A unified three-dimensional model of the lithospheric structure at the subduction corner in southeast Alaska: Summary results from STEEP

The southern Alaska orogen in the Mount St. Elias and Mount Logan region (Fig. 1) has a long list of superlatives including: (1) the rise from sea level to the summit of Mount St. Elias is often given as the highest coastal relief on earth; (2) near world-record exhumation rates (Enkelmann et al., 2015a, 2015b); (3) near world-record sedimentation rates measured offshore (e.g., Gulick et al., 2015); and (4) the largest existing glaciers in North America. Although these superlatives demonstrate why the region is of geologic interest, it is one of the least studied areas of North America that has been shaped by active tectonics.

Advances were made in the past decade due to a project called the ST. Elias Erosion/tectonics Project (STEEP). STEEP was a multidisciplinary project to study the St. Elias segment of the southern Alaska orogen, and particularly, the interactions between regional tectonics and glacial erosion. This paper is a synthesis of large-scale structural results from STEEP, with additions from more recent work. A focus of the paper is a three-dimensional model that we have constructed of the crust and lithosphere based on data from STEEP and other sources. This model is georeferenced and available as part of the Supplemental Materials¹ to this paper. Some parts of this model are well constrained by available data, while other parts are little more than an educated guess. Nonetheless, we emphasize the models are not a cartoon and are an important contribution of this paper. We hope they will be of use to future geoscientists working in this area as a tool to test hypotheses related to their work. This is particularly important given that the USArray is currently deployed across all of Alaska, and new results on crust and lithospheric structure that exploit these data are beginning to appear. The kinematic model we supply with this paper in digital form can provide a valuable framework for anyone working in the region.

The process of building a lithospheric-scale model for the region has yielded several new insights that are additional contributions of this paper:

1. Results from GPS data combined with traditional fault kinematics produce new constraints on Yakutat microplate motion over at least the past six million years. When we account for rotation in the past 6 m.y., the Fairweather fault (Fig. 1) aligns closely with Pacific–North America motion during the 10–6 Ma stage. We demonstrate that during the 10–6 Ma stage, the Fairweather and Queen Charlotte faults may have been a continuous strike-slip boundary parallel to Pacific–North America plate motion.

2. We clarify a poorly constrained issue of the geometry of the eastern edge of the subducting Pacific/Yakutat lithosphere. We present a range of feasible solutions for this problem that will be testable with imaging results from the USArray.

3. We develop a novel concept for the geometry of crustal faults in the vicinity of Mount St. Elias and Mount Logan; we refer to this concept as the middlebuster model.

4. We develop a comprehensive, regional framework of crust-mantle interaction for the Yakutat microplate linking seismicity, GPS velocity measurements, and structure at a range of scales.

This paper makes extensive use of embedded 3D graphic figures. With this new technology, we hope to reduce misconceptions in future work created by 2D thinking applied to this 4D problem.

There is little question today that the overwhelming process controlling modern Alaskan geology is the subduction of the Pacific plate and its appendage called the Yakutat microplate. The system today is the end result of a long history of a margin dominated by the subduction processes for at least 200 m.y. (Plafker et al., 1994; Amato et al., 2013). In this paper, however, the relevant time interval is the past 30 m.y. and particularly the past 5–15 m.y., when the Yakutat microplate has influenced the system strongly. Although many details of the origins of the Yakutat microplate remain unanswered, the plate tectonic history driving the system is well known and provides strong constraints on any model. The model we present here makes three fundamental assumptions. (1) The Yakutat microplate was transported from the south along Queen Charlotte transform fault (Fig. 1) (Bruns, 1983). The Queen Charlotte system today runs from the southern tip of the Yakutat microplate to an intersection of the ridge system off the northern tip of Vancouver Island. The southern tip of the Yakutat microplate is a triple junction where three predominantly strike-slip faults intersect (Fig. 1): the Queen Charlotte fault, the Transition fault, and the Fairweather fault. We assume the eastern boundary between the Yakutat microplate and North America is the dominantly strike-slip Fairweather fault. At the surface, the Yakutat terrane is defined by the outcrop limit of rocks of Yakutat affinity (Figs. 1 and 2), but we define the Yakutat microplate as the components of lithosphere that have been transported coherently with rocks observed at the surface. (2) At present, the Yakutat microplate is in a tectonic corner and defines a zone of transition from pure subduction in the southwestern Alaska to predominantly strike-slip motion along the Queen Charlotte system. Refining the geometry of that corner is a key objective of this paper. (3) Plate models demand that this corner has been present for a minimum of 20–30 m.y., but the amount of time the microplate has been a component of the corner is debated. There is no doubt, however, that the Yakutat microplate has influenced the corner for the past 5–15 m.y. (e.g., Bruns, 1983; Plafker et al., 1994; and Enkelmann et al., 2015a) and probably longer.

With those assumptions, Figure 3 contains all the components of a lithospheric-scale model that is a central element of this paper. The data, assumptions, and methods used to generate this model are described in the Supplemental Materials (^{footnote 1}). The data files used to generate the model and special files that can be used to construct a custom 3D scene with the open source visualization package, called paraview (http://www.paraview.org; last access 1 November 2017), are supplied in the data supplement hosted externally by IUScholarWorks (https://doi.org/10.5967/K8QC01N6).

This model has five components that are best understood by manipulating the figure to get a 3D perspective and turning different elements on and off with the object selector. From the surface downward the model surfaces are:

1. Topography rendered in true spherical geometry. That surface was produced from etopo5 (https://www.ngdc.noaa.gov/mgg/global/etopo5.HTML).

2. We define a top of slab surface that is an enhanced version of a similar surface described by Bauer et al. (2014) described in the Supplemental Materials (^{footnote 1}). Throughout this paper, the top of slab horizon is illustrated as flow lines defined by the relative motion between the Pacific and North America plates. The flow lines are computed as three-dimensional curves using the technique described by Pavlis et al. (2012a). We used the plate model of Doubrovine and Tarduno (2008) for motion of the Pacific plate relative to North America.

3. We define the crust-mantle boundary (Moho) surface with two different surfaces that overlap. For interior Alaska, we use the crustal thickness model of Wang and Tape (2014), which has the tag “Interior Alaska Moho surface” in Figure 3. We masked their surface south of the position where the top of slab surface intersects the interior Alaska Moho surface. The other half of the Moho model has the tag “Yakutat-Pacific Plate Moho Surface.” That surface is comparable to the surface shown by Bauer et al. (2014), but with an eastward and down-dip extension described in the Supplemental Materials (^{footnote 1}).

4. Three models for the eastern limit of mantle lithosphere linked to the Yakutat-Pacific system are illustrated in Figure 3. All three are shown as a 100-km-thick vertical ribbon with a rainbow color map to provide a depth scale. The ribbon-shaped surfaces are best thought of as the easternmost limit of Yakutat lithosphere for different estimates of the relative motion of the Yakutat microplate.

5. We illustrate a virtual surface best thought of as the base of the Pacific-Yakutat lithosphere (tag “LAB surface”). This surface, however, must not be viewed literally as the lithosphere-asthenosphere boundary (LAB). It should be viewed as a simplified model of the base of the mantle lithosphere used to show the geometry of the lithospheric mantle linked to the Pacific plate and the closely related Yakutat microplate. The surface is “virtual” because it is based completely on projections from a combination of the top of slab surface and a model of the eastern edge of the subducting mantle described in the Supplemental Materials (^{footnote 1}).

