Abstract

Tomographic studies of the mantle of southern California (USA) commonly found evidence for seismically high speed material, known as the Isabella anomaly, extending from near the base of the crust of the southwestern Sierra Nevada foothills into the asthenosphere. This anomaly has been interpreted to mark downwelling lithospheric material that had been removed from the southern Sierra Nevada. Using data from the Sierra Nevada EarthScope Project (SNEP) array, we investigate the lithosphere of the Sierra Nevada and surrounding region to better understand the process by which batholiths form dense lithospheric roots that become unstable and founder. Inverting phase velocities of fundamental mode Rayleigh waves for shear wave speeds provides observations of the distribution of high and low wave-speed anomalies, which correspond to portions of the batholith that formed an intact lithospheric root, and where seismically slower shallow asthenosphere marks areas where lithosphere has been removed. Our results corroborate previous observations that the southern Sierra Nevada has thin crust underlain by shallow asthenosphere. High shear wave velocity (Vs) material in the mantle beneath the southwestern foothills marks the location of the Isabella anomaly, to the east of which is a region of low Vs mantle where asthenosphere has risen to replace the delaminating root. Farther north, near the latitude of Long Valley, low velocities at shallow depths beneath the high elevations of the eastern Sierra indicate the presence of asthenosphere close to the base of the crust. Thicker high-speed material, however, underlies the western foothills of the Sierra Nevada at this latitude and dips to the east where it extends to depths of ∼100 km or more, giving it the appearance of a portion of lithosphere that has detached from the east but remains attached to the west as it is currently peeling off. The structure of the Sierra Nevada changes near the latitude of Lake Tahoe, where thinner lithosphere extends between depths of 40 and 80 km, but does not reach greater depths. It appears that the lithospheric material of the Sierra Nevada from latitudes close to Lake Tahoe, and continuing to the north, is not being removed, indicating a change between the structure and evolution of the southern and northern Sierra Nevada.

INTRODUCTION

The process of lithospheric foundering or removal, involving the detachment of negatively buoyant portions of the lithosphere that then sink into the underlying mantle, remains poorly documented. Lithospheric removal has been proposed to have played a role in the development of a wide range of regions, including expansive areas such as the Tibetan Plateau (England and Houseman, 1989), the Altiplano (Kay and Mahlberg-Kay, 1991; Beck and Zandt, 2002), and the Alboran Sea (Seber et al., 1996; Calvert et al., 2000; Dündar et al., 2011), and more focused locations such as the Basin and Range (Platt and England, 1994; Humphreys, 1995; West et al., 2009), the Colorado Plateau (Bird, 1979; Karlstrom et al., 2008; Levander et al., 2011), the Wallowa Mountains in northeastern Oregon (Hales et al., 2005), and the southern portion of the Sierra Nevada batholith (Ducea and Saleeby, 1996; Saleeby, 2003; Zandt et al., 2004; Le Pourhiet et al., 2006; Saleeby et al., 2012). The process may also play a significant role in continental crust becoming more felsic as denser mafic material recycles back into the mantle (Kay and Mahlberg-Kay, 1991; Saleeby et al., 2003). The Sierra Nevada is of particular interest because there is evidence that some portion of its uplift is young. If its uplift were related to the loss of a dense residue root during the past 10 Ma, the lithospheric foundering process would have occurred recently enough (and may be ongoing) that evidence related to the removal process likely remains. Studying the structure of the Sierra Nevada could provide insight into the signature of material being removed from the continental lithosphere and the conditions under which this process occurs.

Here we present a model of shear wave velocities of the crust and upper mantle of the Sierra Nevada based on phase velocities of teleseismic Rayleigh waves. Using a dense array of seismometers in eastern California to record these Rayleigh waves allows us to produce a model that includes variations in velocities over short distances. As Rayleigh waves are sensitive to structures in the lower crust and upper mantle they are particularly well suited for studying lithospheric foundering. Because colder lithosphere has higher shear wave velocities than warm asthenosphere, differences in shear wave velocity anomalies provide insights into which locations of the Sierra Nevada have higher speed lithospheric material and where lithosphere has been replaced by lower speed, warm buoyant asthenosphere. This distinction helps illuminate how pervasive lithospheric foundering has been across the Sierra Nevada. Identifying patterns in the velocity structure and comparing those to geologic observations (e.g., Ducea and Saleeby, 1996; Manley et al., 2000; Cassel et al., 2009; Cecil et al., 2012; Saleeby et al., 2013) and dynamic models (e.g., Le Pourhiet et al., 2006; Saleeby et al., 2012) may help to discriminate whether the removal process occurred over a wide area during a single event or if it gradually progressed across the Sierra Nevada. Answers to these questions will constrain the conditions under which lithospheric foundering occurs.

SIERRA NEVADA BACKGROUND

As the Farallon plate subducted beneath North America during the Mesozoic, leading to an active arc along the margin of North America, accretion of island arcs welded volcanic and sedimentary rocks onto the continental margin. Plutonic bodies later intruded these accreted terranes, forming the Sierra Nevada batholith (Dickinson, 1981; Schweickert, 1981; Dickinson, 2008). A dense concentration of granitic plutons composes the southern and central sections of the Sierra Nevada batholith; more sparsely distributed plutons are to the north (Bateman and Eaton, 1967; Bateman, 1988). The composition of the batholith varies from quartz diorite or granodiorite in the east to more mafic lithologies in the west (Bateman and Eaton, 1967). A large part of the magmatism responsible for the formation of the southern portion of the batholith was primarily confined to two distinct episodes, 160–150 and 100–85 Ma (Ducea, 2001). Between those episodes, the westernmost zone of the Sierra Nevada was emplaced during the Early Cretaceous from 140 to 130 Ma along the transition between the western foothills of the Sierra Nevada and the Great Valley (Saleeby, 2007). Subsequently, the Fine Gold Intrusive Suite, one of the largest intrusive complexes in the Sierra Nevada batholith, was emplaced between 124 and 105 Ma (Lackey et al., 2012). During that time there were pulses of magmatism in the northern Sierra between ca. 130 and 100 Ma (Cecil et al., 2012). These events serve to illustrate that the Sierra Nevada batholith formed continuously over an extended time period through the Cretaceous (Cecil et al., 2012).

The overall composition of the Sierra Nevada changes northward as the proportion of intruded plutons within the country rock decreases. In addition, the tectonic setting into which the batholith was emplaced changes from continental rocks in the south to oceanic rocks in the north. The northern part of the Sierra Nevada has a lower mean elevation than the south, and the deepest portions of the batholith are exposed in the south (Ague and Brimhall, 1988; Cassel et al., 2009). The pattern of heat flow aligns with topography; areas of high heat flow are within the higher elevations, while lower elevations correspond to areas of lower heat flow (e.g., Saltus and Lachenbruch, 1991; Blackwell and Richards, 2004). Paleozoic and Mesozoic metavolcanic and metasedimentary roof pendants, which are present in the central Sierra, and Cenozoic volcanic rocks, which are well preserved in the north, are all absent in the south.

During the formation of granitic batholiths like the Sierra Nevada, in which granitoid plutons compose the majority of the upper crust, there must be a complementary production of a thick, deeper section of garnet pyroxenites that form as residues of the felsic granitoids (Ducea, 2001). Xenolith data indicate that the residue material becomes significantly denser at ∼1.5 GPa (depths of ∼45 km) as it transitions from a granulite to an eclogite facies assemblage (Ducea, 2002). Therefore only sufficiently thick arcs should form an unstable lithospheric root, while thinner arcs would not. The term “arclogite,” developed from the work of Anderson (2005), refers to the combined eclogitic cumulates and interlayered spinel-garnet peridotites that form within arcs. The existence of an arclogite root has been demonstrated for the southern Sierra from garnet-rich xenoliths entrained in 12–8 Ma basalt. No lavas younger than 4 Ma in the southern Sierra have been found to host arclogite xenoliths, suggesting delamination of the dense lower crust between 10 and 4 Ma (Ducea and Saleeby, 1996). The timing, consequences, and dynamics of lithospheric removal have been explored in a series of thermomechanical models by Le Pourhiet et al. (2006) and Saleeby et al. (2012) that were constructed for the specific case of the Sierra Nevada. The seismic observations presented here can be compared to predictions of these models to gain insight into the processes that led to the current distribution of seismic velocities.

