We compute high-resolution seismic images from scattered wavefield to detect discontinuities beneath the High Lava Plains (HLP), using data recorded at a dense broadband array. Our images of the HLP and surrounding regions reveal (1) a prominent Moho discontinuity with varying depth, with thinnest crust of 35 km beneath the volcanic track, and thickened crust of ∼45 km beneath the Owyhee Plateau (OP); (2) distinct intracrustal velocity reversals beneath regions of pre-2.0 Ma volcanism and within the OP; and (3) intermittent negative velocity discontinuities in the uppermost mantle beneath regions of Holocene volcanism and volcanic centers near Steens Mountain and Newberry volcano. These features exhibit remarkable similarity with those seen in the surface wave tomography and Ps receiver functions. We fail to find evidence for a ubiquitous regional lithosphere-asthenosphere boundary (LAB). In concert with petrological constraints on the equilibration depths of primitive basaltic melts, our results suggest that the present-day HLP mantle lithosphere is thin or absent, perhaps a consequence of episodes of extensive mantle inflow, lithospheric extension, and possibly melting induced by rapid slab rollback and trench retreat. It remains possible, however, that strong E-W seismic anisotropy reported across this region may reduce the effective S-wave velocity contrast to render the LAB less detectable. In contrast, the Owyhee Plateau exhibits a clear LAB, consistent with it being a block of older preexisting lithosphere. Our images demonstrate the complexity of mantle dynamics in the Cascadian back-arc and the close casual link between subduction-related processes and the origin of HLP volcanism.
 The High Lava Plains volcanic lineament of central and eastern Oregon remains among the more puzzling enigmas associated with the intense phase of tectonomagmatic activity that has characterized the Cascadian back-arc for nearly the past 20 Ma, from mid-Miocene time to the present day (Figure 1). While the back-arc has seen many prior episodes of subduction-related volcanism, the massive outpourings of flood basalts that began ca. 16.6 Ma in the Steens Mountain (SM) region near the Oregon-Idaho-Nevada border [Brueseke et al., 2007] opened a major new chapter in the region's long history of tectonomagmatism. Following the initial outpouring of flood basalts at Steens, activity migrated rapidly northward along a series of N-S trending rifts on the Oregon-Idaho border to settle in the Washington/Oregon/Idaho border region from which the main volume of Columbia River basalts (CRB) were erupted. Flood volcanic activity waned markedly after 16 Ma and by 14 Ma had virtually ceased.
 Silicic volcanism roughly contemporaneous with the early Steens flood basalts is first observed in northern Nevada along the southwestern edge of the Owyhee Plateau [Brueseke et al., 2007]. By 12 Ma, focused silicic volcanism had begun to follow two temporally migrating paths away from the Owyhee Plateau area, but in opposing directions. One track formed the Snake River Plain-Yellowstone (SRP/Y) volcanic terrane [Pierce and Morgan, 2002; Smith et al., 2009], with silicic volcanism propagating to the northeast at a rate and in a direction consistent with North American plate motion. The second time-progressive track formed the High Lava Plains volcanic lineament, but with a NW trend toward the Holocene Newberry volcano in the eastern Cascades, sharply oblique to the direction of plate motion [Jordan et al., 2004]. While the silicic volcanism along both tracks exhibits contemporaneous age progression as revealed by 40Ar/39Ar dating [Jordan et al., 2004], widespread basaltic volcanism in both provinces occurs with no temporal or spatial correlation along the length of the tracks well into the Holocene [Draper, 1991].
 The dynamics behind these many interrelated tectonomagmatic events remains a matter of active debate. Even the plume model for the origin of the flood basalts and the SRP/Y hotspot track [e.g., Camp and Ross, 2004] has been challenged in the recent literature [e.g., Faccenna et al., 2010; James et al., 2011; Liu and Stegman, 2012], whereas subduction models provide plausible alternative explanations to the plume hypothesis. Alternative hypotheses and speculations for the origin of other features in the back-arc abound, including Basin and Range extension and rotation [Cross and Pilger, 1978], Cascadia back-arc extension [Christiansen and McKee, 1978], asthenospheric inflow induced by the rollback of the subducting Juan de Fuca plate [Carlson and Hart, 1987], and lithospheric delamination [Hales et al., 2005]. None of these explanations has thus far shed much light on the processes that may have formed the HLP volcanic lineament that is the focus of this paper.
