The Shumagin seismic gap structure and associated tsunami hazards, Alaska convergent margin

Along the Alaska convergent margin, three of four earthquake rupture zones represent areas where aftershocks occurred following great instrumentally recorded earthquakes. From NE to SW, these are the Kodiak segment that ruptured in the 1964 great earthquake, the Semidi segment that broke in a 1938 earthquake, and the Unimak segment that ruptured in 1946. In contrast, the Shumagin segment has no historic great earthquake and is constrained by its neighboring segments rather than its own aftershocks. Since Sykes (1971) first drew attention to the Shumagin segment, it has been a proposed seismic gap. Earthquakes in the adjacent segments have produced tsunamis. To the southwest, during the 1946 M8.6 earthquake, the Unimak segment generated the largest Alaskan tsunami ever recorded in the era of instrumental seismology. This earthquake featured low rupture velocities, a long duration, and a lack of high frequency radiation. It was the first recognized tsunami earthquake in that it produced an outsized tsunami wave compared to its moment magnitude (Kanamori, 1972). Northeast of the Shumagin segment is the Semidi segment, which ruptured in the 1938 M8.2 earthquake and produced only a small tsunami because of the rupture’s great depth. Geodetic studies indicate the Semidi segment is currently locked (Fournier and Freymueller, 2007). Meanwhile, geodetic studies in the intervening Shumagin segment have failed to show a stress buildup, and to explain these observations, Fournier and Freymueller (2007) proposed that this segment is creeping rather than releasing stress in large earthquakes. The Shumagin segment’s behavior is inconsistent with the rest of the Alaska-Aleutian subduction zone, but this phenomenon has not been investigated with comprehensive geologic analyses since the reports of Bruns et al. (1987) and Lewis et al. (1988).

The Shumagin earthquake rupture segment extends northeastward from Sanak Island to a short distance past the Shumagin Islands (Fig. 1). We reprocessed legacy seismic data acquired by the U.S. Geological Survey (USGS) vessel Lee in 1981 and 1982 (L8-81-WG and L12-82-WG), which were reported in a comprehensive initial marine geological study (Bruns et al., 1987). Additional data from a 1994 transect by RV Ewing (EW9409) were reprocessed as well (Miller et al., 2014). We also processed data along three lines acquired during the ALEUT project cruise of RV Langseth in 2011 (MGL1110); these data were reported recently with a focus on the plate interface (Li et al., 2015; Bécel et al., 2017; Li et al., 2018). Here, we focus on the tectonics and extent of an out-of-sequence or splay fault along the lower slope and an extensional fault zone along the upper slope. The splay fault is a continuation of a structure in the adjacent Unimak segment (Miller et al., 2014; von Huene et al., 2014) and similar to a splay fault in the Semidi segment (von Huene et al., 2016). A landward-dipping extensional (LDE) fault zone (Bécel et al., 2017) is an element of the upper-slope extensional zone. The fault’s imaged extent to the plate interface is extraordinary. It occurs along an upper-slope ridge that appears to continue SW as Unimak Ridge.

The Shumagin segment’s SW boundary with the Unimak segment corresponds with a major upper-plate tectonic unit. It contains two tectonic features atypical of the Alaska margin, namely the diagonal NW-trending Sanak Basin and the Unimak Seamount (Fig. 1) (Bruns et al., 1987). It is discussed in more detail below. The NE boundary with the Semidi segment is apparent from bathymetry, but with little seismic data, the underlying boundary structures are unknown.

In our study, we apply current seismic processing software to legacy data, with a focus on two structures associated with potential tsunami generation in the Shumagin segment (Supplemental Figs. S1 and S2¹). Integration of bathymetric data with seismic-reflection data shows the extent of the splay fault zone along the lower slope (Fig. 1). With more seismic images and enhanced resolution, we show the extent and character of the splay fault identified in ALEUT line 6 (Bécel et al., 2017). Higher up on the margin, the zone of extensional deformation across the foot of the upper slope reveals structural features that are commonly obscured at the upper- to mid-slope juncture in other Alaskan margins seismic images.

