Volume 107, Issue B7 p. ETG 8-1-ETG 8-19
Geodesy and Gravity/Tectonophysics
Free Access

Lower crustal deformation beneath the central Transverse Ranges, southern California: Results from the Los Angeles Region Seismic Experiment

Nicola J. Godfrey

Nicola J. Godfrey

Earth Science Department, University of Southern California, Los Angeles, California, USA

Now at Landmark EAME, Leatherhead, UK.

Search for more papers by this author
Gary S. Fuis

Gary S. Fuis

U.S. Geological Survey, Menlo Park, California, USA

Search for more papers by this author
Victoria Langenheim

Victoria Langenheim

U.S. Geological Survey, Menlo Park, California, USA

Search for more papers by this author
David A. Okaya

David A. Okaya

Earth Science Department, University of Southern California, Los Angeles, California, USA

Search for more papers by this author
Thomas M. Brocher

Thomas M. Brocher

U.S. Geological Survey, Menlo Park, California, USA

Search for more papers by this author
First published: 25 July 2002
Citations: 18

Abstract

[1] We present a P wave velocity model derived from active source seismic data collected during the 1994 Los Angeles Region Seismic Experiment. Our model extends previously published upper crustal velocity models to mantle depths. Our model was developed by both ray tracing through a layered model and calculating travel times through a gridded model. It includes an 8-km-thick crustal root centered beneath the surface trace of the San Andreas fault, north of the highest topography in the San Gabriel Mountains. A simple mass balance calculation suggests that ∼36 km of north-south shortening across the San Andreas fault in the central Transverse Ranges could have formed this root. If north-south compression began when the “Big Bend” in the San Andreas fault formed at ∼5 Ma, 36 km of shortening implies a north-south contraction rate of ∼7.1 mm/yr across the central Transverse Ranges. If, instead, north-south compression began when the Transverse Ranges formed at 3.4–3.9 Ma, 36 km of shortening implies a contraction rate of 9.2–10.6 mm/yr. North of the San Andreas fault, the Mojave Desert crust has a low-velocity (6.3 km/s) mid and lower crust and a 28-km-deep Moho. South of the San Andreas fault, beneath the Los Angeles and San Gabriel Valley basins, there is a fast (6.6–6.8 km/s), thick (10–12 km) lower crust with a 27-km-deep Moho. Farther south still, the lower crust of the Continental Borderland is fast (6.6–6.8 km/s) and thin (5 km) with a shallow (22 km deep) Moho.

1. Introduction

[2] Southern California is actively deforming. One region of interest is the Big Bend region of the San Andreas fault, where there is a left step in the right-lateral transform system (Figure 1a). This restraining bend, in principle, results in approximately north-south compression across the San Andreas fault [e.g., Atwater, 1970; Bird and Rosenstock, 1984], although Weldon and Humphreys [1986] point out that little contraction is actually currently occurring across the San Andreas fault itself. Nevertheless, the Transverse Ranges, located on both sides of the San Andreas fault in the region of the Big Bend, responded, and are still responding, to north-south compression by thrust faulting, folding, and block rotations [Bailey and Jahns, 1954; Luyendyk et al., 1980; Jackson and Molnar, 1990; Luyendyk, 1991]. Deformation in the upper crust is believed to be decoupled from that of the lower crust and mantle [Yeats, 1981; Nicholson et al., 1994; Shen et al., 1996; Ryberg and Fuis, 1998; Kohler, 1999; Fuis et al., 2001a]. Until recently, it was thought that the San Gabriel Mountains (central Transverse Ranges) had little or no crustal root [Hadley and Kanamori, 1977; Hearn, 1984; Fuis and Mooney, 1990; Shearer and Richards-Dinger, 1997]. Kohler and Davis [1997], however, modeled a significant crustal root beneath the surface trace of the San Andreas fault in the San Gabriel Mountains using relatively high-resolution teleseismic earthquake data collected during the passive phase of the (1994) Los Angeles Region Seismic Experiment. The Transverse Ranges are also associated with a vertical, tabular body of anomalously fast mantle velocities 60–80 km wide that extend 100–250 km into the mantle lithosphere [Hadley and Kanamori, 1977; Raikes, 1980; Humphreys et al., 1984; Humphreys and Clayton, 1990; Zhao et al., 1996; Kohler, 1999]. Kohler [1999] mapped the high-velocity mantle anomaly as a near vertical feature directly beneath the crustal root. A high-velocity, and presumably high-density, mantle anomaly may be gravitationally unstable. It may sink into the lithosphere, drawing lower crustal material down with it, thereby enhancing crustal contraction [Fleitout and Froidevaux, 1982; Humphreys and Hager, 1990; Jackson and Molnar, 1990; Kohler, 1999].