 The puzzle of complex intraplate volcanism in the Pacific Northwest has inspired wide-ranging seismic studies drawn from the extraordinary data set provided by the EarthScope USArray Transportable Array (TA). The regional focus on the velocity and discontinuity structure of the crust and upper mantle beneath this complex region is variously based on teleseismic body waves [Burdick et al., 2008, 2009; Roth et al., 2008; Obrebski et al., 2010; James et al., 2011; Sigloch, 2011; Sigloch et al., 2008], surface waves [Lin et al., 2011; Pollitz and Snoke, 2010; Wagner et al., 2010], ambient noise correlation [Gao et al., 2011; Hanson-Hedgecock et al., 2012; Wagner et al., 2012; L. S. Wagner and M. D. Long, Distinctive upper mantle anisotropy beneath the High Lava Plains and Eastern Snake River Plain, Pacific Northwest, USA, in press, Geochemistry Geophysics Geosystems, 2013], and receiver functions [Eagar et al., 2011; Gilbert, 2012; E. Hopper et al., The lithosphere-asthenosphere boundary and the tectonic and magmatic history of the northwestern United States, submitted to Earth and Planetary Science Letters, 2013]. While these studies have yielded important new information on regional 3-D velocity structure, they reveal little, for example, about the nature and depth of the lithosphere-asthenosphere boundary (LAB), or discontinuities that may be associated with melt zones in the uppermost mantle.
 The 118-station HLP broadband experiment (Figure 1b) produced a data set with sufficient density and geographic extent to make high resolution, essentially unaliased, studies of discontinuity structure practical. Eagar et al.  used receiver function analysis to produce detailed maps of crustal thickness, crustal velocities, and Poisson's ratios across the HLP region, and in the analysis that follows we capitalize on that same high-density data set from the HLP broadband seismic experiment to image discontinuity structures within both the crust and the uppermost mantle. Our results are remarkably consistent with uppermost mantle features seen in a range of other seismic studies of the greater High Lava Plains region.
 In this paper, we tackle directly the question of the enigmatic origin of the HLP. Our studies are motivated and guided by recent work, much of it still unpublished, in which a preliminary understanding of the dynamical processes central to HLP has begun to emerge from the joint study of physical modeling experiments coupled with measurements of mantle fabric [Long et al., 2012]. As we shall develop in the course of this paper, the dynamical restructuring of the Farallon/Juan de Fuca subducting plate resulting from the southerly truncation of the subduction zone by the northward migration of the Mendocino Triple Junction, coupled with a rapid increase in causally related trench retreat and slab rollback that took place around the time of flood basalt volcanism, provides a framework in which to interpret the seismic images that we obtain of discontinuity structure beneath the HLP volcanic track. Our high-resolution images derived from 2-D teleseismic migration of the dense HLP data set reveal detailed lithospheric structure beneath the HLP and provide important constraints both on the tectonomagmatic processes that drive the volcanism in this region and on the low-velocity structures in the mantle associated with Holocene volcanism within the High Lava Plains province.
 To detect discontinuity structure in the subsurface material properties, we use an imaging technique that migrates scattered waves in the coda of teleseismic P-waves recorded at dense seismic arrays [Bostock et al., 2001; Rondenay et al., 2005]. This technique considers both forward-scattered and backscattered wavefields generated by volumetric perturbations in P-(δα/α) and S-wave velocities (δβ/β) within a smoothly varying, isotropic, 1-D background medium. In this section, we briefly describe the data preprocessing steps and the imaging method used in this study.