The southern edge of the subducting Juan de Fuca plate migrated northward past the southern Sierra Nevada at 20–15 Ma (Atwater and Stock, 1998), the time when the ancestral Cascade arc became active in the area to the north of Yosemite National Park and south of Lake Tahoe (Fig. 1), as indicated by the westward sweep of basaltic andesite and dacite magmas (Busby et al., 2008; Cousens et al., 2008; Busby and Putirka, 2009). The southern edge of the Juan de Fuca plate would have migrated past the central Sierra by 5 Ma, and currently is to the north of Lake Tahoe (Atwater and Stock, 1998). Once the subducting Farallon plate migrated north, the lower crust and lithosphere of the southern Sierra Nevada would no longer be supported from below and could mechanically detach into the mantle with buoyant asthenosphere then rising to replace it. With the northward migration of the southern edge of the Juan de Fuca plate, the west coast of California transitioned from a convergent to a transform margin. Eruptions of basaltic magmas beginning in the Miocene in the southern and central Sierra and adjacent Basin and Range support the presence of shallow, upwelling asthenospheric material in the area (e.g., Manley et al., 2000). Episodes of potassium-rich volcanism during the Pliocene in the central and eastern Sierra may indicate melting and remobilization of the Precambrian mantle lithosphere. Alternatively, the melting of metasomatized peridotite beneath the lithosphere has also been proposed to produce the potassic volcanism (Elkins-Tanton and Grove, 2003), which was followed by basaltic and bimodal eruptions across the eastern Sierra and Walker Lane (Manley et al., 2000; Farmer et al., 2002).

The changing composition of xenoliths that erupted over time provides strong evidence that some portions of the Sierra Nevada lithosphere were removed. Xenoliths that erupted within the central part of the range during the Late Miocene indicate that, at that time, the lithosphere had a thick section of arclogite. This arclogite would have formed at depths of 40–90 km as a residue, or cumulate, of the granitoids that compose the upper portions of the batholith. The composition of xenoliths entrained in younger volcanics that erupted along the eastern side of the range in the Pliocene and Pleistocene differs from the older xenoliths in that they do not contain garnet pyroxenites, instead having high-temperature spinel peridotites that originated from similar depths (Ducea and Saleeby, 1996, 1998). Such a compositional change could result from the younger xenoliths sampling warm asthenospheric material instead of lithosphere. Questions persist regarding how the lithosphere beneath the Sierra Nevada has been modified to lead to the differences in xenolith composition.

Asymmetry of the Sierra Nevada structure is evident at the surface as well as at depth. High elevations along the southeastern portion of the mountain range abruptly drop into the much lower Owens Valley. Conversely, on its western side elevations gradually diminish, with a near constant slope from the high Sierra down to the Great Valley. Well cores from the Great Valley reveal that the batholithic basement extends at least half the width of the valley (Saleeby et al., 2009). Controlled source seismic data collected across the southern Sierra Nevada show low crustal velocities to depths of ∼30 km and have been interpreted to indicate that the granitic plutons found at the surface are deep seated (Fliedner and Ruppert, 1996; Ruppert et al., 1998; Fliedner et al., 2000). The signature of the Moho changes from a sharp boundary at 30 km beneath relatively thin crust on the eastern side, to a thick gradual transition at 50 km beneath the western foothills (Zandt et al., 2004; Frassetto et al., 2011). Low mantle velocities have been identified beneath the eastern Sierra along its margin with Owens Valley (e.g., Jones, 1987; Schmandt and Humphreys, 2010). Tomographic studies of the mantle (Aki, 1982; Biasi and Humphreys, 1992; Benz and Zandt, 1993; Boyd et al., 2004; Yang and Forsyth, 2006a; Burdick et al., 2009; Biasi, 2009; Schmandt and Humphreys, 2010) found evidence for a high-velocity body, known as the Isabella anomaly (Fig. 1), extending from near the base of the crust of the southwestern Sierra foothills to >200 km deep. Magnetotelluric studies similarly identified a resistive region that extends from the base of the crust to depths of ∼250 km near the boundary between the Sierra Nevada and the Great Valley (Park, 2004). This high-velocity, resistive body has been interpreted to mark downwelling mantle and lithospheric material that had been removed from the southern Sierra (Boyd et al., 2004; Zandt et al., 2004). The Isabella anomaly has also been interpreted as a dehydrated remnant piece of the subducted Farallon plate (Forsyth et al., 2011). A similarly high velocity anomaly, originally called the Redding anomaly, appears in the upper mantle of northern California beneath the Sierra and northernmost Great Valley. Body wave tomography shows that it is at the southern end of the subducting Juan de Fuca slab (e.g., Benz and Zandt, 1993; Burdick et al., 2009). However, in Jones et al. (2004) it was noted that the increased magnitude of this anomaly compared to other portions of the slab farther to the north could result from the Redding anomaly also containing descending Sierra Nevada lithosphere.

The lack of thick crust beneath the southeastern portion of the Sierra Nevada suggests the need for buoyant mantle to explain much of the ∼150 mGal negative Bouguer anomaly associated with the high Sierra (Oliver, 1977). Early geophysical studies suggested a shallow lithosphere-asthenosphere boundary beneath the Moho of the eastern Sierra Nevada based on seismic and heat-flow observations (e.g., Crough and Thompson, 1977; Jones, 1987). Explanation of these and other geophysical signatures in the southern and central Sierra Nevada involved the removal of an arclogite root from beneath the southern Sierra Nevada since ca. 10 Ma, leading to the observed tectonic and magmatic response. However, the details of the lithospheric removal process remain unclear: what conditions contribute to the formation of an unstable root? Does the lithosphere delaminate and peel away, or does it drip off as a Rayleigh-Taylor type instability? Has removal occurred along the entire Sierra Nevada or is it confined to the southern part? If removal does extend beyond the southern Sierra, did it occur all at once, or is it a time-transgressive feature?

The Sierra Nevada EarthScope Project (SNEP) array was designed to investigate lithospheric structure of the Sierra Nevada and further our understanding of the lithospheric removal process. Using data from the SNEP array, we identify which parts of the batholith formed a lithospheric root that remains intact and where lithosphere has been replaced by shallow asthenosphere. The dense station spacing of the SNEP array permits identifying variations in seismic structures in more detail than would be possible with an array of more sparsely spaced stations. We are particularly interested in distinguishing between areas of cold, dense, seismically fast lithosphere and arclogite from seismically slower, warm, buoyant asthenosphere.

SURFACE WAVE DATA

We investigate lithospheric structure using fundamental mode Rayleigh waves from more than 150 earthquakes with magnitudes >5.0 recorded by various arrays in eastern California with epicentral distances between 30° and 120°. The Rayleigh waves were recorded by stations from permanent networks and past temporary arrays that spanned the Sierra Nevada, including the Sierra Paradox Experiment (Jones et al., 1994; Boyd et al., 2004) and SNEP (Gilbert et al., 2007) temporary arrays and the neighboring Transportable Array component of the EarthScope USArray. The SNEP array comprised more than 50 broadband stations, the majority of which occupied 2 locations during a 30 month recording interval between the spring of 2005 and the fall of 2007. The average spacing between stations across the SNEP array area is ∼25 km and spans the area covering the Sierra Nevada batholith from Mount Lassen in the north to just north of the Garlock fault in the south and extends between the Great Valley and the Basin and Range (Fig. 1).

