Crustal structure and signatures of recent tectonism as influenced by ancient terranes in the western United States Open Access

The terranes that would later become the North American continent grew between 1.8 and 1.3 Ga as Proterozoic island arcs of a variety of affinities and ages (Hoffman, 1988) accreted to the southern margin of the Wyoming province (Bowring and Karlstrom, 1990; Karlstrom et al., 2001). Neoproterozic to early Cambrian rifting of the resulting continent established the limits of the modern North American craton (e.g., Hoffman, 1988), which was followed by a passive margin extending along its west coast (Speed and Sleep, 1982) (Fig. 1). Shortening of the continent occurred during the Antler, Sonoma, and Nevadan orogenies from the Mississippian to the Jurassic as various terranes accreted onto North America (Saleeby and Busby-Spera, 1992). The Mesozoic deformational history of the western United States includes the Sevier-Laramide orogeny, which persisted from Cretaceous to early Tertiary time (Hamilton, 1969; Burchfiel and Davis, 1975; Dickinson and Snyder, 1978). Early stages of this orogeny (ca. 120–80 Ma) were characterized by the development of a magmatic arc along the west coast of North America and foreland fold-thrust belts that were mainly located within the Sevier fold-thrust belt (Fig. 1) (Armstrong and Oriel, 1965; Lawton, 1985). From ca. 80 to 75 Ma, the deformation front propagated eastward, leading to Laramide uplifts as far as eastern Wyoming. These uplifts contributed to the formation of the Rocky Mountains (Fig. 1); the overall shortening across the Rocky Mountains is estimated to be close to 10% (Stone, 1993; Brown, 1993). The stress field changed following the Laramide, giving way to Cenozoic volcanism, and extension, including the metamorphic core complexes (Coney and Harms, 1984) and the Basin and Range (Zoback and Thompson, 1978). The culmination of these events has led to a complex pattern of topography and continued deformation within western North America.

The present-day strain in the western United States extends well beyond the boundary between the North American and Pacific plates, where varying styles of deformation characterize different provinces. These distinct tectonic provinces that compose the western portion of North America include the San Andreas fault, the active Cascadian arc, the highly extended Basin and Range, the Laramide contracted and unextended Colorado Plateau, the Snake River Plain, the Rocky Mountains, and the tilted and intact Great Plains. The long history of strain recorded by the faulting and surface deformation has created the diverse landscape, yet the origins of the stresses responsible for producing this pattern of strain are not fully understood (e.g., Jones et al., 1996; Flesch et al., 2000; Fay and Humphreys, 2005; Roy et al., 2009). Well-resolved models of crustal thickness are necessary for improving our understanding of the evolution of the tectonic regimes within the western United States. The thickness of the crust serves as an input to geodynamic models to constrain the distribution of densities in the lithosphere. In addition, tomographic models include variations in crustal thickness to account for the effect of seismically slower crust on observed traveltime anomalies and remove it to determine wave-speed variations in the mantle (e.g., Roth et al., 2008; Sigloch et al., 2008; Schmandt and Humphreys, 2010; Burdick et al., 2010; Obrebski et al., 2010). Well-resolved constraints on the thickness of the crust are therefore necessary for producing reliable geodynamic and seismological models, which can then be used to further study continental evolution.

Determining the lithospheric structure across the boundaries of terranes and tectonic provinces that have been affected by specific events allows for identifying the degree to which those events modified crustal structure. Data collected by the EarthScope Transportable Array (http://www.earthscope.org/observatories/usarray) in the western United States allow for sampling of a sufficiently large area that structures can be compared between various tectonic provinces. The crustal structures presented here from a uniformly processed data set provide a means to compare previous focused studies in the larger context of the evolving portion of western North America. The resulting map offers a framework to study continental evolution by providing insight into the degree to which inherited lithospheric properties contribute to crustal structure and subsequent events have reshaped the crust. Studying the crust in this way highlights relationships between crustal thickness and topography and how that relationship varies across the western United States. Improved estimates of crustal thickness can then be used to determine how various tectonic events shaped the crust, or inversely, the role played by structures in the crust governing the extent of tectonic processes.

Previous investigations have employed various techniques investigating crustal structure across the western United States in order to identify links between structures and tectonic environments. In areas spanning the west coast of North America to the High Plains, these studies found specific tectonic provinces and terranes exhibiting structures with large variations in crustal thickness and changes in seismic velocities across the Moho (e.g., Braile et al., 1989; Prodehl and Lipman, 1989; Catchings and Mooney, 1991; Snelson et al., 1998; Bostock et al., 2002; Gorman et al., 2002; Li et al., 2002; Magnani et al., 2004; Keller et al., 2005; Frassetto et al., 2006; Lerch et al., 2007; Rumpfhuber et al., 2009; Hansen and Dueker, 2009; Buehler and Shearer, 2010; Cox and Keller, 2010; Eagar et al., 2011; Hansen et al., 2011). Data from the Transportable Array can be used to investigate the crust across extensive distances to compare and connect the observations from these previous localized studies.

