en echelon to the Transition fault, consistent with the idea of Gulick et al. (2007) and Gulick et al. (2013) that the Transition fault may now be taking up a portion of strike-slip motion as plate motion reorganization occurs within this region.

Holtkamp and Ruppert (2015) also observed off-fault aftershock clusters in the 2013 Craig aftershocks, including thrust events, which are not common along this part of the Queen Charlotte fault system, and strike-slip events with nodal planes rotated by ∼45° from the Queen Charlotte fault. Unfortunately, we do not have enough first-motion data or any waveform information that would be required to compute reliable focal mechanisms. This prevents us from determining whether some of the north off-fault aftershocks could have involved reverse or strike-slip faulting similar to that observed in 2013. Body waveform modeling of a M_w 5.8 aftershock by Doser and Lomas (2000) located at the southernmost end of the 1972 rupture zone also indicates a slight rotation (∼7°) of the fault plane relative to the main Queen Charlotte fault trace.

The total length of the 1972 aftershock zone is ∼160 km; this is slightly smaller than the rupture length of 180 km estimated by Schell and Ruff (1989). It is possible that aftershocks continued to the south but were not detected by enough stations to be adequately located. The mainshock epicenter also is located near the midpoint of the aftershock zone, consistent with bilateral rupture.

In order to determine if the regions where maximum slip occurred in 1972 were seismically quiescent before and after the mainshock, we examined background seismicity from 1919 to 1972 relocated by Doser and Rodriguez (2010) using regional and teleseismic phase data (blue circles, Fig. 6) and from 1973 to 2015 using network phase data from the Alaska Earthquake Center (2016) (green circles, Fig. 6). We also examined the aftershock distribution of the 2013 Craig earthquake (Holtkamp and Ruppert, 2015) (yellow circles, Fig. 6). Note that instrumental coverage of the region was very poor prior to installation of the local network in 1973, and coverage continues to be hampered by the lack of offshore seismographs.

Seismicity following the 1972 Sitka sequence (Fig. 6, green circles) is similar to the pattern of aftershocks in 1972 (squares). Background seismicity continues to occur in the vicinity of two aftershock clusters, one north of 57°N latitude and the other near the mainshock epicenter, with an ∼25 km gap between the mainshock and the northern cluster of seismicity. Post-1972 seismicity is also found at the southernmost end of the 1972 rupture. Seismicity within most of the 1927 rupture zone is limited to M ≤2 events, while there is abundant seismicity within the 1958 Fairweather rupture zone.

With the exception of the southernmost part of the 1972 rupture zone, only events with M ≤5 occurred within <25 km of the Queen Charlotte fault between 1919 and 1972 (blue circles, Fig. 6). The pre-1972 events of M >5 at the southernmost end of the 1972 rupture zone are located west of the Queen Charlotte fault in the region where Walton et al. (2015) suggest the Aja fracture zone may serve as a barrier to rupture. Aftershocks of the 2013 Craig earthquake (yellow circles) suggest little overlap between the ruptures of the 1972 and 2013 earthquakes. In summary, these observations suggest that the regions of maximum slip in the 1972 mainshock were seismically quiescent both before and after the mainshock.

We high-pass filtered free-air gravity residual data (Sandwell and Smith, 2009) with a pass band starting at wavelengths of 120 km to compare the free-air anomalies to seismicity and fault structure. The resulting map (Fig. 7) shows that the Queen Charlotte fault is associated with high free-air anomaly values (red and pink), but these highs are discontinuous.

The 2013 Craig aftershock sequence (X symbols, Fig. 7) occurs along a free-air anomaly high that is 40–60 km wide and extends from 55.1°N to 56.1°N. The anomaly extends along both sides of the fault zone but is two to three times as wide on the eastern side of the fault. The termination of the free-air anomaly high at 56.1°N occurs at the point where the Aja fracture zone intersects the Queen Charlotte fault, a feature that Walton et al. (2015) indicate was a major barrier to northward rupture in 2013. The free-air anomaly high associated with the 2013 aftershocks is paired with an anomaly low (∼60 mGal) to the west of the Queen Charlotte fault. Using a combination of seismic, magnetic, sonar, and gravity data (processed differently than in our study), Walton et al. (2015) noted that the flexure of the Pacific plate south of 56°N produces the thicker sediment deposits that correlate with the offshore low.

The spatial distribution of the aftershocks of the 2013 Craig earthquake correlates well with the width of the free-air anomaly high, although a small cluster of aftershocks at 54°N is located to the west of the Queen Charlotte fault on the negative free-air anomaly. Holtkamp and Ruppert (2015) indicate thrust mechanisms were associated with this small cluster. Post-1972 background seismicity (plus symbols) appears to surround the free-air anomaly low.

