Ten years of teleseismic earthquakes recorded by broadband seismic instruments from the Anza network–USArray stations around the San Jacinto fault were used to create P receiver function images of the lithospheric structure beneath this major strike-slip fault. Analyses of back-azimuthal variation and location of the conversion points near the fault suggest an ∼8 km vertical offset of the Moho directly beneath the San Jacinto fault. This implies that the fault extends through the entire crust and into the mantle lithosphere, supporting the idea that the strain in the lower crust is localized within a narrow zone. The Moho offset and surface trace of the San Jacinto fault zone are coincident with a compositional boundary in the Peninsular Ranges batholith previously identified in potential field geophysical data and Sr isotope analyses. The position of the offset with respect to this relict geologic feature, which predates the pluton emplacement that formed the batholith, may be a controlling factor in strain location and plate-boundary fault initiation.
Within oceanic plates, deformation along transform faults is constrained to a narrow zone, but the degree to which continental strike-slip boundaries are localized at depth continues to be debated. This leads to unanswered questions, such as: What is the strength of the lithosphere; how do faults develop; how does the crust transition from the brittle upper crust to the ductile lower crust; and how do plate boundaries evolve? The plate boundary between the North American plate and the Pacific plate in southern California is expressed across a 200-km-wide zone that includes the San Andreas fault, San Jacinto fault zone, and Elsinore fault. The total rate of relative motion between the plates is ∼55 mm/yr (DeMets, 1995; DeMets and Dixon, 1999), and ∼45 mm/yr of that motion is equally distributed between the San Andreas and San Jacinto faults (Fialko, 2006). Of these three major faults, the San Jacinto fault zone is the most seismically active (Thatcher et al., 1975) and is expressed as a complex zone of overlapping strike-slip fault segments, as shown in Figure 1. The relative structural complexity of the San Jacinto fault zone likely reflects the recent (1.0–1.5 Ma) development of the fault zone and the accommodation of only ∼25 km of total displacement (Wesnousky, 1988; Rockwell et al., 1990; Janecke et al., 2011; Dorsey et al., 2012).
Regional tomography studies (Magistrale and Sanders, 1995; Hauksson, 2000; Tape et al., 2009) show broad seismic velocity contrasts across the San Jacinto fault zone in our study region, where the material to the northeast has a higher seismic velocity than the rocks to the southwest. Allam and Ben-Zion (2012) used double-difference tomography to image the upper crust and detailed fault zone structure, where they also found velocity contrasts across the fault. They identified the largest velocity contrasts in the upper 5–10 km along the northern end of the San Jacinto fault zone closest to the northernmost extent of the Peninsular Ranges batholith north of Anza (Allam and Ben-Zion, 2012). Their images also show a slight velocity contrast in the basement depth rocks (15–20 km), where the rocks northeast of the San Jacinto fault zone have a higher velocity than those to the southwest, similar to the results of the regional tomographic studies. These studies, however, did not concentrate on velocities nor structure at the depth ranges of the lowermost crust and upper mantle and so consequently did not image variations in crustal thickness across the region.
