Using a network of temporarily deployed broadband seismometers, we characterize an unusual region of crustal earthquakes in the west-central Sierra Nevada, California (USA). We locate 131 earthquakes, which occurred from 3.1 to 47.1 km deep during June 2005 to May 2006. We detect more events, at greater depths, than are present in the Northern California Seismic Network catalog during this period. Most of the events occur at depths of 20–35 km and cluster into two distinct groups. In addition, some of the events appear to be repeating due to the similarity of their waveforms and locations. We calculate focal mechanisms for 52 of these events, and about half exhibit reverse faulting, which represents a state of horizontal compressional stress that is distinct from the regional stress field. From first arrivals, we calculate a one-dimensional model of crustal P-wavespeeds, which resolves a gradational increase from 5.8 km/s near the surface to 6.7 km/s at 35 km depth. The events overlie a significant variation in the character of the Moho, and two long-period events occur near the seismically imaged Moho at nearly 40 km depth. We suggest that these earthquakes could be the seismogenic response of the crust to active foundering of mafic-ultramafic lithosphere and resultant asthenospheric upwelling beneath the central Sierra Nevada.

Deep intraplate earthquakes occur within the lower crust and upper mantle of the western United States (Wong and Chapman, 1990) in areas with a wide range of geologic and tectonic histories, including multiple regions within the Colorado Plateau, near zones of Sevier-Laramide deformation in the Rocky Mountains and Wyoming craton (Frohlich et al., 2014; O’Rourke et al., 2016), and across a plate boundary system in California. Much of this seismicity is located below 20 km depth and coincides with areas of locally low heat flow, which is indicative of temperatures low enough to allow for brittle deformation at unusually great depths (Wong and Chapman, 1990). In the eastern San Joaquin Valley and western Sierra Nevada foothills of central California, small (ML ≤3.2) earthquakes are found to ∼40 km depth (Wong and Savage, 1983; Miller and Mooney, 1994). These events have been inferred to occur on crustal faults but appear to be unrelated to tectonically active features at the surface, and their cause has been heretofore unknown.

The Sierra Nevada EarthScope Project (SNEP) (Owens et al., 2005) consisted of an array of three-component broadband seismometers with a north-south extent of >400 km north from about Fresno, California, and a west-east extent of >150 km from the Great Valley to the Basin and Range (Fig. 1). The primary purpose of the SNEP was to image the deep structure of the Sierra Nevada batholith in order to better characterize the location and disposition of mafic and ultramafic lower crust and upper mantle beneath the Sierra Nevada. This root is inferred to be gravitationally unstable within the surrounding lithosphere and foundering in places beneath the Sierra Nevada (Ducea and Saleeby, 1996, 1998b; Jones et al., 2004; Zandt, 2003), a process typified by the “Isabella anomaly,” which is evident in regional seismic tomography (Benz and Zandt, 1993; Biasi and Humphreys, 1992; Jones et al., 1994). The uniform spacing (∼25 km) of the SNEP stations, when combined with the larger framework of the EarthScope Transportable Array (TA) and permanent regional monitoring stations, provides the opportunity to study the unusual earthquakes of the western Sierra Nevada in a manner not previously possible. Here we evaluate this seismicity and explore its potential causes.

The Sierra Nevada consists primarily of metamorphic and plutonic rocks that separate the Great Valley forearc basin on its western edge from the Basin and Range extensional domain to its east (e.g., Bateman and Wahrhaftig, 1966). Its crust records Paleozoic- to Mesozoic-age terrane accretion and batholith-producing crustal melting as subduction of the Farallon plate fueled an extensive Cordilleran arc (e.g., Ducea, 2001; Dickinson, 2008), with most of the batholith being emplaced by 85 Ma (e.g., Ducea, 2001). Subduction-related volcanism reemerged during the Miocene between present-day Yosemite National Park and Lake Tahoe (Busby and Putirka, 2009; Cousens et al., 2008). Small volumes of magmas began to erupt across the southern and central Sierra Nevada beginning ca. 12 Ma (Manley et al., 2000; Moore and Dodge, 1980; Dodge and Moore, 1981). Xenoliths in the oldest phase of these eruptions record near-Moho eclogitic garnet- and pyroxene-rich material present as recently as 10–8 Ma beneath the central Sierra Nevada batholith (Ducea and Saleeby, 1996). This mafic-ultramafic material, often termed arclogite, dates contemporaneously with the overlying felsic plutons and represents an unusually dense (∼3500 kg/m3) root of complimentary residue and cumulates left over by the extraction of felsic melt (Ducea and Saleeby, 1998a; Lee et al., 2001).