A number of investigations focused on specific portions of the crust at boundaries between terranes (e.g., Henstock et al., 1998; Snelson et al., 1998; Gorman et al., 2002; Crosswhite and Humphreys, 2003; Magnani et al., 2004; Gilbert et al., 2007; Rumpfhuber and Keller, 2009) and found differences in crustal structures across terrane boundaries. However, the focused nature of several of these studies limits the degree to which their observations can be extrapolated to be representative of the boundaries between terranes. In a study that assembled data from the Continental Dynamics of the Rocky Mountains experiment (CD-ROM) and Deep Probe seismic arrays, Rumpfhuber et al. (2009) expanded on previous observations of the Cheyenne belt between the Wyoming craton and accreted Proterozoic terranes (e.g., Snelson et al., 1998; Gorman et al., 2002). Similar to these previous investigations, this study identified the Archean Wyoming craton as having a thick fast lower crustal layer that terminates ∼100 km north of the surface expression of the terrane boundary marked by the Cheyenne belt in southwestern Wyoming. The location where the high-velocity lower crust terminates results in the crust thinning toward the southern edge of the Wyoming craton and an abrupt increase in crustal thickness to the south in the Proterozoic terrane in northwestern Colorado across the Cheyenne belt (e.g., Snelson et al., 1998; Crosswhite and Humphreys, 2003). Further evidence for the preservation of structures related to accretion comes from the boundary between the Yavapai and Mazatzal terranes farther south in central New Mexico. Oppositely dipping arrivals in seismic reflection data sampling the boundary between the Yavapai and Mazatzal provinces have been interpreted as a bivergent zone that formed during continental assembly (Magnani et al., 2004). The boundary appears to have remained a zone of weakness and has served as a conduit for magmas within the Jemez Lineament.

Patterns of crustal thicknesses have also been identified that relate to the boundaries of tectonic provinces that formed well after continental growth. Utilizing picks of Pn arrivals from regional earthquakes recorded by the same USArray stations that contributed data to this study, Buehler and Shearer (2010) determined crustal thicknesses and P wave velocities in the uppermost mantle across the western United States. Their study found variations in crustal thickness from ∼20 km to >45 km, many of the changes in thickness occurring at boundaries between different tectonic provinces. The thickest crust, ∼50 km, identified by this study is in the Rocky Mountains in southern Wyoming, while the thinnest crust is in the southern Basin and Range along the southern Arizona border.

Several previous investigations that have focused on specific areas identified the general distribution of crustal structure of the western United States. A sampling of studies of crustal thickness in the western United States demonstrates the trend, albeit with a number of exceptions, that crustal thickness generally decreases to the west of the Rockies toward the west coast. Thicknesses >50 km have been identified in the Rockies, where there is a general increase in crustal thickness along the Laramide front (Li et al., 2002; Keller et al., 2005). However, detailed investigation of the Rockies in southern Colorado found thinner crust beneath areas of high topography than beneath the topographically lower adjacent High Plains (Hansen et al., 2011). Thin crust within the Basin and Range has been found to relate to extension (e.g., Catchings and Mooney, 1991; McCarthy and Parsons, 1994; Benz and McCarthy, 1994; Lerch et al., 2007). Based on increased seismic velocities of portions of the lower crust of the Basin and Range that could result from mantle-derived materials entering the crust, it also appears that underplating accompanied crustal thinning (Catchings, 1992). The crust of the Colorado Plateau between parts of the Rockies and Basin and Range exhibits a broad transition over a wide depth interval between the crust and mantle, with crustal thicknesses between 40 and 45 km (Wilson et al., 2005). Along the western margin of the Basin and Range, the crust thickens into the Sierra Nevada to as much as 40 km or more (e.g., Ruppert et al., 1998; Fliedner et al., 2000; Frassetto et al., 2011). Crustal thicknesses close to 40 km continue farther to the north of the Sierra Nevada into the active Cascade volcanic arc, to the east of which the crust thins in the backarc to close to 30 km within the High Lava Plains (Cox and Keller, 2010; Eagar et al., 2011).

Studies of the margin of the North American plate identified variations in crustal structures in the area of the Mendocino Triple Junction. The plate margin at this location transitions from the Cascadia subduction zone, where the Gorda and Juan de Fuca slabs descend beneath the North American plate to the north of the triple junction, to a transform margin along the San Andreas fault to the south of the triple junction. The thickness of the crust has been elusive to detect within the forearc along the western portions of Washington and Oregon. This is thought to result from the serpentinization of the mantle wedge above the subducting slab that reduces the velocity contrast between the crust and mantle (e.g., Bostock et al., 2002; Brocher et al., 2003). To the south, where the plate margin transitions from a subduction zone to a strike-slip fault, the North American crust thins in the area of the slab window to the south of the Mendocino Triple Junction zone (Hole et al., 2000). On the Pacific plate, to the west of the San Andreas fault, is crust that is thinner than that found on the North American plate to the south of the triple junction (Verdonck and Zandt, 1994; Beaudoin et al., 1996).

The degree to which the results of these, and other, investigations can be interpreted to be representative of the tectonic environment they sample has not been determined. However, the expansive, nearly regular 70 km sampling of the Transportable Array component of the EarthScope USArray provides a framework in which smaller aperture studies, which generally have focused on specific tectonic areas, can be compared over greater length scales. The data from the Transportable Array can also be complemented with data from other broadband seismic stations operating in the region, producing a database that both samples a large area and, where sufficient data are available, resolves smaller structures. Patterns of crustal structures identified by the Transportable Array in the western United States can then be compared with other studies of crustal thickness (e.g., Bassin et al., 2000) and seismic wave speeds (Bensen et al., 2009) that cover the western United States.