Langenheim et al. (2004) used potential field data to explore the geologic structure of the region. They did not observe a major contrast in basement rock properties across the San Jacinto fault zone, except at the southern end of the Clark fault (see Fig. 1 for location), where there are higher gravity values northeast of the fault and a slight magnetic gradient across the entire length of the fault. The majority of basement rock beneath this region is part of the Peninsular Ranges batholith, which is composed of plutonic rocks intruded into metasedimentary and metavolcanic rocks (Silver and Chappell, 1987). These plutons of Jurassic and Cretaceous age have a compositional range between gabbro and granite, with distinct western and eastern terranes identified by different geophysical, geochemical, and petrological signatures, suggesting different sources (Magistrale and Sanders, 1995; Langenheim et al., 2004, and references therein). The western terrane is older and is generally more mafic, suggesting an oceanic origin, whereas the eastern terrane is younger and more silicic, suggesting a continental origin (e.g., Silver and Chappell, 1987). The Peninsular Ranges compositional boundary between these two terranes is truncated at the northern end of the San Jacinto fault zone, where Allam and Ben-Zion (2012) identified the largest cross-fault velocity contrast, but the rest of the boundary runs roughly subparallel and ∼50–75 km west of the San Jacinto fault zone (Fig. 1). Although this compositional boundary is not coincident with the San Jacinto fault zone, the modeled magma compositions indicate a distinct chemical variation across the fault (Baird and Miesch, 1984). These variations are not identifiable in the field, like the compositional/petrologic boundary between the eastern and western terranes of the Peninsular Ranges batholith (Peninsular Ranges compositional boundary), yet isotopic signatures do show a discontinuity across the fault. Specifically, strontium isotopic ratios illustrate a change in Sri values that coincides with the San Jacinto fault zone as the 0.0706 isopleth, where southwest of the fault, Sri values are lower, and northeast of the fault, Sri is higher (Kistler et al., 2003; Langenheim et al., 2004). Langenheim et al. (2004) suggested that the Sri gradient is a geologic feature, which they term the San Jacinto discontinuity, which separates blocks of different origin (Fig. 1). The differences in affinity could be from variation in the amount of sedimentary material in the source region prior to the emplacement of the plutons, which suggests that the San Jacinto discontinuity could be a crustal flaw that may have influenced the development of the fault zone.
We have combined mapping of the crust-mantle boundary from converted P-to-S waves, the crustal velocity and properties, and the geological history of the region to provide insight into the lithospheric structure and evolution of the relatively young San Jacinto fault zone. Together, these observations are used to infer the way in which strain is localized along this strike-slip fault and how it is has evolved at this location, both of which are important considerations for understanding the dynamics of the plate boundary between the Pacific and North American plates.
DATA AND ANALYSES
We computed P receiver functions (Langston, 1977; Vinnik, 1977) for 145 teleseismic events from 2000 to 2011 recorded by 11 broadband seismic stations from the Anza network (Fig. 1; Table 1) for earthquakes with magnitudes greater than 6.0 with epicentral distances between 30° and 95° (Fig. 2, inset). Three-component seismograms for each of these stations in the Anza network around the San Jacinto fault zone were processed by first removing the mean and trend, and then they were band-pass filtered by a fourth-order Butterworth filter with corner frequencies of 0.02 and 1.5 Hz. The two horizontal components were rotated to be parallel and perpendicular to the “great circle path,” and then the vertical component was deconvolved from the radial component with a Gaussian filter width of 1.5 using an iterative, time-domain deconvolution algorithm following Ligorria and Ammon (1999) and implemented by Bailey et al. (2012).
The individual receiver functions, for each station-event pair with a high signal-to-noise ratio (>10) and with its largest amplitude signal corresponding to the direct P arrival, were depth converted using a velocity model extracted for that unique raypath from the Southern California Earthquake Center (SCEC) Three-Dimensional (3-D) Community Velocity Model (Kohler et al., 2003; Plesch et al., 2009) for the crust and mantle lithosphere, and the Tectonic North America velocity model (Grand and Helmberger, 1984) for the deeper mantle. The complex geology of the region is clearly reflected in the seismic velocities, which have up to 5%–15% Vs contrast in the crust (Plesch et al., 2009). By using these unique estimates of the velocity structure based on a priori velocity information for each station-event pair, we reduced the error and mislocation of the mode conversions determined from the receiver functions compared to when an average one-dimensional (1-D) model is used.