In order to measure phase velocities at a range of frequencies, we apply a series of narrow band-pass filters centered at frequencies between 10 and 50 mHz (periods between 20 and 100 s) to vertical seismograms that have been recorded by the seismometers used in this study. Because a heterogeneous collection of seismometers recorded the data analyzed here, all of the waveforms were transferred to the instrument responses of an STS-2 broadband seismometer. Following visual inspection of each of the waveforms, we only retain traces for which the signal:noise ratio of the Rayleigh wave was at least 10:1. More details of the data selection and processing procedures were given in Yang and Forsyth (2006a). Maps of ray coverage illustrate the dense sampling achieved in this study from all backazimuths, leading to numerous crossing rays within and along the perimeter of the array of stations (Fig. 2). Following removal of noisy seismograms, the sampling of the clean data set varies from period to period, the longest and shortest periods having fewer events with sufficient signal to make reliable phase velocity measurements (Table 1). Differences in the ray coverage between different frequencies illustrate the variations in sampling from the moderately well sampled shorter period data (30 s), to the well-sampled intermediate periods (50 s), and the more sparsely sampled longer periods (100 s) (Fig. 2).

Waveforms from surface waves that have traversed tectonic boundaries or through tectonically active areas exhibit changes in amplitude across the array of stations. These amplitude differences can result from multipathing, scattering, or focusing and defocusing of the wavefield as a result of topography or velocity heterogeneities along the ray path. Figure 3 illustrates seismograms recorded by Transportable Array stations for an earthquake that occurred ∼9800 km southwest of the study area. The effects of focusing and defocusing of the wavefield can be observed in the changing amplitudes of the waveforms that vary between smaller arrivals to the north and larger arrivals in the south.

To account for complexities of the incoming Rayleigh waves, the processing employed here treats the incoming wavefield as two interfering plane waves; this accounts for the wavefield arriving off its projected great circle path and the interference effects that produce amplitude variations. We construct our two-dimensional phase velocity maps following the methods in Yang and Forsyth (2006b; methods built on those in Forsyth and Li, 2005), wherein the amplitude and phase of Rayleigh waves are inverted for lateral variations in phase velocity and azimuthal anisotropy on a grid of nodes distributed across the study area. We use two-dimensional sensitivity kernels that account for the finite frequency effects on both the phase and amplitude variations of the surface waves; this leads to increased resolution (Yang and Forsyth, 2006b). The kernels account for the concentrated sensitivity of the Rayleigh waves within the first two Fresnel zones that decreases as the kernels broadens along the ray path with increasing distance from the station. Higher order Fresnel zones are relatively unimportant because the windowing of the seismograms leads to a finite effective bandwidth at each frequency producing destructive interference in the outer zones. The phase velocity map is constructed through an iterative process starting with a grid of nodes initially spaced by 1° and refined during successive iterations to 0.8°, 0.6°, and 0.4°, or ∼44 km. The grid of nodes extends beyond the array of stations such that any traveltime anomalies from outside the array that are not accounted for in the two-plane wave approach can be mapped onto these outer nodes. The grid nodes used to solve for azimuthal anisotropy terms remain fixed with a spacing of 1°. Because the dominant anisotropic terms for Rayleigh waves are the 2θ variations (Smith and Dahlen, 1973), in which θ is the direction of propagation, we restrict the inversion for azimuthal anisotropy to these terms.

DISTRIBUTION OF PHASE VELOCITIES

Rayleigh wave phase velocities provide information of the distribution of heterogeneous seismic velocities across the Sierra Nevada. To accurately account for rapid large lateral changes in seismic wave speeds associated with variations in crustal thickness or other heterogeneities, we construct an initial three-dimensional starting model within the study area based on previous seismological studies of crustal velocities (Moschetti et al., 2010) and thicknesses (Frassetto et al., 2011; Gilbert, 2012). Using this initial model, we calculate predicted phase velocities for each period at each node location throughout the region, assuming the mantle everywhere to uniformly have the structure of the TNA (Tectonic North America) model (Grand and Helmberger, 1984). The phase velocity maps are then calculated using these predicted phase velocities as the starting model at each node location for each period (Rau and Forsyth, 2011). This procedure minimizes inaccuracies resulting from mapping crustal features into the mantle and helps identify structural changes over shorter distances than could be identified using a uniform starting model during the phase velocity inversion (Wang et al., 2009; Rau and Forsyth, 2011).

Because the lateral resolution for Rayleigh waves at longer periods is not as fine as at shorter periods, the longer periods are effectively more heavily damped when inverted on a fine grid. Subsequent inversions for vertical shear velocity structure can then yield artifacts in anomalous regions because the relative amplitude of the phase velocity anomalies at different periods will be altered. Phase velocities are therefore first determined on a grid of nodes spaced 1° apart with linear interpolation of velocities between nodes. At this spacing all periods have roughly the same resolution as measured by the rank of the resolution matrix. The phase velocities are inverted point by point to construct a new three-dimensional shear velocity model with the initial model serving as the starting model. The phase velocities are then iteratively refined on grids for which the spacing between nodes decreases to 0.8°, 0.6°, and 0.4° (Fig. 4). The starting phase velocity model at each subsequent step is the set of phase velocities predicted for the new shear velocity model calculated at each step, rather than the actual observed phase velocities of the previous step. Using predicted rather than observed velocities as starting models reduces the period to period oscillations that can accumulate in an iterative inversion, effectively emphasizing the better resolved portions of the model. A minimum length constraint and a strong minimum curvature constraint that penalize variations from the starting model are imposed during both the phase and shear velocity inversions. At each step, the phase velocities are retained only for those periods with similar lateral resolution (defined as having rank ≥80% of the maximum rank for phase velocities at any period). Consequently, the longer periods are progressively eliminated as the grid is refined, although information from the longer periods is retained through its influence on the refined starting models. At the final grid spacing of 0.4°, only periods of 50 s and less are retained. This iterative approach has the advantages of largely eliminating underestimates of velocity anomalies caused by damping and of increasing lateral resolution at shallow depths constrained by the shorter periods. However, it must be recognized that both vertical and lateral resolution decay with increasing depth.

The effects of closer grid spacing and the use of iteratively refined starting models on the phase velocity inversions can be seen comparing the 30 s phase velocities inverted on a grid with nodes spaced from 1.0°, 0.8°, 0.6°, and 0.4° (Figs. 4A–4D). Smaller features that do not appear in inversions using more largely spaced grids (Fig. 4A) can be detected in the grids using finer spacing and the velocity gradients sharpen, such as in the vicinity of the Garlock fault (Fig. 4D). The large-scale trend of higher phase velocities to the west of, and within, the Sierra Nevada compared to the east remains.

Variations in phase velocities that are greater than approximately two times their standard deviation from the starting model can be viewed as statistically significant. We limit our interpretation of the phase velocity maps to areas that are sufficiently well sampled that the standard deviation of velocities for each grid node is <0.05 km/s for the well-sampled 30 s phase velocity map produced using the final 0.4° grid (Figs. 5A–5C). For periods sampled by a large number of Rayleigh wave paths, the phase velocity inversions have regions that satisfy our standard deviation cutoff that encompass an area extending beyond California into Nevada (Fig. 5B). However, for the shortest and longest periods used here, where larger amounts of noise and scattering lead to less useable data, the area with acceptable standard deviations is within a smaller region centered on the densely sampled Sierra Nevada (Fig. 5C). The standard deviations increase with increasing fineness of the grid, which means that in the iterative inversion for shear velocity structure, the a priori starting models are effectively given progressively more relative weight. Because the standard deviations do not account for correlated noise between stations, which could particularly be the case in areas sampled by the closely spaced SNEP stations, the standard deviation values presented here may slightly underestimate the actual standard deviations (e.g., Wagner et al., 2010).

For the entire study area, the average phase velocities range between 3.43 km/s at 20 s period and 3.97 km/s at 100 s (Table 1; Fig. 6A). At shorter periods, the average values measured here are smaller than the average values measured for southern California (Yang and Forsyth, 2006a), but our measurements approach similar velocities to those of southern California at periods >∼50 s. At a period of ∼50 s, the dispersion curve changes slope switching from concave down to concave up (Fig. 6A). Changes in the slope of phase velocity plots can be indicative of a shift in sensitivity from lithospheric to asthenospheric structures.