The extensive and even sampling of the Transportable Array affords the opportunity to investigate the structure of the crust over the entire western United States. The array started recording along the west coast of the United States in 2004 and is composed of ∼400 broadband stations deployed 70 km apart for 2 years. After two years of recording, the station is picked up and redeployed at the eastern edge of the array, such that stations within the array have been leapfrogging eastward (http://www.usarray.org). Data recorded by the Transportable Array prior to the middle of 2010 by stations located as far east as 100°W have been used to calculate receiver functions that provide the observations of crustal structure presented here. To increase sampling where available, we also included data from past temporary broadband deployments (Fig. 2). The large footprint of the Transportable Array allows for observations from smaller aperture arrays with focused targets across the western United States to be tied together. Combining data from dense arrays and the Transportable Array stations presents a challenge in determining the appropriate level of structures that can be resolved. Smaller detailed crustal features with horizontal scales of ∼50 km can be identified in areas with close station spacing, while only larger scale structures that extend for ∼100 km in the crust can be identified in regions only sampled by the more broadly spaced Transportable Array. As the area sampled by the Transportable Array continues to expand, and data become available from other deployments within the contiguous United States, the data set used here can be updated to produce a crustal thickness map over a larger area in a manner similar to the updated tomographic models of Burdick et al. (2010).

Receiver functions reveal discontinuities in seismic velocities beneath seismic stations as teleseismic P waves encounter the discontinuity where a transmitted P wave and converted S wave continue to the station. The difference in arrival times between these two phases allows for the depth of the discontinuity to be determined by assuming a velocity model. The P waves from teleseismic earthquakes with magnitudes >6.0 at distances between 30° and 95° from the seismic stations contribute the signal for calculating the receiver functions. Prior to further processing, the seismic records from these events are checked to ensure the presence of adequate signal. High pass filtering of the three component broadband traces at 0.02 Hz removes low-frequency noise. The traces are then rotated into their great circle coordinates for each station-event pair. The receiver functions are then calculated from these rotated traces using an iterative deconvolution method (Ligorria and Ammon, 1999) that isolates P to S conversions on the radial component. Following deconvolution, we only retain those receiver functions that meet a variance reduction cutoff of 70%. This resulted in 49,738 acceptable receiver functions calculated using a Gaussian width of 2.5, which corresponds to a frequency content of ∼1.2 Hz. Assuming a vertical sensitivity of one-quarter wavelength and crustal shear wave velocities of 4 km/s, data in this frequency range would only be from sensitive layers thicker than ∼0.8 km. As illustrated in the observations presented here, the amplitude of the converted S wave is proportional to the change in seismic velocities and densities across the discontinuity where the converted wave originated.

In order to address the issue of combining receiver functions recorded by stations within the evenly distributed Transportable Array with other data sets that used a range of station spacing, we adopted a stacking approach with bins whose radii depended on the density of stations sampling an area (e.g., Owens, et al., 2000). This method starts similarly to other common conversion point stacking procedures (e.g., Dueker and Sheehan, 1997; Gilbert and Sheehan, 2004) with a grid of evenly spaced bins spanning 2250 km in the north-south direction, 2100 km in the east-west direction, and to a depth of 110 km. The grid used here has a bin every 45 km in the north-south and east-west directions and 1 km vertically. The 45 km horizontal bin spacing was chosen after testing bins spaced between 25 and 75 km to determine the size capable of imaging small-scale features in areas where closely spaced stations permit, yet maintain sufficient separation to limit overlap while still imaging a continuous area. The radius of each bin depends on the sampling density where it is located (Fig. 3). Bins with shorter radii facilitate identifying smaller scale structures in locations sampled by closely spaced stations. However, in regions sampled only by stations spaced at greater distances, larger bins allow for sufficient sampling. The radius of each bin is based on the minimum size required for it to be sampled by 40 traces. This ensures enough data for stacking to enhance signal and limit the effects of noise. Each bin starts with a radius of 45 km and the number of traces sampling that bin is determined. If a bin is not sampled by at least 40 traces, its radius is increased by increments of 11.25 km (one-quarter of the bin spacing) until it contains at least 40 traces. The maximum allowable size for each bin is 135 km. Bins sampled by <40 traces at their maximum size, that primarily are on the perimeter of the array, are not included in further analysis. The size of the stacking bins and the spacing of the stations contributing data control the lateral resolution of our stacks. In well-sampled areas, where the sampling criteria are met with bins at their minimum radius, we can resolve structures that vary over distances of 90 km, or the diameter of the bin. Structural variations over shorter distances would be averaged together and therefore could not be resolved. In areas where the radii of the stacking bins need to be increased, the minimum extent of structures that can be resolved increases accordingly.

Picking the thickness of the crust across the study area requires determining which signal on the receiver function stacks corresponds to the arrival from the Moho. Multiple peaks within the potential depth range of the Moho can be seen on the cross sections of the receiver function stacks (Fig. 4). To facilitate picking the Moho and identifying the most laterally continuous surface, peaks on receiver function stacks within each column of bins are automatically identified within 20 km depth intervals starting at depths of 10, 20, 30, and 40 km. For bins with more than one peak in these depth ranges, the distribution of the peaks on the receiver function stacks in adjacent bins are used to identify the Moho. For a specific bin, the differences in depth between each of the peaks in that bin and peaks in adjacent bins are calculated. The receiver function peak whose depth is closest to the depths of the peaks in adjacent bins is identified to be the Moho (Fig. 4). Across this large study area, complex receiver function signals lead this automating picking approach to identifying some spurious picks as the Moho. Therefore, following the automated picking, the receiver function picks are visually inspected to insure the most likely peak is identified to be the Moho.