Tsunami hazards due to faults that branch from the plate interface and follow a path behind the frontal prism were first appreciated in studies of the Nankai convergent margin (cf. Moore et al., 2007). In Nankai, the frontal prism is separated from an older rock mass by a seaward-verging thrust fault zone termed a splay fault (Moore et al., 2007). The fault zone provides a shortened path for slip propagation from the seismogenic zone to the lower-slope seafloor compared to the plate interface path toward the trench axis. Because the splay’s dip is steeper and its path shorter than the plate interface path, a given amount of slip results in greater seafloor uplift (Wendt et al., 2009). To retain the original sense of the Nankai margin fault zone and separate it from thrust faults termed splays on the Alaskan shelf (cf. Liberty et al., 2013), we termed these splays “backstop splay fault zones” (BSFZs). They occur in very deep water (>3 km), and thus they can generate much larger tsunamis than thrust faults on the shelf. Since BSFZ is a new term, we explain before proceeding to an explanation of data.

Backstop is a term often applied to a lower-slope area where stratal reflections lose coherence and image reflectivity is scattered. This term is commonly assumed to indicate consolidated complexly deformed accretionary strata that have also lost impedance contrasts. Here we show that along the Alaska margin, the backstop is a transition from the seaward end of the margin framework to the landward end of the frontal prism. Framework rock corresponds with the acoustic basement that is unconformably overlain by a stratified cover. In all Alaskan earthquake segments, this unconformity returns a characteristic high-amplitude reflection. That reflection extends from the shelf downslope. In transects with less deformed middle slopes, it is imaged to the backstop area. On the Shumagin shelf, the unconformity can be traced from the insular subaerial outcrops across the shelf and down into the mid-slope area (Bruns et al., 1987; their figure 8, horizon A). Here, wide-angle seismic data generally indicate an acoustic basement velocity greater than 4.6 km/s (Bruns et al., 1987), and 19 Sonobuoy records give an average velocity contrast across the unconformity of 2 km/s. This explains the robust reflectivity that makes it a geologic marker. The diffuse reflectivity of basement is different from the deformed strata of the frontal prism; so the transition from one to the other can be recognized from reflective character when seismic resolution allows.

The frontal prism is characterized by deformed stratal reflections. Seismic images of trench axis strata merge into an imbricated frontal prism that extends up slope to a lower or middle-slope seafloor juncture. That juncture is marked by a landward-dipping zone of short and diffuse landward-dipping reflections and is commonly a complexly structured transition between the frontal prism and margin framework basement. On the landward side is a disrupted down-slope end of the Oligo-Miocene unconformity. The unconformity reflection can be traced to the backstop area in several Alaskan margin seismic images (Fig. 2). It is assumed that the unconformity on acoustic basement is also the top of continental basement rock. This is not peculiar to the Alaskan margin since shallow-water sediment was recovered from the lower slope of the Peru and Central American margins.

The acoustic basement unconformity is also an inferred time horizon. Beneath the shelf, framework basement is correlated with the Cretaceous Shumagin Formation, named for offshore insular outcrops along the Alaska Peninsula (Burk, 1965; Moore, 1972). The Shumagin Formation is part of an extensive regional belt of similar Cretaceous metamorphic rock studied most extensively along the Kodiak Islands shelf. There Cretaceous basement is overlain unconformably by Eocene to Recent strata, dated in dredged and drilled sections (Winston, 1983; Turner et al., 1987; von Huene et al., 1987). Oligocene rock that is 1500 m thick on shore is missing in three of four industry exploratory drill holes, indicating a period of profound erosion. The unconformity is probably a time-transgressive temporal marker reflection. Following Bruns et al. (1987), we refer to it as the Oligocene/Miocene unconformity, recognizing a period of time to the youngest sampled age of sediment deposited upon it (Turner et al., 1987).

In regional bathymetric maps of the Shumagin segment, a BSFZ ridge like that of adjacent segments extends from the Unimak segment to the Shumagin Islands that then diminishes to become almost imperceptible near the Semidi segment (Figs. 1 and S3 and S4 [^{footnote 1}]). GLORIA sidescan images (https://pubs.usgs.gov/of/2010/1332/htmldocs/gak/gak_indexmap.html) are consistent with bathymetry, resolving irregular peaks and saddles along the BSFZ ridge, with local sharp outcrops of erosion-resistant beds (Paskevich et al., 2010). Where seismic lines cross a peak, they show a sudden steep slope, while the slope remains almost unchanged across saddles. The BSFZ ridge is 6–7 km wide and 10–14 km from the deformation front. Slope failure scars and mass-wasting deposits are numerous.