Until recently, the strongest constraint on the geometry of the subducting Pacific plate came from seismicity. It is thus useful to compare the surfaces in our lithospheric-scale model to seismicity. To do this, we used catalog location estimates from the Alaska Earthquake Center (AEC) catalog. We used only data from 1990 through 1 November 2015, because coverage prior to 1990 was sparse, creating potentially large event mislocations and a detection bias. That catalog has 447,288 earthquakes with 64,971 events larger than magnitude 2.5.

Animation 1 is a three-dimensional visualization of seismicity we use to show how our lithospheric model is related to seismicity. It uses two visualization techniques. First, the hypocenters of earthquakes larger than 2.5 are displayed as small spheres at their true position in space. We use a slicing plane in the animation to produce what is best thought of as a continuous series of cross sections that change with every frame of the animation. A fundamental problem with viewing only hypocenter locations, however, is that the eye tends to see only areas with the largest number of events and ignore areas with lower but nonzero rates. For this reason, we also present the seismicity with an alternative display method. We plot the seismicity rate as a three-dimensional field. The rate displayed is the total number of earthquakes in the 1990 to late 2015 period normalized by the averaging volume. The averaging volume used is a sphere with a radius of 20 km. We computed the normalized earthquakes per unit volume metric in a 200 × 140 × 60 grid with a nominal grid size of 10 × 10 × 5 km. The nodes were defined with the georeferenced grid methods introduced by Fan et al. (2006).

Animation 1 demonstrates two things:

1. In most of this area, our top of slab surface is systematically deeper than the AEC catalog event locations. Our model is the same as the U.S. Geological Survey (USGS) slab model (https://earthquake.usgs.gov/data/slab accessed 2015) in the area where mantle seismicity is present. Their model is a compilation of diverse data fit to a surface as described by Hayes et al. (2012). The misfit to AEC seismicity is to be expected because the USGS slab model is based on a different earthquake catalog. A different earth model and a different mix of data were used for the USGS catalog, so differences are expected. On the other hand, the systematic difference illustrates the degree of uncertainty in the inferred position of the top of slab surface.

2. Fuis et al. (2008) proposed a tear along the part of the subducting slab separating Yakutat and Pacific lithosphere. One of the reasons they developed that argument can be seen in Animation 1. The mantle seismicity rate falls off dramatically east of the flow line in our model that originates around Kayak Island (around the 3 s mark in Animation 1). The rate is, however, not zero but falls off by a factor of 100. That makes the problem of defining the slab geometry from the seismicity alone subject to large uncertainties in the east. Receiver function imaging results given by Bauer et al. (2014) show that the tear hypothesis is not supported by receiver function images to at least the latitude of the coverage provided by the STEEP seismic network. Animation 1 shows mantle earthquakes have been located north of the area discussed by Bauer et al. (2014). The observed seismicity, however, is consistent with a northern continuation of our top of slab model (Fig. S.2 and related text in Supplemental Materials [^{footnote 1}]). The surface shown is a simple, continuous surface. The tear suggested by Fuis et al. (2008) is an unnecessary complexity that is not consistent with this geometry. Furthermore, using a gap in seismicity to infer a tear requires one to explain a contradiction. A much better documented tear in the Nazca plate in Colombia (e.g., Vargas and Mann, 2013) shows the opposite characteristic. That is, in Colombia, the seismicity rate is higher in the vicinity of the tear as opposed to a gap. A final argument against the tear hypothesis is the recent observation by Wech (2016) that tremor is continuous across the hypothesized tear.

Figure 4 provides an alternative way to demonstrate the change in seismicity rate along the strike of the subducting plate. Figure 4 shows the same metric used in Animation 1 (number of earthquakes per unit volume) but sectioned following the curved horizon defined by our top of slab surface instead of the series of vertical planes used in Animation 1. The counts are produced by searching for the largest value within ±20 km of each point defining the top of slab surface. We claim this figure has two implications.

1. This map is a simpler demonstration that mantle seismicity is not zero in the lithosphere linked to the Yakutat microplate, but the rate is roughly two orders of magnitude lower than the region to the west.

2. This figure suggests a correlation between the flow line labeled as “back projection of west edge of Yakutat” and the large contrast in mantle seismicity rate. That curve was defined as the westernmost flow line that intersects the area where Yakutat rocks are currently exposed at the surface (Fig. 2).

The physical process that creates this lateral change in seismicity rate is not established. Many have used the observation of an eastward cutoff in seismicity to interpret that position as the boundary between Yakutat and Pacific lithosphere (e.g., Eberhart-Phillips et al., 2006; Fuis et al., 2008). The coincidence of this seismicity edge with a plate flow model and the present western limit of Yakutat exposures, however, has not been recognized previously. We explore the implications of this observation in the discussion.

Wech (2016) recently discovered tectonic (nonvolcanic) tremor is present on the subduction interface over a large portion of Alaska (Fig. 4). The eastern edge of the zone of tectonic tremor takes a sharp turn to the south at the point where the mantle seismicity drops off dramatically. Wech also notes major changes in the characteristic of tectonic tremor from west to east. He suggests slip may be continuing to the east, but the slip is continuous and generates neither regular earthquakes nor tectonic tremor. We note that the change in location characteristics that he identifies correspond with the center of the flow line highlighted in Figure 4. Wech’s observation and the parallel seismicity variation emphasize that this seismicity edge is a first-order feature of southern Alaska geodynamics.

STEEP collected an extensive marine active-source data set including ∼1250 km of multichannel seismic-reflection data and two ∼250 km wide-angle reflection and/or refraction profiles (Figs. 5 and 6). The refraction profiles constrained the seismic structure of the offshore Yakutat microplate. These data show seismic velocities and crustal thickness consistent with that of a wedge-shaped oceanic plateau (Christeson et al., 2010; Worthington et al., 2012). The Yakutat crust of oceanic plateau affinity increases in thickness from 17 km thick where it is subducted beneath North America near the Bering Trough to >30 km thick southwest of the Dangerous River Zone (Worthington et al., 2012). This geometry is illustrated in three dimensions in Figure 6 as a rectangular shape (distorted by spherical geometry) with a diagonal line through it marking the interpreted basement position. Inversion of travel time data from the marine shots recorded by the onshore STEEP broadband stations shows that the depth of Yakutat Moho rapidly increases from ∼30 km offshore to 40–45 km beneath the Chugach–St. Elias Mountains (Christeson et al., 2013). The thickest portion of Yakutat crystalline crust enters the St. Elias orogen north of the Malaspina Glacier where the orogen displays its highest relief and its highest exhumation rates (Worthington et al., 2012). The Dangerous River Zone is expressed as an ∼8-km-thick, low-velocity (<6 km/s) crustal cap overlying the eastern third of the offshore Yakutat microplate (Worthington et al., 2012). The location, geometry, and seismic velocity of the crustal cap are consistent with the metasedimentary Yakutat Group. This assemblage has been interpreted as a remnant accretionary prism from an earlier history of the Yakutat microplate (Worthington et al., 2012).