The stacking and picking approach described here is used to generate the map of crustal thickness using a uniform S wave velocity, V_S, of 3.8 km/s and a V_P/V_S of 1.74. This model allows for the conversion of P to S wave arrival times to depths and tracing the path of each ray to identify the stacking bins to assign it. However, because the large sampling area of the western United States encompasses regions of heterogeneous wave speeds, variations in seismic velocities from the uniform starting model lead to the arrival times of the converted phases from the Moho being mapped to incorrect depths. Therefore corrections to account for the effects of lateral heterogeneities in seismic velocities are applied using the shear wave velocity model of Bensen et al. (2009) and the thickness of sedimentary basins from global crustal model CRUST2.0 (Bassin et al., 2000). The details of applying these corrections and their effects are further discussed in the Supplemental File¹.

The general trends in crustal thickness imaged here (Fig. 5) follow patterns that have previously been identified by several of the investigations mentioned above that found changes in crustal structure between various physiographic provinces. It is important to note that the lateral averaging and frequency content of the data used here allow for imaging crustal structure on a continental scale in a single coherent image. Focusing on this scale leads to smaller scale features, which are present in focused studies, sometimes being below the level of resolution of this study. However, the results presented here retain sufficient detail to provide insight into the tectonic evolution of the western United States. The observations presented here serve as a means to link many of these previous observations and to compare how crustal structure changes in a manner that reflects the specific tectonic history of each of the provinces.

The primary features identified by this compilation of data across the western United States include the large amount of variability in crustal thickness, from >50 km to <30 km. Of additional interest are areas of weak Moho signal in the Pacific Northwest and Colorado Plateau, where the impedance contrast between the crust and mantle appears to be diminished (Fig. 4). The crust thickens in the active Cascade volcanic arc and the Sierra Nevada batholith to the south compared to the thinner crust along the coast. The Basin and Range exhibits thinner crust than surrounding areas; the southern portion of this province has some of the thinnest crust within western North America. The Snake River Plain exhibits distinctively thickened crust compared to neighboring areas. Crustal thicknesses appear to gradually increase across the Colorado Plateau between the thinner crust of the Basin and Range and the thicker crust in the Rocky Mountains. It is interesting that below portions of the Rockies, specifically within Montana, the crust thins to <40 km. Thick crust continues to the east of the Rockies into portions of the High Plains. Particularly abrupt transitions appear between some of the regions sampled here, such as across the Cheyenne belt and between the Rocky Mountains and the Basin and Range along the northern most portion of the Wasatch Front. Details of crustal structure presented here show variations within physiographic provinces that have not been detected by previous studies investigating large-scale crustal structure in the western United States. As described here, these structures provide insight into how these provinces developed.

The nearly regular 70 km sampling of the Transportable Array provides data that can create a model of crustal structure that includes structures smaller than the 2° bins of CRUST2.0 (Fig. 6A; Bassin et al., 2000). Similar overall trends in crustal thickness are apparent between the results presented here and the crustal thicknesses in CRUST2.0, with thick crust >50 km in the Rockies and thin crust, <30 km, in the southern Basin and Range and along the west coast (cf. Figs. 6A, 6B). Figure 6C presents a comparison between the thicknesses determined by the two models. The observations presented here illustrate changes in the thickness of the crust within physiographic provinces that do not appear in CRUST2.0. An inherent difference in the crustal thicknesses that occur over limited distances presented here is that they cannot be identified in the 2° by 2° grid of the CRUST2.0 model. We therefore filter our results to identify differences over length scales that exist between these observations and those of CRUST2.0. Within CRUST2.0 the crust thickens in the Rockies to as much as 50 km as far north as Montana and remains thick to the east into the High Plains, while this study finds the crust of the Rockies to be thinner in Montana. In addition, the variations in crustal thickness identified here in the Colorado Plateau and Basin and Range do not appear in CRUST2.0.

The observed crustal thicknesses do not correlate with topography across the western United States, illustrating that this region does not follow expectations based on simple Airy-type isostasy. There is not a strong correlation between thick crustal roots and areas of high elevations (Fig. 7); instead, the area exhibits a complex relationship between crustal thickness and surface elevations. Crustal thicknesses vary from 30 to 50 km in areas of high elevations within the Rocky Mountains between northern Colorado and western Montana. In addition, the crust varies from ∼30 km to >40 km thick across the relatively flat Colorado Plateau. Changes in crustal thickness that do not follow trends in surface elevations require lateral variations in the density of the crust, the mantle, or both for the surface topography to be compensated.

Comparing the patterns of crustal structure observed to the distribution of gravity anomalies provides insight into the density distribution within the lithosphere and the source of buoyancy that supports surface topography. The pattern of Bouguer gravity anomalies, as shown in Figure 8 (http://gis.utep.edu/rgsc/index.php?option=com_content&view=article&id=197%3Agdrp-home&catid=51%3Amain-site&Itemid=59; Keller, 2005), follows a general trend that is anticorrelated with topography such that areas of higher topography coincide with locations of more negative anomalies. This relationship illustrates the effects of missing mass resulting from low-density crustal material supporting areas of higher elevation. Following this trend, we find the most negative values across the western United States in the region of the Rocky Mountains in Colorado, which also has some of the thickest crust. Similarly, low-lying areas in the southern Basin and Range have some of the least negative Bouguer gravity anomaly values and thinnest crust. Identifying deviations from this trend between crustal thickness and gravity (Fig. 9) helps discriminate between topographic support from the crust or mantle. One area of deviation is in central Nevada, where the Bouguer anomaly becomes more negative in an area where the average surface elevations do not vary greatly from the surrounding locations.