2.1. Data Preprocessing
 Careful preprocessing of original total wavefield data to extract scattered wavefield is critical to the success of the imaging algorithm. We employ a multichannel approach similar to that of Bostock and Rondenay  to isolate the scattered wavefield from the total recorded wavefield. The advantage of such multichannel scheme over single station-based deconvolution lies in more robust estimation of incident wavefield; hence, more stable deconvolved waveforms to higher frequencies. The approach involves the following steps [see Rondenay et al., 2005, for details]: (1) transform recorded three-component data from N-E-Z to upgoing P-SV-SH using the free-surface transfer matrix [Kennett, 1991]; (2) align the wavefield relative to the incident P wave by multichannel cross correlation [VanDecar and Crosson, 1990]; (3) estimate the incident P wavelet (source-time function) by eigenimage decomposition [Ulrych et al., 1999]; and (4) construct a source-normalized scattered wavefield by subtracting the estimated P wavelet from the P wavefield and deconvolving the P wavelet from the residual P, SV, and SH wavefield.
 Note that in the deconvolution step we adopt an improved deconvolution algorithm developed by Pearce et al. , in which an optimal damping parameter (i.e., water level) is independently determined for each station component, in contrast with the conventional way using ad hoc, uniform damping for all components. Following Pearce et al. , we chose the smallest damping value such that unstable oscillations are restricted below a prescribed energy threshold (e.g., ∼0.01% of the undamped energy). Alternatively, if this criterion cannot be reached, the damping is fixed at 5% of the peak in the source wavelet's power spectrum. This procedure alleviates over-damping and excessive low-pass filtering of the scattered wavefield, and yields more stable scattered signals than those obtained with uniform damping.
2.2. Teleseismic Migration
 The migration approach used in this study considers the interaction of an incident, planar teleseismic wavefield from arbitrary back azimuths with 2-D structure (local perturbation), assuming high-frequency (i.e., ray theoretical) single scattering, within a smoothly varying 1-D reference model. The inversion/back projection approach is formulated based on generalized Radon Transform (GRT) [Bostock et al., 2001] and is thus commonly referred to as the 2-D GRT inversion [Rondenay et al., 2005]. The imaging process involves construction of weighted diffraction stacks over all sources and receivers to yield estimates of scattering potential at a given imaging point in the subsurface, with the weights determined by the analogy between the forward-scattering equation and the GRT [e.g., Miller et al., 1987]. The scattering potential is then linearly inverted for velocity perturbations at this point. Finally, the inverse problem is solved for all points in the model space to generate a 2-D image of velocity perturbations. A full theoretical derivation of the 2-D GRT inversion is described in Bostock et al. .
 The 2-D GRT inversion operates on a series of individual forward-scattered and backscattered modes, including: (1) forward scattering of incident P wave to S wave (Pds); (2) free-surface-reflected P wave to backscattered P wave (Ppdp); (3) free-surface-reflected P wave to backscattered S wave (Ppds); and the free-surface-reflected S wave to (4) backscattered Sv (Psds|v) and (5) backscattered Sh (Psds|h). The contribution and weighting of each mode is calculated based on analytical expression for the travel times and amplitudes of the relevant combination of incident and scattered waves [Rondenay et al., 2001]. Then the individual modes are combined to produce the final composite image.
 The resolving power of 2-D GRT depends largely on the frequency content of the scattered signal and on the source/receiver distribution [Bostock et al., 2001; Shragge et al., 2001; Rondenay et al., 2005, 2008; Rondenay, 2009]. The maximum volume resolution is approximately equivalent to a quarter of the wavelength (λ/4) of the backscattered signal, which offers resolution by roughly a factor of two higher than that of the forward-scattered signal (λ/2), due to the longer travel times of the former at a given interface depth [Rondenay et al., 2005]. In this study, the high cut-off frequency is 0.5 Hz, yielding maximum volume resolution on the order of 2–3 km for structure in the lower crust and upper mantle.