Differences between four dispersion curves created for locations interpolated between nodes on the 0.4° grid for periods up to and including 50 s, on the 0.6° grid for periods of 59–87 s, and on the 1.0° grid for 100 s period illustrate the effects of structural heterogeneity on phase velocities within the Sierra Nevada (Fig. 6A). Several signatures indicative of that heterogeneity can be identified based on these dispersion curves. Phase velocities that are greater than those observed in other locations characterize the western foothills for periods between 33 and 67 s. Both the Great Valley and the western foothills have phase velocities that are higher than those found in the central Sierra Nevada or Owens Valley for periods <50 s. Between 30–40 s, which are most sensitive to the 40–60 km depth range, the low phase velocities in Owens Valley are reduced by more than ∼3%–4% compared to the higher velocities in the western foothills. The dispersion curves converge at longer periods, indicative of the decreasing level of lateral variations at depth.

The distribution of high and low phase velocities produced on the 0.4° grid for periods of 20, 27, 40, and 50 s and on the 0.6° grid for periods of 67 and 77 s provides insight into crustal and upper mantle structure within and surrounding the Sierra Nevada. At short periods between 20 and 27 s, the pattern of phase velocities displays high values along the western side of the Sierra Nevada and lower phase velocities along the eastern Sierra Nevada. The higher phase velocities along the western portion of the batholith extend into the Great Valley (Figs. 7A, 7B). The northern extent of the batholith can be identified at shorter periods where higher phase velocities marking the northern Sierra give way to lower phase velocities in the active arc near Mount Lassen (Figs. 4D and 7B). The high-velocity anomaly located to the north of the southern end of the Sierra Nevada at 36°N, 119.5°W straddles the boundary between the Sierra Nevada and the Great Valley at periods to 77 s (Fig. 7F). This anomaly extends to the west into the Great Valley and continues to the coast at shorter periods (Figs. 7A–7E).

The high phase velocities present in the western foothills of the central Sierra at shorter periods shift toward the northern and southern ends of the Sierra Nevada for periods >50 s (Figs. 7E, 7F). The locations of high phase velocities at these longer periods straddle the boundary between the western Sierra Nevada and the Great Valley, toward the Gorda slab and Redding anomaly in the north and the Isabella anomaly in the south (Fig. 7E). The location of the northern anomaly shifts to the east for periods >67 s, reaching 121°W at 77 s (Fig. 7F). The eastward shift of elevated phase velocities with increasing periods appears to indicate that the high-velocity material responsible for producing these increased phase velocities dips to the east. The location of the highest phase velocities associated with the Isabella anomaly also shifts to the east for periods >40 s, the small eastward drift indicating only a slight dip of the shear velocity anomaly in that direction (Figs. 7C–7F).

In contrast to the batholith, low phase velocities mark the eastern boundary of the southern Sierra, where reduced phase velocities appear to wrap from the eastern side of the southern Sierra around to the west. Several areas that have undergone recent volcanism to the east of the Sierra Nevada are located in areas of reduced phase velocities in Owens Valley close to the area of the Long Valley Caldera (Figs. 7A, 7B). Lower phase velocities are to the north of the Sierra Nevada at shorter periods (e.g., Figs. 7A, 7B) and shift eastward with increasing period (Figs. 7C, 7D). The manner in which these low velocities remain east of the zone of high phase velocities that likely mark the Gorda slab suggests that they may correspond to low-velocity material above the slab.

In the Great Valley greater than average phase velocities continue from the western foothills into the valley and reach the coast at short periods of 20 and 25 s (Figs. 7A, 7B). The elevated phase velocities in this region at shorter periods actually appear to increase to the west across the Great Valley into the central portion of the Coast Ranges. High phase velocities are also present in the area south of the Garlock fault and continue southward into the Mojave province for periods between 20 and ∼50 s (Figs. 7A–7D) and into the Transverse Ranges at greater periods (Figs. 7E, 7F).

OVERVIEW OF FEATURES IN SHEAR WAVE VELOCITY MODEL

Because of the sensitivity of phase velocities to the integrated Earth structures sampled by Rayleigh waves, phase velocities need to be inverted for shear wave speeds to obtain direct information of physical properties of the Earth at specific depths. Our phase velocity maps for 14 periods between 20 and 100 s serve as input to invert for shear wave velocity, Vs, at each grid node across the study area. Because the Rayleigh waves used in this study are primarily sensitive to vertical shear wave speeds (Vsv), we are only able to solve for Vsv and are insensitive to horizontal shear wave speeds (Vsh). To account for the slight sensitivity of Rayleigh waves to crustal compressional wave velocity, Vp, and density variations, we assume that these parameters vary proportionally to Vs scaled by factors of 1.2 and 0.3 (kg m−3)/(km s−1), respectively, instead of leaving them as poorly resolved free parameters (Rau and Forsyth, 2011). Shear wave speeds are determined by interpolating the phase velocities observed at each period onto a more finely spaced grid of 0.25°. The dispersion curve for each grid node is then inverted for the Vs profile at that location using the model that was used to generate the predicted phase velocities at that period as the starting model. The inversion determines the necessary perturbations to the Vs starting model that will provide the best fit between observed and predicted phase velocities (Saito, 1988). The rank of the resolution matrices and the resolution kernels for the shear wave inversion using the 1.0° spaced grid provide a rough indication of the vertical resolution of the shear wave model presented here. In the well-sampled area in the vicinity of the SNEP stations, the rank of the resolution matrices is between 5 and 6, with the Rayleigh wave data contributing 2 pieces of information about crustal structure and 3–4 about mantle structure. The regions of peak sensitivity of the resolution kernels are ∼20–30 km wide at a depth of 50 km and the kernels widen to ∼50 km for 110 km depth.

The variations observed in dispersion curves for the locations along a profile highlighted here lead to changes in the velocity profiles from the Vs inversion (Fig. 6B). The velocity profile for the TNA model (Grand and Helmberger, 1984) is plotted for comparison; it has a shallow high-velocity lid where Vs increases to >4.4 km/s at depths between 40 and 60 km, below which is a low-velocity zone where velocities reduce to 4.2 km/s between depths of 100 and 150 km. A similar high-velocity lid can be found in the upper mantle of the profiles for the Great Valley and the western foothills. However, these are clearly different from the profiles for Owens Valley and the eastern Sierra Nevada, which do not have the same high-velocity lid. The Vs in the upper mantle of Owens Valley is <4.2 km/s to depths of 70 km and gradually increases to a maximum of 4.25 km/s at 100 km depth, where it becomes indistinguishable from the other example profiles. All 4 Vs profiles presented here have low-velocity zones where Vs decreases to a minimum of ∼4.1 km/s at depths between 140 and 180 km. This comparison illustrates the relatively low wave speeds of the Sierra Nevada and California compared to the TNA model at the top of the mantle and between depths of 150–200 km.

Comparing the Vs values observed here to those observed by seismic investigations elsewhere and measurements of various lithologies allows us to discriminate between areas underlain by cool lithospheric material and those underlain by warm asthenosphere. The Vs in the lithosphere is >4.5 km/s at ∼100 km depth in the IASP91 (International Association of Seismology and Physics of the Earth’s Interior) seismic velocity model (Kennett and Engdahl, 1991). However, in the seismically slower tectonically active portion of North America, Vs in the lithosphere reaches just >4.4 km/s (TNA; Grand and Helmberger, 1984). Lithospheric shear velocities are further reduced in southern California where Yang and Forsyth (2006a) identified average velocities of ∼4.3 km/s above a lower velocity asthenosphere where velocities decrease to <4.1 km/s at 125 km depth. Similarly low shear wave velocities in the upper mantle of California have also been found by previous investigations of arrivals from local events (e.g., Savage et al., 1994) and surface wave studies (e.g., Moschetti et al., 2010; Lin et al., 2011). Shear velocities in the mantle <∼4.3 km/s are difficult to generate as a result of only temperature or compositional variations (e.g., Wang et al., 2009), unless the sheer quality factor (Qμ) for the area is 30 or less. On average, this low of a value of Qμ is not found in the asthenosphere beneath southern California (Yang and Forsyth, 2008) or beneath the East Pacific Rise (Yang et al., 2007), which has similarly low Vs velocities. This pattern of low shear velocities with moderate attenuation (Qμ values of 50–60 in the asthenosphere) suggests that the velocities are affected by attenuation outside the band of seismic frequencies used here, most likely by the melt squirt mechanism, which would require <∼1% melt (Hammond and Humphreys, 2000). Either melt (Stixrude and Lithgow-Bertelloni, 2005) or a solid-state mechanism that would lead to relaxation of the elastic moduli outside the seismic frequency band could explain the observed low mantle velocities.