The distinct increase in crustal thickness between the coast and the arc can be identified along the western boundary of the North American plate. However, as indicated by comparing the extent of thickened crust on the crustal thickness map to the location of the San Andreas fault, the effects of lateral averaging during stacking become apparent as the zone of thin crust extends beyond the Pacific plate (Fig. 5). It is therefore necessary to limit our interpretations to structures with a greater lateral extent that can be identified despite the lateral averaging employed here. Many of the smaller scale features identified along the west coast that are referenced here are too small to be identified. The thickness of the crust cannot be clearly identified in the region of active subduction in the Pacific Northwest from southern Washington to central Oregon due to the lack of a clear P to S converted phase to pick (see the western portions of cross-sections A and B in Fig. 4). The small amplitude, or even lack, of the Moho signal and low Pn velocities observed in previous investigations in this area (e.g., Bostock et al., 2002; Brocher et al., 2003) has been interpreted to result from the release of water from the subducting Juan de Fuca slab, leading to the serpentinization of the mantle wedge. The slow shear wave speeds of the serpentinized wedge reduce, or in some cases invert, the impedance contrast across the Moho, making it difficult to detect using receiver functions.

Observations presented here corroborate previous crustal thickness studies that detected thinner crust within the Basin and Range than in other regions of the western United States. Normal faulting during extension of the western portion of North America produced the characteristic evenly spaced mountain blocks separated by basins of the northern portion of the Basin and Range across much of Nevada (e.g., Zoback et al., 1981). Extension-related deformation modified the crust with seismically reflective silica-rich material marking the effects of shearing in the mid-crust of central Nevada (Holbrook et al., 1991). The crustal thicknesses presented here for the northwestern part of the Basin and Range, which extends from southern Oregon into northern Nevada, are close to 35 km. The crust thins to ∼30 km to the south of the northern margin of the Basin and Range within an east-west swath that extends across the northern half of Nevada eastward toward Salt Lake (Fig. 5). The crust thickens sharply at the eastern boundary of the Basin and Range and Rocky Mountains in northern Utah. In the northeastern portion of the Basin and Range, 30-km-thick crust transitions to >50 km in the Rocky Mountains over a distance of <100 km close to the Uinta Mountains along the northern portion of the Wasatch Front.

Observations from the 1986 PASSCAL experiment that sampled from northwestern to central Nevada provide insight into the complex lower crustal structure beneath the northern Basin and Range in northwestern Nevada (Catchings and Mooney, 1991). This study identified layering within the crust with variations in the thickness of these layers leading to changes in crustal thickness. A layer of high-velocity lower crust beneath the northwestern portion of their study area pinches out as crustal thicknesses increase to ∼35 km toward central Nevada (Catchings and Mooney, 1991). This location of crustal thickening coincides with the zone of ∼35 km thick crust identified here near 39°N, 117°W in central Nevada (Fig. 10). The western portion of the transect presented in Figure 9 coincides with the location of the northwest-southeast–trending line from the 1986 PASSCAL experiment. The addition of mafic material to the crust from the mantle during regional extension (e.g., Catchings, 1992), or in response to lithospheric removal (e.g., Elkins-Tanton, 2005; West et al., 2009), could be responsible for producing the layer of higher seismic velocities observed at the base of the crust by Catchings and Mooney (1991).

Both controlled source and earthquake investigations identified crust >35 km thick in the portion of the Basin and Range near the northwestern corner of Nevada, the crust thinning to ∼30 km to the east away from the western margin of the province (Lerch et al., 2007; Gashawbeza et al., 2008). These studies in the northwestern portion of the Basin and Range did not identify the same discrete zone of high velocities that was reported to the south by Catchings and Mooney (1991). This difference may result from the greater amount of extension away from the northwestern corner of Nevada (Colgan et al., 2006), leading to the addition of a greater amount of underplated material to the lower crust. The crustal thickness map presented here follows a similar trend of thicker crust in the northwestern corner of Nevada, and thinner crust directly to the south with the crust thickening toward central Nevada. A negative Bouguer anomaly within the Basin and Range is within central Nevada near the area where the crust thickens to the south of the swath of thin crust (Fig. 5). The negative Bouguer gravity anomaly of ∼–200 mGal suggests an increase in crustal thickness leading to a loss of mass. However, the observations here, presenting crust that thickens to ∼35 km in an area where surface elevations average 2 km in this portion of the Basin and Range, suggest the need for low-density mantle material that would support the high elevations.

The swath of thin crust across the northern portion of the Basin and Range from northern Nevada that extends eastward to Salt Lake coincides with the location where movement along a large-scale detachment surface near the base of the crust has been proposed to accommodate surface deformation (Wernicke et al., 2008). The detachment surface would then accommodate movement observed in this location of the northern Basin and Range that appears in geodetic observations of ground motion that started to deviate in the middle of 1999 from expectations based on velocities observed prior to 2000 (Wernicke et al., 2008). Displacement on this detachment surface could then contribute to crustal thinning in northern Nevada. This mechanism offers a potential explanation to the zone of thin crust observed here. Upwelling mantle would rise up into the area where the crust has thinned and contribute to the high elevations in an area with thin crust. While not envisioning the same mechanism of crustal thinning on a crustal-scale detachment fault, Sonder and Jones (1999) suggested that extension in the northern portion of the Basin and Range has been driven by upwelling mantle following lithospheric thinning.