A decreasing height and roughness of seafloor features characterize the morphology between line 213/219 and ALEUT 04 (Fig. 1). Northeast of ALEUT 04, the smoother morphology is cut by deep canyons in a transition to the characteristic morphology of the Semidi segment.

Legacy seismic images in Bruns et al. (1987) were processed with 1980s software. Our reprocessing with current software included trace editing, multiple suppression, deconvolution, velocity analysis, scaling, post-stack migration, and for some lines, pre-stack depth migration (Figs. S1 and S2 [^{footnote 1}]). Processing is further described in the Supplemental Information (^{footnote 1}).

Five seismic images across the BSFZ are concentrated in the SW half of the Shumagin segment (Fig. 1). ALEUT 06 crosses a morphological notch along the boundary between the Unimak and Shumagin segments. Bécel et al. (2017) took note of a BSFZ in ALEUT 06 and mentioned its tsunami potential. At the NE boundary, line ALEUT 04 is in a transition to the Semidi segment (Fig. 1). The SW group of five images illustrates the amount of structural variability along strike (Figs. 2–6 and S3 and S4 [^{footnote 1}]).

We begin with legacy line 217 (Fig. 2), which shows a basic structure with greater clarity than in other images. At the frontal prism landward limit, the BSFZ cuts the flank of strata arched over subducted lower-plate relief. These truncated frontal prism strata and those on the opposite side along the margin framework are interpreted as the limits of the BSFZ fault damage zone. The 8.5–9.5 km depth of the connection between the BSFZ and the plate interface is relatively shallow. This facilitates seismic resolution of this critical junction, owing to shortened acoustic travel paths and the absence of a seafloor multiple. The BSFZ reflectivity (Fig. 2) continues through the junction with the plate interface subducted sediment cutting off the megathrust sector beneath the frontal prism. It can become a principal source of fragmented fault material input to the plate interface sediment layer rather than trench sediment at the deformation front (Fig. 2). In the overlying hanging-wall slope basin, sediment strata are tilted in sets of increasing dip with depth, indicating periodic tilting over time from thrusting on the BSFZ fault (Fig. 2). The break in slope is characteristic of active BSFZs.

About 15 km southwest of line 217, line 1235 images a different seismic character (Fig. 3). The BSFZ boundaries are not clearly marked, and the BSFZ appears to be arched over subducted relief at the plate interface. Clarity of the structure may be affected by the seismic line position across the sloping side of seafloor relief, which hampers good two-dimensional imaging. The seafloor multiple obscures structure below a depth of 12 km. Hanging-wall basin strata are irregular (Fig. 3), and below the basin’s unconformable floor, the continental framework acoustic basement appears somewhat stratified similar to local simply structured basement outcrops on the Shumagin Islands (Moore, 1972). Hanging-wall uplift is ∼800 m, and the fault zone reflectivity is 1.2–1.8 km thick. This unusual thickness may be apparent and result from out-of-plane reflections because structure strikes at an angle to the line of section.

ALEUT line 05 (Fig. 4) crosses the BSFZ amidst small canyons that probably generate side reflections. Strata in the upper 1.5 km of the fault zone have poor coherence (Fig. 4). Below 7 km depth, however, the BSFZ contains strong reflections. The BSFZ appears to locally truncate frontal prism strata. Plate interface reflections with poor coherence beneath the frontal prism at 8–9 km depth may indicate lower-plate roughness. The downdip juncture with the plate interface is partly obscured by the seafloor multiple, but weak plate interface reflections and the 1.5-km-thick fault zone of the BSFZ appear to merge. Li et al. (2018) report subduction channel thicknesses 15 km from the trench that are ∼200 m less than in our image; those measurements could be more precise than ours because of the more thorough velocity analyses they performed on the channel reflectivity.

Legacy line 219 (Fig. 5) crosses the BSFZ ridge where it is still morphologically obvious but its height has diminished. A modest seafloor notch marks the BSFZ. Its junction with the plate interface is complicated by out-of-plane reflections above the plate interface. Our interpretation of the BSFZ is equivocal. Seaward of the BSFZ and plate interface junction, interface roughness appears to increase with depth in the subduction zone as noted in other images.

ALEUT line 04 (Fig. 6) is located along the SW end of aftershocks from the 1938 Semidi earthquake; these aftershocks mark the end of that rupture segment (Fig. 1). Here, the BSFZ ridge is hardly noticeable in available bathymetric compilations, and the seafloor shows only a low slope break at the BSFZ. The multibeam swath acquired during shooting of the ALEUT seismic transect shows a rough low ridge that ends not far away against a deep canyon. The lower-slope bathymetry also changes immediately to the northeast, where the widening area of accreted prism ridges in the Semidi segment begins.