The cessation of subduction beneath southwestern North America, opening of a slab window, and establishment of a transform boundary (e.g., Dickinson, 1979) appear to have significantly affected the Sierra Nevada during the mid- to late Tertiary. The Sierra Nevada block experienced normal faulting on its eastern margin, which has commonly been associated with uplift and, in the central and northern Sierra, westward tilting (Christensen, 1966), though the timing of Sierran uplift remains disputed (e.g., House et al., 1998; Mulch et al., 2006). The subsequent eruption of more typical, asthenospheric spinel peridotite mantle xenoliths in the eastern Sierra Nevada suggests that the dense root foundered and was replaced beneath parts of the range between ca. 10 and 3 Ma (Ducea and Saleeby, 1996, 1998b). Evidence for uplift in the southern Sierra Nevada and an increase in extensional strain rates in the adjacent Basin and Range ca. 3.5 Ma indicates ongoing detachment during this period. Volcanism in the eastern Sierra (e.g., Long Valley caldera) may result from the subsequent inflow of asthenosphere to replace the foundered material (Jones et al., 2004).

Any lingering arclogite and old mantle lithosphere would manifest as high-wavespeed upper mantle in tomography results (Gilbert et al., 2012; Jones et al., 2014) and in receiver functions, which clearly show that the crust-mantle boundary (Mohorovičić discontinuity or Moho) deepens and becomes much less distinct west from the high Sierra Nevada toward the Sierra foothills and Central Valley (Zandt et al., 2004; Frassetto et al., 2011). A westward increase in crustal wavespeeds reduces the wavespeed contrast across the Moho, making it difficult to resolve (Frassetto et al., 2011). The region where the dense residuum still resides appears to be limited to the western portion of the Sierra Nevada south of Lake Tahoe (Frassetto et al., 2011; Gilbert et al., 2012). These western Sierra foothills have experienced both Quaternary subsidence to the south (Saleeby and Foster, 2004; Sousa et al., 2017) and uplift to the north (Jones et al., 2014; Saleeby et al., 2013).

Our data from the SNEP array focus on the first year of deployment between the springs of 2005 and 2006, when SNEP stations were in the vicinity of the events of interest. We used waveforms from 46 broadband seismometers recording continuously at 40 samples per second (sps) (Owens et al., 2005). SNEP data are archived at the IRIS Data Management Center ( The stations used extended from Fresno to just north of Yosemite National Park (Fig. 1; Table 1). Three of the backcountry stations within the SNEP array (SNP14, SNP65, SNP75) recorded at 20 sps during part or all of their deployment. Data from other temporary and permanent stations in the region were also included in our analysis. We ran a STA/LTA (ratio of short-term signal average to long-term average to identify seismic arrival) event detector in the Antelope software package (Boulder Real Time Technologies; to identify potential events. Although the SNEP array recorded extensive seismicity within the Owens Valley and Long Valley caldera regions, in this paper we focus on the less-well-understood events located in the western portions of the SNEP array. P- and S-wave arrivals detected and quality checked in Antelope were manually picked in SAC (Seismic Analysis Code; from vertical (P) and horizontal (S) seismograms high-pass filtered above 5 Hz. Initial locations were derived from these picks using HYPOCENTER software (Lienert and Havskov, 1995) and a five-layer wavespeed model adapted from Miller and Mooney (1994; Fig. 2, blue line).

Only earthquakes with clear body waves at no less than five stations and S-P times of 10 s or less were analyzed. Events typically had a 3–5 s S-P time at the nearest station. Including arrival times of S-waves together with those of P-waves greatly reduced the uncertainty of hypocenter locations (Gomberg et al., 1990). Locations showed little improvement by adding arrival times from stations that were more distant than the nearest five stations. Most events were small and could only be reliably picked at an average of 12 stations at distances within ∼75 km of the events; only six were recorded on >30 stations including distances >150 km from the event.

We used the one-dimensional (1-D) hypocenter and wavespeed inversion program VELEST (Kissling et al., 1994) to derive a series of models from the earthquake arrival times (Fig. 2). During this process, P- and S-arrival picks were further refined, and the events were relocated with HYPOCENTER software using the preferred VELEST model discussed below. Focal mechanism solutions were calculated from P- and S-wave arrivals using FOCMEC (Snoke, 2003) run within SEISAN software (Havskov and Ottemöller, 1999).

Crustal Wavespeed

The station coverage from the SNEP deployment improves upon the few permanent seismometers and less dense TA deployment in the Sierra Nevada, thus providing the opportunity to use local earthquake arrivals to derive models of crustal wavespeed over a broader region of the Sierra Nevada than has been previously studied (e.g., Miller and Mooney, 1994). After discarding events with azimuthal gaps >180° and root mean square (RMS) travel-time residuals >1 s, we used 1225 arrival times from a total of 92 earthquakes to derive a 1-D wavespeed model using VELEST (Kissling et al., 1994). The HYPOCENTER-located events used in the inversion range from depths of 3.1–47.4 km and lie in the vicinity of the two main earthquake clusters (Foothills and Yosemite clusters; described in more detail below). In order to maintain a broad depth distribution of earthquakes, we did not attempt to create separate models for each cluster.