At <30 km thick, crustal thicknesses for the southern Basin and Range display some of the thinnest crust identified by this study in the western United States. These observations align with those of the Pacific to Arizona Crustal Experiment (PACE) profile (McCarthy et al., 1990), which was designed to determine lithospheric structure along a transect spanning the southern Basin and Range, near the Gulf of Mexico, to the northeast into the Colorado Plateau. Thin crust that increases from ∼28 km in the southern Basin and Range (McCarthy et al., 1991) to 40–45 km beneath the Colorado Plateau was similarly observed along the PACE profile (McCarthy and Parsons, 1994). Additional sampling of the southern Basin and Range provided by USArray presented here indicates that similarly thin crust to that imaged along the PACE profile extends eastward throughout the southern Basin and Range toward the Rio Grande rift (Fig. 5).

Complex crustal structure characterizes the Colorado Plateau, which is bound to the northwest and southeast by thinned crust of the St. George and Jemez lineaments. Along its eastern margin with the Rocky Mountains, the crust of the Colorado Plateau is almost 50 km thick, and thins eastward to <40 km along its eastern border with the Basin and Range. This gradient makes it appear to be a transitional region between the thicker crust of the Rockies to the east and the thinner Basin and Range crust to the west. Localized zones of thickened crust that deviate from this general trend are located along the eastern and southwestern portions of the plateau (Fig. 5). The thickness of the crust within the interior of the Colorado Plateau is difficult to determine based on receiver function observations as the data recorded in this area exhibit multiple low-amplitude receiver function arrivals between the depths of 30 and 50 km (Figs. 4D and 11).

Previous observations of Colorado Plateau structure provide insight into how it resisted deformation and why the receiver function signal of the Moho may be difficult to detect. The presence of a layer characterized by high seismic wave speeds within the Colorado Plateau has previously been proposed to result from mafic material that has underplated the Colorado Plateau (e.g., Wolf and Cipar, 1993; Zandt et al., 1995). If seismic velocities in the lower crust gradually increase toward mantle velocities, no abrupt velocity contrast would be present. Such could be the case in portions of the Colorado Plateau, where only small amounts of converted energy corresponding to the Moho appear on receiver functions (Fig. 11). In addition, the top and bottom of this high-speed layer could explain the multiple arrivals at depths of 30 and 50 km.

Differences in the reflectivity and layering observed by the PACE array within the middle and lower crust of the Basin and Range and the transition between it and the Colorado Plateau have led to differing hypotheses of the tectonic evolution of the region. The identification of a continuous mid-crustal layer extending from the southern Basin and Range to the edge of the transition to the Colorado Plateau led to suggestions that material was added to the crust during Basin and Range extension (McCarthy and Parsons, 1994). Alternatively, the lack of observations of continuous layering within the crust of both the Basin and Range and Colorado Plateau may indicate that the crust of the Basin and Range represents an extended version of the same crust that makes up the Colorado Plateau with only localized areas of magmatic intrusions (e.g., Wolf and Cipar, 1993). The receiver functions presented here exhibit stronger converted signals in the Basin and Range than in the Colorado Plateau, indicative of a smaller velocity contrast between the mantle and the base of the crust of the plateau. The reduced velocity contrast could result from the addition of mafic material to the crust of the Colorado Plateau; this would increase its rigidity and the seismic velocity of its lower crust (Zandt et al., 1995).

Comparing crustal thicknesses presented here for the Colorado Plateau with the Bouguer gravity anomalies of the western United States illustrates the coincident thinning of the crust toward the southeastern and northwestern margins of the plateau with locations where the Bouguer anomalies become more negative (Fig. 8). A negative Bouguer anomaly in an area of high elevation would typically be associated with the presence of a low-density crustal root; however, it appears here in an area where the crust thins, suggesting that the anomaly results from the presence of lower density mantle material. Reduced densities in the mantle beneath the Colorado Plateau have been proposed to originate from thermal warming of the thicker and more iron-depleted lithosphere of the plateau over the 30–40 m.y. following removal of the Farallon slab (Roy et al., 2009), which could then provide a source of buoyancy for the plateau relative to the surrounding areas. An alternative means by which the mantle densities around the plateau could be reduced is that upwelling asthenosphere progressively infiltrates and replaces the lithosphere, as suggested by the chemistry of young basalts that erupted along the perimeter of the plateau (Crow et al., 2011). This would be most effective in areas where Proterozoic compositional boundaries exist, such as the Jemez and St. George lineaments (Magnani et al., 2004).

Investigations of the lithospheric structure of the Colorado Plateau reported features within the upper mantle at depths of 70–90 km that may represent a portion of the lithosphere that has been removed from the plateau (Levander et al., 2011) and replaced by inflowing warm asthenosphere (e.g., Schmandt and Humphreys, 2010). The results of modeling refracted arrivals recorded by the PACE array also identified a discontinuity in seismic velocities in this depth range (Benz and McCarthy, 1994). The removal of lithosphere would then provide a mechanism to explain the uplift of the Colorado Plateau and pattern of volcanism encroaching the central portion of the plateau from its margins. The removal of Proterozoic mantle lithosphere material could also produce edge-driven convection, leading to the uplift of the margin and giving the plateau its characteristic bowl shape (Van Wijk et al., 2010). Other than indicating the need for additional buoyancy from the mantle, the crustal thicknesses presented here cannot discriminate between these models of mantle buoyancy.