The upper-slope extensional fault zone was initially formed at an earlier time represented by strata beneath the slope sediment cover. One age limit can be drawn from the relation of structural trends. The Unimak-Shumagin earthquake segment boundary is formed by the diagonal-trending Sanak Island platform and adjacent basin (Fig. 1). Sanak Island (Moore, 1972) and Sanak Basin (Bruns et al., 1987) strike 40° to the regional trend to the Alaska margin and parallel the Beringian margin (Fig. 1). The long-standing explanation is inheritance from a margin configuration preceding development of the Aleutian arc when the Alaska and Bering margins were joined (cf. Moore, 1972; Bruns et al., 1987; Scholl, 2007). The feature’s sizes are substantial in that Sanak Basin is ∼170 km long, as deep as 7 km, and the associated insular basement platform is ∼40 km wide. These features limited aftershocks of the 1946 Unimak tsunami earthquake (Lopez and Okal, 2006). The Aleutian arc is a later feature, the trends for which cut across the remnant diagonal Beringian trends. The oldest rocks dredged from the arc have an Early Eocene age (Jicha et al., 2006; Scholl, 2007). The Alaska/Aleutian margin parallel bathymetry includes the upper-slope extensional zone marked by a ridge (Fig. 1). The western part contains the 130-km-long Unimak Ridge on which the ∼3-km-high Unimak Seamount erupted. East of the seamount, a less prominent ridge continues the Aleutian/Alaska trend. We refer to this as Shumagin Ridge, and it continues from the indentation northeastward to the Semidi segment boundary (Fig. 1). The 19.8 ± 1.0 Ma age of a sample from Unimak Seamount (Bruns et al., 1987) brackets an initial age of the Unimak and Shumagin ridges as between Early Eocene and Early Miocene.

Unimak Ridge, shown in the legacy line 205 seismic image, brings basement to a near-surface exposure (Fig. 7). The ridge is formed by an LDE fault that elevated and tilted the basement block. Eocene limestones and shale dredged from the ridge indicate its basement provenance, consistent with its elevated seismic velocity (Bruns et al., 1987). Minimum vertical displacement on the LDE fault is 2.5–3 km, based on dislocation of the Oligocene/Miocene basement unconformity. The LDE fault zone of disrupted reflections is ∼2 km wide at its deepest imaged parts. Unimak Seamount age is consistent with the Oligocene–Miocene inferred age of the unconformity flooring the associated basin (Winston, 1983; Bruns et al., 1987; Turner et al., 1987). A flank of Unimak Seamount might be displaced along the LDE fault, as indicated in a bathymetric profile acquired during dredging (Bruns et al., 1987). However, displacement on the LDE fault system is inferential because the dredging was performed prior to GPS navigation; so deviation of the ship’s course was not recorded.

As mentioned previously, the shelf edge is indented northeast of Unimak Seamount where the Sanak and LDE fault systems intersect. Across that intersection and buried beneath the arched stratal reflections of Shumagin Ridge, legacy line 1235 (Fig. 8) reveals a 4-km-high basement ridge similar to that of Unimak Ridge in line 205 (Fig. 7). The seaward flank of the buried Shumagin Ridge is formed by a seaward-dipping extensional (SDE) fault zone and associated slope basin. The LDE and SDE faults converge beneath the seafloor to form a peak under the morphological Shumagin Ridge. Basin strata dip increases with depth, indicating growth faulting on both flanks. The basin-filling strata are cut by younger and smaller landward- and seaward-dipping normal faults with displacement of only hundreds of meters. About 10 km east of line 1235, legacy line 217 reveals a bifurcation into two 5-km-high ridges flanked by extensional faults (Fig. 9). Weak fault reflections merge at ∼10 km depth below the ridges. A fault scarp above the seaward ridge but not the landward one occurs above the LDE and SDE fault convergence. ALEUT line 05 is 15 km farther east (Fig. 10); here, the Shumagin Ridge is broad and only 3 km high, consistent with diminished morphological expression in the overlying bathymetry. Both the LDE and SDE faults are broken along several crossing extensional fractures, some of which run into the overlying basin-filling strata. About 12 km NE of line ALEUT 05, in legacy seismic line 215 (Fig. 11), the Shumagin basement ridge is only 2–2.5 km high and flanked by the LDE and SDE faults. However, above the basement ridge, the LDE displaces sediment strata and emerges at a small scarp on the seafloor. Line 213 also shows a clear subsurface ridge ∼1.5 km high overlain by undeformed strata, but this line is not shown here.