Details are in the caption following the image
(a) Index map of California showing the location of the LARSE experiment (black box). CA is California, MX is Mexico, NV is Nevada, GF is the Garlock fault, MTJ is the Mendocino triple junction, and SAF is the San Andreas fault. (b) Enlargement of black box shown in Figure 1a showing the location of the three LARSE profiles. Shading is topography; white is sea level. (c) Detailed map of LARSE line 1 showing the sources and receivers used for the model presented in this study. Geological features are labeled. Shading is topography; white is sea level. EF is the Elsinore fault, GF is the Garlock fault, HF is the Hollywood fault, NIF is the Newport-Inglewood fault, PF is the Punchbowl fault, PH are the Puente Hills, PVF is the Palos Verde fault, RF is the Raymond fault, SAF is the San Andreas fault, SaMoF is the Santa Monica fault, SFV is the San Fernando Valley, SGF is the San Gabriel fault, SGV is the San Gabriel Valley, SJF is the San Jacinto fault, SMF is the Sierra Madre fault, SMM are the Santa Monica Mountains, VT is the Vincent thrust fault, and WF is the Whittier fault. Earthquakes close to line 1 are shown by black stars where 1 is the 1991 M5.8 Sierra Madre earthquake, 2 is the 1987 M5.9 Whittier Narrows, 3 is the 1933 M6.3 Long Beach earthquake, 4 is the 1971 M6.7 San Fernando earthquake, and 5 is the 1994 M6.7 Northridge earthquake.

[3] This paper presents results for the lower crust and mantle beneath line 1 of the 1994 active phase of the Los Angeles Region Seismic Experiment (LARSE) (Figure 1). In a companion paper [Fuis et al., 2001a], we discuss the upper and middle crust. In this paper we investigate what the geometry of the lower crust and Moho can tell us about lower crustal deformation in this active region, as well as how it may relate to the previously documented mantle anomaly. We use active source data from LARSE line 1, and gravity data from the same profile, to model velocity and density variations in the crust along a transect that extends from the offshore Continental Borderland, across the Los Angeles basin and San Gabriel Mountains, into the Mojave Desert. We use our velocity and density models to determine how the transition from the offshore Pacific plate to the undeformed North American plate is accommodated in this complexly deforming region. The lower crust has deformed by thickening into an 80-km-wide root centered beneath the surface trace of the San Andreas fault. We believe that this thickening in the lower crust is most likely a ductile response to north-south compression across the Transverse Ranges, where ductile lower crustal deformation is decoupled from brittle upper crustal deformation.

2. Tectonic History

[4] At ∼30 Ma the Pacific and North American plates came into contact, when the Pacific-Farallon spreading ridge collided with the North American plate margin [Atwater, 1970; Atwater and Molnar, 1973; Atwater, 1989]. Two triple junctions formed as a result of the collision [Atwater, 1970, 1989], and they migrated away from one another due to their geometry [McKenzie and Morgan, 1969], allowing the San Andreas transform fault system to develop between them. Between 24 and 18 Ma, a period of extensional tectonics began. Extension manifested itself in mid-Miocene rifting, widespread volcanism and high heat flow [Henyey, 1976; Dokka, 1989; Tennyson, 1989; Wright, 1991]. Crustal extension coincided with clockwise block rotations that may have been related to the shallowing of Farallon subduction [Luyendyk, 1991] and/or microplate capture [Nicholson et al., 1994].

[5] At ∼5 Ma, the opening of the Gulf of California caused the Pacific-North American plate boundary to migrate from offshore southern California onshore to the east side of the Los Angeles basin [Atwater, 1970, 1989; Nicholson et al., 1994]. The right-lateral San Andreas transform fault either inherited the “Big Bend”, or developed and/or exaggerated this shape, after the migration of the plate boundary [Cox and Engebretson, 1985; Atwater, 1989]. Uplift of the Transverse Ranges began between 3.4 and 3.9 Ma in the vicinity of the Big Bend in the San Andreas fault [Woodford et al., 1954; Wright, 1991].