The southern 30 km of the 1972 rupture extends along a narrow (<10-km-wide) free-air anomaly high along the Queen Charlotte fault. This area was one of the two regions of the fault that experienced maximum moment release during the mainshock (see Figs. 2 and 3).

At 56.5°N, the free-air anomaly high changes strike, broadens (∼50 km), and begins to span both sides of the fault (Fig. 7). This portion of the fault contains the 1972 epicenter and a cluster of aftershocks (asterisk symbols). Post-1972 background seismicity also occurs within this zone (Fig. 7).

There is a slight westward bend in the free-air anomaly high at ∼57°N, with a corresponding ∼5° change in its strike. This bend occurs near the northern edge of a zone maximum moment release and the region associated with the strong energy pulse in the 1972 mainshock (Figs. 2 and 3). In addition, this bend lies just east of Mount Edgecumbe (Fig. 7, black triangle). Brew (1994) suggested a localized perturbation in the stress field must be occurring near Mount Edgecumbe in order to produce the observed Holocene volcanism. Multi-beam bathymetry data (Brothers et al., 2020) show a small (∼10-km-long) pull-apart basin along the Queen Charlotte fault at this latitude, indicating divergence is occurring. There is little change in the strike or width of the free-air gravity anomaly high from the northern end of the 1972 rupture zone (Fig. 7) through the northern end of the 1927 rupture zone to the termination of the high near Cross Sound. Two free-air anomaly lows observed west of the 1972 rupture zone (Fig. 7) indicate thicker sediment is being deposited offshore in regions that appear to parallel the strike of the free-air anomaly highs located to the east. Background seismicity occurs between the southern low (−30 mGal) that strikes north-northwest and the smaller magnitude (−20 mGal) northern low that strikes north-south (Fig. 7). The large free-air anomaly high associated with the Yakutat terrane appears to intersect the Queen Charlotte fault near the northern end of the 1927 rupture zone. This high is likely due to the shallow depth of water over the Yakutat terrane.

Free-air anomaly lows indicate sediment accumulation on the Pacific plate may be related to fault segmentation. Walton et al. (2015) suggested the free-air low associated with 2013 aftershock region records Pacific plate flexure caused by oblique convergence along the western side of fault before it moved northward into the strike-slip regime of the northern fault. It is possible the lows associated with the 1972 and 1927 rupture zones record past periods of plate flexure that would cause variations in the sediment thickness on the Pacific plate along the fault zone leading to variations in coupling across the fault.

We observe a number of correlations between seismicity and fault morphology as revealed by the multi-beam bathymetry (Fig. 2). There are similar water depths on either side of the fault trace extending ∼20 km from the southern end of the 1972 rupture zone, roughly corresponding to one of the zones of maximum moment release (dashed rectangle) in 1972. The 1972 mainshock epicentral region and a portion of the second zone of maximum moment release also are located in a region where similar water depths occur on either side of the fault. Although there is little change in the free-air gravity anomaly at the northern end of the 1972 rupture zone (Fig. 7), the bathymetry (Fig. 2) indicates another region of similar water depth occurs on both sides of the fault at this point.

Brothers et al. (2020) suggest that fault obliquity controls fault geometry, rupture segmentation, and asperity development along the Queen Charlotte fault. They compute divergence (maximum of 5 mm/yr) along the southernmost portion of the 1972 rupture zone; this divergence decreases to near zero ∼30 km south of the 1972 mainshock epicenter. This region of divergence corresponds well with the southern zone of maximum moment release (Fig. 2) in 1972. Immediately north of the 1972 mainshock, they compute an ∼17-km-long zone of convergence (maximum of 3 mm/yr). This roughly corresponds to the southern half of the northern zone of maximum moment release in 1972 (Fig. 2). Few aftershocks or background seismicity have occurred in this region (Fig. 6). The northern 60 km of the 1972 rupture lie within a zone of divergence (maximum of 3 mm/yr) (Brothers et al., 2020); this zone spans the northern region of aftershocks (Fig. 6). Obliquity decreases to near zero in the 1927 rupture zone (Brothers et al., 2020). An increase in convergence (maximum of 3 mm/yr) then occurs along the northernmost portion of the fault between the 1927 and 1958 rupture zones.

The association of free-air anomaly highs with fault segmentation also indicates that density variations, related to changes in plate thickness or rock type, control rupture behavior along the fault zone. A major change in North American plate geology occurs at 56°N where the Alexander terrane (to south) abuts the Chugach terrane (Plafker et al., 1994). Based on seismic velocity studies, the Alexander terrane appears to be more mafic in composition than average continental crust and has a faster velocity than the young Pacific plate crust in the Craig aftershock region (e.g., Walton et al., 2019). This faster, denser terrane could explain the wideness of the free-air anomaly high associated with the east side of the Queen Charlotte fault in the 2013 aftershock zone.

