We use data from the 118-station High Lava Plains (HLP) seismic experiment together with other regional broadband seismic data to image the 3D shear wave velocity structure in the Pacific Northwest using ambient noise tomography. This extensive data set allows us to resolve fine-scale crustal structures throughout the HLP area in greater detail than previous studies. Our results show 1) a high velocity cylinder in the crust and average velocities in the upper mantle beneath the Owyhee Plateau; 2) a mid-crustal high velocity anomaly along the Snake River Plain that also extends south into Nevada and Utah; 3) a low velocity anomaly directly beneath Yellowstone throughout the crust; and 4) low velocities beneath the HLP both in the crust and uppermost mantle, possibly indicating very thin or absent mantle lithosphere in the area. These features provide important constraints on possible models for Miocene to recent volcanism in the Pacific Northwest.
 The relationship between the various mid-Miocene to present volcanic episodes in the Pacific Northwest with each other and with coeval tectonic processes is not well understood. Extensive mid-Miocene volcanism in the Cascadia back-arc region began ∼17 Ma with the eruption of the voluminous Steens/Columbia River flood basalts (Figure 1). In a time span of ∼1.5 Ma, over 200,000 km3of basalt erupted from N-S trending dike swarms located within the accreted terranes along the western boundary of Precambrian North America [Camp and Ross, 2004]. The initiation of silicic volcanism at the southwest edge of the Owyhee Plateau (OP) followed the cessation of flood basalt volcanism. By ∼12 Ma, two distinct tracks of time-progressive silicic volcanism had developed, both originating in the vicinity of the OP: the Yellowstone-Snake River Plain (YSRP) track that trends to the northeast parallel to apparent plate motion [Pierce and Morgan, 1992] and the High Lava Plains (HLP) volcanic lineament that trends to the northwest, terminating at present near Newberry volcano [Jordan et al., 2004]. Basaltic eruptive activity that is not time progressive persists into the Holocene along both the HLP and YSRP [Camp and Ross, 2004, and references therein].
 Seismic imaging can help constrain the relative importance of various tectonic structures in the formation of the HLP. Recent studies of shear wave velocity structure determined from teleseismic surface waves show significant differences in the upper mantle beneath the HLP and YSRP [Warren et al., 2008; Wagner et al., 2010]. However, these studies primarily image regions below the crust. Ambient noise studies can be used to examine shallower structures within the crust and uppermost mantle, but recent broad-scale studies of the entire western US do not have sufficient resolution to examine the HLP in detail [e.g.,Bensen et al., 2008; Yang et al., 2008, 2011; Moschetti et al., 2010a, 2010b] and the more detailed regional studies do not focus on the HLP [e.g., Stachnik et al., 2008; Gao et al., 2011; Porritt et al., 2011]. Although other seismic methods have been used to examine the structure of the HLP, this is the first study to specifically focus on the fine-scale features of the crust and shallow mantle beneath the HLP and surrounding regions using ambient noise tomography.
2. Data and Methods
 Data for this project come from 361 broadband seismic stations deployed across the Pacific Northwest from January 2006 to September 2009 (Figure 1). We used records from the 118 broadband seismometers that comprised the High Lava Plains seismic experiment [Eagar et al., 2011] and 243 broadband stations from the EarthScope/USArray Transportable Array (TA) (rows D-Q and columns 1–19). An additional 43 stations from the other regional networks supplement the more extensive TA and HLP stations.
2.1. Phase Velocity Models From Ambient Noise Tomography
 We determine surface wave dispersion curves from cross-correlations of ambient noise [Bensen et al., 2008, and references therein]. We invert these phase velocity measurements to obtain 2-D phase velocity models via the method ofBarmin et al. . We remove data with travel-time residuals greater than three standard deviations from the mean following the criteria ofBensen et al. . From the remaining data we calculate final phase velocity maps for periods at 8, 10, 12, 20, 25, 30, 33 and 40 s on a 0.25° × 0.25° grid (see Figures S1 and S2 and Text S1 in the auxiliary material for details).