Crustal Structure of the Sierra Nevada and Surrounding Areas

In the upper crust, there are two primary features revealed by the Rayleigh wave tomography: high velocities in the western Sierra Nevada and low velocities in the sedimentary basins of the Great Valley. We show velocities in our model at 10 km (Fig. 8A), but note that given the depth resolution, these lateral velocity anomalies represent an integrated effect from the surface to ∼15 km. Some of the model structure at this depth comes from the initial starting model that incorporates information from surface waves at shorter periods than we measure obtained by ambient noise analysis (Moschetti et al., 2010). The anomalies around the periphery of the region are somewhat questionable, because at this depth, the primary control comes from the shortest period data, which are less numerous (Fig. 2), more scattered, and less well distributed in azimuth. The disruption of the Great Valley low-velocity anomaly at ∼37°N is due to thinning of the sediment cover at that latitude. The high-velocity Sierra Nevada block terminates to the north in the vicinity of the Mount Lassen volcanic field. In the south, it is truncated by the Garlock fault. Lower velocities lie east of the Kern Canyon fault where numerous extensional features have deformed the basement rocks of the Sierra Nevada and the seismicity indicates ongoing extension (Jones and Dollar, 1986; Mahéo et al., 2009). The highest velocity gradients generally are on the eastern side in the vicinity of the normal faults just to the east of the crest of the Sierra Nevada. The high velocities of the shallow crust in the western Sierra Nevada can be attributed to lack of sediments and an intact, largely unfaulted crustal block with metamorphic and plutonic rocks exposed at the surface. The Vs variations within, and surrounding, the Sierra Nevada can be viewed in cross sections perpendicular (Figs. 9A–9L) and parallel (Figs. 10W–10Z) to the axis of the mountain range.

At a depth of 25 km, representing the middle to lower crust, an area of high Vs (>3.9 km/s) extends along the western foothills of the central portion of the Sierra Nevada and continues across the Great Valley and Coast Ranges (Fig. 8B). Higher Vs values in the western foothills extend from the shallow crust to the mid-crust, where Vs is >3.9 km/s in locations of the central and northern foothills between lat ∼36.5°N and 39°N (Fig. 8B). Within the Sierra Nevada there are increased Vs values in areas where crustal thicknesses are >50 km beneath the south and central portions of the western foothills (Frassetto et al., 2011; Figs. 9B–9G). Receiver functions sampling the area of reduced velocity contrast resulting from the juxtaposition of high Vs lower crust overlying mantle with similar Vs would exhibit a diminished Moho signal (e.g., Figs. 9C–9F), as observed for the region along the western side of the Sierra (e.g., Zandt et al., 2004; Frassetto et al., 2011). In the mid-crust the region of elevated Vs continues to the west of the Sierra Nevada, through the Great Valley, and into the Coast Ranges, where Vs is >4.0 km/s (Fig. 8B). Elevated Vs values continuing from the western foothills into the Great Valley support previous observations that the Sierra Nevada batholith extends beneath the Great Valley (Saleeby et al., 2009).

While we are confident that the zone of high crustal Vs continues some distance into the Great Valley, the crust thins to the west to 30 km or even less. Because Rayleigh waves cannot resolve sharp discontinuities and have a vertical resolving length of ∼20 km at this 25 km target depth, the high velocities in the lower crust in the western Great Valley and Coastal Ranges could be due to blurring of mantle structure into the lower crust. The importance of this blurring depends on the correctness of the depth and velocity contrast at the Moho in the initial starting model. The uncertainty in structures identified to the west also increases due to the sparse station coverage outside of the Sierra Nevada.

The north-south extent of the 3.95 km/s contour along the western edge of the Sierra Nevada closely matches the north-south extent of the area Frassetto et al. (2011) identified as the petrologic Moho formed in the Mesozoic (Fig. 8B), characterized by a weak velocity contrast between the high-velocity, high-density mafic lower crust and the mantle. The Mesozoic petrologic Moho in this area would result from a sufficient amount of material being emplaced during batholith formation to produce a thick arclogite layer. East of the area where Vs in the mid-crust is >3.9 km/s, much of the eastern Sierra has low Vs, corresponding to the areas where the mountain ranges reach higher elevations and the denser, high-velocity mafic material has probably been removed. The lack of an arclogite root in this area supports the modeling predictions of Saleeby et al. (2012) that the root was removed from this area. Farther to the north along the western foothills, the northern limit of the 3.9 km/s contour appears to coincide with the termination of the region Frassetto et al. (2011) labeled as having Mesozoic tectonic Moho, which they interpreted as corresponding to the original Moho prior to the intrusion of the Sierra Nevada batholith. The correlation between the location where the original Moho can still be identified and the location of elevated Vs suggests that the crust in that area may be representative of the crustal structure that the batholith intruded into and was not subsequently modified by extension or lithospheric removal.

The other major features at 25 km are the pronounced low-velocity anomalies in the vicinity of recent magmatism (Fig. 8B), including the southern Cascades and the Long Valley, Big Pine, Golden Trout, and Coso volcanic fields east of the Sierra Nevada. The crust in these areas thins to between 30 and 40 km. Subdued low velocities continue to the south of Coso beneath the southernmost Sierra Nevada. The southern extent of the zone of low wave speeds continues along the eastern side of the Sierra and spreads to the southwest, where a zone of low Vs underlies areas of high elevation in the southern Sierra Nevada. Lake Tahoe marks the northern extent of the area of reduced Vs along the eastern portion of the Sierra Nevada. The east-west asymmetry becomes less pronounced to the north of Lake Tahoe where the northern Sierra lacks the signature of high Vs to the west and low Vs to the east (Figs. 9H, 9I). The northern Sierra does not have thick crust characterized by high wave speeds similar to that found in the western foothills farther south.

Upper Mantle Structure of the Sierra Nevada

The mantle beneath the Sierra Nevada also has asymmetric structures between the eastern and western sides of the mountain range. Similar to the crust, differences in mantle structures vary from being pronounced in the southern Sierra to subtler in the northern Sierra Nevada. High upper mantle velocities where Vs is >4.4 km/s underlie the zone of thickened crust along the boundary between the southern Sierra and the Great Valley (Figs. 9B–9D). This is the location of the Isabella anomaly, which appears prominently in this Vs model and has zones of increased Vs that are >4.5 km/s at depths near 80 km beneath the western foothills. The high velocities associated with this anomaly extend from the base of the crust to depths of ∼125 km beneath the southern Sierra between ∼36.5°N and 37.5°N (Fig. 11E). The magnitude of the anomaly diminishes with depth and the highest velocities shift slightly to the east with increasing depth. Along the western part of the southern portion of the Sierra Nevada, these high velocities appear to mark thick, high-wave-speed material that remains in contact with the overlying batholith. In the 55 km depth range, a zone of high Vs extends northward from the southern Sierra along the western foothills to close to the latitude of Yosemite National Park (∼38°N), where it shifts northward and extends toward the central portion of the batholith to latitudes just north of Lake Tahoe (Figs. 10Y, 10X, and 11A). The northward extent of the Isabella anomaly at depths between 55 and 70 km reaches ∼37.5°N, as can be seen in the northwest-southeast–trending cross section of the depth to the 4.3 km/s contour (Fig. 10Y) and in map view (Figs. 11A, 11B). Waveform modeling of P (compressional) waves from regional events indicates low velocities in the mantle beneath the southern Sierra Nevada and Walker Lane underlain by an abrupt transition to higher velocities at a depth of 75–100 km (Savage et al., 2003), thus a separation in that region between the batholith and high-velocity lithosphere.