 The assumptions inherent to the migration method, in particular the 2-D geometry of the imaging target and the isotropy of material properties, pose limitation on the robustness of the resulting migration image. That is, the degree to which the actual structure in a given study area satisfies these assumptions determines how robust the resulting image can be. Therefore, a critical measure in the imaging is to determine an optimal projection line that represents the principal direction, or azimuth, of the 2-D structure to be imaged, along which we project the stations and generate the migration image. Previous HLP studies already show that the main lithospheric structures in the HLP region align approximately with the general direction of the volcanic track [Wagner et al., 2010; Eagar et al., 2011]. To determine the optimal projection line for migration, we project the HLP stations on a series of straight lines with varying azimuths passing through the center of the HLP array, yielding corresponding GRT images. The optimal azimuth is the one that returns the most focused signals of lithospheric structure, e.g., Moho, with least distortion and migration artifacts such as migration smiles. This exercise yields an optimal projection direction of N115°E ± 10°, which matches well with the HLP volcanic trend and thus the main direction of the HLP array.
 It is worth noting that the 2-D GRT inversion used in this study has been previously applied only to imaging of subduction zones [Suckale et al., 2009; MacKenzie et al., 2010; Rondenay et al., 2010; Pearce et al., 2012], due to the method's advantage in better resolving dipping and/or irregular, layered structure over other multichannel approaches. In the context of our study, while the HLP region is close to the Cascadia subduction zone, previous seismic tomographic studies show no evidence of any high velocity, dipping slab present in the uppermost 100 km of the HLP mantle [e.g., Wagner et al., 2010; James et al., 2011, and the references therein]. As such, this study presents a first attempt to apply 2-D GRT to a tectonic region where no major dipping structure is expected to exist within the imaging depth range.
 We analyze data collected from the seismic experiment as part of the multidisciplinary HLP project [Carlson et al., 2005]. The HLP seismic array consisted of 118 broadband seismic stations (Figure 1b), with average station spacing of 15–20 km, operating between 2006 and 2009 and spanning areas across eastern Oregon, northern Nevada, and western Idaho. The majority of instruments were deployed along the HLP volcanic track, while the rest were installed in the immediate vicinity of the track. The array geometry is designed to yield dense data coverage for maximum resolution along the HLP track, with the surrounding stations providing broader regional control.
 To select useful events for this study, we use the following criteria [see, e.g., Rondenay et al., 2001, 2005; Suckale et al., 2009; Pearce et al., 2012]: (1) epicentral distances between 30° and 90° from the center of the array, to avoid triplicated phases from the mantle transition zone and diffraction at the core-mantle boundary; (2) magnitude (mb) greater than 5.5; (3) an incident P-wave arrival that can be identified across the entire array; (4) minimum contamination of the coda from secondary phases (e.g., PcP, PP); and (5) no overlaps with foreshocks and aftershocks. As a result, we obtain over 70 high-quality events that are deemed amenable to 2-D GRT application (Figure 2).
 Here we present the 2-D GRT images of the High Lava Plains crustal and uppermost mantle structure (Figure 3). The composite images are computed by simultaneous inversion of all selected events with combined contribution of different scattering modes, along the optimal projection line (A-A′ in Figure 1b). The images reveal velocity perturbations of P (δα/α) or S (δβ/β) waves. The transition in the color-scale marks the presence of velocity discontinuities: low-to-high downward (positive) velocity contrast is denoted by red-to-blue; high-to-low (negative) velocity contrast is denoted by blue-to-red.
 As in previous applications of this migration method, we focus our interpretation on the S-velocity (δβ/β) profile, as δβ/β images have shown to be more robust than the δα/α images [Rondenay et al., 2005; Rondenay, 2009]. The δβ/β images prevail because (1) they are based on a series of scattered modes, whereas the δα/α image relies on only one backscattering mode (Ppdp) and (2) the S-scattered waves are more accurately separated from the full wavefield than P waves; this is especially the case for signals scattered at horizontal discontinuities, which are dominant features beneath the HLP.
 From the δβ/β composite image in Figure 3, we make the following main observations:
 1. A prominent positive discontinuity extends across the entire profile at an average depth of 40 km, which represents the Moho discontinuity. The Moho is at ∼40 km depth immediately east of the Cascades, thins to ∼35 km depth beneath the HLP track, gently dips southeasterly beneath the Jordan Crater, and flattens at ∼45 km depth beneath the Owyhee Plateau. The Moho across the HLP appears to be strikingly sharp and flat.