2.2. Shear Wave Velocity Model
 We invert for 1-D shear wave velocity structure at each point in map view. We use a simplified starting model and adjust the crustal thickness at each point according to the receiver function results ofEagar et al.  (see Figure S3 and Text S1 in the auxiliary material). Following the method of Weeraratne et al. , we invert for the 1-D shear wave velocity model at each point in map view that most accurately predicts the suite of phase velocities observed at that location. We then combine these 1-D shear wave models to produce a 3-D shear wave velocity model. The RMS misfit between the observed phase velocities at each point and the phase velocities predicted by our shear wave velocity models can be seen inFigure S7. Given the range of periods used in this inversion, our best resolution ranges between 5 and 50 km depth, though we are able to recover anomalies from depths up to 70 km. We also performed error analyses to investigate the sensitivity of our shear wave models to error in the phase velocities, which can be found in Figure S8 and Text S1 in the auxiliary material.
3. Results and Discussion
 The results of our shear wave velocity inversions are shown in Figures 2, 3, S4–S6, and S12–S15. We recover many of the major features seen in previous ambient noise tomography studies, giving us confidence in the robustness of our results. The major features include the shallow low velocity Columbia Basin anomaly in northern Oregon and Washington that is likely due to an extensive sedimentary basin in the area [Gao et al., 2011; Moschetti et al., 2010a], the N-S trending high velocity anomaly associated with the ancestral Cascades [Porritt et al., 2011; Gao et al., 2011], the high velocity Wyoming Craton and Siletzia terrane [Gao et al., 2011], and the broad low velocities associated with the Basin and Range, the Coastal Ranges of northern California, and the Intermountain Belt [Moschetti et al., 2010a; Yang et al., 2008, 2011]. In addition to these features, our study reveals a number of additional structures that have either not been identified or not been discussed in previous papers.
3.1. Owyhee Plateau Anomaly
 The Owyhee Plateau (OP) is a relatively undeformed region [Shoemaker, 2004] characterized by somewhat thicker crust than the surrounding HLP and Basin and Range province [Eagar et al., 2011]. The central portion of the Plateau is visible in our models as a roughly cylindrical high velocity anomaly that extends through the upper and mid crust (Figures 2a and 2b, anomaly 1). The lowermost crust and upper mantle beneath this anomaly appear to have more average velocities (Figures 2c, 2d, and 3a) in contrast to the very low velocities of the adjacent High Lava Plains and Basin and Range.
 The genesis of the high Vs anomaly in the upper crust of the Owyhee Plateau is uncertain. The thickened lithospheric block that became the OP appears to have formed by the thrusting of accreted Mesozoic lithosphere over older Precambrian cratonic lithosphere during the Sevier orogeny [Shoemaker, 2004, and references therein]. The onset of mid-Miocene flood basalt volcanism ca. 17 Ma in the region of the OP was largely limited to its structurally weaker margins, with major eruptive centers located along a narrow array of N-S trending vents in the vicinity of Steens Mountain to the west. At ∼11 Ma, Basin and Range extension resulted in the production of mafic basalts along the margins of the OP [Shoemaker, 2004; Shoemaker and Hart, 2002], but by 5 Ma volcanism on the Plateau itself ceases with only small volcanic centers resuming in the past 1 Ma at the northernmost margins.
 One possible interpretation of our observations is that the region of average velocities in the upper mantle beneath the OP could represent older Precambrian lithosphere that has been further depleted during the flood basalt stage [Shoemaker, 2004]. The depleted lithosphere would provide stability to, and limit Basin and Range extension within, the Owyhee Plateau. The cylindrical high velocity anomaly in the upper and mid crust could be evidence of a residue left behind from the more limited mid-Miocene volcanism that occurred within the margins of the Plateau.
3.2. High Velocity Mid-crustal Anomaly Beneath the SRP
 While we are not the first to observe the mid-crustal high velocity anomaly of the SRP [e.g.,Smith et al., 1982; Sparlin et al., 1982; Priestley and Orcutt, 1982; Stachnik et al., 2008], our results resolve important distinctions in the lateral extent of this anomaly. In particular, we observe that the southwestern half of this high velocity anomaly extends south of the SRP into Nevada and Utah. The southward extension of this high velocity anomaly is seen in a number of previous tomography studies, though it has not been discussed explicitly before [Pollitz and Snoke, 2010; Lin et al., 2011; Yang et al., 2011]. Of notable exception are the results of Moschetti et al. [2010a, 2010b] who see a high velocity anomaly constrained solely within the limits of the SRP. The methodology of Moschetti et al. [2010a, 2010b] differs from those of the previous tomography studies in their inclusion of Love waves and radial anisotropy. While this may suggest that radial anisotropy is responsible for the southward extension of the high velocity anomaly in our results, we see no corresponding feature in the crustal radial anisotropy maps of Moschetti et al. [2010b]. Further research is needed to determine the cause of this model discrepancy.