3. Geology

[6] LARSE line 1 crosses several geological terranes and tectonic features and passes close to the epicenters of three major earthquakes. A magnitude 6.3 earthquake (Long Beach earthquake) occurred on the Newport-Inglewood fault in 1933 [Richter, 1958; Hauksson, 1987], a magnitude 5.9 earthquake (Whittier Narrows earthquake) occurred on a blind thrust in 1987 [Hauksson et al., 1988], and a magnitude 5.8 earthquake (Sierra Madre earthquake) occurred in 1991 [Hauksson, 1994] (Figure 1c). Line 1 crosses the Los Angeles region approximately perpendicular to the major mapped faults [Jennings, 1977; Murphy et al., 1996]. Below we briefly describe the major tectonic elements (from north to south) relevant to LARSE line 1.

3.1. Mojave Desert

[7] The pre-Tertiary basement of the Mojave Desert consists mainly of Mesozoic plutonic and volcanic rocks, and Paleozoic sedimentary rocks [see Jennings, 1977]. The plutonic rocks are the remnant of a continental Mesozoic magmatic arc, continuous with the batholiths of the Sierra Nevada and Peninsular Ranges. Generally, shallow basins filled with Tertiary and Quaternary sediments overlie the Mesozoic batholith [Dibblee, 1968; Jennings, 1977].

3.2. San Andreas Fault

[8] The modern San Andreas fault in southern California became active at ∼5 Ma (see summary of Powell [1993, pp. 49–50]. Between the Mendocino triple junction and the Gulf of California, the San Andreas fault in most places has a trend of ∼330° (Figure 1a). This is approximately parallel to relative plate motion between the Pacific and North American plates. In the Transverse Ranges in southern California, however, the trend of the San Andreas fault changes to 295°, a more east-west trend, which is oblique to the relative plate motion. In this section of the San Andreas fault, known as the “Big Bend”, the San Andreas fault system consists of a broad belt of northwest-trending strike slip faults, of which the San Andreas fault itself is the northeasternmost [e.g., Powell, 1993] (Figure 1). Older strands include the Punchbowl and San Gabriel faults (Figure 1c) [Crowell, 1952; Ehlig, 1981; Powell, 1993].

3.3. San Gabriel Mountains (Central Transverse Ranges)

[9] The Transverse Ranges are a late Cenozoic feature resulting from a combination of compression across the left-stepping bend in the San Andreas fault [Bailey and Jahns, 1954], and upper crustal block rotations [Luyendyk et al., 1980; Jackson and Molnar, 1990; Luyendyk, 1991]. The Transverse Ranges include the western Transverse Ranges, the San Gabriel Mountains or central Transverse Ranges, and the San Bernardino and Little San Bernardino Mountains or eastern Transverse Ranges. They are east-west trending mountain ranges that cut across the predominantly northwest-southeast trending structures of the surrounding provinces [Bailey and Jahns, 1954]. The Transverse Ranges are undergoing convergence as is evident, (1) from faulting [e.g., Crowell, 1968; Morton and Yerkes, 1987], (2) from the occurrence of moderate historical reverse-fault earthquakes (the 1971 M6.7 San Fernando, 1991 M5.8 Sierra Madre, and 1994 M6.7 Northridge earthquakes) (Figure 1c), and (3) from modern north-south compression measured from geodetic data [Feigl et al., 1993; Shen et al., 1996]. The country rocks making up the San Gabriel Mountains south of the San Andreas fault are divided into two plates by the Vincent thrust fault (Figure 1c). The upper plate consists of Precambrian metasedimentary rocks, and anorthosite and Mesozoic plutonic rocks [Ehlig, 1981]. The lower plate consists of Pelona Schist, a chiefly mica-quartz-albite schist [Ehlig, 1968, 1981]. For a discussion of the geometry of these two plates, the reader is referred to Fuis et al. [2001b].