 2. In the SE portion of the profile (distance 400–520 km in Figure 3), a pronounced subhorizontal negative discontinuity at ∼60 km depth underlies a ∼15 km-thick high velocity zone over the extent of the OP uppermost mantle from the edge of our image to below the Jordan Crater. From there, weaker signals of negative discontinuity appear to continue and dip toward northwest, flattening and terminating at ∼75 km depth beneath Diamond Crater.
 3. Beneath the sharp and shallow Moho of HLP, we see a weak negative discontinuity at ∼50 km depth, much less prominent than similar features observed in the uppermost mantle of the adjacent Owyhee Plateau and Newberry. Apart from this weak boundary and the low velocity packet at 75 km depth mentioned above, no other coherent structure stands out in the HLP uppermost mantle.
 4. Beneath the Newberry volcano region, we observe a negative discontinuity, which resides at 55 km depth at distance of 30 km and gently shallows to 45 km depth at distance of 120 km. This discontinuity appears to dip abruptly, and perhaps continues to connect with the weak negative discontinuity under the HLP mentioned above.
 5. Within the crust, we identify two notable features. We observe a high velocity zone between ∼15 and 25 km depth but without a clearly defined boundary in the OP mid-crust. We also image a weaker velocity reversal in the HLP crust at 20–25 km depth that extends from ∼50 km east of the Newberry region to Diamond Crater. This gradational discontinuity resumes beneath Jordan Crater and appears to extend into the western margin of the OP.
 To detect possible structure variations perpendicular to the projection direction, we investigate back azimuthal contributions to the seismic images by constructing migration profiles using events from restricted back azimuthal quadrants, as shown in Figure 4. Each back azimuthal quadrant contains a significant number of events (>10) with incident waves illuminating the subsurface from directions both along and oblique to the HLP track, allowing us to isolate contributions of scattered waves coming from different angles. In general, obliquely incident waves such as those used in SW and NE quadrants strike the imaging target at a perpendicular offset from the array; the offset increases with increasing target depth [Rondenay et al., 2005, 2010]. From previous analytical analysis [Rondenay et al., 2005], the maximum perpendicular offset of scattered waves contributing to the SW and NE quadrant images is approximately 20 km for targets at 30 km depth, and 40 km for targets at 80 km depth. This exercise shows that the depth variations of Moho are well resolved and remarkably consistent in all four profiles and with the composite image. The coherence indicates that the Moho in this region is dominantly 2-D with most variations parallel to the HLP track. Significant coherence is also observed for the negative velocity discontinuities beneath the Newberry volcano and the Owyhee Plateau. Beneath the HLP, despite the fact that the images are somewhat noisier, we are still able to trace consistent signals of the weaker negative gradients in the mantle. We discuss the implications of these structures in the following section.
 Our results show several prominent subhorizontal seismic discontinuities in the crust and uppermost mantle to depths of ∼75 km. Below that depth, we do not identify any coherent discontinuities, and migration artifacts (e.g., crustal multiples) become problematic. In the discussion that follows, we focus on the regional crustal structure, an analysis of the Newberry caldera, and variations in regional lithospheric structure to a depth of 75 km. We compare our results with previous seismic studies as well as with recent petrological constraints based on the minimum equilibration depth of primitive Holocene basalts (Figure 5). The correlation between these observations sheds light on the mantle dynamics and the evolution of the HLP lithosphere.
5.1. Differences in Regional Crustal Structure
 Variations of crustal thickness across the region imaged in this study are highly consistent with those determined from receiver function common-conversion-point (CCP) stacking and H-k depth estimates [Eagar et al., 2011; Figure 5b], Kirchhoff migration images [Liu and Levander, 2013], as well as from ambient noise tomography [e.g., Moschetti et al., 2010; Hanson-Hedgecock et al., 2012; Wagner et al., 2012]. The variations correlate well with geological province boundaries, including crustal thickness of ∼40 km beneath the general region of the Newberry caldera, thinner crust (∼35 km) along the HLP volcanic track, and thickened crust (∼45 km) under the Owyhee Plateau. The intracrustal complexity observed in our images (Figure 3) coincides with previous seismic observations as well. The diffuse negative velocity discontinuity in the HLP mid-crust (distance 100–270 km) is also seen in CCP stacked receiver functions as a low velocity pinch zone [Eagar et al., 2011], as well as in ambient noise and joint ambient noise/surface wave tomography [Hanson-Hedgecock et al., 2012; Wagner et al., 2012]. Notably, both tips of this low velocity zone taper out beneath active Holocene volcanic centers of Newberry caldera and Diamond Crater, respectively, but further study of detailed receiver function modeling will be needed to better delineate the character of this low velocity zone.