The initial VELEST inversion started from the Miller and Mooney (1994) crustal model extended with additional layers down to 50 km. After five iterations for P- and S-wavespeeds, the output model was used to derive 15 new models by perturbing the P- and S-wavespeeds. Next, each of the 15 models was run through five inversion iterations for P- and S-wavespeeds, and six models were selected from the 15 based on high values in their resolution matrix, low standard deviations associated with calculation of model parameters, and low overall RMS residuals. Finally, the six models were each run through two final VELEST iterations inverting for P-wavespeeds only (with a constant Vp/Vs ratio of 1.73) and using output of the previous run as input for the next iteration. We derived Vp/Vs from Wadati diagrams generated with first-arrival picks from the deepest events (Wadati, 1933; Frassetto et al., 2011). Due to sampling limitations, we did not independently invert for S-wavespeeds. S-wave picks are down-weighted relative to the P-wave picks for all VELEST inversions, and the six models show only slight variations between the final P- and S-wave inversions and the final P-wave-only inversions. While each of the six models exhibit distinct differences, they converge on a similar wavespeed structure following the final VELEST iterations (Fig. 2).

The final models reveal a gradational increase of wavespeed with depth. In most of the models, wavespeeds in the upper 10 km of the crust range from 5.8 to 6.1 km/s. A smooth transition from 6.1 km/s to 6.7 km/s exists from 10 km to ∼35 km, while a larger transition from 6.7 km/s to 7.6 km/s is observed at lower crustal depths. For a majority of well-resolved layers, the diagonal values of the resolution matrix of the models exceed 0.8. Layers at depths >32 km are less well resolved due to only seven earthquakes (35 rays) sampling below this depth. This lack of resolution is also reflected in the large differences in wavespeeds for the six models below ∼35 km (Fig. 2). Wavespeeds of the models presented here are as much as 0.5 km/s lower than that of the model of Miller and Mooney (1994). This difference likely results from our sampling including a larger portion of the granitic batholith, while the earlier model sampled more mafic material that underlies the eastern Great Valley and the Foothills metamorphic belt.

Station delays for most of the six models of crustal wavespeed, which were solved for simultaneously during inversion iterations, range from –0.71 to 0.49 s (Table 1). This broad range is unsurprising given that the SNEP array spans geologically heterogeneous sections of the San Joaquin Valley, Foothills metamorphic belt, and Sierra Nevada batholith. Overall RMS values from final models decreased from ∼0.50 s in the initial iteration to ∼0.15 s in the final iteration, while the average P-wavespeed model standard deviation also decreased. Epicenters shifted minimally between the initial HYPOCENTER location and the relocation with our preferred model, although depths deepened by an average of 3.2 km across all events after the relocation.

Event Waveforms and Spectra

Earthquake waveforms and corresponding spectra demonstrate the general characteristics of these events across the central Sierra Nevada. We used the GISMO data-analysis toolbox (Thompson and Reyes, 2018) for time-series processing and display; each vertical component seismogram was demeaned and detrended, then high-pass filtered (two passes, two poles) with a 0.5 Hz Butterworth filter to remove microseismic noise. Spectrograms were calculated using a fast Fourier transform length of 3 s with 1.5 s of overlap. Instrument response is flat within the displayed frequency range, and we applied a sensitivity correction to show time-series units as velocity (μm/s). Examples from different stations for each event are displayed over a common time frame starting shortly before the P-arrival (Fig. 3).

For events of all depths in the Foothills and Yosemite clusters, the arrivals of P- and S-waves are clearly observed at adjacent stations. These waveforms stand out over any remaining background noise and have consistent durations of 15–20 s at up to 30 km distance. The spectra of these earthquakes are especially prominent between 5 and 15 Hz, peaking in most cases at >70 dB above the ambient noise level (Figs. 3A and 3B). Observed variations may result from different levels of attenuation along travel paths or the azimuth of the station relative to the source mechanism.

In addition, we observe notably different waveform and spectra for two long-period event sequences. Deep long-period (LP) earthquakes have been observed at the southern edge of the Yosemite cluster (Pitt et al., 2002), and the Northern California Seismic Network (NCSN) (NCEDC, 2014) reported LP events on 14 September 2005 and 20 October 2005. These events exhibit positive, emergent P- and S-phase arrivals and plainly lower-frequency content. The first event has a 30–35-s-long waveform and may represent the comingling of several sources (Fig. 3C). The second is two-minute-long sequence that clearly delineates into distinct earthquakes (Fig. 3D), resembling the spasmodic earthquakes seen in the Long Valley caldera to the east (Ryall and Ryall, 1983). The NCSN cataloged these as two crustal earthquakes and two LP earthquakes. However, in all cases, the spectra are almost entirely lacking in energy above 10 Hz, and the signals concentrate between 1 and 6 Hz. These earthquakes were not widely viewed across the SNEP array, but one from each sequence was identifiable at enough surrounding stations to relocate.