3.4. Los Angeles and San Gabriel Valley Basins

[10] The Los Angeles basin is a fault-bounded basin whose bounding faults include the Palos Verde, Santa Monica, Hollywood, Raymond, and Whittier faults (Figure 1c). The San Gabriel Valley basin, a subbasin of the Los Angeles basin, is bounded on the southeast by the Puente Hills and on the north by the Sierra Madre fault and the San Gabriel Mountains (Figure 1c). The present-day Los Angeles basin began evolving in the Late Miocene as a result of subsidence between the right-oblique Whittier and Palos Verde fault zones and the left-oblique Santa Monica fault system [Wright, 1991]. Since the mid-Pliocene, the Los Angeles basin has been deforming by crustal shortening accommodated by blind thrusts beneath the basin, such as the thrust fault that moved in the 1987 Whittier narrows earthquake [Hauksson et al., 1988]. The sediments in the Los Angeles basin are more than 10 km thick [Yerkes et al., 1965; Fuis et al., 2001a]. They form a northwest-southeast elongated synclinorium, with its flanks folded and cut by Quaternary active faults such as the Whittier-Elsinore fault system, and the Newport-Inglewood and Palos Verde faults [Ziony and Yerkes, 1985; Wright, 1991].

3.5. Peninsular Ranges

[11] The Peninsular Ranges batholith is exposed southeast of the Los Angeles basin (Figure 1c), and it may extend northwest to floor at least part of the Los Angeles basin. The Peninsular Ranges batholith consists of Mesozoic plutons (tonalite, granodiorite and gabbro), and Mesozoic and older metamorphic rocks [e.g., Jennings, 1977; Silver et al., 1979].

3.6. Continental Borderland

[12] The Continental Borderland region is broken up into northwest-trending blocks separated by strike slip faults [e.g., Gibson et al., 1985]. The northeasternmost block, which underlies LARSE line 1, consists of Cenozoic volcanic and sedimentary rocks, and Catalina Schist, which is thought to be coeval and cogenetic with the Franciscan accretionary prism assemblage of the Coast Ranges and Pelona Schist [Sorensen, 1984].

4. Data

[13] In 1994, the first active source phase of the Los Angeles Region Seismic Experiment was carried out [Brocher et al., 1995; Okaya et al., 1995; Fuis et al., 1996; Murphy et al., 1996; ten Brink et al., 1996]. Three multichannel seismic (MCS) reflection lines were recorded: two (lines 1 and 2) trending approximately north to northeast and one trending approximately east-west (line 3) (Figure 1b). Onshore receivers collinear with the MCS profiles recorded air gun shots generated offshore by the R/V Ewing (Figures 1 and 2) resulting in an onshore-offshore reflection/refraction data set. Lines 1 and 2 also included ocean bottom seismometers (OBSs) beneath the MCS profiles. Line 1 was augmented by an onshore reflection/refraction survey utilizing explosive sources recorded by closely spaced receivers. The onshore explosions were not recorded by the OBSs. We present a model along LARSE line 1 that incorporates all three data sets: OBSs recording air gun shots, onshore recorders recording air gun shots, and onshore recorders recording land shots.

Details are in the caption following the image
(a) Example of onshore-offshore data from station 39 at model coordinate 165 km with picked phases labeled. The data have been band-pass filtered and are shown as true amplitude plots. Example of OBS data is presented in Figure 4 of ten Brink et al. [2000]. Example of onshore data is presented in Figures 2 and 3 of Fuis et al. [2001b]. (b)–(k) Travel time plots for a selection of stations used in the model. Solid lines are model arrival times and dashed lines are times picked from the data. (b) Ocean bottom seismometer C4 (model coordinate 17 km) recording Ewing air gun shots. (c) Ocean bottom seismometer A1 (model coordinate 76 km) recording Ewing air gun shots. (d) Onshore-offshore data recorded at station 21 (model coordinate 132 km). (e) Onshore-offshore data recorded at station 39 (model coordinate 165 km). Shaded box shows extent of seismic data shown in Figure 2a. (f) Onshore-offshore data recorded at station 47 (model coordinate 182 km). (g) Onshore-offshore data recorded at station 64 (model coordinate 216 km). (h) Explosion 9450 (model coordinate 103 km) recorded on onshore stations. (i) Explosion 8050 (model coordinate 152 km) recorded on onshore stations. (j) Explosion 8302 (model coordinate 178 km) recorded on onshore stations. (k) Explosion 8360 (model coordinate 184 km) recorded on onshore stations.