 We also note that Eagar et al.  observe high Poisson's ratios (>0.29) along the zone of thinnest crust (distance 100–300 km), indicative of a mafic crustal composition and consistent with evidence for a generally primitive crustal section [e.g., Jordan et al., 2004; Hart et al., 1984]. Interestingly, the regions with the highest Poisson's ratios (∼0.32) appear to flank both sides of the HLP main volcanic track, suggesting higher—possibly near-melting—temperatures in the crust beneath those regions [Mueller and Massonne, 2001]. These observations led Eagar et al.  to conclude that the HLP track represents a magma “drainage” zone where basaltic melt was transported to the surface in a focused fashion, perhaps facilitated by major faults along the Brothers Fault Zone [e.g., Wagner et al., 2012]. While our image covers only a 2-D cross-sectional area away from the high Poisson's ratio zones, our results suggest the presence of a low-velocity zone in the central HLP where Holocene volcanism is absent, a finding not inconsistent with Eagar's melt drainage zone hypothesis.
 The Owyhee Plateau, which straddles southeastern Oregon, southwestern Idaho, and northern Nevada (the so-called ION region), appears to be an isolated block of older continental lithosphere characterized by thickened crust with low Poisson's ratio [Eagar et al., 2011] and significantly higher intracrustal S-wave velocities than those that characterize the continental crust further west [Hanson-Hedgecock et al., 2012; Wagner et al., 2012]. The transition from shallow Moho of the HLP to a deeper Moho beneath the Owyhee Plateau occurs along our 2-D transect around the area of Jordan Crater, which marks the approximate western margin of the Owyhee Plateau. The irregular high velocity anomaly we image in the Owyhee mid-crust is similar to that observed in ambient noise tomography [Hanson-Hedgecock et al., 2012; Wagner et al., 2012; Figure 5d], although in their models the high velocity zone spans most of the Owyhee Plateau crust. While our 2-D GRT imaging is not fully capable of resolving the extent of the high velocity layering in the Owyhee crust given its depth and location at the eastern edge of the imaging array, we note that there are indications of high velocity layers, as shown in Figure 3.
5.2. Newberry Caldera
 The upper mantle beneath Newberry caldera region exhibits a mantle wedge structure suggestive of a melt-producing subduction environment. Surface wave tomography [Wagner et al., 2010, 2012; Figure 5d] reveals a confined low velocity zone in the mantle wedge above the descending Juan de Fuca plate beneath Newberry. The irregular negative discontinuity shown in Figure 3 that starts beneath the Newberry volcano and which dips to the southeast, matches well in its spatial extent with the upper boundary of the surface-wave imaged low velocity zone, a region that in this part of the subduction zone is plausibly a manifestation of warm, probably hydrated, mantle beneath the volcanic arc. We note that the surface projection of the eastern edge of this negative velocity feature (distance 115 km) coincides roughly with the surface expression of a pattern change in rhyolitic volcanism that occurred around 5 Ma [e.g., Wagner et al., 2010]. While difficult to verify, the rather abrupt change in the nature of the rhyolitic volcanism ca. 5 Ma could reflect a transition between the westward migrating HLP volcanic track and the on-going Cascadian arc volcanism associated with Newberry. Beneath the Newberry caldera itself, the depth of the negative discontinuity coincides with the basalt equilibration depth, consistent with the possibility that it could be the mechanical lithosphere-asthenosphere boundary [e.g., Till et al., 2013].