The Transition fault delineates the southern boundary of the Yakutat microplate. It is a near-vertical, strike-slip fault separating Yakutat crust and Pacific crust (Christeson et al., 2010). Across the Transition fault, the Moho steps abruptly down from ∼12 km below sea level on the Pacific plate to ∼30 km below sea level underneath the Yakutat crust (Christeson et al., 2010). Figure 6 shows this geometry in three dimensions when rotated to the right perspective. The Transition fault may accommodate ∼8 mm/yr of differential motion between the Pacific plate and Yakutat microplate (Elliott et al., 2010). Active deformation at the sea floor is expressed as a deformation zone encompassing three fault strands near the intersection with the Aleutian trench, narrowing to one fault strand near the intersection with the Fairweather fault (Figs. 1, 2, and 5) (Gulick et al., 2013).

The STEEP01 refraction line indicates that the sedimentary cover overlying the Yakutat basement is thickest at the western end of STEEP01, reaching ∼15 km thickness near the western end of the line (Fig. 6). This thick sedimentary succession is subducted and accreted at the Pamplona Zone fold-thrust belt (Bruns and Schwab, 1983; Worthington et al., 2012; Van Avendonk et al., 2013). The décollement location was not directly imaged from the active source data but was inferred as a low-velocity zone within the sediments located at either the top or bottom of the Poul Creek Formation (Van Avendonk et al., 2013). East of the frontal faults of the Pamplona zone, the shelf sediments experience porosity loss due to lateral compaction over ∼60 km, but active faulting is absent outside of the fold-thrust belt (Worthington et al., 2008; Van Avendonk et al., 2013). The sedimentary cover thins to less than 1 km at the Dangerous River Zone (Fig. 1) and eastward (Worthington et al., 2012). Although the sedimentary package is tapered from west to east, a striking feature revealed by these data is that the Moho of the Yakutat microplate appears to be at a nearly constant depth of 30 km. Thus, an overall perspective of Yakutat microplate crust illustrated in Figure 6 is that basement is 30 km thick at its eastern edge. The basement itself thins to only ∼15 km at the western side but is overlain by as much as 15 km of sediment. Hence, the total crustal thickness stays nearly constant at 30 km across the largely undeformed section between the Pamplona Zone and the Transition fault.

We utilized the results from the STEEP active source experiments and the receiver function imaging results of Bauer et al. (2014) as a foundation for the regional Moho model illustrated in Figure 3. The Yakutat-Pacific Moho surface is an extension of that given earlier by Bauer et al. (2014). Details on how this surface was generated can be found in the Supplemental Materials (^{footnote 1}). Data used to construct this surface are:

1. The offshore portions of the surface are constrained by STEEP active source data.

2. The geometry near the coastline is controlled by the onshore-offshore data analysis by Christeson et al. (2013).

3. The Moho surface north of the coastline between Prince William Sound and Yakutat Bay is constrained by the P and S receiver function data described by Bauer et al. (2014).

4. West of Prince William Sound the surface is a downward projection of the top of slab surface that assumes the subducting oceanic crust has a constant thickness.

5. The Yakutat-Pacific Moho surface is truncated at the Fairweather fault on the eastern side of the Yakutat block. For the Queen Charlotte fault south of the triple junction, we define an offset of the Moho moving from oceanic crust of the Pacific plate to continental crust onshore. We note this part of the model is completely unconstrained by data and is only a downward projection of the mapped surface fault positions. Nonetheless, that part of the Moho surface provides a valuable approximate perspective of the eastern limit of Yakutat and Pacific lithosphere.

A final detail is that our Moho model tapers the end of control from the refraction data illustrated in Figure 6 to a downward projection of our top-of-slab surface in the west. At the same time, the sedimentary wedge is assumed to thin westward to near zero south of the western shore of Prince William Sound. This can be seen in Figure 6 as the westward taper in the offset of the Moho surface at the southern edge of the Yakutat microplate. We stress that taper is purely an interpolation of available data and may not reflect reality.

STEEP dramatically expanded the number of global positioning system (GPS) measurements in the St. Elias region. The estimates of surface velocities relative to stable North America by Sella et al. (2007) are illustrated in Figure 7. That figure compares GPS velocity vectors to plate motions predicted by the three plate motion models for the Yakutat microplate used in this paper. The GPS velocities west of Prince William Sound are from the data set of Freymueller et al. (2008) with model predictions of the postseismic effects from the 1964 Alaska earthquake (Suito and Freymueller, 2009) and glacial isostatic adjustment (Elliott et al., 2010) removed. Between eastern Prince William Sound and Yakutat Bay, the GPS velocities come from the data set of Elliott et al. (2013) with models for postseismic effects of the 1964 earthquake, the 2002 Denali earthquake, and glacial isostatic adjustment applied. For the area east and south of Yakutat Bay, the GPS velocities are from the data set of Elliott et al. (2010) and have had model predictions for glacial isostatic adjustment applied.

We suggest that Figure 7 illustrates the following:

1. Velocity vectors in the region of Yakutat Bay are closely aligned to the motion predicted by the Yakutat–North American pole estimated by Elliott et al. (2010). Because those data provided the primary constraint on that plate motion, this is expected. This is noteworthy only because that is not true for all other GPS velocities measured on rocks along the deforming northern margin of the Yakutat microplate.

2. There are large changes in surface motion on both edges of the Yakutat microplate. West of Mount St. Elias, velocity vectors are rotated counterclockwise from the motion predicted by any of the three large-scale plate motions illustrated in Figure 7, but vectors become closest to the end member shown in Figure 7B. West of Prince William Sound, the directions change to be more closely aligned with that expected from motion of the Pacific plate.

3. West of Yakutat Bay, velocities change over short distances, reflecting the high strain rate in this region. Marechal et al. (2015) used simplified fault geometries from Elliott et al. (2013) to correct an expanded GPS data set for effects of elastic strain accumulation and then used the corrected data to derive a regional strain rate field. They suggested their results were most consistent with distributed deformation around an indenter corner. Earlier work by Elliott et al. (2013) used a block model to fit the entire region north and west of Mount St. Elias (Fig. 7). Their inversion model simultaneously solved for long-term tectonic block motions and the effects of interseismic (elastic) strain accumulation along the block-bounding faults. One particular aspect of that model is that much of the upper plate west of Mount St. Elias (a region they called the Elias block) exhibits a counterclockwise rotation. This rotation was similar in direction to the rotation Fletcher (2002) inferred for the southern Alaska block from GPS data near the Denali fault. The westerly surface motion linked to this block is one explanation for why the GPS vectors do not align with any of the plate models examined in this paper. Figure 7D shows the slip directions for the low-angle faults defined in Elliott et al.’s (2013) model. Slip on the model faults is due to relative motion between the upper plate and the down-going plate (assumed to be Yakutat); so the estimated slip vectors are slightly smaller in magnitude and more northerly than predicted from Yakutat–North America motion.

4. Animation 2 illustrates a different and noteworthy property of Elliott et al.’s (2013) model. The low-angle faults required to fit the GPS data have a much shallower dip than the top-of-slab model, which in this region is well constrained by receiver function imaging (Bauer et al., 2014). Consequently, all the surfaces estimated from GPS data lie well above the top of slab surface. This is due to an assumption in the GPS fault model that the dominant process controlling surface deformation measured by GPS data is elastic strain accumulation along the model interface. If the dips of the model fault planes are increased, the surface signal from strain accumulation will decrease quickly downdip. In order to fit the observed GPS velocities, the block motion of the upper plate would need to exceed full plate convergence rates to compensate for the drop in elastic deformation with distances. In addition, a change in dip also impacts the rate at which strain decreases landward from the trench. Strain accumulation along a shallow slip interface will result in a large surface signal over a broad region, such as that exhibited by the GPS velocity field east of Prince William Sound. A fault with a steeper dip will result in a more rapid decrease in strain away from the trench, similar to the observed GPS velocity field west of Prince William Sound.