Earthquake Locations

During the first phase of the SNEP deployment (May 2005–June 2006), we detected and located 129 earthquakes at depths between 3.1 and 47.4 km (Table 2; see also Supplemental Material1). Of these, 88 earthquakes were between 20 and 35 km depth (Fig. 4A). We observe no temporal pattern in the events, which occurred at a rate of 8–17 per month, but most events were part of two distinct spatial clusters. The Foothills cluster straddles the boundary between the Great Valley and the western foothills of the Sierra Nevada north of Fresno and southwest of Yosemite National Park (Fig. 1). This group includes 56 events, nearly all of which occur at depths >20 km with an average depth of 28.9 km. The other group of events straddles the southeastern boundary of Yosemite National Park, and the 48 events in this Yosemite cluster were distributed over a wide range of depths with an average of 16.7 km. Earthquakes outside of either cluster are mostly scattered in the foothills and mostly occur below 20 km depth.

We can compare this event compilation with the NCSN catalog, which has 4845 earthquakes in this region (119.2°–120.5°W and 36.6°–37.8°N) from 1 January 1980 to 31 January 2017. This catalog includes 85 events during the time period of this study, 57 of which passed our detection criteria and were relocated. In addition, we identified 72 events that were not contained in the NCSN catalog for this time period. The NCSN catalog assigns magnitudes, which range from 1.21 to 2.65 (average 1.94) for events also located by the SNEP. The epicenters for these events match well, with the mean horizontal difference being 4.1 km (median of 1.9 km, minimum of 0.1 km, and maximum of 33.2 km). However, most NCSN depth estimates are considerably shallower than those determined here. Overall, events located in this study using SNEP stations are anywhere from 0.04 to 44.2 km deeper (mean of 10.7 km, median of 7.5 km) than in the NCSN catalog. The shallower event depths in the NCSN catalog compared with this study (Fig. 4B) probably result from the sparser and more asymmetrical coverage of NCSN stations and the use in the construction of the NCSN catalog of a different model of crustal wavespeed that includes a shallower (36 km) Moho than used here (Fig. 2; Oppenheimer et al., 1993; Waldhauser and Schaff, 2008).

In addition, we examine how the locations for the LP events changed. The NCSN detected 171 LP events from 1980 to 2005, most near the Yosemite cluster. Over 80% occurred between 1999 and 2002. In addition, 109 of the 168 events that predated the SNEP were located within 15 km of the LP events analyzed in this study. The other 59 LP events away from the main locus of LPs may be mislocated or mischaracterized. Until the early 2000s, most events were identified only by a handful of stations, yielding hypocenters that were anywhere from above sea level to nearly 38 km deep. For the two LP earthquakes that we reanalyzed, the NCSN originally located these events using 15–25 stations, with coverage weighted heavily toward ∼22 short period stations from the NCSN that were sited in and around Mammoth Lakes. These events were originally placed at 37.8 and 36.3 km depth. Relocating with SNEP stations and a new model for crustal wavespeed resulted in minor changes to both hypocenters. Both hypocenters only deepened by 1.1–1.9 km and shifted laterally by 1.2–6.3 km. As such, the modern coverage of monitoring networks postdating the SNEP deployment may provide reasonably accurate locations of LP events in this region.

Previous studies (Wong and Savage, 1983; Miller and Mooney, 1994) noted the concentration of deep earthquakes in the area of the Foothills cluster but found comparatively few around the Yosemite cluster. This may have resulted from poor station coverage in the High Sierra. Both studies determined that seismicity in the foothills occurs less frequently outside of the main Foothills cluster. These studies also expressed uncertainty regarding the geographic extent of the lower-crustal earthquakes. The NCSN catalog displays both clusters, but also includes many events scattered outside of the clusters (Fig. 1).

We made a concerted effort to not overlook any events that met our minimum criteria throughout the western portion of the SNEP footprint. We located a handful of events east of the Foothills cluster and south of the Yosemite cluster. In addition, we also observed a few events north and west of the main zones of seismicity. Overall, the SNEP failed to detect 28 events captured by the NCSN during this period (Fig. 1), half of which occurred when the network was still being installed. These events appear in both clusters and are scattered within the San Joaquin Valley beyond the western edge of the SNEP footprint, where detections are difficult using our criteria. The aperture of the SNEP and general agreement with the distribution of seismicity seen by the NCSN suggests that few appreciable events were missed.