[14] The offshore OBS data set recorded along line 1 consists of eight OBSs located beneath the line 1 MCS profile [ten Brink et al., 1996] (Figure 1c). These OBSs recorded air gun shots from the 8470 cubic inch tuned air gun array of the R/V Ewing. These data show upper crustal refractions (Pg) out to maximum offsets of 60 km. The southernmost two instruments also recorded Moho reflections (PmP). Examples of the OBS data are given by ten Brink et al. [2000, Figure 4]. These data primarily provide information about the upper crust beneath the offshore region.

[15] The onshore explosion data set recorded along line 1 consists of shot points every kilometer across the northern San Gabriel Valley and San Gabriel Mountains, and shot points spaced at 5–50 km elsewhere [Murphy et al., 1996] (Figure 1c). Shot sizes generally ranged from 5 to 907 kg, although a 2722-kg shot was used at the northernmost end of line 1. Seismic recorders were closely spaced (100 m) in the region of dense shot locations and more sparsely spaced (250–1000 m) elsewhere [Murphy et al., 1996]. A total of 649 instruments were used. Examples of the onshore explosion data are given by Fuis et al. [2001b, Figures 2 and 3]. These data show crustal refractions out to maximum offsets of 160 km, and crustal and Moho reflections. This study is mainly concerned with lower crustal structure, and we have not included the crustal reflections in the model presented here. A detailed upper crustal model, including the crustal reflections, is given by Fuis et al. [2001a].

[16] The onshore-offshore data set recorded along line 1 consists of the R/V Ewing air gun shots recorded by 84 onshore seismometers [Okaya et al., 1995]. The R/V Ewing made multiple traverses along each MCS profile. Some traverses were carried out at night to minimize cultural noise, some had large shot intervals (60 s) appropriate for OBS recording, while others had short shot intervals (20 s) appropriate for higher-fold MCS acquisition. We examined all the data and picked the best traverse in terms of signal-to-noise ratio for each station, trying where possible to use the 20-s shot interval data. Signal-to-noise ratios on the line 1 onshore-offshore gathers are very low in most of the Los Angeles basin and San Gabriel Valley, and are also low in parts of the Mojave Desert, where offsets are large. We were only able to correlate phases with confidence on 29 out of the 84 stations (Figures 1c and 2a). Data from these stations and the land shots are used to create the model of the lower crust presented in this study. The line 1 onshore-offshore data set contains crustal refractions (Pg) out to maximum crossover distances of 170 km. Pn is observed out to offsets of 215 km, and strong PmP is observed on most of the 29 gathers used.

5. Modeling

[17] We modeled line 1 as a 270-km-long, 2-D model, extending from San Clemente Island (model coordinate 0 km) to the Mojave Desert (model coordinate 270 km). We use a forward modeling approach that utilizes both conventional ray tracing (using MacRay [Luetgert, 1992]), and travel time calculations through a gridded model (based on code developed by Hole [1992]). A single model of the upper crust along the entire line 1 transect was obtained as follows. For the onshore region, the model of Fuis et al. [2001a] was used. For the offshore region, a starting model was taken from ten Brink et al. [2000] and T. M. Brocher (written communication). The composite model was checked to make sure it was consistent with all three data sets (offshore OBS data, onshore explosion data, and onshore-offshore data) by interactive ray tracing, resulting in slight adjustment to the offshore part of the model. The range of uncertainties for the travel time picks are ±20 ms for Pg in the onshore explosion data and offshore OBS data, ±30 ms for PmP in the onshore-offshore data, ±40 ms for Pn in the onshore-offshore data, ±40 ms for PmP in the onshore explosion data and offshore OBS data, and ±75 ms for Pg in the onshore-offshore data.

[18] We extended the upper crustal model to upper mantle depths by including travel times (Pg, Pn and PmP) picked from data recorded by the 29 best stations in the onshore-offshore data set. Moho reflections (PmP) from all three data sets (southernmost two OBSs, 13 of the onshore-offshore stations and data generated by 13 of the onshore explosions) were used to constrain the depth to the Moho and, to some extent, the velocity of the lower crust.