The most important thing the GPS data add to understanding this orogen is that one or both of two processes must be going on in the St. Elias orogen to reconcile the GPS data with plate kinematics: (1) the upper crust in southeast Alaska is not moving in the same direction as the mantle lithosphere of the subducting plate, and/or (2) the Yakutat microplate lithosphere is moving as a block that is not locked to the Pacific plate, and our inferences are all impacted by edge conditions. At this point, none of the available data can unambiguously address which of these two processes dominate.

Since George Plafker’s pioneering work (Plafker, 1967, 1987), it has been known that some of the most intense active deformation in the southern Alaska orogen is concentrated in the St. Elias region. Structural and stratigraphic observations by Plafker (1987) in the Samovar Hills recognized angular unconformities within the Cenozoic section and complex fault cross-cutting relationships. STEEP studies further analyzed these geologic complexities with details reported in a pair of companion papers (Chapman et al., 2012; Pavlis et al., 2012b). Here we summarize some of those results in the context of the larger-scale, regional tectonic relationships that are the focus of this paper. In particular, we focus on the processes that occur within this complex “corner” where the system transitions from a slip-partitioned transpressional system east of Mount St. Elias to a fully convergent fold-thrust system to the west (e.g., Bruhn et al., 2004; Pavlis et al., 2004; Koons et al., 2010; Marechal et al., 2015). The relationships discussed here are the basis for the fault geometries used in development of our kinematic model for the St. Elias region.

The clearest surface manifestations of the structural complexity in the Mount St. Elias region is the ∼90° change in strike of rock units and faults when following the mountain front through the bend from Yakutat to Cape Yakataga (Fig. 2). This bend is now recognized as an analog to the Himalayan syntaxes (Elliott et al., 2013; Koons et al., 2013; Enkelmann et al., 2015a; Zeitler et al., 2015) both in terms of structural style (Chapman et al., 2012) and localization of uplift and exhumation (e.g., Enkelmann et al., 2010, 2015a, 2015b). The changes that occur around this bend are manifestations of the 3D strain inherent in such corners (e.g., Koons et al., 2010, 2013). Figure 8 illustrates the geometry of the structures we infer for this corner.

South of Yakutat Bay, the active tectonics are comparatively simple. All evidence indicates that area is a slip-partitioned, transpressional system (Bruhn et al., 2004). The strike-slip component is taken up on the Fairweather fault (labeled such in Fig. 8), and the contractional component is taken up on steeply to moderately dipping blind thrust faults along the Yakutat foothills (Bruhn et al., 2004; Plafker and Thatcher, 2008; Elliott et al., 2010). Plafker and Thatcher (2008) modeled the thrust systems they linked to uplift related to the 1899 earthquake sequence (surface in Fig. 8 with the tag “Thatcher and Plafker fault model”). We used their surfaces and Estabrook et al.’s (1992) model of the 1979 St. Elias earthquake as the foundation of our interpretations. Details on how the surface with the tag “Middlebuster surface projected” in Figure 8 was constructed from this collection of earthquake rupture surfaces can be found in the Supplemental Materials (^{footnote 1}). Longer-term deformation is not as well understood, but recent work by Enkelmann et al. (2015a, 2015b) and Falkowski et al. (2014) supports preliminary evidence from Sisson et al. (2003) that the inferred modern-day full slip partitioning is not consistent with the long-term record indicated from exhumation. Specifically, these data indicate the region east of the Fairweather fault has experienced long-term exhumation. That implies a significant component of east-side-up displacement on the Fairweather fault system. This conclusion is not necessarily surprising given the high terrain in this region (e.g., ∼5000 m Mount Fairweather) but is important here because it indicates long-term crustal shortening is occurring east of the Fairweather fault.

At the latitude of Yakutat Bay, the orogenic front takes a sharp turn from a NNW trend to a NW trend (Fig. 2). We infer that this segment marks the southern edge of the complex strain associated with the “corner.” Between Yakutat Bay and the Seward Glacier outlet, the Mount Cook block (Fig. 2) represents a transpressional pop-up bounded to the north by the continuation of the Fairweather fault and to the south by a system of poorly defined thrust faults (e.g., Bruhn et al., 2004, 2012; Pavlis et al., 2004). Several significant features within the Mount Cook block demonstrate it is the nexus of deformation in the corner.

1. There is a dramatic increase in peak heights relative to the Yakutat foothills to the south. Mount Cook rises abruptly from sea level in Yakatat Bay to a height of 4196 m.

2. It is now well established that exceptionally high exhumation rates are centered on the Mount Cook block and extend across the northern extension of the Fairweather fault (Spotila and Berger, 2010; Enkelmann et al., 2015a, 2015b, 2016; Falkowski et al., 2014). The exhumation data suggest a transpressional pop-up of the Mount Cook block is a prime factor but it is also part of a broader uplift.

3. The highest peaks in the St. Elias Mountains are immediately north of the Mount Cook block (i.e., Mount Logan, Vancouver, King George, Kennedy, and Hubbard).

4. The collisional suture defined by the Chugach–St. Elias fault (CSEF; Fig. 2) carries high-grade metamorphic rocks of the Mount St. Elias massif on top of Yakutat rocks. That structure terminates on the east side of Mount St. Elias within the Mount Cook block.

Collectively, these observations indicate the Mount Cook block has experienced major uplift and exhumation. Enkelmann et al.’s (2015a, 2015b) work indicates much of the exhumation is very young. The exhumation data also suggest that the absence of the suture within the Mount Cook block is probably related to erosional removal of the hanging wall instead of the a strike-slip juxtaposition idea that was favored by Bruhn et al. (2004) and Pavlis et al. (2004).

At the Samovar Hills (Fig. 2), the orogenic front takes another strong turn from a NW to a NE trend. This area represents the heart of the structural syntaxis. It is probably no coincidence that this area also contains the largest, most erosive glacier in the orogen (Headley et al., 2013). The Seward glacier outlet, which feeds the large piedmont ice lobe (Malaspina Glacier) to the south, sits in this position. Chapman et al. (2012) and Pavlis et al. (2012b) described several features of the geology of this segment of the orogeny; these features are important to review here.

1. Structural relationships in the Samovar Hills demonstrate a complex history as material was transported into the corner. They infer the early NW-trending structures seen across the area formed along the transpressional boundary prior to being transported into the corner. These earlier structures were refolded as the rocks passed into the corner. The exact age of this transition is not well constrained due to poor age control on Yakataga Formation rocks that lie above an angular unconformity that marks the boundary with the first stage of deformation. Nonetheless, the structural relations indicate material has been transported into the corner over an extended period with large changes in structural style with time.

2. Although the geology of the Mount St. Elias massif and adjacent high ridge is not well constrained due to the extreme terrain and ice cover, an important general observation is that metamorphic grade in the hanging wall of the CSEF increases systematically from west to east until the suture disappears beneath ice cover west of the Seward glacier outlet. Together with the topography and thermochronology data (Enkelmann et al., 2015a), this observation suggests that uplift and exhumation increase from west to east along the suture.