The statistical uncertainty in earthquakes’ epicenters and depths are reported in Table 2; on average, event 1σ epicenter errors are estimated to be 2.6 km and depth errors 2.8 km. Such error estimates ignore uncertainty in the wavespeed structure. With depths largely constrained by upgoing rays to nearby stations, to first order the depth would increase directly in proportion to any underestimate in Vp. From the scatter in the VELEST models (Fig. 2; Table 3), it is plausible that our preferred model is as much as 0.1 km/s slow, which would mean our depths would be ∼2%–6% too shallow, which is smaller than the uncertainties reported in Table 2. The poorer constraints on velocities below our deepest earthquakes (shaded area, Fig. 2) affects Pn (P-wave refraction from the top of the mantle) arrival times at distant stations, which are down-weighted and have little impact on our earthquake depths.

“Repeating” Earthquakes

Repeating earthquakes identified in this study occurred within small subgroups of events, over a broad vertical and lateral zone. The time separation of the repeating events ranges from hours to months with no discernible temporal pattern (Table 2). The vertical waveforms for these events are commonly nearly identical (Fig. 5), but similarities also appear when waveforms are rotated into the radial and tangential components. Sets of repeating earthquakes typically contain two or three events with similar waveforms. Events occurring very close in space or time may look identical, while other similarly coincident event pairs share no comparable characteristics. The Foothills cluster contains two sets of repeating earthquakes, while the Yosemite cluster contains four sets. The presence of these events suggests that future analysis of the region might benefit from the application of some style of match filter (e.g., Gibbons and Ringdal, 2006) or subspace detector approach (e.g., Harris, 2006).

Focal Mechanisms

To better understand the stress state in the crust, we examined focal mechanisms for the events in this region. We extracted 341 events from the NCSN catalog with uniquely determined focal mechanism solutions, which exhibit a full range of faulting styles (Fig. 6A). Using the classification scheme of Zoback (1992), 36% are normal or oblique normal, 25% are reverse or oblique reverse, and 14% are strike-slip. The remaining 25% are null, which represent either extremely low-angle or extremely high-angle faulting. Wong and Savage (1983) previously found seven fault-plane solutions around the Foothills cluster but none near the Yosemite cluster. They found three strike-slip, three oblique-reverse or reverse, and one oblique-normal. Fault plane solutions from the Foothills cluster are generally consistent with a north-northeast–striking greatest principle horizontal stress direction, roughly between trends found in the San Andreas fault and Walker Lane areas (Bellier and Zoback, 1995; Zoback and Zoback, 1980; Heidbach et al., 2016; Shelly et al., 2016). The events near southern Yosemite National Park and the High Sierra hint at normal faulting typical of the Basin and Range, but predominantly exhibit a myriad of mechanisms more consistent with a heterogeneous stress state. Additional work would be needed to see if these mechanisms also reflect non-double-couple behavior such as that seen around the active Long Valley caldera just to the east (Foulger et al., 2004).

For determining first motions from SNEP seismograms, we used bandpass filters (1–5 Hz, 5–10 Hz, and 10–15 Hz with custom filters used as needed) to determine the polarity of the P-wave arrival; this is mainly to remove ambiguity introduced by the finite impulse response (FIR) filters used in recording the data. Horizontal components were rotated into radial and tangential orientations in order to best identify these arrivals. We omitted any noisy station-event pairs that did not provide definitive picks. Polarity picking was done using the MULTPLT (trace plotting, phase picking, and spectral analysis) portion of the SEISAN seismic software package (Havskov and Ottemöller, 1999; Ottemöller et al., 2017) on the first clear arrival to avoid any potential signal aliasing from downsampling the data. Due to their small magnitudes, many events simply lacked a sufficient number of picks to yield a well-constrained solution.

Once calculated in FOCMEC, the fault plane solutions were graded by the similarity of solutions to one another and number of polarity errors. For each event, we investigate combinations of orientation of the seismological P- and T-axes space with a granularity of 5° on the focal sphere, i.e., not to the extent where the type of solution may change. Events with a grade of A have only solutions within 5° of one another on the focal sphere and one or fewer polarity errors. Grade B have only solutions within 5° of one another on the focal sphere and two to three polarity errors, which are typically reduced by discarding or refining a questionable pick. Grade C mechanisms may have either solutions that differ by >5° on the focal sphere, more than three polarity errors, or both; grade C mechanisms are omitted from further analysis. This resulted in 52 accepted focal mechanisms (51 grade A, one grade B) from 95 candidate events (Table 4); a subset of mechanisms spanning the range of polarity picks is shown in Figure 7.