Higher wave-speed material (Vs > 4.2 km/s) is in the uppermost mantle at depths of 60 km and greater beneath the western foothills under the area of thickened crust to the area north of the southern Sierra (Figs. 9B–9D and 11A–11E). The eastward dip of the Isabella anomaly can be discerned from the area of increased Vs reaching farther to the east at greater depths and shifting from the Great Valley to beneath the high elevations toward the central portion of the Sierra Nevada at depths >90 km for latitudes south of 37°N (Figs. 9C–9E and 11A–11C). Lower wave-speed mantle with Vs just >4 km/s is to the east of the Isabella anomaly and underlies the region of high elevations in the eastern Sierra from the base of the crust to depths shallower than 70 km (Figs. 9B–9F, 11A, and 11B). Higher wave-speed mantle underlies the area of shallow lower wave speeds in the mantle along the eastern portion of the southern and central Sierra Nevada to as far north as Lake Tahoe (Figs. 10W and 11B–11E). These zones of lower wave speeds suggest the presence of partial melt and appear to correspond to the location of shallow asthenosphere that replaced portions of the lithospheric root following its removal. This distribution of velocities appears similar to predictions from thermomechanical modeling of the distribution of lithosphere and asthenosphere following the removal of the dense root of the Sierra Nevada. Following removal, the root would be beneath lower velocity asthenosphere that ascended to replace it (Le Pourhiet et al., 2006; Saleeby et al., 2012). Within the northern portion of Owens Valley at depths of 70 km and more, Vs remains slightly >4.2 km/s, while the area of lower Vs that is reduced to <4.2 km/s does not extend beneath the entire length of Owens Valley as it does at 55 km, and instead is restricted to its southern end (Figs. 11B–11E). The presence of high Vs values at depths near 70 km beneath a body with lower Vs may correspond to lithospheric material at this depth range that is not present at shallower levels. The presence of descending portions of lithospheric material that have been removed and are now sinking into the asthenosphere (e.g., Le Pourhiet et al., 2006) could be responsible for this distribution of wave speeds.

A change in lithospheric structure coincides with the location where the Kern Canyon fault would project into the mantle, which appears to mark the western limit of a region of extension in the eastern Sierra (Mahéo et al., 2009) (Figs. 1, 8B, 9B, 9C, and 11A). This location correlates with the point where the velocity contrast between the crust based on receiver function amplitudes changes from the large contrast observed in the eastern Sierra Nevada to the diminished contrast to the west (Frassetto et al., 2011). To the east of the Kern Canyon fault Vs diminishes in both the crust and uppermost mantle (Figs. 9B, 9C), suggesting a link between the shallow extension and alteration of the mantle lithosphere.

To the north of Yosemite, close to the latitude of Lake Tahoe, the difference in shear wave speeds between the eastern and western sides of the Sierra Nevada diminishes. The Vs increases to >4.2 km/s at depths to 70 km on both the western and eastern sides of the northern portion of the Sierra (Figs. 9I, 9J, 11A, and 11B). Similar to the pattern of Vs observed in the crust, the lack of low velocities in the mantle on the eastern side of the northern Sierra offers evidence that processes that shaped the structure of the southern Sierra Nevada did not operate in the same manner in the northern Sierra Nevada. The lack of low Vs in to the east of the northern Sierra is consistent with the decreased level of recent magmatic activity in the area to the north of Lake Tahoe. In addition, the high Vs lithospheric material in the mantle to the north does not extend as deep as in the south, extending to only ∼90 km at its deepest (Figs. 10W, 10X).

Toward the northern portion of the Sierra Nevada, areas of higher Vs in the mantle coincide with the location of the Gorda slab (Figs. 9J–9L and 11B–11E). The anomaly associated with the Gorda slab remains consistent with increasing depth (Figs. 9K, 9L, and 11A), in contrast to the Isabella anomaly, which appears as a stronger Vs anomaly at shallow depths and weakens in amplitude with greater depth. To the east of the high Vs anomaly associated with the Gorda slab is a region of decreased Vs that consistently shifts to the east as the position of the slab progresses to the east (Figs. 11C–11E). The flux of volatiles from the slab into the overlying mantle wedge could be responsible for producing this low Vs region. The areas of high Vs imaged in the mantle here appear to be related to the Gorda slab and not to other high-speed material associated with another body that has been proposed to produce the Redding anomaly.

Structures Surrounding the Sierra Nevada

Toward the top of the upper mantle, to the south of the Sierra Nevada, the area of high Vs associated with the Isabella anomaly in the southern part of the Great Valley continues southward through the Mojave province and into the Transverse Ranges at a depth of 55 km (Figs. 10Z and 11A). The high Vs values also continue to the northwest from the Isabella anomaly into the Coast Ranges. High-wave-speed material has previously been identified in the upper mantle beneath the Transverse Ranges (e.g., Hadley and Kanamori, 1977; Humphreys et al., 1984; Kohler et al., 2003; Yang and Forsyth, 2006a). Similar to the results in Yang and Forsyth (2006a), the anomaly beneath the Transverse Ranges presented here diverges into two separate bodies that extend toward the eastern and western ends of the range with increasing depths. The eastern anomaly migrates to beneath the Mojave province at greater depths. The form of the Transverse Range anomaly presented here has a gap as the regions where the anomaly reaches its greatest amplitude diverge into two separate bodies at depths >70 km (Fig. 11B). This form of the Transverse Range anomaly supports the idea that it represents localized drips that could entrain, and remove, lithospheric material from the surrounding area (Yang and Forsyth, 2006a).

High-wave-speed material extends northward from the Transverse Range and across the Mojave province at depths between ∼50 and 70 km (Figs. 10W, 10X, and 11A). Seismic observations of high subcrustal wave speeds beneath the Mojave province have been suggested to result from an increasing amount of oceanic crust within the North American lithosphere with depth (e.g., Fuis et al., 2003; Yan et al., 2005). Mantle xenoliths from the central Mojave province also suggest that its lower lithosphere represents tectonically subcreted and imbricated material from an oceanic protolith (Luffi et al., 2009). This subcreted material could be a portion of the oceanic Farallon slab that accreted during low-angle subduction and currently composes part of the lithosphere beneath the Mojave province (Luffi et al., 2009). The region to the north of San Francisco Bay is also an area of high Vs at the top of the upper mantle; however, the distribution of stations used here has better resolution to the east of the Great Valley. Still, it is interesting to note the location of this anomaly, as results from earlier body wave tomography also identified high seismic velocities in this area and attributed them to the presence of a small portion of the Pacific plate that subducted beneath the Coast Ranges (Benz et al., 1992).

The Vs images presented here indicate areas of higher Vs extending westward from the Isabella anomaly at the southern end of the Great Valley (e.g., Figs. 11A, 11B), particularly compared to those observed farther north in the valley (e.g., Figs. 11B–11D). Though the high velocities surrounding the Isabella anomaly do not appear to continue to the coast in the same manner as the high Vs region to the north of San Francisco Bay and the area associated with the Gorda slab, they appear to spread across the San Andres fault at depths of 70 km or less (e.g., Figs. 11A, 11B). The lack of sampling and diminished resolution along the coast (Figs. 5A–5C) prohibit us from determining whether the zone of high Vs extending to the west of the Isabella anomaly is a stalled segment of subducting lithosphere (Forsyth et al., 2011). It is notable, however, that the stalled slab explanation for the high Vs material does not explain the stratigraphic history observed in the southern Great Valley (Saleeby and Foster, 2004; Saleeby et al., 2013) that fits well with the foundering root hypothesis. Connecting the crustal structure of the Great Valley to the Franciscan complex in the adjacent Coast Ranges suggests that a significant slab of Franciscan eclogite and blueschist accreted against the western margin of the arclogitic root and was subsequently drawn westward back out of the base of the Late Cretaceous subduction zone (Saleeby et al., 2012). The presence of Franciscan eclogite and blueschist material beneath the crust of the Great Valley offers an alternative explanation for the shallow zone of high Vs extending to the west of the Isabella anomaly.