[19] The upper crust is quite complex, with large lateral velocity variations beneath the entire length of the model (Figure 3a) resulting from the fact that major faults juxtapose rocks of significantly different velocities. In the offshore region, substantial bathymetry results in large velocity variations in the uppermost 2–3 km of the model. Using MacRay [Luetgert, 1992], it was not always possible to generate rays through the complex layered model that reached the surface at all the offsets represented in the data. By rasterizing the layered model into a grid with 100 m by 100 m cells, however, we were able to generate travel times through the gridded model using code modified from Vidale [1988] and Hole [1992] to compare with the observed travel times from the three data sets.

Details are in the caption following the image
Velocity model. Solid lines are velocity boundaries, dashed lines are velocity contours. Numbers are velocity in km/s. Vertical exaggeration is 2, except in Figure 3b, which has a vertical exaggeration of 4. LAB is the Los Angeles Basin, NIF is the Newport-Inglewood fault, PVF is the Palos Verde fault, SAF is the San Andreas fault, SGF is the San Gabriel fault, SGM are the San Gabriel Mountains, SGV is the San Gabriel Valley, SMF is the Sierra Madre fault, VT is the Vincent Thrust, and WF is the Whittier fault. (a) Upper 10 km of the final model constrained by crustal refractions generated by offshore air gun shots recorded by ocean bottom instruments, and onshore stations recording onshore explosions. Heavy lines are bathymetry, topography and the ocean surface. The upper crust between model coordinates 0 and 100 km is modified from ten Brink et al. [2000]. The upper crust between 100 and 270 km is slightly modified from Fuis et al. [2001a]. Circles are ocean bottom seismometer locations, triangles are onshore-offshore receiver locations; stars are explosion locations for gathers recording PmP arrivals. Thick shaded lines represent reflectors modeled from wide-angle reflections in the explosion data set [Fuis et al., 2001a]. (b) Forward “layered” model. Light and medium shaded regions in the upper crust are constrained by the Pg phase in the OBS data (I), the onshore-offshore data (II) and the explosion data (III). The dark shaded region (IV) shows the constrained part of the mantle (onshore-offshore Pn data). The heavy line on the Moho shows the parts of the Moho constrained by PmP bounce points from all data sets. Box with dashed outline shows region of poorly constrained velocities. (c) Velocity model showing the lower crust (shaded) divided into five regions (A–E), which are discussed in the text. The decoupling zone between brittle and ductile lower crustal deformation is shown with a heavyweight dashed line. Vertical dashed line at model coordinate 50 is the boundary between the Inner and Outer Continental Borderland. (d) Ray diagram for PmP from the two OBSs that have this phase. (e) Ray diagram for PmP from the onshore-offshore data. (f) Ray diagram for PmP from the onshore explosion data.

[20] Our final model for the whole crust is shown in Figure 3b. The onshore explosion data constrain the uppermost 10 km between model coordinates 100 and 270 km. This upper crustal part of the model is changed little from that of Fuis et al. [2001a]. At the base of this constrained region velocities reach 6.2 km/s in the Mojave Desert and about 6.0 km/s in the San Gabriel Mountains. In the offshore region (model coordinates 0–100 km), the eight OBSs constrain the uppermost 8 km of the offshore region, and velocities of 6.0 km/s are reached at 8 km depth. The onshore-offshore data provide midcrustal constraints from crustal refractions (Pg), and require velocities of 6.6 km/s at 15 km depth beneath the center of the model (model coordinates 90–130 km). Examples of model fit to the data are shown in Figure 2.

[21] Lower crustal velocity information is obtained indirectly from mantle refractions (Pn) in the onshore-offshore data set, and Moho reflections (PmP) in all three data sets. Arrivals through the mantle and to the Moho have to travel through the lower crust on their downward and upward paths. This does not constrain velocity in the way that rays turning in the lower crust would, but it does provide some measure of average lower crustal velocity. For simplicity, we assume that there are no velocity inversions with depth below the regions constrained by the upper crustal refractions. We estimate the error in lower crustal velocity by perturbating the lower crustal velocities until the misfit between the modeled travel times and the picked travel times exceeds the picking error.