These seismic images show the buried basement structure of an 80-km-long segment of Shumagin Ridge basement that underlies the seafloor ridge in conventional bathymetric maps (Figs. 1 and S3 [^{footnote 1}]). About 60 km farther NE, in the boundary area between the Shumagin and Semidi earthquake segments, line ALEUT 04 (Fig. 12) contains a low basement ridge that is formed by an LDE and SDE of similar displacement. Here, the upper part of the ridge includes the Early Miocene sediment section rather than basement. And rather than an overlying seafloor ridge, the basement ridge is overlain by a basin formed by the opposing half graben on both flanks. Sediment ponding in the basin produces a ∼20-km-wide terrace, and strata are only slightly arched at depth above the subsurface ridge.

At the SW end of Shumagin Ridge, ALEUT line 06 (Fig. 13) illustrates structure in the shelf edge indentation at the seaward end of Sanak Basin. It shows a ridge formed by converging LDE and SDE faults, buried beneath ∼2-km-thick strata and a 5-km-deep basin upslope. The seafloor scarp indicates current faulting above the SDE fault zone. Where the LDE fault zone crosses the Sanak Basin’s structural trend, bathymetry is irregular, indicating a disruption of the ridge across the seaward end of Sanak Basin indentation. However, a continuation of Unimak Ridge structure into Shumagin Ridge was probably complicated by the intersection with this major crossing structure.

An unconformity (B in Bruns et al., 1987, their figures 8 and 17) divides the basin sediment along Shumagin Ridge into a lower basin fill and a slope cover. The basin fill is thought to be of Early to Late Miocene age, and the slope cover is Latest Miocene to Quaternary. The inferred age of the upper unconformity allows some estimates of average rates of faulting after ca. 6 Ma based on horizon B vertical displacement.

Within the upper sediment section of line ALEUT 05, annotated extensional growth faults bound either side of a narrow graben (Fig. 10). Unconformity B is displaced 350 +50 m vertically, and across both faults of the graben the unconformity is displaced ∼300 m. These graben faults are interpreted to transmit extensional slip traveling from the plate interface to the seafloor, where bathymetry shows a 5-m-high fault scarp (Bécel et al., 2017). Similar grabens occur without a physical connection to an LDE fault zone (Figs. 7–9), and a number of them have similar displacement and seafloor scarps.

The implied dynamics of the Shumagin segment set it apart from other Alaska margin segments. Creeping behavior has been suggested to explain the lack of a great earthquake in historic time (Sykes, 1971) and later, strain accumulation, in both standard surveying (Savage et al., 1986) and GPS (Fournier and Freymueller, 2007). However, with only land-based geodetic observations, this inference remains uncertain, and its causes are not understood. Dissimilar tectonic structure might help clarify differences between the Shumagin and adjacent segments, as suggested by Bécel et al. (2017). Geophysical data compiled here show previously unreported tectonic similarities and differences with adjacent segments.

A similar feature is the continuous BSFZ along the Shumagin segment observed in seismic images linked with bathymetry (Figs. 1 and 4; Fig. S3 [^{footnote 1}]). The Shumagin BSFZ ridge is less prominent than the Unimak BSFZ ridge, especially in the NE. In the NE, erosional channels are continuous across the BSFZ, whereas in the SW, channels commonly stop at the BSFZ ridge. Evidently, morphological development is more rapid in the SW. BSFZ tectonic activity is inferred from the character of seafloor morphology in the absence of more direct uplift data. Sharp angular relief is considered younger than eroded and rounded bed forms. The sharpest morphology occurs in the SW, whereas to the NE of line 219, the BSFZ ridge becomes almost imperceptible without multibeam bathymetry (Figs. 1 and S4 [^{footnote 1}]). In contrast with seismic lines to the SW, ALEUT 04 has no hanging-wall basin (Fig. 6) that records repeated vertical displacements along the BSFZ (Figs. 2–5). The morphological change from SW to NE is relatively abrupt, and rates of tectonism differ.