Along a swath marked by the Snake River Plain, southwest of the Yellowstone caldera, the crust thickens to ∼40 km. The Snake River Plain includes a sequence of caldera fields that initiated 16 Ma in southwestern Idaho and northern Nevada at McDermitt caldera and propagated eastward to the most recent activity within the area of Yellowstone caldera (Armstrong et al., 1975; Smith and Braile, 1994). The origin of voluminous volcanism along the Snake River Plain involves the emplacement of large amounts of basaltic magma into the crust. This led to the melting of crustal material and an initial eruptive phase of rhyolite lavas and ignimbrites followed by a later phase of basaltic activity (Leeman et al., 2008).

Previous focused seismic and gravity data within the Snake River Plain identified the presence of a dense reflective layer at depths between 10 and 20 km that has been interpreted as a gabbroic sill resulting from the emplacement of basalts derived from upwelling mantle material (Sparlin et al., 1982; Smith and Braile, 1994; Peng and Humphreys, 1998; Stachnik et al., 2008). Crustal thicknesses here along the Snake River Plain include a continuous zone of thickened crust, suggesting that basaltic material may have thickened the crust along the length of the region (Fig. 5). The increased densities of the mid-crustal sill would provide the negative buoyancies, resulting in the reduced elevations within an area of thickened crust of the Snake River Plain (Fig. 12).

The Bouguer gravity anomaly becomes less negative within the Snake River Plain, indicating less low-density crustal material; this also argues for the underplating of high-density material into the crust. This pattern differs from the structure observed within the High Lava Plains, which largely follow the structures identified by the focused High Lava Plains deployment (Eagar et al., 2011), where no similar increase in crustal thickness or gravity anomaly exists. Differences between the structures of these two regions suggest that the same underplating process did not affect the crust of the High Lava Plains.

Basement rocks that have been uplifted during shortening characterize the Rocky Mountains, which mark the eastward extent of Laramide deformation (e.g., Dickinson and Snyder, 1978). The structure of the crust within the Rocky Mountains imaged here varies along the north-south extent of the range, from Montana to northern New Mexico. The crust thickens to >50 km in the Rockies along the border between Colorado and Wyoming, which marks some of the thickest crust observed in North America. The crust in the Rockies generally thins to the north and south of this region, reaching only ∼40 km elsewhere in Wyoming, Montana, and New Mexico (Fig. 5). These observations agree with previous seismic studies that also identified the crust as ∼50 km or more beneath the Rockies in northern Colorado, and thinner to the north and south of this area (Henstock et al., 1998; Snelson et al., 1998; Dueker et al., 2001; Gorman et al., 2002). Similar to the structure identified here, sharp increases in crustal thicknesses have been noted along the western portion of the Colorado-Wyoming border. Both refraction and earthquake studies identified crustal thickening from <40 km in the southern part of the Archean Wyoming craton to >50 km beneath the Proterozoic province to the south of the Cheyenne belt (Prodehl and Lipman, 1989; Snelson et al., 1998; Crosswhite and Humphreys, 2003; Poppeliers and Pavlis, 2003; Levander et al., 2005; Hansen and Dueker, 2009). If this abrupt change in thickness occurred during the Laramide, it would indicate a change in the crustal response to deformation between the Archean craton and the Proterozoic terranes. However, the expanded sampling here indicates that thick crust spans the Cheyenne belt beneath the uplifts further to the east along the Laramide front, suggesting that the Archean craton and Proterozoic terrane deformed in a similar manner in that portion of the Rocky Mountains.

Away from the southern end of the Archean craton, a 20-km-thick welt of material with high seismic wave speeds has been imaged by the Deep Probe refraction line from an area ∼150 km north of the southern Wyoming border, extending through much of Montana (e.g., Nelson, 1991; Henstock et al., 1998; Snelson et al., 1998; Gorman et al., 2002; Rumpfhuber et al., 2009). This layer was interpreted to result from the addition of material from the mantle that underplated the crust (e.g., Nelson, 1991; Henstock et al., 1998; Snelson et al., 1998; Gorman et al., 2002; Rumpfhuber et al., 2009). There is no evidence for the volume of either Proterozoic or Phanerozoic magmatic activity in this area needed to produce this amount of underplated material during those times, suggesting that the layer formed during the Archean (Snelson et al., 1998). The high-velocity lower crustal layer is consistent with mafic garnet granulite or hornblendite compositions, similar to the compositions of xenoliths from northern Montana (Reed et al., 1993). This layer would reduce the impedance contrast between the crust and mantle, leading to a smaller receiver function arrival from the Moho.

The Moho signal on the receiver functions presented here is difficult to detect in central Wyoming between the Wind River and Big Horn uplifts within the Archean craton (Fig. 4B; between 109°W and 105°W). Correspondingly, this is a region of large uncertainty on the crustal thickness map (Fig. 5). The Moho has previously been difficult to identify in southeastern Wyoming in the region of the Laramie Mountains (Allmendinger et al., 1982; Gohl and Smithson, 1994; Morozova et al., 2005) possibly as a result of a gradational change in velocities, a large amount of scattering associated with the boundary, or because the Moho is marked by a small impedance contrast (Hansen and Dueker, 2009). Positive receiver function arrivals at 42 and 61 km depth beneath the Archean craton within the Laramie Mountains suggest the presence of an imbricated Moho to the north of the Cheyenne belt as a result of Proterozoic lower crust underthrusting the Archean crust (Hansen and Dueker, 2009). An imbricated Moho could also lead to a gradational increase in seismic velocities with depth.