These new focal mechanisms differ from the NCSN trends; for the SNEP, nearly half the events (47%) are characterized by reverse faulting versus 25% of NCSN mechanisms. Normal mechanisms account for only 12% and strike-slip and oblique slip mechanisms for 22% (Fig. 6B). Null mechanisms account for only 20% of the total, slightly less than the baseline from the NCSN. We find 23 fault plane solutions from the Foothills cluster, 20 from the Yosemite cluster, and nine from outside those groups (Fig. 8). The geometries of the P- and T-axes determined for these fault plane solutions vary spatially and with depth across the region (Fig. 9). P- and T-axes in the Foothills cluster tend to group well, while the mechanisms in the Yosemite cluster are more chaotic. In the Foothills cluster, P-axes are usually nearly north-south for both NCSN and SNEP mechanisms, while T-axes are steep and largely fall in an east-west band. In the Yosemite cluster, NCSN mechanisms show no organization at all, but the SNEP mechanisms tend toward near-vertical T-axes with shallowly plunging P-axes.

Crustal Deformation within the SNEP Footprint

In a regional context, the Sierra Nevada and adjacent Great Valley represent a transition between northwest-southeast–oriented right-lateral shear and northeast-southwest–oriented shortening to the west, right-lateral shear and extension in the Walker Lane bordering the eastern Sierra Nevada, and east-west–oriented extension farther east in the Basin and Range (e.g., Zoback and Zoback, 1980; Kreemer et al., 2012). The Sierra Nevada itself generally behaves as a near-rigid unit. However, localized analysis of the strain field using a reduced deformation rate tensor calculated from local seismicity (Unruh and Hauksson, 2009; Unruh et al., 2014) reveals finer-scale patterns of deformation within the Sierra Nevada. Strain in the southern Walker Lane belt appears to be simple shear, accommodating northwest translation of the Sierra Nevada block with coincident crustal thinning. However, the southeastern Sierra is shown to be elongating and flattening horizontally south of ∼36.5°N and especially due east of the Isabella anomaly (Unruh et al., 2014). This is coincident with thinned crust that is observed with receiver functions above inflowing mantle asthenosphere (Frassetto et al., 2011; Jones et al., 2014). This can be interpreted as a consequence of lithosphere thickening and foundering to the west within the Isabella anomaly (Unruh and Hauksson, 2009). Unruh et al. (2014) speculated that a rotation of the strain field in this part of the Sierra Nevada foothills and adjacent San Joaquin Valley is the response of the crust to rebound in the late stages of lithospheric foundering.

The focal mechanisms calculated using the SNEP provide a better picture of the crustal strain-rate field northwest of the proposed location of foundering in the lower lithosphere. The Foothills cluster is within the Sierran Foothills wavespeed anomaly; both clusters are ∼50 km from the northern edge of the Isabella anomaly as resolved in recent teleseismic P-wave tomography (Jones et al., 2014). We calculated the reduced deformation rate tensor for these events (Twiss and Unruh, 1998; Unruh and Hauksson, 2009), which yields a transpressional strain-rate field with significant components of net vertical thickening for both groups of earthquakes. Maximum shortening trends are ENE-WSW for the Yosemite cluster and NNE-SSW for the Foothills cluster, in both cases differing from the regional trends of north-south shortening for the Walker Lane and northeast-southwest for the San Andreas transform. Intriguingly, the Foothills cluster mechanisms suggest triaxial deformation, with shortening on both NNE-SSW and ENE-WSW orientations and vertical lengthening. Both groups of events would benefit from more complete, robust event data sets in order to refine these patterns.

Lithospheric Structure and Crustal Seismicity

Moderate to large teleseismic earthquakes recorded by the SNEP were used to map discontinuities within and between the crust and upper mantle with teleseismic receiver functions, which allows us to place local seismicity in structural context (Frassetto et al., 2011). We focus on one profile (Fig. 10) oriented ENE-WSW that runs just south of the events in the Yosemite cluster and bisects the Foothills cluster (Fig. 1). Events occurring 15 km north and south of this transect are projected onto it, roughly matching the lateral sensitivity and averaging of the receiver functions.

The Foothills cluster earthquakes occurred entirely between 19 and 40 km depth (average 28.9 km), overlying the region of a diffuse Moho conversion described by Frassetto et al. (2011). The events appeared in a few distinct pockets, none of which coincide with any structural features resolved by receiver functions. We did not locate a single earthquake above 19 km depth in this cluster, indicating that this seismicity might lie under a décollement. This is very similar to the depth distribution of earthquakes in this area determined by Wong and Savage (1983), which was restricted to between ∼15 and 40 km depth. The persistent absence of shallow events separates this deep cluster from other more-ephemeral deep seismicity that is then followed by shallower events (e.g., Smith et al., 2004, 2016). The absence of seismicity in the surrounding region near 20 km depth suggests that this cluster is not the deeper manifestation of shallower seismicity elsewhere, a suggestion reinforced by the different shortening directions when compared with surrounding seismicity. Thus, this seismicity might be independent of the surface tectonics.