5.3. Regional Variations in Lithospheric Structure
 Regional variations in lithospheric structure across the greater High Lava Plains and surrounding regions provide important insight into the dynamics of lithospheric modification and evolution tied to regional trends of tectonomagmatism. The HLP crust and upper mantle exhibit two compelling characteristics: (1) the Moho discontinuity is shallow, sharp, and flat; and (2) the mantle lithosphere is thin or perhaps absent, but in either case is not detectable in our GRT images. In contrast, the Owyhee Plateau Moho steps to increased depths over lateral distances of <20 km, and the base of a mantle lithosphere is clearly evident at depths of ∼60 km.
 The surprisingly featureless HLP uppermost mantle suggests the lack of a prominent mantle lithosphere that is seismically distinct from the asthenosphere. In our Ps migration images, we do not see coherent scattered signals derived from velocity contrasts indicative of the presence of an LAB. This result is consistent with the general absence of strong velocity gradients with depth as imaged by Wagner et al. . We note, however, that a recent study using Sp receiver function stacking appears to detect a decrease in velocity with depth in the range of 50–75 km beneath the HLP (Hopper et al., submitted manuscript, 2013), suggesting the presence of a relatively shallow LAB. These seemingly conflicting results could be due to a variety of factors, including a difference in the dominant periods of the seismic waves used to illuminate the velocity gradient and/or seismic anisotropy. For instance, the dominant period for both forward-scattered and backscattered Ps phases in this study is ∼5 s, whereas the dominant period for Sp converted phases is ∼10 s in the Hopper et al. (submitted manuscript, 2013) study. The differences in these dominant periods could suggest that the LAB beneath the HLP is a more gradational, rather than sharp, transition in S-wave velocity. Indeed, in a region such as the HLP where thermal effects play a major role in the tectonic evolution, the LAB could be more gradational. Finally, azimuthal seismic anisotropy could play a role in masking a sharp LAB from Ps imaging. This effect may be possible due to the very strong and uniform SKS splitting that exhibits E-W fast polarization directions (L. S. Wagner and M. D. Long, submitted manuscript, 2013), the effect of which would be to reduce the S-wave velocity contrast across the LAB seen by Ps phases subparallel to this fast direction. Further, new surface wave results by Wagner et al.  suggest that peak azimuthal anisotropy beneath the HLP appears to occur at ∼50–60 km depth, coincident with the depth of a velocity reversal with depth observed by Sp phases, but not by Ps phases. Further study beyond the scope of this paper is needed to sort out these effects.
 In contrast, the Owyhee Plateau lithosphere and sublithosphere are well imaged by multiple methods. While surface wave tomography shows that the Owyhee Plateau is underlain by a relatively thin layer of high velocity lithospheric mantle, it also reveals a much more striking low velocity anomaly in the uppermost mantle [Wagner et al., 2010]. Laterally, the Owyhee low velocity zone coincides closely with that of our imaged negative velocity discontinuity, suggesting that the velocity reversal represents the upper boundary of the surface-wave imaged low velocity zone. The combined surface wave and GRT observations strongly indicate that the Owyhee Plateau possesses a characteristic, albeit thin, continental lithosphere, with a normal velocity continental crust and a flat Moho underlain by a thin lid of high velocity mantle lithosphere. The velocity reversal we observe beneath the lithospheric block thus marks a sharp lithosphere-asthenosphere boundary. We note that while very little volcanism has occurred within the Plateau itself, abundant volcanism occurs around the margins, including a cluster of primitive basaltic samples in the vicinity of Jordan crater that exhibit minimum magma equilibration depths of 45–65 km [Till et al., 2013; Figure 5]. This northwestern margin of the Owyhee Plateau is the region in which the Moho transitions from shallow beneath the HLP to deeper beneath the Plateau, suggesting the possibility that the strong gradient in Moho topography may serve as a channel for migration of basaltic melts into the crust.