Many similar characteristics can be seen by comparing the crustal structure presented here along a north-south transect following the Deep Probe profile to the transects produced from the Deep Probe data presented by Snelson et al. (1998) and Gorman et al. (2002) (Fig. 13). By comparing the receiver function arrivals to the shear wave velocity model of Bensen et al. (2009), it becomes apparent that the receiver functions sampling the zone of high wave speeds within the Archean craton (between 43°N and ∼46.5°N) exhibit low-amplitude Moho arrivals. The low amplitude of the Moho arrival likely results from a reduction in the velocity contrast between the crust and mantle due to the presence of high-speed lower crustal material. It is also noteworthy that the lower crustal velocities increase toward the southern end of the zone of higher velocity lower crust in much the same manner as that identified by Gorman et al. (2002). Viewing the same region from an orthogonal perspective illustrates that the amplitude of the Moho arrival on receiver functions diminishes beneath the eastern edge of the Rocky Mountains compared to areas to the east and west, which is also the location of high-speed lower crustal material (between 108°W and 105°W; Fig. 14). Following the seismic observations and xenolith data that suggest the presence of mafic lower crust (e.g., Nelson, 1991), its potential influence on the pattern of deformation along the eastern Rockies in northern Wyoming can be considered. The mafic component of the thickened crust may be responsible for strengthening that portion of crust and leading it to resist internal deformation while focusing deformation along its margin. The Big Horn uplifts would mark a zone of deformation toward the margin of the area of strong mafic lower crust. This scenario implies that the strong underplated material within the Wyoming craton contributes to the pattern of deformation within the Rockies.

The crust thins to ∼30–35 km across much of the Rockies in Montana north of the Yellowstone caldera. The contrast of this thinner crust with the thicker crust in the High Plains to the east highlights the lack of a correlation between crustal thickness and topography in this part of the western United States (Fig. 15). The crust thickens to the east of the Rockies to between 45 and 50 km within a portion of the High Plains, creating a pattern of thicker crust underlying areas of lower elevations. East of the Rocky Mountains, the crust is mostly between 45 and 50 km thick beneath much of the High Plains, with localized areas of thickening corresponding to uplifts in eastern Wyoming and western South Dakota. Additional density associated with the high-velocity lower crustal layer beneath the Wyoming craton could explain how thick crust underlies areas of lower elevations beneath the High Plains.

A general trend emerges with areas of thicker crust in the Rocky Mountains, particularly beneath the high elevations within Colorado, being located with the most negative Bouguer gravity anomaly (Fig. 8). The Bouguer anomaly becomes more negative in the area of thickened crust toward southwestern Colorado, illustrating a relationship between low densities and areas of high topography. However, crust with thicknesses between 40 and 45 km within the Rockies does not explain the present regional elevations >3000 m. Additional buoyancy from the upper mantle, which appears to be warm based on lower velocities, may provide the necessary support for the high topography (Sheehan et al., 1995; Hansen et al., 2011). As the Rocky Mountains transition southward toward the Rio Grande Rift, the area of thicker crust shifts to the east side of the rift. Differently thinner crust is beneath the west side of the Rio Grande Rift and, as discussed above, extends beneath the southeastern side of the Colorado Plateau. This asymmetric pattern of crustal thickness with thin crust and a more negative Bouguer anomaly along the west side of the rift suggests that buoyant mantle material may be compensating the topography along the western side of the Rio Grande Rift and the southeastern portion of the Colorado Plateau.

Imaging crustal structure across the western portion of the United States clearly delineates distinct regions corresponding to tectonic provinces in the western United States. Although many of the large-scale structural trends presented here were previously known, this investigation provides new details, including a zone of magmatic underplating leading to crustal thickening along the Snake River Plain that is distinct from the High Lava Plains and the surrounding Rocky Mountains and Basin and Range. Observations presented here indicate that the Rocky Mountains have thin crust in much of Montana. The Rockies also have a large zone of reduced impedance contrast across the Moho resulting from fast lower crust toward the southern portion of Archean crust in southern Wyoming. This is in the same area that has been identified to have high-velocity lower crust (e.g., Snelson et al., 1998; Gorman et al., 2002), and suggests that preexisting structures influence more recent tectonic processes and persist through subsequent episodes of deformation. The northern Basin and Range generally exhibits thin crust nearly 30 km thick across much of Nevada, but the crust thickens toward the northwestern corner of the Basin and Range, a zone that has undergone less extension. The crust also thickens to >35 km in the southern portion of Nevada. The Moho within the Colorado Plateau exhibits a low-amplitude receiver function arrival, supporting ideas that it has a high-velocity lower crust that provides strength and allows it to resist internal deformation compared to surrounding areas. Comparing the crustal thickness of the Colorado Plateau to the Bouguer gravity anomaly suggests that a source of buoyancy comes from the mantle to support the elevation of the plateau, especially around its perimeter. The observations presented here provide necessary constraints for future geodynamic and seismic models, in order to answer questions related to the role of the crust in continental evolution.

This work benefited from thorough reviews by Matt Fouch and anonymous reviewers as well as from helpful discussions with Ken Dueker, Tom Owens, and George Zandt. National Science Foundation grants EAR-0454558 and EAR-0454554 supported this research. I appreciate the efforts of several investigators who contributed data from temporary deployments (archived at the IRIS [Incorporated Research Institutions for Seismology] Data Management Center) used in this study. The collected data are available through the IRIS Data Management Center. The National Science Foundation supports the facilities of the IRIS consortium under Cooperative Agreement EAR-0004370.