The Yosemite cluster maps into a compact, linear trend through the crust within the Sierra Nevada batholith, ranging from 25 km depth to near the surface (average 16.7 km). The deep LP events underlie the Yosemite cluster along the Moho at 38–39 km depth, at the point separating a shallow and bright Moho receiver-function conversion to the east from the deeper and more diffuse conversion to the west. Although scattered, two-thirds of the older LP events in the NCSN catalog are located within 15 km of these events, albeit with highly varying depths. As noted before, we suspect that these older depths are questionable owing to the known challenge of constraining depth in an earthquake location from a distant network. Most of these LP events are just to the north of the two we analyzed, associating even more strongly with our observations in the Yosemite cluster.

Long-period earthquakes occur beneath several active volcanic centers in the southern Cascade Range–Sierra Nevada including Mount Shasta, Mount Lassen, and Mammoth Mountain. These earthquakes locate at depths from 10 to 35 km, and are usually found to coincide with shallower clusters of small-magnitude earthquakes. These LP events have been suggested to occur in relation to periods of magmatic migration and magmatic-hydrothermal interactions in the crust (Pitt et al., 2002). The LP earthquakes here situate at least 12 km deeper and 10 km south of the Yosemite cluster, but we cannot ascertain a direct connection between the two styles of deformation without additional observations.

Sequences of unusual lower-crustal seismicity near Lake Tahoe (Smith et al., 2004, 2016; von Seggern et al., 2008), which consisted of two dense, linear clusters of deep crustal (>20 km) earthquakes, may offer insights into the factors driving seismicity within the Yosemite cluster. The first Lake Tahoe cluster occurred beneath northwestern Lake Tahoe in 2003 and was followed by a shallower swarm active from 2003 to 2005. The second cluster happened near Sierra Valley in 2011 and was preceded by a larger deep-crustal earthquake located between the swarms. Smith et al. (2004, 2016) suggested that the deep swarm originated from a magmatic intrusion into the lower crust, and the corresponding shallow swarm is a stress-triggered response. Focal mechanisms for many of the shallower Lake Tahoe events exhibit P- and T-axes consistent with the regional stress field. However, the mechanisms for the earthquakes in the deeper cluster display scattered P- and T-axes. The distribution of events in the Yosemite cluster resembles the planar swarm of earthquakes beneath Lake Tahoe (von Seggern et al., 2008). Similar to those for the deeper Lake Tahoe events, the focal mechanisms for the Yosemite cluster are inconsistent with regional stress patterns. The similar characteristics of these clusters of seismicity may indicate that they resulted from a common mechanism.

Previous studies have not considered the role of crustal seismicity, particularly the unusually deep or LP events, when suggesting that lithospheric foundering occurs only in the southern Sierra Nevada (e.g., Zandt, 2003) or along the entire batholith (Jones et al., 2004). With the exception of reservoir-induced events, seismicity along the western Sierra Nevada and its foothills north of ∼38°N is largely absent in historical records (NCEDC, 2014). The limited spatial extent of deep and LP events suggests that a geographically focused process is responsible for the earthquakes identified here. The diminished level of seismicity north of ∼38°N coincides with a transition in surface geology from prominently batholithic crust to the south to primarily metamorphic rock to the north.

Several lines of evidence suggest that the Foothills cluster most likely represents some phase of lithospheric foundering. The absence of any connection to surface deformation suggests a local source of stress acting under some kind of detachment. The different strain-rate orientation compared to other seismicity suggests a local source of stress. This seismicity is occurring in high-wavespeed and thus probably high-density lower crust, an inference compatible with the fairly high-isostatic-gravity anomalies (0 to +10 mGal) in this area (Oliver et al., 1993). Localized thickening of dense lower crust and/or uppermost mantle appears to be occurring. An alternative could be impingement of some deeper body against the base of the lithosphere (e.g., a northern extension of the proposed Monterey plate; Wang et al., 2013), but the proposed position of such a body is both well to the south and much too deep. Additionally, any impingement would presumably be stronger where the body was shallower, yet similar seismicity is absent in the surrounding areas. Finally, teleseismic tomography (Jones et al., 2014, their figure 17) and magnetotellurics (Ostos and Park, 2012) in this area suggest that some dynamic process is affecting the lithosphere near the Moho just to the east, which might also explain the LP events and the Yosemite cluster, thus possibly providing the means for the lower crust or lithospheric mantle to become gravitationally unstable.