DISCUSSION

As outlined above in the description of features in the Vs model, the Isabella anomaly imaged here extends to ∼125 km depth and appears to dip to the east (Figs. 9C–9E), 11B, and 11C. Along the vertical profile in the Vs model where the Isabella anomaly is close to the crust at shallow depths, the velocities actually decrease at greater depths (Figs. 6B and 9B). While there are concerns of oscillating patterns of high and low velocities when inverting phase velocities for Vs, that does not appear to be the case here; the dispersion curve for the western foothills in this area shows that phase velocities for periods >40 s do not increase as rapidly with increasing period as they do for shorter periods (Fig. 6A). It therefore appears that Vs does not continue to increase with increasing depth below the shallow portion of the Isabella anomaly, suggesting a dip to the anomaly.

Observations of the dip of a the shallow portion of the Isabella anomaly do not agree with P- and S-wave tomography that identified the portion of the Isabella anomaly where the magnitude of the anomaly is greater to be more of a vertical structure to depths of ∼150 km and dip to the northeast at greater depths (e.g., Schmandt and Humphreys, 2010). A heterogeneous distribution of anisotropy characterizes southern California along the boundary between the Sierra Nevada and the Great Valley. Changes from high to low SKS splitting times (Bastow et al., 2006; Zandt and Humphreys, 2008) and discrepancies between the velocities of short-period Rayleigh and Love waves (Moschetti et al., 2010) indicate that this region has pronounced variations in azimuthal and radial anisotropy. Additional evidence for anisotropy in the mantle comes from phase velocities of longer period Love and Rayleigh waves in southern California that also exhibit a clear discrepancy that cannot be explained by an isotropic model (Polet and Kanamori, 1997). The presence of laterally varying anisotropy could contribute to different results between isotropic Vp and Vs inversions.

The regional Vs values observed, i.e., 4.3–4.4 km/s for the lithosphere, can be compared to the lower Vs values expected for melt-rich asthenosphere. These differences in Vs facilitate discriminating between high-speed areas of the upper mantle that likely correspond to lithosphere or arclogite that can be contrasted with areas of lower Vs, that correspond to locations of inflowing asthenosphere. Calculating seismic properties for rocks with compositions of the garnet-bearing xenoliths from the southern Sierra Nevada, Frassetto et al. (2011) determined that Vs of arclogite would be between ∼4.3 and 4.4 km/s at pressures corresponding to the 30–50 km depth range. Exposed cumulates in the lower crust of the Talkeetna arc of southern Alaska (Behn and Kelemen, 2006) have a composition analogous to that expected for the arclogite layer underlying the Sierra Nevada granitic batholith, and can provide estimates of the range of Vs expected for the arclogite. These rocks exhibit a large change in velocity at pressures corresponding to the depth range of 25–75 km where the gabbro-eclogite phase change occurs and Vs increases from 4.0 to 4.4 km/s. Guided by these values and observations of the distribution of shear wave speeds, it is possible to identify areas where the lithospheric root of the Sierra Nevada remains and where it was replaced by melt-rich asthenosphere.

Where Lithosphere Was Removed

Low shear velocities (<4.2 km/s) extend along the eastern portion of the southern Sierra Nevada at depths from the base of the crust to depths >55 km. The low velocities that characterize much of the eastern Sierra Nevada indicate the lack of significant amounts of lithosphere or arclogite in the upper mantle. Instead, Vs values near 4 km/s, closer to expectations for asthenosphere, are directly beneath the crust. The area of low wave speeds below 4.2 km/s between the Moho and the ∼80 km depth range beneath the eastern Sierra extends from Long Valley and continues south through Owens Valley. No area of similarly low velocities is at the top of the upper mantle beneath the northern Sierra Nevada. This suggests that if the removal process is responsible for producing zones of low-velocity material in the crust and upper mantle, then removal has been more extensive in the southern Sierra Nevada than in the north.

Within the area of low shear wave speeds along the eastern Sierra Nevada, focused zones where Vs is further reduced in the crust and upper mantle coincide with locations of recent volcanism (e.g., Long Valley, Golden Trout, Big Pine, and Coso; Manley et al., 2000) (Figs. 8B and 11A). The low velocities could mark the presence of partial melt, magma chambers, or fluids associated with volcanic activity. Partial melt would likely accompany inflowing, upwelling asthenosphere. Comparing observations from the Long Valley caldera to those from the Yellowstone caldera, the low Vs values observed here, in which Vs is ∼4 km/s at the Moho, do not extend as deep as those found beneath Yellowstone, where Vs decreases to 3.9 km/s at 70 km depth (Schutt et al., 2008). This difference could result from lower temperatures in the upwelling mantle beneath Long Valley leading to shallower onset of pressure release melting than beneath the Yellowstone caldera.

The modeling of both Le Pourhiet et al. (2006) and Saleeby et al. (2012) found the removal of the Sierra Nevada mantle lithosphere as a Rayleigh-Taylor instability to be a necessary condition for the arclogite root to subsequently delaminate. This makes the removal of arclogite from the base of a batholith a two-stage process. This modeling further predicts that fragments of arclogite and lithosphere could then potentially reascend while entrained in the asthenosphere that is rising in the Rayleigh-Taylor return flow. The presence of reascended lithospheric material offers an explanation for the isostatic modeling results of Schulte-Pelkum et al. (2011), who identified intact lithosphere present beneath the California-Nevada border, while the seismic images presented here show asthenosphere above the lithosphere in that area. As illustrated by the modeling of Saleeby et al. (2012), the majority of descending and reascending material becomes laterally displaced relative to where it was located prior to being removed. This can explain why the high-velocity anomalies in the central Sierra Nevada appear to dip to the east and extend beneath Owens Valley, as imaged here and by previous investigations of seismic velocities (e.g., Savage et al., 2003; Boyd et al., 2004; Yang and Forsyth, 2006b).

Summary of Where the Lithosphere Remains Intact

Xenoliths that erupted within the central part of the range during the Late Miocene indicate that the lithosphere then had a thick section of arclogite. These arclogites would have formed at depths of 40–90 km as a residue, or cumulate, of the granitoids that compose the upper portions of the batholith. Based on observations from xenoliths that erupted during the Late Miocene indicating the presence of a thick section of arclogite, areas in the mantle with Vs of 4.2 km/s or more appear to correspond to either portions of residuum or where preserved lithosphere remains intact.

Locations where high Vs material remains connected to the overlying crust include the western foothills and localized regions near the center of the southern and central Sierra Nevada and the central and western portions of the northern Sierra Nevada (Figs. 11A, 11B). The thickness of the remaining lithosphere or arclogite varies along the length of the mountain range, reaching depths of ∼100 km or more in the southern part of the western foothills (Figs. 10X, 10Y, and 11D). The continuation of the high Vs anomaly beneath the southern Great Valley at depths between 55 and 90 km make it appear that the batholith underlies portions of the Great Valley, in agreement with inferences based on basement cores from the southern part of the valley (Saleeby et al., 2009).

Northward, the zone of higher wave-speed material within the upper mantle thins beneath the central and northern Sierra Nevada, where velocities only exceed 4.2 km/s to depths of ∼70–80 km to the north of Lake Tahoe near 39.5°N (Figs. 10W, 10X, 11B, and 11C). Geochemical evidence indicates that at least part of the northern Sierra Nevada batholith equilibrated with arclogite cumulates, although the northern batholith is more spatially dispersed, and was emplaced into thinner preexisting crust; this appears to have inhibited the accumulation of a substantial arclogite layer (Cecil et al., 2012). Uncertainties remain regarding the continuation of the arclogite root into the northern Sierra due to differences between the northern and southern Sierra Nevada. Because the width of the Sierra Nevada arc increases in the north, it is unlikely that it would have formed a thick arclogitic root similar to that of the narrower southern Sierra Nevada. Instead the increased width to the north could have led to the production of no arclogitic root or a thinner root distributed over a broader area. In addition, the greater proportion of metamorphic rocks in the northern Sierra Nevada may have caused the more sparsely emplaced granites to produce a smaller amount of arclogite during the formation of a root (Bateman and Eaton, 1967). We observe high velocities from the base of the crust to 80 km depth in the northern Sierra Nevada (Figs. 9G–9I, 11C, and 11D); this could indicate that a root formed beneath a limited portion of the northern Sierra to the west of Lake Tahoe and that it has not yet been removed. Perhaps it lacks the negative buoyancy necessary for the removal process to initiate due to its reduced thickness. Alternatively, the more sparsely emplaced batholith during the Cretaceous may have intruded the crust in the northern Sierra in such a diffuse manner that it did not disrupt significantly the mantle wedge in this area. In this scenario the observed high velocities beneath the crust in the northern Sierra Nevada would mark a remnant of the peridotitic mantle wedge and a lack of significant arclogite cumulates.