5.4. Lithospheric Evolution of the High Lava Plains
 A broad range of conceptual models for the origins of Pacific Northwest intraplate tectonomagmatism exist, including a mantle plume [Camp and Ross, 2004; Jordan et al., 2004], lithospheric delamination [e.g., Camp and Hanan, 2008; Wells and Hoisch, 2008], and asthenospheric inflow from the east and south due to Juan de Fuca slab rollback [e.g., James et al., 2011; Long et al., 2012]. The first two models—a mantle plume and lithospheric delamination—are difficult to reconcile, given the structures we observe and the known history of regional subduction. Further, an upwelling mantle plume would be impeded by the intervening Juan de Fuca plate itself and the regional mantle flow field it induces. Lithospheric delamination, on the other hand, is typically associated with regional extension and mantle downwelling [e.g., Zandt et al., 2004; Hales et al., 2005; West et al., 2009]. Models including strong mantle downwelling would predictably result in significantly reduced azimuthal anisotropy in the region [e.g., the Great Basin; West et al., 2009]. To the contrary, however, the HLP region exhibits the largest and most coherent SKS splitting values of anywhere in the western United States [Long et al., 2012], as well as strong and shallow azimuthal anisotropy measured by surface waves [Wagner et al., 2012]. It remains possible that, however, some amount of destabilization of the HLP lithosphere (through thermal weakening and/or mechanical removal) could play a role in regional tectonomagmatism.
 The seismic observations we report here fit well with the conceptual model proposed by Long et al.  for the mantle dynamics of the Pacific Northwest. In their model, a combination of subduction-related processes, including slab rollback, trench retreat, back-arc extension, and slab steepening, induce complex patterns of asthenospheric flow that are consistent with patterns of HLP volcanism. The long history of back-arc extension in the greater HLP region explains the presence of a thin crust and a flat Moho. The widespread volcanism in the region may well have induced melting and/or deformation in the lower crust [Mooney and Meissner, 1992], which eventually evolved and crystallized to form a modified, sharp Moho boundary. This Mooney and Meissner  conceptual model has been proposed previously to explain the lack of Moho relief in the Kaapvaal craton in southern Africa [James et al., 2001], and the evolution of the extensional Basin and Range province in western US [Mooney and Meissner, 1992].
 In this paper, we have presented high-resolution scattered-wave migration images that reveal detailed discontinuity structures in the lithosphere of the High Lava Plains and surrounding regions. We observe varying Moho topography across southern and eastern Oregon, with thin crust of ∼35 km thick beneath the HLP volcanic track transitioning to ∼45 km under the Owyhee Plateau. The Moho gradient at the margin of the Owyhee Plateau may facilitate migration of basaltic melts to HLP crust. Our images also suggest the presence of intracrustal velocity discontinuities, the extent of which lie between, but not beneath, active Holocene volcanic centers at surface. In the uppermost mantle, velocity reversals beneath Newberry Caldera and the Owyhee Plateau indicate the presence of lithosphere-asthenosphere boundary at depths of 50 km and 65 km, respectively. On the other hand, the absence of such signals under the HLP volcanic track suggests a thin or perhaps absent mantle lithosphere. In this regard, however, we note that strong azimuthal seismic anisotropy observed in this region could act to hinder the detection of S-wave velocity contrasts at LAB depths. Our results are remarkably well correlated with seismic observations based on joint inversion of ambient noise and Rayleigh waves and CCP stacked Ps receiver functions. On the basis of these combined seismic results, we conclude that the most plausible driving mechanism for generating intraplate volcanism in this region is related to asthenospheric inflow driven by the post 20 Ma increase in the rate of slab rollback and trench retreat.
 Data used in this work come from the High Lava Plains seismic experiment, funded by NSF Continental Dynamics EAR-0507248 (MJF) and EAR-0506914 (DEJ). CWC was supported in part through a Carnegie Harry Oscar Wood Fellowship and Taiwan's National Science Council (NSC) under grant NSC 101–2611-M-002–020-MY2. We thank Jenda Johnson, Steven Golden, and the many people who helped install and service the 118+ stations in this deployment. We also thank the many ranchers and landowners who generously and freely agreed to host seismic stations for the HLP experiment on their property. Special thanks to Fred Pearce for numerous insightful discussions on the data processing and imaging techniques, and to Christy Till and Rick Carlson for discussions on the petrological aspect of the HLP structure. We thank Editor-in-Chief Thorsten Becker, David Eaton, and an anonymous reviewer for constructive comments.