 The high velocities observed beneath the SRP have been attributed to a mid-crustal layer of mafic sill intrusions [e.g.,Peng and Humphreys, 1998; Shervais et al., 2006; Stachnik et al., 2008; DeNosaquo et al., 2009]. DeNosaquo et al.  use regional gravity anomalies along with other geophysical data to constrain the density structure of these mafic intrusions. The observed gravity anomaly referenced in their study also extends to the south into Utah and Nevada, roughly coincident with our observed high velocity anomaly. This suggests that the same type of sill structures inferred to exist within the SRP may produce the high velocity anomaly south of the SRP. If that is the case, then the broadening of these sill structures along the earlier portions of the SRP volcanic track could indicate a broadening of the mantle upwelling responsible for their formation. This has important implications when considering whether or not the entire SRP track was formed by passage of a narrow plume tail, or whether a more regional upwelling is required to explain the broader anomaly associated with the earlier portions of the volcanic sequence. Evolving stress in the extending lithosphere may also have focused intrusions farther south for a period of time, regardless of the details of the specific mantle upwelling involved.
3.3. Low Velocities in the Lower Crust Beneath Yellowstone
 We observe a significant low velocity anomaly directly beneath Yellowstone (Figure 2, anomaly 3) that appears to extend vertically from the surface to the upper mantle (Figure 3). This is consistent with the results of Moschetti et al. [2010a, 2010b], but stands in contrast to other previous results that find high shear wave velocities in the lower crust directly beneath Yellowstone, and a SW dipping low velocity anomaly [e.g., Stachnik et al., 2008]. Given that this anomaly is near the edge of our model, we have performed a number of recovery tests to demonstrate the robustness of our results (see Figures S10 and S11 in the auxiliary material).
 The low velocity anomaly under Yellowstone broadens to the southwest in the lower crust and upper mantle roughly consistent with earlier results (Figure 3b) [Smith et al., 1982, 2009; Schutt et al., 2008; Stachnik et al., 2008]. Holocene basaltic volcanism in the SRP is roughly co-located with this lower crustal low velocity zone (Figures 2c and 2d, anomaly 3; Figure 3b), suggesting a possible causal link between the two. The nature of such a causal link is beyond the scope of this paper, but would be a valuable subject for future research.
3.4. High Lava Plains Low Velocity Zone
 The HLP region exhibits widespread low velocities throughout the crust and upper mantle (Figure 2, anomaly 4). A few features stand out within this region of generally low velocities. At shallow depths (Figure 2a, anomaly 4), the lowest velocities are present beneath the active arc and Newberry caldera. Low velocities are also observed beneath Diamond and Jordan Craters in eastern Oregon. Along the HLP volcanic track, pronounced low velocities are observed from mid-crustal depths into the upper mantle (Figure 2b, anomaly 4; Figure 3a).
 Of particular interest is the apparent absence of mantle lithosphere beneath the HLP (Figure 3). Low velocities directly below the Moho are observed in the region between the subducted Juan de Fuca plate and the mantle beneath the OP. These low velocities could indicate either increased temperatures within a persistent chemically distinct mantle lithosphere, or they could indicate the absence of virtually all mechanical mantle lithosphere. Recent work by Till et al. (C. Till, personal communication, 2011) suggests that <10.5 Ma primary melts are generated from within these very low velocity regions at depths that are directly beneath the Moho, favoring the latter interpretation. Regardless, the low shear wave velocities suggest a very weak layer underlying the HLP crust, which could help to explain the formation and persistence of volcanism across southeastern Oregon.
 We are very grateful to the many people involved in making both the HLP experiment and Transportable Array deployment a reality. Many thanks also to the PASSCAL Instrument Center and the IRIS Data Management Center for their ongoing support, which are supported through the GEO Directorate through the Instrumentation and Facilities Program of the National Science Foundation under Cooperative Agreement EAR-0552316. The High Lava Plains project (http://www.dtm.ciw.edu/research/HLP) was funded through NSF award EAR-0507248 (MJF) and EAR-0506914 (DEJ). SHH's and LW's participation was supported by NSF award EAR-0809192. We thank M. Ritzwoller, M. Barmin, and W. Shen for providing the ambient noise tomography codes.