A similar trend is observed in the upper-slope fault system. Mid-slope terraces are common to convergent margins, but their underlying structure is not usually as well imaged as in the Shumagin segment upper-slope extensional fault system. Since terraces separate the mid and lower slopes from upper slopes and the shelf, their role as indicators of strain appears significant. The Unimak and Shumagin ridge basement is structured similarly. An LDE fault zone forms Unimak Ridge, and SDE faults are minor (Fig. 7). In the SW Shumagin segment, both LDE and SDE fault zones have a km or more vertical displacements on the basement unconformity (Figs. 8–10 and 13). In the NE, however, the opposing LDE and SDE faults imaged in line ALEUT 04 have much less displacement. Faulted basins filled with sediment form the flat surface of terraces (Figs. 7 and 13) where the rate of sedimentation overwhelms fault displacement. The basement structure seen in ALEUT line 04 is repeated in the SW Shumagin segment, but with greater fault displacement relative to sedimentation. Here, rather than a terrace over the basement high, a ridge runs along the SW part of the segment—despite no indications of a great change in rate of sedimentation. Rates of fault zone displacement appear to have been greater in the SW than in the NE halves of the Shumagin segment.

The abrupt decline of morphological expression in both the upper-slope extensional zone and the BSFZ suggests a common tectonic system (Figs. 14 and S4 [^{footnote 1}]). Upper-slope extension appears compensated by BSFZ or frontal prism contraction. Bécel et al. (2017) mention a similar tectonic scheme.

A Unimak segment tsunami investigation included a diagram relating movement in the mid-slope area of extension to thrusting on the BSFZ (von Huene et al., 2016). With a few modifications, the diagram is relevant to the Shumagin segment (Fig. 14). An important implication of both diagrams is an apparent mechanical isolation of the upper-plate wedge seaward of the mid-slope extensional area. How does the strain monitored geodetically on land differ from strain across the continental slope wedge? Despite being separated by a zone of major extension, the upper and middle slope also display many small normal faults (Figs. 9–13). Contractile deformation begins in proximity to the BSFZ. Regardless of separation from the upper slope, the mid-slope block is strong enough to stand firm against contractile deformation during the earthquake cycle. Its rigidity may be similar to that of the upper slope and shelf and significantly more than that of the frontal prism.

Against this background of mostly similar features, what are some differences? Tectonic dissimilarity between the Shumagin and other segments builds on the former connection of the Alaska and Bering margins and their separation during Aleutian Arc development. In the scenario for the origin of the Aleutian–Bering Sea region of Scholl (2007), this happens between 50 and 55 Ma. The diagonal Sanak tectonic zone derives its tectonic trend from a position at the locus of arc capture when the Bering-Alaska juncture was disrupted to become the Alaska-Aleutian arc (cf. summary of Scholl, 2007, and citations therein). If Bering and/or Alaska structural trends survived, one could question whether Bering basement also underlies the Shumagin segment. In bathymetric maps, the current Shumagin shelf edge projects ∼20 km farther seaward than adjacent segments (Fig. 1), which is consistent with a greater seaward extension of basement. The location of the outer Shumagin Islands in proximity to the shelf edge is consistent with shallow basement near the shelf edge (Figs. 7–10) and with positive free-air gravity anomalies (Sandwell and Smith, 1997).

The plate interface dip depicted in Slab 1.0 (Hayes et al., 2012) varies from shallow beneath the Sanak Island platform to steeper opposite the Shumagin Islands and then shallower again northeast of the islands. Refraction seismic transects across the Unimak (Lizarralde et al., 2002) and Kodiak segments image a plate interface dip that is ∼6 m shallower than that of ALEUT line 05 across the Shumagin segment (Bécel et al., 2017). The agreement between dips in refraction transects and Slab 1.0 lends confidence in Slab 1.0 deviation around the Shumagin Islands.

The unusually narrow frontal prism noted by Bécel et al. (2017) is typical of the Shumagin segment. The width of frontal prism segments undisturbed by subduction of lower-plate relief averaged ∼20 km in the Kodiak segment, 24 km in the Semidi segment, 13 km in the Unimak segment, and only 10 km in the Shumagin segment. The difference in width is difficult to attribute to an unusual character of the incoming ocean crust or volume of trench sediment. GLORIA sidescan images show no major relief in the ocean basin, and they show uniform development of bend faults at an angle to the deformation front. However, in the subduction zone, the plate interface images contain considerable subducting relief that is ∼2 km high. Subducting relief beneath the frontal prism is puzzling considering the lack of any seafloor relief on the incoming plate. Bend faulting of the lower plate after subduction is a possible explanation. The effects of variable subducting sediment thicknesses along the plate interface have recently been studied by Li et al. (2018). They relate the effect of sediment thickness on pore pressure that may be linked to fault behavior.