Our results show a strong variation in Moho depth across the region, with a deeper Moho found beneath the southwestern side of the San Jacinto fault zone (∼33 km) and a shallower Moho to the northeast (∼25 km), which is in general agreement with previous results (Baker et al., 1996; Lewis et al., 2000; Zhu and Kanamori, 2000; Yan and Clayton, 2007). P receiver functions from Lewis et al. (2000) and Ichinose et al. (1996) suggest that the Moho has a 10°–20° westward dip across the Peninsular Ranges batholith and a range in depths of 27–37 km. With the additional years of data for the 11 stations nearest the San Jacinto fault zone (Table 1), we were able to investigate the back-azimuthal variation and Moho conversions at a range of sampling points around the stations (Fig. 1; supplementary Fig. S11) in order to better establish the three-dimensional variation in crustal thickness in more detail, especially across the fault.
We analyzed the receiver functions by back azimuth and estimated the location of the conversion points for the Moho signal (Pms) for the different event locations for each station (Fig. S1 [see footnote 1]). Since some of the stations are located very close to the surface trace of the fault, the conversion points can occur on either side (or within) the fault zone. The change in lithology, and therefore seismic velocity, on either side of the fault is important to identify (Schulte-Pelkum and Ben-Zion, 2012), as are the lateral locations of the conversion points when interpreting the Moho signal. The receiver function gathers (Fig. 2) are plotted by back azimuth and are colored corresponding to the earthquake distribution plot shown in the inset, where orange is the NE quadrant, green is NW, red is SE, and blue is SW. The first positive arrival for each trace is the direct P arrival, and the second positive arrival is the converted wave (Pms) signal from the Moho. As seen from the three selected stations in Figure 2 (and the other stations in supplementary Fig. S2 [see footnote 1]), the depth of Moho signals varies for each station, but they have broad similarities within each back-azimuth quadrant.
For stations that have conversions that occur just beneath the fault trace, such as the events for the SE quadrant for station SND and the NE quadrant for station CRY (Figs. 2A–2B), the signals become very complicated. There are additional phases before and after the converted signal from the Moho at ∼30 km, which can be attributed to the complex structure within the fault zone, and these add to the complexity of picking a clear Moho signal. For stations located farther away from the fault zone, such as PFO, the delay times of Pms with respect to the direct P are mostly consistent within each back-azimuth bin, as all are generally sampling one side of the fault (Fig. 2C). However, the delay times of Pms with respect to direct P have variations as large as 25% for the different stations analyzed, specifically those on either side of the San Jacinto fault zone. For stations (and conversions) located on the west side of the fault, such as CRY, the delay times of Pms with respect to the direct P are ∼30 km (Fig. 2A). For stations east of the fault, such as PFO, the delay times are ∼25 km (Fig. 2C). When these receiver functions are depth converted for the station-event using the extracted velocity model from the SCEC Community Velocity Model (Plesch et al., 2009), these differences result in up to an 8 km difference in Moho depth across the fault (Fig. 3; Table 1). The Moho depths were picked from individual depth-converted receiver functions and plotted at their conversion points to accurately position the picks and illustrate the complexity of the fault zone structure (Fig. 4). Although the picks have some variation in them, we show both the picks and the resulting contour map to best illustrate the variability in the observations with the consistent change in crustal thickness across the fault. In addition, common conversion point (CCP) stacks were created by mapping the time-domain signals to depth and laterally in space to their approximate conversion point (Dueker and Sheehan, 1997). Figure 3 shows the CCP stack as a profile (location shown in Fig. 1), which clearly images an ∼8 km difference in Moho depth across the fault.