3. Restoration of cross sections along Icy Bay can only account for a limited amount of shortening (∼150 km). That restoration suggests strongly that large amounts of material have been underthrust beneath the margin (Pavlis et al., 2012b).

4. Field data west of Icy Bay suggest the presence of several out-of-sequence thrusts as well as significant underplating of sediments beneath the frontal thrusts of the Yakataga fold-thrust belt system (Pavlis et al., 2012b). The active fault systems within the system form a distinct en echelon array (Chapman et al., 2008; Pavlis et al., 2012b), suggesting dextral oblique convergence within the thrust belt. These observations are consistent with GPS data that show contraction and a westerly rotation of velocities compared with Yakutat motion seen to the east (Fig. 7).

Collectively, these observations support the concept that over a large fraction of the orogen’s history, material has been carried northward along the transpressional Fairweather system. When this material arrives at the corner, it experiences a major kinematic change reflected in the structures that shape the surface geology seen in Figure 2. Deformation west of the corner, however, is also almost certainly influenced by this localized effect.

Figure 8 illustrates the subsurface geometry we infer for the fault systems that are now present within this corner. The key elements of this 3D model are projections of faults beneath the Mount Cook block toward the west and connected to the NE-striking faults that form the frontal fault between Icy Bay and the Samovar Hills. The near-surface geometry of the latter is locally well constrained by a drill hole and restorations of the frontal thrust (Pavlis et al., 2012b), but the geometry of the fault beneath the Mount Cook block is poorly constrained by surface geology. In any case, a robust feature of the inferred subsurface geometry are the two distinct fault trends that converge beneath the Samovar Hills, forming a ∼NS-trending intersection line that plunges northward beneath a point near the top of the Malaspina Glacier. We noticed that the geometry indicated by the geology is consistent with the orientation of the two phases of rupture estimated from waveform inversion of seismograms recorded for the 1979 St. Elias earthquake by Estabrook et al. (1992). Our geometric model (Fig. 8) connects mapped surface fault locations with the 1979 earthquake fault planes to produce a single, continuous surface. We note here that this geometry is unusual, resembling a farm implement called a “middlebuster plow.” Thus, we refer to this concept as the “middlebuster model” for the syntaxis.

We analyzed the applicability of the “middlebuster model” to the eastern syntaxis area by constructing a 4D kinematic model with the components described in the Supplemental Materials (^{footnote 1}). In this model, the plow is driven northward in the direction defined by plate motion. That model is undoubtedly an oversimplification of the motion because geologic data (e.g., Chapman et al., 2012; Pavlis et al., 2012b) suggest complex fault interactions within the corner. Nonetheless, the model provides useful first-order predictions of uplift and/or exhumation patterns and the kinematic evolution of the system (Fig. 9). We ran this model with the three different plate motion vectors used in this paper (tracks in Fig. 3). The results were sufficiently similar, however, that we show only the Pacific–North America (PAC-NA) motion model (track A). This similarity arises primarily because of the geometry of the fault model. The intersection line between the two-fault system (axis of the top of the middlebuster surface) trends approximately N-S, whereas all motion models are oblique to this axis, with a more northwesterly trend.

The principal result of this kinematic model is a prediction of uplift patterns that would result from the interpreted subsurface fault geometry (Fig. 9). Specifically, the model predicts an uplift pattern that is consistent with observed topography where uplift is centered on what are now the Mount St. Elias and Mount Cook massifs. The model also predicts a structural divide between these massifs at the site of what is now the Seward Glacier (Fig. 2). This geometry is robust for any subsurface fault geometry with this “middlebuster” configuration, provided the intersection line between the two fault surface trends more northerly than the convergence vector. We suggest that this result supports our interpretation of the subsurface fault geometry model illustrated in Figure 8. In addition, the model provides an explanation for a previously enigmatic geomorphic observation for this region: the location of the Seward Glacier outlet. That is, the location of this glacial stream cuts directly across the structural grain with no obvious structures controlling its location. Our model suggests the location of this divide is not coincidental, but it is a natural consequence of the kinematics and dynamics of the corner. As the topography developed, ice would have spilled naturally over the topographic low between two emergent highlands to form the glacial outlet stream of the modern Seward Glacier. Furthermore, this model also predicts localized extension to the north and east of the axis of the “middlebuster,” which is consistent with sub-ice structural interpretations beneath the Seward Glacier (Bruhn et al., 2012).

There is a long-standing controversy on the motion of the Yakutat microplate and how lithosphere linked to the Yakutat microplate has or has not been subducted. It has long been known that the eastern Aleutian arc was characterized by flat slab subduction (e.g., Isacks and Barazangi, 1977), but close association of the flat-slab area with both the Yakutat collision and Wrangell volcanic arc segments confused the association of flat-slab processes with regional tectonics for years. Work in the past decade has been strongly influenced by the tomography results of Eberhart-Phillips et al. (2006). A polygon of the outline of Yakutat lithosphere that appears in Figure 2b of their paper has been shown in numerous papers (e.g., Fuis et al., 2008; Worthington et al., 2012; Wang and Tape, 2014; or Martin-Short et al., 2016). Although that polygon has been heavily reproduced, it is important to realize that (1) the eastern edge of the Yakutat polygon they develop is an interpretation from the tomography model developed in that paper, and (2) the eastward extent of that interpretation was based on very sparse data at the time. Most authors who have reproduced this polygon fail to note that the same figure shows a dashed line with the tag “Limit of Potential Slab,” which is more in line with the edge models shown in our Figures 3 and 8. Our three-dimensional model, however, differs significantly from that polygon as illustrated in Figure 10. Figure 10 shows back projections of two end members (Tracks A and C in Fig. 8) of the Yakutat-Pacific–North America triple junction. Both projected edges are well east of the limit indicated by the polygon of Eberhart-Phillips et al. (2006). There are two primary reasons for this disconnect. First, Eberhart-Phillips et al. (2006) pinned the eastern edge of their model on the St. Elias syntaxis, but STEEP studies and details reported here show that this reference point is west of all feasible projections of Yakutat basement. For reasons described at length in the Supplemental Materials (^{footnote 1}), we pin our edge model to the triple junction formed by the Fairweather-Transition–Queen Charlotte fault intersection, which Figure 10 shows has moved parallel to Pacific–North America motion for nearly 30 m.y. Since the triple junction is more than 300 km from the St. Elias syntaxis and offset from Pacific–North America plate motion, this produces a large difference in any of our “Limit of Potential Slab” estimates and from the comparable estimate shown by Eberhart-Phillips et al. (2006). The second reason for the large difference of our polygon is that Eberhart-Phillips et al. (2006) used the older plate motion model of Stock and Molnar (1988) and a cruder approximation of plate motion. Specifically, they used a conic map projection and approximated the motion in the three latest plate motion stages as straight-line vectors for three stage poles given by Stock and Molnar (1988). By comparison, we use a more precise method tracking motion along three-dimensional flow lines (see Supplemental Materials [^{footnote 1}] for detail). The polygon illustrated as a separate layer in Figure 10 is our proposed replacement for the Eberhart-Phillips et al. (2006) polygon. Points defining this polygon can be found in the data portion of the Supplemental Materials (^{footnote 1}).