Places Where Lithosphere Is Currently Being Removed—Peeling Off

The eastern portion of the central Sierra Nevada lacks high-velocity material at the base of the crust, suggesting that lithospheric material that had previously been present was removed. Instead a layer of lower Vs material extends to depths between 55 and 70 km beneath Owens Valley along the eastern Sierra and extends southward to the southeast portion of the batholith. The distribution of low velocities along the eastern Sierra largely follows the extent of the high velocities along the western Sierra Nevada. However, zones of higher Vs exist at depths >70 km that connect to bodies of high Vs at the base of the crust beneath the western foothills (Figs. 9B–9D and 11A–11C). This appears to be the location of the high-speed arclogite that has been removed from the eastern part of the central Sierra Nevada. The low Vs material above the high Vs layer would correspond to buoyant asthenosphere that rose and replaced the lithosphere that was removed; this agrees with modeling predictions (Le Pourhiet et al., 2006; Saleeby et al., 2012). Though partially removed, the remaining arclogite root along the western foothills near the latitude of Yosemite National Park appears to be actively delaminating. Stresses originating from delamination in this area offer a potential explanation for the enigmatic lower crustal earthquakes occurring beneath the western foothills in this area at depths as great as 38 km (Gilbert et al., 2007; Frassetto et al., 2011). The interlayering of asthenosphere and mantle lithosphere between lat 37°N and 38°N in Figures 10W and 10X appears to be similar to the complex interlayering structures predicted by the modeling. However, note that these layers are nearly 25 km thick, near the limit of the resolution of the Vs model in this depth range.

A consistent result in all the modeling runs of both Le Pourhiet et al. (2006) and Saleeby et al. (2012) is that the higher topography of the Sierra Nevada and Nevada Plano relative to the Great Valley led to their greater gravitational potential driving lower crustal flow westward. Following this pattern of flow, the arclogite along the eastern Sierra Nevada could then delaminate in the central portion of the Sierra Nevada and, once removed from the base of the batholith, would lead to the area of low Vs in the shallow mantle along the eastern Sierra between 36°N and 38°N (Figs. 11A, 11B). Saleeby et al. (2012) speculated that a portion of this delaminating arclogite eventually necked off and dropped into the mantle. This would leave partially delaminated arclogite beneath the western Sierra Nevada and explain the zone of high Vs beneath the western foothills that dips to the east (Fig. 9C–9F). Material peeling away in this manner aligns more with expectations of lithospheric delamination (Le Pourhiet et al., 2006; Saleeby et al., 2012) than with those of a Rayleigh-Taylor type convective instability, which would predict material descending as vertical drips (e.g., Elkins-Tanton, 2005).

Due to the lack of high topography to the south of the Sierra Nevada, no gravitational potential drove northward-directed lower crustal flow, which thereby inhibited an initial south-north progression in delamination. This would also explain why delamination did not follow the opening of the slab window (Saleeby et al., 2012). It is likely that the high Vs body imaged here as the Isabella anomaly also followed an east-west path while initially delaminating. However, following the foundering of a portion of arclogite from the central Sierra Nevada, the root in the south would be able to peel away from south to north. The distribution of low Vs material to the south and east of the Isabella anomaly at depths >70 km is nearly collocated with the location of the thermal transient and distribution of ca. 4 Ma or younger volcanic rocks presented in Saleeby et al. (2012). The locations of the high and low Vs regions presented here, combined with their additional geomorphic and stratigraphic evidence, further support the suggestion of Saleeby et al. (2012) that the southern part of the Sierra Nevada is a location of active east to west and south to north delamination.

CONCLUSIONS

The shear velocity model presented here includes features within the Sierra Nevada corresponding to various stages in the process of the removal of a lithospheric root. Areas with very little or no high-speed lithospheric material present represent the final stage of removal; these are places where the lithosphere has been removed and replaced by asthenosphere, as in the eastern Sierra Nevada spanning from its southern limit to ∼38.5°N. The shallow low velocities found directly below the base of the crust in the southern and eastern Sierra Nevada indicate the lack of high-velocity lithosphere. The inflowing asthenosphere in these areas would provide the necessary buoyancy to support the high elevations of the southern and eastern Sierra Nevada.

The juxtaposition of high-velocity mantle material along the western portion of the Sierra Nevada with lower velocities in the shallow mantle to the east that are underlain by higher velocities marks the location of ongoing lithospheric removal. The high-velocity material at greater depth to the east corresponds to lithosphere that was removed and now underlies inflowing asthenosphere that replaced descending dense material. The removal process at this stage appears to be progressing westward and to the north for the southern portion of the Sierra Nevada. This would predict removal of the high Vs material currently observed beneath the western foothills of the Sierra Nevada in the future.

Thinner high Vs lithospheric material underlies the northern Sierra Nevada than that found beneath the western foothills along the central portion of the mountain range. Without additional information it is difficult to determine the origin of this thinner lithosphere. One possibility is that because granitic plutons have a more limited distribution in the northern portion of the batholith than in the south, the northern portion of the batholith would correspondingly form a thinner arclogite root. This portion of lithosphere would then lack the necessary negative buoyancy to be removed. Another possibility is that the high-velocity material marks lithosphere that was present prior to the emplacement of the batholith. Regardless of the origin of the lithospheric material in the northern Sierra Nevada, there is no clear evidence that lithospheric removal occurred to the north of Lake Tahoe. There are no regions of pronounced low velocities in the shallow crust or at the top of the upper mantle, as in the southern Sierra Nevada. Comparing the distribution of high Vs crustal material to variations in Moho signal, it appears that the extent of the petrologic Moho marking where a thick arclogite root formed, and remains intact, mirrors the north-south limits of low crustal Vs beneath the eastern Sierra Nevada. This would suggest that the removal of the thickened root is required for the upwelling low Vs material to rise up in the mantle and produce basaltic melt that intrudes the crust.

We are grateful for thoughtful reviews by David Hill, Brandon Schmandt, and Jason Saleeby that improved this manuscript. We thank those involved in the Sierra Nevada EarthScope Project (SNEP) field program and the private landowners and public officials who provided access to the land that hosted SNEP stations. Grants EAR-0454554, EAR-0454524, and EAR-0454535 from the National Science Foundation (NSF) EarthScope program to the Universities of Arizona, Colorado, and South Carolina, and Purdue University helped support this research. This is contribution 203 of the ARC (Australian Research Council) Centre of Excellence for Core to Crust Fluid Systems (http://www.ccfs.mq.edu.au) and contribution 845 of the GEMOC (Geochemical Evolution and Metallogeny of Continents) Key Centre (http://www.gemoc.mq.edu.au). The instruments used in the SNEP field deployment were provided by the EarthScope FlexArray, which is managed by the Incorporated Research Institutions for Seismology (IRIS) through the PASSCAL (Program for Array Seismic Studies of the Continental Lithosphere) Instrument Center. The collected data are available through the IRIS Data Management Center. The NSF under cooperative agreement EAR-0004370 supports the facilities of the IRIS Consortium. Data from the Transportable Array network were made freely available as part of the EarthScope USArray facility supported by the NSF Major Research Facility program under cooperative agreement EAR-0350030.