Our results show a systematic change in crustal thickness on either side of the fault along the extent of the fault where there are data, which is further supported by other receiver function studies that use additional data at greater distances from the fault (Baker et al., 1996; Ichinose et al., 1996; Lewis et al., 2000; Zhu and Kanamori, 2000; Yan and Clayton, 2007). The San Jacinto discontinuity (SJD in Fig. 1) that is shown by the potential field data (Langenheim et al., 2004) demonstrates that this feature extends down to at least 20 km depth beneath the fault zone. Our receiver function observations indicate that this feature extends much deeper, crossing the Moho, and down into at least the uppermost part of the mantle lithosphere. These data add to the growing body of evidence showing that major strike-slip faults can extend through the entire crust, and possibly the entire lithosphere, as discrete features (e.g., Stern and McBride, 1998; Wittlinger et al., 1998; Zhu and Helmberger, 1998; Parsons and Hart, 1999; Hole et al., 2000; Zhu, 2000; Weber et al., 2004; Wittlinger et al., 2004; Veenstra et al., 2006; Niu et al., 2007; Lekic et al., 2011; Levander and Miller, 2012).
We have considered the possibility that the observed step in crustal thickness is the result of only a seismic velocity contrast across the fault. This was tested by computing synthetic seismograms (supplementary Fig. S3 [see footnote 1]) and then processing the receiver functions for a model with crustal velocity differences of 10% and 20% across the fault, which were the maximum perturbations found by recent double-difference tomography (Allam and Ben-Zion, 2012) in the middle to upper crust. Although there is clearly a difference in arrival times for the Pms phase in the synthetic receiver functions, which is comparable with our observations, the 20% velocity contrast across the fault would have to extend throughout the entire crust to 30 km. Since the basement rock beneath the entire region is part of the Peninsular Ranges batholith, which does not have significant velocity perturbations within it, even at midcrustal depths (Magistrale and Sanders, 1995; Tape et al., 2009; Allam and Ben-Zion, 2012), it is not realistic to assume such a large, extensive velocity difference across the fault to explain the variable depth Moho we observe. The possibility that we are assuming an unrepresentative 1-D velocity model for either side of the fault, which could also mislocate the signals, was eliminated by using existing 3-D velocity models of the region (Plesch et al., 2009) and extracting the raypaths for each event-station pair to map the converted wave signals from time into space.
Since the observed Moho offset is located directly beneath the trace of the San Jacinto fault zone, and previous results have found a change in properties in the basement rock (Kistler et al., 2003; Langenheim et al., 2004) that is coincident with the fault, we propose that the San Jacinto discontinuity played a role in fault localization. Langenheim et al. (2004) noted that the position of the near-vertical San Jacinto discontinuity and the way in which it influenced the location and evolution of the San Jacinto fault could not be determined with their geophysical data. Their potential field–based models do not extend to Moho depths (only 20 km), but our new images of the Moho offset from receiver functions provide evidence that this relict structure within the Peninsular Ranges batholith extends through the lower crust and into the upper mantle. Furthermore, studies of the ways in which the Moho and basement features control seismicity indicate that these can concentrate stress, resulting in seismicity and faulting (i.e., Stuart et al., 1997; Dentith et al., 2009; Molnar and Dayem, 2010). Therefore, we propose that the San Jacinto fault zone developed along a near-vertical crustal discontinuity within the Peninsular Ranges batholith, which extends through the entire crust. This relict, pre–pluton emplacement structure, which is defined by both geophysical and geochemical data (e.g., Silver and Chappell, 1987; Magistrale and Sanders, 1995; Kistler et al., 2003; Langenheim et al., 2004), may have influenced the localization of strain and development of the relatively young San Jacinto fault zone into its present location. We have imaged this boundary as a step in the Moho directly beneath the San Jacinto fault zone using an a priori 3-D velocity model and back-azimuth variation in receiver function signals from the dense seismic network that spans across the region.
We would like to thank David Okaya for his assistance with the synthetics and comments on the manuscript. This project was funded by the Southern California Earthquake Center (SCEC). SCEC support came from National Science Foundation Cooperative Agreements EAR-0529922 and EAR-1033462 and by U.S. Geological Survey Cooperative Agreements G07AC00026 and G12AC20038. Data were requested from the IRIS Data Management Center, and figures were made with Generic Mapping Tools (Wessel and Smith, 1991). This is SCEC contribution 1755.