We illustrate our reconstruction of block rotation for the Yakutat microplate over the past 6 m.y. in Figure 11. This reconstruction suggests an important idea that seemingly had not been recognized previously. Notice that the strike of the Fairweather fault for the projected polygon with the layer tag “Translated GPS polygon” aligns closely with the track of the triple junction south of its current location. Because the triple-junction track matches the modern strike-slip plate boundary in that region, it suggests that the modern Fairweather was once a continuous strike-slip boundary that ran parallel to plate motion during the 10–6 Ma stage of Pacific–North America motion. The geometry illustrated in Figure 11 would predict that the Transition fault had little to no motion in the 10–6 Ma period because no motion independent of the Pacific plate is required if the ancestral Fairweather fault was oriented parallel to Pacific–North America plate motion.

If full plate motion was taken up on the Fairweather fault prior to 6 Ma, how did the system respond when plate motion changed? If we assume basement and cover, or mantle and crust, have moved separately during the past 6 m.y., the reconstruction in Figure 11 has an important implication. Specifically, if Yakutat mantle followed a North America–Pacific track but the upper crust responded to the plate motion change, then the apparent rotation shown in the reconstruction would transfer as both westward motion (relative to North America) of the actively deforming fold-thrust belt as well as contraction along the Fairweather transform (transpression) during this time interval. Both of these motions are, in fact, known from both surface geology and GPS data based on the following observations.

1. Significant evidence exists for slip partitioning of strike-slip and thrust systems along the Fairweather fault. That observation is consistent with transpression in the past 6 m.y. (e.g., Bruhn et al., 2004, Plafker and Thatcher, 2008, or McAleer et al., 2009) and is confirmed by GPS measurements (Elliott et al., 2010).

2. GPS velocity vectors are rotated counterclockwise relative to current North America–Pacific (NA-P) motion within the fold-thrust belt west of Mount St. Elias (Fig. 7). The vectors return to a direction closer to NA-P parallel motion to the west of Prince William Sound. This can be explained by block motion of the Yakutat microplate relative to the Pacific and/or deformation of the upper crust.

3. Onshore active thrusts in the fold-thrust belt have an en echelon geometry that suggests distributed dextral shear (Chapman et al., 2008; Pavlis et al., 2012b). That structural relationship is consistent with cover rocks moving westward relative to North America.

Further support that the Fairweather fault is an ancestral, through-going plate boundary is shown in Figure 12. That figure shows that if we propagate the Yakutat microplate back 6 m.y., we can project a southern coastline for Alaska that is consistent with the edge of slab imaged by Kim et al. (2014). The reconstruction suggests that the southern coastline of Alaska may have been much straighter 6 m.y. ago.

To understand the new constraints our results place on this 4D problem, it is helpful to review the current state of knowledge of this region. Three results from STEEP active source data provide new constraints:

1. Seismic-reflection data show that the bulk of shortening in the Yakutat microplate occurs north and west of the offshore Pamplona Zone fold-thrust belt (Worthington et al., 2008, 2012; Van Avendonk et al., 2013). The same data suggest that the interior of the Yakutat microplate south and east of the Pamplona zone is largely undeformed.

2. Seismic-reflection data collected across the Transition fault show little evidence of shortening (Gulick et al., 2013). An exception is the western end of the fault where Gulick et al. (2013) argue that the corner is evolving as an unstable triple junction that is now deforming Yakutat crust internally and may eventually accrete part of the Pacific crust to North America.

3. Structural modeling informed by seismic-reflection data and onshore geology show that the Pamplona Zone is characterized by thin-skinned, décollement style deformation. Collectively, these observations demonstrate detachment of cover from subducted basement throughout the system.

Bauer et al. (2014) used the insights from the active source data to build a regional-scale model of the top of the subducting slab that has been extended in this paper (Fig. 3). Three insights from Bauer et al. (2014) provide additional constraints on the nature of crust-mantle coupling.

1. They noted that the aerial extent of the Pamplona Zone fold-thrust belt narrows from ∼75 km offshore of the Bering Glacier to zero at the Seward Glacier. This narrowing corresponds exactly with an along-strike change in sediment thickness from more than 15 km in the western Yakutat microplate to near zero at the Dangerous River Zone (DRZ in Fig. 1). They used three-dimensional P and S wave receiver function images to show that there is a parallel change in subsurface structure. In the western part of the Yakutat microplate, the system is a classic deforming sedimentary wedge. It narrows eastward to form a sharp deformation front in the vicinity of Mount St. Elias where the crust thickens abruptly from 30 to 50 km.

2. The Moho of the Yakutat microplate can be traced northward to the limit of coverage by the STEEP array in the southern edge of the Wrangell Mountains. Their results show an increasing dip of this surface from west to east. The model we produced here extends that projection northward.

3. Bauer et al. (2014) argue that there is no evidence in the Moho geometry for a tear in the slab as suggested by Fuis et al. (2008). Bauer et al. (2014) noted, however, that this inference was not unambiguous due to large holes in coverage in the vicinity of Prince William Sound. We add here that this conclusion is also limited by the finite resolution of the S-wave receiver function data that was the main constraint they used for that inference. An offset of as much as 10 km could easily be present and obscured by inadequate horizontal and vertical resolution. This ambiguity should be resolved soon with new data from USArray.

These observations provide some answers to how the system responded when Pacific–North America plate motion changed at 6 Ma. Figure 13 is a map comparing track A and track B but with an origin at the western edge of the Yakutat microplate. These two flow lines illustrate our best estimate of convergence perpendicular to plate motion from independent motion of the Yakutat microplate. The map shows the cross-flow convergence is ∼30 km, which would translate to a rate of ∼5 mm/yr. The predicted convergence for track C is roughly twice these numbers. The implications for where deformation is likely to occur for different models, however, are very different. In the track A scenario (Yakutat moving coherently with the Pacific), crustal shortening would have to be accommodated by strain on all edges of the microplate except the Transition fault. The reason is that in that scenario no motion on the Transition fault is allowed. Track C, in contrast, demands no convergence across the Fairweather fault. In that scenario, convergence would only occur north and east of the point where the Fairweather fault bifurcates east of Yakutat Bay (Fig. 2). Observations at the surface are not consistent with either of the end members of tracks A and C. There is strong geologic evidence for transpression along the Fairweather south of Yakutat Bay (e.g., Bruhn et al., 2004; Pavlis et al., 2004, McAleer et al., 2009); this evidence contradicts the predictions of track C. Track A is inconsistent with the GPS data, which define our track B in Figure 3. Track B is thus the only result consistent with the geologic data and the GPS data, which is why we illustrate only track B in Figure 13.

From a mantle perspective, the different models also have variable implications. If we assume rigid plate motion is defined by motion of the mantle lithosphere, track C requires ∼10 mm/yr of mantle-lithosphere shortening perpendicular to Pacific plate motion, track B ∼5 mm/yr, and track A requires none. We know of only two ways to accommodate convergence of mantle lithosphere: internal deformation of the mantle lithosphere, which would likely be focused at the boundaries, or bending of the entire plate. In the Appendix, we test the plate-bending hypothesis by computing shortening that could be accommodated solely by that mechanism. That analysis indicates that only 10–20 km of shortening could be accommodated by this mechanism alone without requiring the slab to be folded to a near vertical dip. That is roughly half the shortening required for track B but only about one-fourth of that required by track C. Hence, plate bending alone is probably not sufficient to take up the relative motion of track B in the mantle, let alone track C, which requires twice the convergence. New tomographic imaging of the slab with USArray data should provide constraints on the actual curvature of the subducting slab. When that geometry is better known, it will be possible to estimate the relative importance of plate bending and internal deformation of mantle lithosphere.