[22] Beneath the northern part of the model (model coordinates 180 to 270 km), velocities are relatively low. We assume velocities in the mid and lower crust in this region to be greater than or equal to 6.2 km/s, the deepest-constrained upper crustal velocity. Velocities producing the best fit to Pn are 6.3±0.1 km/s for the lower crust, and 7.8±0.1 km/s for the uppermost mantle (Figure 3b). In the center of the model (model coordinates 80 to 160 km), we assume lower crustal velocities are at least 6.6 km/s, the deepest-constrained midcrustal velocities. The best fitting velocity at the base of the lower crust in this region is 6.8±0.05 km/s. The southern part of our model shows low to intermediate (6.0–6.6 km/s) velocities for the upper and midcrust, similar to that of ten Brink et al. [2000]. We model a thin (5 km) layer of intermediate-to high-velocity (6.6±0.05–6.8±0.05 km/s) material beneath the entire Continental Borderland. In contrast, ten Brink et al. [2000] model high velocities (6.75 km/s) at the base of the crust only beneath Santa Catalina Island, with lower (6.5 km/s) velocities at the base of the crust elsewhere. Our model is constrained by more data, including the onshore-offshore data, than that of ten Brink et al. [2000]. The lower crustal region between model coordinates 160 and 180 km (dashed box, Figure 3b) is poorly constrained, as there are no PmP arrivals from the Moho in this region (Figures 3e and 3f). Extending the uppermost-constrained velocity of 6.0 km/s down to the Moho in this region results in the same model travel times (within picking errors) as those generated using a velocity of 6.3 km/s extrapolated from the lower crust to the north, or even 6.5 km/s. We assign this region a velocity of 6.25±0.25 km/s (dashed box, Figure 3b) This part of the model is discussed in more detail in section 6.

[23] Using the best fitting lower crustal velocities of 6.3 km/s in the northern part of the model, and 6.6–6.8 km/s in the central and southern part of the model, we modeled the depth to the Moho using the travel times of Moho reflections (PmP). We modeled the PmP arrivals from all three data sets. The data require the lower crust to have a “root” between model coordinates 160 and 240 km, which reaches a maximum depth of 35 km in a region centered beneath the surface trace of the San Andreas fault (Figure 3). This root represents a deepening of 6–8 km compared with crustal thickness to the north and south. In addition, the Moho becomes shallower still south of model coordinate 100 km, reaching a depth of 22 km beneath the offshore region southwest of Santa Catalina Island (model coordinate 50 km) (Figure 3). The exact depth to the Moho is, or course, dependent on the lower crustal velocities used, which, as discussed above, are not well resolved. Lower crustal velocities as much as 0.2 km/s greater than modeled, the maximum velocities allowed by the data, would result in a 1-km-increase in the depth to the Moho. The general shape of the Moho, however, is fairly robust.

[24] Using the Pn arrivals from the onshore-offshore data set, and the PmP arrivals from all three data sets, we can obtain a reasonably well constrained model with regards to the lower crust and Moho. Despite the uncertainties in our modeling, there are two persistent features of the lower crust and Moho. First, there is the crustal root, which is located beneath the surface trace of the San Andreas fault. Second, there is a difference in lower crustal velocities and thickness between the southern (fast, thin), central (fast, thick) and northern (slow, thick) parts of the model (Figure 3b).

[25] In the onshore region (model coordinates 100–270 km), our velocity model is based on the upper and midcrustal parts of the model of Fuis et al. [2001a]. It does not, therefore, differ significantly from this model. Like the model of Fuis et al. [2001a], the model of Lutter et al. [1999] was derived from the onshore LARSE explosion data, and is also quite similar to our model (Figures 4k to 4o). In the offshore region (model coordinates 0–100 km), the mid and upper crust of our model is generally similar to that of ten Brink et al. [2000], except in the Outer Continental Borderland, where it is somewhat faster (Figure 4b), and beneath Santa Catalina Island, where it is somewhat slower. Velocity models used by Ryberg and Fuis [1998], Kohler and Davis [1997], and Kohler [1999], which are based on preliminary analysis of LARSE refraction data, are superceded by our model.