Crustal Wavespeeds

Past investigations of crustal wavespeeds of the Sierra Nevada region found a wide range of results (Bateman and Eaton, 1967; Holbrook and Mooney, 1987; Miller and Mooney, 1994; Savage et al., 2003; Fliedner et al., 1996, 2000; Ruppert et al., 1998; Prodehl, 1979; Spieth et al., 1981). Miller and Mooney (1994) interpreted this heterogeneity to be a marker of the many different tectonic provinces and crustal compositions that characterize the area. Our model most resembles the wavespeed structure from Fliedner et al. (1996) under the western Sierra Nevada foothills to the south; we do not see the rapid increase in wavespeed in the upper crust found farther west (Fliedner et al., 2000; Miller and Mooney, 1994).

The smoother transition between layer wavespeeds in our model compared to that of Miller and Mooney (1994) may be from the slightly different regions sampled by each study. The earthquakes in the Foothills cluster likely record higher wavespeeds associated with metamorphic rocks found beneath the western foothills. In contrast, Yosemite cluster earthquakes record the intermediate wavespeeds associated with the granitic composition of the batholith itself. A vast majority of earthquakes used by Miller and Mooney (1994) were from the Foothills cluster region, and the stations recording these events were exclusively to the north in the metamorphic belt of the western Sierra Nevada. Therefore, the raypaths were nearly entirely outside the batholith, which explains why they observed much higher wavespeeds and more abrupt transitions from 4 to 12 km depth than we see in our preferred model over the same depth interval. Results from Ruppert et al. (1998) showing upper-crustal wavespeeds of 6.0–6.4 km/s and a lower-crustal wavespeed of 6.6 km/s beneath the southern Sierra Nevada batholith are much more consistent with our findings, as are results from Fliedner et al. (2000) that find wavespeeds of 5.9–6.3 km/s below the batholith. Our model loses resolution below 35 km, but forward modeling of receiver functions suggests that a gradual increase in wavespeed persists to at least 70 km depth (Frassetto et al., 2011). The lack of any sharp boundary is consistent with an arc root beneath the region that is structurally intact.

With the observations from the SNEP deployment, we place in a broader, geodynamic context the unusual seismicity distributed across the central High Sierra and adjacent western foothills. Small, commonly very deep crustal earthquakes in two distinct clusters show a unique and distinct tectonic process occurring within the Sierra Nevada batholith. Previous studies suggested the presence of a partially removed and actively foundering eclogitic root beneath the southern Sierra Nevada crust, and we demonstrate that the observed earthquakes are a potential seismic consequence of this. Events occurring in the Foothills cluster may be a response to stressing of the lower crust due to active root foundering. Those in the Yosemite cluster beneath the High Sierra may be related to inflow of adjacent mantle asthenosphere stressing the overlying crust or magmatic activity within the crust itself.

A one-dimensional P- wavespeed model for the south-central Sierra Nevada created by inverting crustal earthquake data suggests a very gradual wavespeed increase with depth, from ∼5.8 km/s at the surface to ∼6.7 km/s at ∼35 km. If the presence of deep-crustal earthquake activity is an indication of active root removal, as we suggest, then the extent of removal in the southern Sierra Nevada does not extend northward of ∼37.5°N. It is important to note that the seismic data used in this study spans only the central portion of the batholith, which precludes our ability to make any conclusions regarding root removal farther north or south. Existing seismic catalogs, however, can be interpreted to indicate that there appears to be no seismogenically active, large-scale root removal along the entire strike of the batholith.

We thank the SNEP field teams for installing, operating, and removing the array. We are also especially grateful to the private landowners who allowed us to deploy on their land. Edi Kissling provided support for VELEST and offered comments on our initial results. PASSCAL Quick Look (PQL) and Seismic Analysis Code (SAC) were used throughout this analysis. We also thank Glenn Thompson and Jackie Caplan-Auerbach for technical assistance with GISMO and spectrogram plotting, and David Shelly and Mike Brudzinski for assistance accessing the NCSN catalog and working with the fault plane solutions. Two anonymous reviews strengthened this presentation. This work was supported by U.S. National Science Foundation (NSF) EarthScope grants EAR-0454554, EAR-0454524, and EAR-0454535 to the Universities of Arizona, Colorado, and South Carolina. The seismic instruments were provided by the Incorporated Research Institutions for Seismology (IRIS) through the PASSCAL Instrument Center at New Mexico Tech. Data from the TA network were made freely available as part of the EarthScope USArray facility, operated by IRIS. The facilities of the IRIS Data Management Center were used for access to waveforms, related metadata, and/or derived products used in this study. These facilities are supported by the NSF under Cooperative Agreement EAR-1261681 and the U.S. Department of Energy National Nuclear Security Administration.

1Supplemental Material. ASCII text file with arrival time picks in SEISAN format. Please visit or access the full-text article on to view the Supplemental Material.
Science Editor: Andrea Hampel
Associate Editor: Huaiyu Yuan
Gold Open Access: This paper is published under the terms of the CC-BY-NC license.