How then does the crust move in response to the geometry of Yakutat microplate, which seems to have some independent motion? The GPS data provide strong constraints at the surface, but how the displacement field varies with depth depends on inferences from models like those of Elliott et al. (2010, 2013) and Marechal et al. (2015). Furthermore, crustal motions are linked to motions of blocks of a wide range of scales from giant plates such as the Pacific to points surrounded by arrays of active structures that make the deformation field more easily modeled as distributed strain (e.g., the results in Marechal et al., 2015). To examine this issue, it is useful to consider regional observations from the largest to smaller scales including:

1. A number of authors have suggested that the collision of the Yakutat microplate has driven deformation well inboard of the suture (e.g., Mazzotti and Hyndman, 2002; Leonard et al., 2007). Numerous authors have attempted to develop geodynamic models of this system with and without a Yakutat contribution (Soofi and Wu, 2008; Koons et al., 2010, 2013; Finzel et al., 2011; Jadamec et al., 2013). Freymueller et al. (2008) analyzed GPS data from the entire Alaska region and found evidence for widespread deformation inboard of the main plate boundary region. Marechal et al. (2015) use the concept of an indenter to explain the GPS data. In contrast, Jadamec et al. (2013) get large-scale motion by coupling of the subducting plate to the lithosphere of Alaska and treating the Denali fault as a zone of weakness. In summary, there are multiple competing hypotheses to explain the deformation field, and it is not yet clear what processes dominate.

2. Convergence between the Pacific and North American plates has been an important process for the past 6 m.y. in British Columbia as demonstrated by Figures 10 and 11. Hyndman (2015) analyzed data from the 2012 Mw = 7.8 Haida Gwaii earthquake in combination with a multifaceted set of geologic and geophysical data. He notes the same convergence we demonstrate in Figures 10 and 11 and argues the system in the area of Haida Gwaii is partitioned in strike-slip motion on the Queen Charlotte fault and incipient subduction created by the convergence. Haida Gwaii, in fact, is no doubt an island because of that fault geometry. Hence, large-scale convergence related to this bend in the plate boundary may be a larger factor in driving the tectonics of British Columbia and southeast Alaska than is widely recognized.

3. All the evidence indicates that the western edge of the Yakutat microplate is in the vicinity of the western side of Prince William Sound (Eberhart-Phillips et al., 2006; Bauer et al., 2014; Kim et al., 2014). An observed tremor that Wech (2016) associates with the Yakutat slab abruptly terminates downdip of the western side of Prince William Sound (Fig. 4). West of Prince William Sound in the Kenai Peninsula, GPS velocities are most consistent with subduction of the Pacific plate (e.g., Suito and Freymueller, 2009). The complete geometry of that boundary in three dimensions, however, is poorly constrained at this time. Bauer et al. (2014) and Kim et al. (2014) provide some constraints on the position of the boundary in the mantle, but both studies suffer from coverage limitations. In any case, resolving this ambiguity is an important target of seismic imaging with the USArray data.

4. At crustal scale, all the evidence indicates that the region near Mount St. Elias is the nexus of deformation created by the transition from a convergent to transform plate boundary. We argue here that the middlebuster fault structures mark the brittle response of the upper crust to this rapid lateral change in the strain field. The western limb of the plow reflects the termination of the Aleutian megathrust. The eastern limb is the response of the crust to the sliver of crust being transported northward and fed into the corner with motion parallel to the Fairweather fault. Ductile deformation of the lower crust must be a significant factor in shaping the overall deformation field in this corner. The most direct evidence of this claim is the thickened crust under this region inferred from refraction data (Christeson et al., 2013) and receiver function data (Bauer et al., 2014). That general pattern is also consistent with that predicted from geodynamic models by Koons et al. (2010).

Figure 4 and Animation 1 quantify a well-known feature of Alaska seismicity—the dramatic falloff in mantle earthquakes through the center of the state. Our lithospheric model provides some new insights on what processes could be responsible for this dramatic change in seismicity rate within the subducting plate.

We found a strong change in seismicity rate across a flow line that has an origin at the western limit of where Yakutat rocks are currently exposed (Fig. 4). There is significant evidence, however, that lithosphere of Yakutat affinity is present in the subsurface well to the west of that flow line. Bruns (1983) was the first to use magnetic data to postulate Yakutat crust defined the footwall of the megathrust well to the west of where the Yakutat rocks are found at the surface. Seismic evidence from local earthquake tomography (Eberhart-Phillips et al., 2006) and receiver function imaging (Ferris et al., 2003; Abers, 2008; Kim et al., 2014) have provided additional evidence to support the idea that much of the flat slab region of the subducting plate has Yakutat affinity. If most of the lithosphere in the subducting plate in central Alaska has Yakutat affinity, what then causes the dramatic difference in the seismicity rate across the flow line emphasized in Figure 4? A possible answer comes from a synthesis of three ideas found in previous papers.

First, Figure 12 demonstrates the kinematics of how the plate boundary likely evolved over the past 6 m.y. The focus here is the westward extension of the Transition fault used as a marker similar to that used by Bruns (1983). That extension also defines the southern edge of the Eberhart-Phillips et al. (2006) polygon illustrated in Figure 10. Both capture the concept that Yakutat lithosphere is present as far west as Kodiak Island today. Figure 12 adds the insight from Bruns (1983) that the southern boundary of the Yakutat microplate was likely comparable to the Transition fault offshore today, but that part of the Transition fault has now been subducted.

The second observation linked to the seismicity issue is that it is now clear that Yakutat lithosphere that has been subducted in the past 6 m.y. varies strongly but consistently along the strike of the plate boundary. Active source results from STEEP and onland geology show that up to ∼5 km of sedimentary material are being entrained at the top of the subducting Yakutat slab (Van Avendonk et al., 2013). Furthermore, Christeson et al. (2010) argue the basement of the Yakutat microplate is likely derived from an oceanic plateau creating a compositional as well as thickness difference. Since the pioneering work of Plafker (1987), it has become clear that the fold thrust belt containing Yakutat rocks west of Mount St. Elias was created by stripping of most, if not all, of the sediments from the basement. However, to the west, the suture turns south, continuing offshore west of Kayak Island, and merges with the Aleutian trench to the southwest (Figs. 1 and 2). The special flow line in Figure 4 has an origin at the western limit of Yakutat rocks. To the west of this zone (aka Kayak Island zone), it has been suggested that the subduction margin is erosional (Fruehn et al., 1999). We suggest the east-to-west thinning from a thickness of 30–15 km of Yakutat basement provides a possible explanation for that transition from accretion to erosion and the modern extent of Yakutat sediments. That is, we suggest that prior to the initiation of subduction, the sedimentary cover of the Yakutat microplate may have thinned west of the vicinity of Kayak Island. With little to no cover to strip, the subduction zone could be expected to transition to a more normal oceanic crustal subduction zone and even become erosional if the sediments became sufficiently thin.