The Chugach terrane, an accretionary complex in southeastern Alaska, is characterized by a Cretaceous flysch and basalt assemblage locally intruded by lower Miocene to upper Paleocene granites (Plafker et al., 1994). With the exception of the granitic intrusions, this material may also be expected to have higher than average crustal densities giving rise to the free-air anomaly highs. The variation in rock strength between the granitic intrusions and surrounding terrane is also likely to influence fault rupture (i.e., persistent fault asperities are not as likely to form in the weaker granitic material).

Regions where free-air anomaly highs occur on both sides of the fault (Fig. 7) appear to correlate with areas of shallower bathymetry on both sides of the fault (Fig. 2) and often with higher levels of aftershock and background seismicity. This suggests the rock types along both sides of the fault in these locations may be similar, but how this influences observed seismicity patterns requires further investigation.

Magnetic data (Fig. 8) also indicate variations in structure within the Pacific plate that likely influence fault segmentation. The reduced to pole magnetic data were provided by R. Saltus (2017, written commun.). The most distinct change in magnetic patterns is related to the Aja fracture zone, which juxtaposes two portions of the Pacific plate that differ in age by 3 m.y. Walton et al. (2015) previously noted this feature as controlling the end of the 1972 and 2013 rupture zones. Although Colpron and Nelson (2011) do not indicate any major variation in the age of the Pacific plate that forms the west side of the Queen Charlotte fault between the Aja fracture zone and the Transition fault, several variations in the intensity of magnetic patterns within the Pacific plate appear to align (Fig. 8, dark-blue dashed lines). One change in intensity patterns aligns with the epicenter of the 1972 mainshock, and the other aligns with the northern end of the 1972 rupture. In addition, these changes in magnetic field also appear to correlate with edges of the free-air anomaly highs (Fig. 7). These variations could be related to age or compositional changes within the Pacific plate that affect both its density and magnetic properties.

We converted paper copies of P and S phase arrival times for aftershocks of the 30 July 1972 Mw 7.6 Sitka earthquake in Alaska to digital format and successfully relocated 87 aftershocks using a double-difference relocation algorithm. Relocated aftershocks concentrate in two clusters, one at the northern end of the rupture zone where significant moment release occurred during the mainshock and one near the mainshock epicenter. Many aftershocks of the northern cluster appear to be occurring east of the Queen Charlotte fault on features that have a strike similar to that of the Transition fault. The aftershock distribution suggests a rupture length of ∼160 km compared to a rupture length of 180 km from waveform modeling studies by Schell and Ruff (1989). Background seismicity prior to and after 1972 indicates the regions where aftershocks occurred are seismically active at a low level, while very few earthquakes are located in regions experiencing maximum moment release in 1972.

Changes in the width and strike of free-air anomaly highs within the North American plate appear to correlate with the 1972 epicenter region, the ends of maximum moment release in 1972, and the southern end of the 1972 rupture. Regions where anomaly highs are found on both sides of the fault tend to be associated with higher levels of aftershock and background seismic activity. Three free-air anomaly lows west of the 1927 and 1972 rupture zones indicate regions of thicker sediment deposition. These may represent shallow basins created when the Pacific plate was in flexure due to shortening along the plate margin before it was translated northwestward along the Queen Charlotte fault to the region where plate motion becomes translational. Walton et al. (2015) have suggested this mechanism for the creation of the basin associated with an anomaly low located west of the 2013 Craig sequence. Anomaly highs likely reflect differences in rock types within the North American plate, which is composed of a complex mixture of flysch and basalt intruded by younger granites. Regions where anomaly highs are found along both sides of the fault also appear to be associated with fewer changes in bathymetry across the fault and higher levels of aftershocks and background seismicity. Magnetic data indicate several changes in Pacific plate structure west of the Queen Charlotte fault. The most distinctive feature revealed by the data, the Aja fracture zone, likely influences the boundary between the 1972 and 2013 ruptures (e.g., Walton et al., 2015). More subtle changes in magnetic intensity suggest changes in Pacific plate age or composition that also appear to affect rupture behavior along the Queen Charlotte fault.

Fault obliquity variations computed by Brothers et al. (2020) along the northern Queen Charlotte fault also show a striking spatial correspondence to the patterns of historical seismicity observed in our study. This reinforces their idea that obliquity plays a significant role in fault geometry and rupture behavior during M >7 earthquakes.

We thank W. Ellsworth for sending us the original computer printouts of the Sitka aftershock phase data, R. Saltus for the magnetic data, and P. Haeussler for conversations regarding the regional tectonics. We also thank the associate editor, an anonymous reviewer, and L. Worthington for their constructive reviews, which improved the manuscript.