These indications of a distinctive geology do not yet offer an obvious explanation for the Shumagin gap’s creeping behavior. The Shumagin segment’s geology differs from other segments, but a link to its seismic behavior is beyond the resolution of available data. Marine geodetic observations may show whether the seaward part of the margin framework is detached from the landward part. Whether mid-slope extensional structures root in the plate interface (Bécel et al., 2017) is not likely to be determined without RV Langseth–type seismic acquisition capabilities. Seafloor geodesy would be a major contributor to assessment of earthquake hazards.

A key aspect of Shumagin segment geology applicable to other convergent margins is recognition of significant extensional tectonism landward of the lower slope. Extensional structure is more common than has been appreciated in much of the literature. Well-developed mid-slope terraces can be extensional structures. Terraces are commonly more obvious in a margin’s morphology than the ridge formed by an active BSFZ. Along the Shumagin and Unimak margins, extensional structure is not as deeply buried or perhaps as complex as on many other margins, and thus they can serve as a model to consider where conditions obscure deep structure in seismic data.

We integrated reprocessed legacy and more recent seismic data with bathymetry—both multibeam and conventional—and GLORIA images in the Shumagin earthquake segment. Developments in understanding of convergent margin tectonics and the advancement of analytical systems since a first report in 1985 (Bruns et al., 1985) allowed resolution of unreported tectonic features. An Alaskan backstop splay fault zone (BSFZ) extends along the Shumagin margin, filling a gap between BSFZs in adjacent earthquake segments. The BSFZ hanging-wall morphology indicates recent activity in the SW. Morphology bridges a diverse structure observed in seismic images. The SW half of the segment has a rough morphology, whereas the NE half displays a smoother morphology consistent with decreased tectonic deformation in subsurface information. Rates of tectonism differ in two parts of the segment.

In seismic images, the BSFZ fault zone consists of disrupted reflections in a damage zone. Where the BSFZ extends into the plate interface, its reflections cut across much of the subducted sediment strata and merge into it. The BSFZ damage zone could be a source of clastic material input to plate-interface subducted sediment. When the BSFZ is active, slip beneath the frontal prism is diminished. Alternatively, with the BSFZ truncated at the plate interface, slip may propagate to the deformation front. BSFZ activity appears to be linked with subducting relief.

Along the upper slope, a low ridge underlain by an extensional fault zone is similar to a ridge seen in the Unimak segment. Informally named Shumagin Ridge, it forms a 3–4-km-deep sediment-filled basin. Its subsurface and seafloor expression declines in the NE half of the segment, as does the BSFZ. Much of the extensional fault displacement may have occurred prior to 20 m.y. ago, as in the Unimak segment.

Slope sediment dating from Latest Miocene to Holocene covers Shumagin Ridge and is cut by many extensional growth faults. Their vertical displacement is relatively small and usually does not exceed 400 m. Some of these faults connect with Shumagin Ridge LDE and SDE faults. Activity on the connected normal faults can be explained by continued displacement on the basement faults and downslope gravity extensional forces.

The Shumagin Ridge causative faults displace the regional Oligocene/Miocene unconformity. Unimak Seamount’s 20 Ma age provides a probable youngest age, thereby bracketing the extensional fault origin as between Oligocene and earliest Miocene. Interaction between the extensional fault system and the BSFZ appears likely. The Shumagin BSFZ, like those in adjacent segments, is a potential tsunami generator.

We benefited from discussions with David Scholl and thank Peter Dartnell, U.S. Geological Survey (USGS), for illustrations from his previous bathymetric compilations. Greg Moore shared a Nankai splay fault image for comparison with line 217. Reviews by Maureen Walton and Ginger Barth of the USGS, an anonymous journal reviewer, Donna Shillington, and guest editor Laura Wallace were essential in improving earlier versions of this report. RvH received support for travel from Menlo Park to Denver from a USGS emeritus scientist support fund. The assistance and patience of Gina Harlow and GSA staff throughout the evolution of this contribution is greatly appreciated.