Bauer (2014) added an additional insight that may help explain the along-strike change in seismicity rate illustrated in Figure 4. His insight was that if the megathrust is linked to the top of basement, then the initial temperature of the top of the slab and the subducting crustal material below it will be elevated by the geothermal gradient times the sediment thickness. He developed a series of three simple thermal models produced by projecting 2D, analytical, thermal models of subducting slabs (England and Wilkins, 2004; England et al., 2004; England and Katz, 2010) along strike with variable initial temperatures. The models all predict a strong warping of temperature with higher temperatures in sections of the slab that started subduction with a higher initial temperature at the top of the subducting crust. Additional support given by Bauer (2014) for his conjecture was the exceptionally shallow depth of the Wrangell volcanoes. We illustrate this graphically in Animation 3 and Figure 3. Furthermore, Preece and Hart (2004) show that many Wrangell volcanics displayed adakitic geochemical signatures consistent with slab melting. Bauer’s (2014) model demonstrates that elevation of the initial temperature of the subducting slab must strongly warp the temperature profile in the subducting crust linked to the Yakutat microplate. This inference is analogous to the effects of slow subduction of young oceanic crust that has long been used to explain the low rates of intermediate depth earthquakes in Cascadia (e.g., Wiens, 1993). A second process, sediment subduction, could amplify the thermal effect. Pavlis et al. (2012b) used mass balance estimates from the thrust belt together with regional geology to support the hypothesis of Fruehn et al. (1999) that the Yakutat collision has entrained large volumes of forearc crust and sediments. Entrainment of these more silicic rocks along the subduction interface, together with temperature effects, could explain the seismicity variations. These interpretations should be testable as more data accumulate from the Earthscope deployment.

Our synthesis model suggests that the mantle lithosphere is moving differently from the upper part of the crust in much of this region. Several lines of evidence support this conclusion. First, measured GPS velocities west of Mount St. Elias are rotated relative to any feasible rigid plate model for motion of the Yakutat microplate (Fig. 7). In contrast, east of the Seward Glacier, which we here infer to be the top of the middlebuster structure, the measured GPS velocities are well matched by the North America–Yakutat pole estimated by Elliott et al. (2010). This suggests a major change in crust-mantle interactions across the top of the middlebuster structure. Second, there is unambiguous evidence from coastal uplift that slip in the great Alaska earthquake of 1964 extended eastward at least as far as the Suckling Hills, which are well within the core of the Yakutat collisional zone (Plafker, 1969; Chapman et al., 2014). Furthermore, Shennan et al. (2009) argue that earlier megathrust events ∼900 and ∼1500 years ago were even larger, with a continuous rupture of the megathrust to Icy Bay. Hence, there is little doubt that the megathrust, which is synonymous here with our top of slab surface, is continuous from the Aleutians to Icy Bay. Our model suggests the megathrust extends under Mount St. Elias, where it is mapped as the Malaspina fault. We suggest here that the eastern limit of the megathrust is also the western plow of the middlebuster (Fig. 8). Finally, the GPS data from the region west of Mount St. Elias and block model inversions described by Elliott et al. (2013) (Fig. 7) have implications for crust-mantle interactions. Figure 7 demonstrates that GPS velocity vectors are not consistent with any of the three rigid plate models used in this paper. Since two of those models are end members, it means the large-scale plate models do not describe the motion of the crust west of Mount St. Elias. Elliott et al. (2013) found these data could be fit by motion of a block of crust combined with elastic deformation on low-angle faults. As Figure 7D shows, the slip vector estimated on these low-angle surfaces matches the directions predicted from block motion of the Yakutat microplate by Elliott et al. (2010). On the other hand, as noted earlier, these low-angle surfaces are not consistent with the top of slab model inferred from seismic data. Either the top of slab surface is not the megathrust, or the GPS data are responding to a different process.

We suggest that the disconnect between the top of slab surface inferred from seismic data and that used to fit GPS data is a reflection of ductile deformation of the lower crust. Specifically, it can be explained by the existence of an orogenic wedge overlying the subducting megathrust. That wedge is probably a composite crustal section with a ductile lower crust sandwiched between the megathrust and the brittle upper crust. That interpretation is further supported by models of the system developed by Koons et al. (2010). Such a ductile wedge allows the brittle upper crust to move oblique to plate motion as observed in the GPS velocities (Fig. 7). We further suggest that the depth of this brittle-ductile transition may be controlling the depths of the dislocation fault surfaces modeled by Elliott et al. (2013) and illustrated in Figure 7D. Moreover, the coincidence between the model depths of Elliott et al. (2013) with the top of a higher velocity region imaged by Christeson et al. (2013) provides additional support for this conclusion. The map outline of this hypothesized crustal wedge is triangular with the apex of the triangle at Mount St. Elias and the widest part of the wedge near Prince William Sound. Westward flow may be induced in this wedge by the eastward thinning of the wedge as it terminates against the western plow of our proposed middlebuster structure. The resulting flow may be driving the westward deflection of the GPS velocities seen in Figure 7. This conclusion is an illustration of a process that has long been recognized in geologic studies of ancient orogenic systems but rarely documented in active orogens. Oldow et al. (1989) coined the term “orogenic float” to emphasize that the kinematics of shallow crustal deformation are generally distinct from deep crustal flow and that both commonly diverge from the plate motions that drive the deformation. In the case of the southern Alaskan orogeny, the divergence is enhanced because of the crustal thickening created in the tectonic corner at Mount St. Elias.

STEEP was a collaborative project funded by the Continental Dynamics Program of the National Science Foundation. Instrumentation to support the passive seismic components of STEEP was provided by the Incorporated Research Institutions for Seismology, a facilities program of the National Science Foundation. Marine seismic data survey MGL0814 (STEEP; Christeson et al. [2010]; Gulick et al. [2008]) is available at the University of Texas Institute for Geophysics Academic Seismic Portal (http://www.udc.ig.utexas.edu/sdc/), with processed navigation files available on the Lamont-Doherty Earth Observatory Academic Seismic Portal (http://www.marine.geo.org/portals/seismic/). This research was supported in part by the Geophysical Institute, University of Alaska Fairbanks, and by the Office of Alaska State Seismologist. Earthquake data were provided by the Alaska Earthquake Center, GI-UAF. Sincere thanks are given to Sarah Roeske and an anonymous reviewer for slogging through the initial version of this paper. We experimented with an html format that both reviewers helped us realize was a huge mistake. Sincere thanks to Craig Jones and the staff of Geosphere for pointing us to the new technology of 3D pdfs, which provided a huge improvement for understanding 4D concepts compared to animations we used in the first version of this paper.

Independent motion of the Yakutat microplate with respect to the Pacific requires strain in the mantle lithosphere. One way that can happen is by bending of the subducting lithosphere. In this appendix, we quantify the strain that can be achieved through bending for the specific geometry of the Yakutat microplate.

Shortening is then computed as δL = L_w – L.

The exact form of the true deflection function, w(x), is not known; but Figure A.2 shows results for three analytic forms: w(x) = a₁x, w(x) = a₃x³, and w(x) = a₄x⁴. The linear model has a simple analytic solution that follows from basic trigonometry, δL = L(1 – sec ϕ), where ϕ is the dip angle. We used that formula to validate a numerical implementation used for the other two forms. The x³ and x⁴ curves are appropriate models because of well-known analytic solutions for bending of an elastic beam for a point load and uniformly distributed load, respectively (e.g., Turcotte and Schubert, 2014). All three curves are computed assuming L = 500 km, which is approximately the maximum extent of Yakutat lithosphere in the direction perpendicular to plate motion. In all cases, the curves were generated by adjusting the coefficients (a₁, a₃, or a₄) to make the total vertical deflection, which defines the x-axis in Figure A.2, match the assumed value.