Details are in the caption following the image
(a) Velocity model showing the locations (shaded bars and arrows) of velocity-depth profiles shown in Figures 4b–4p. ICB is the Inner Continental Borderland, LAB is the Los Angeles Basin, NIF is the Newport-Inglewood fault, OCB is the Outer Continental Borderland, PVF is the Palos Verdes fault, SAF is the San Andreas fault, SGF is the San Gabriel fault, SGM are the San Gabriel Mountains, SGV is the San Gabriel Valley, SMF is the Sierra Madre fault, and WF is the Whittier fault. Dashed lines are velocity contours (see Figure 3 for velocity values). (b)–(i) Velocity-depth curves for our model compared to previously published models that show a poor match (mostly models earlier than 1977). (j)–(p) Velocity-depth curves for our model compared to previously published models that show a good match. Numbers on velocity depth curves are additional identifiers for model authors (see legend).

[26] Velocity models developed from other data sets include Shor and Raitt [1958] for the offshore region just west of Santa Catalina Island (Figure 1), Hadley and Kanamori [1977] for the San Gabriel Mountains, and Hauksson and Haase [1997] and Hauksson [2000] for the Los Angeles basin, San Gabriel Mountains and Mojave Desert. The models of Hauksson and Haase [1997] and Hauksson [2000] generally agree with our model (Figure 4). The biggest discrepancies between our model and previously published models are seen in the models of Hadley and Kanamori [1977] (Figures 4d to 4i) and Shor and Raitt [1958] (Figures 4b and 4c), both of which are based on data sets that are much sparser than the LARSE data set. Both have significantly shallower Moho's in the San Gabriel Mountains and southern Mojave Desert (model coordinate170–220 km), and both are faster than our model in the San Gabriel Mountains. The model of Hadley and Kanamori [1977] is used for routine earthquake locations in southern California. Their model is based on data from an array of 120 stations within the Transverse Ranges and Mojave Desert. They used Nevada Test Site shots and quarry blasts to record data on three profiles, and they were able to correlate Pg and Pn phases (no PmP), which were used to derive a crustal and uppermost mantle velocity model [Hadley and Kanamori, 1977]. They also created a map of Moho depth based on P wave delay times, which revealed a high-velocity “ridge” in the mantle beneath the Transverse Ranges. Our model is derived from a much denser 2-D array of instruments and many more sources, enabling us to correlate Pg, Pn and PmP, and create a more detailed velocity model in the same region.

[27] Zhu and Kanamori [2000] also derived a map of Moho depth using teleseismic receiver functions recorded at stations in southern California. Their method is independent of P wave velocity, but depends on Vp/Vs ratios in the crust. Their model shows a 37-km-deep Moho beneath the eastern Transverse Ranges, a 31-km-deep Moho beneath the western Transverse Ranges, and a 29–30-km-deep Moho, and no crustal root, beneath the central Transverse Ranges, which is the region of our model. Vp/Vs values are not well known across the whole of southern California [Zhu and Kanamori, 2000], and if P wave velocities vary widely in the upper and lower crust, as demonstrated by our model, S wave velocities may vary considerably too, creating the potential for a larger range of Vp/Vs than that used by Zhu and Kanamori [2000]. We believe that our data set, which is of higher resolution than that used by Zhu and Kanamori [2000], can be used to derive a well-constrained P wave model for the central Transverse Ranges.

6. Density Models Along Line 1

[28] Bouguer gravity data along line 1 [Oliver et al., 1980; Langenheim and Jachens, 1996] were modeled to provide additional information about the nature of the lower crust beneath line 1, in particular, beneath the San Gabriel Mountains (model coordinates 160 to 180 km), which is the most poorly constrained region of the velocity model. Modeling of potential field data is particularly nonunique, and we therefore use the velocity model as a starting point to constrain the density models. We are chiefly concerned with the long-wavelength features of the gravity data that are primarily controlled by deeper density variations in the crust. Short-wavelength misfits should be the result of upper crustal density variations and thus be accounted for by the velocity model. The first density model (Figure 5b) was created by converting the velocity model to a density model using the velocity-density functions of Nafe and Drake [1957] and Gardner et al. [1974] for velocities <5.5 km/s. For velocities >5.5 km/s, the velocity-density relationship of Christensen and Mooney [1995] for “all rocks except monomineralic rocks” was used. The resulting calculated gravity does not match the observed gravity data in the region between the Los Angeles Basin and the Mojave Desert (Figure 5a). The mismatch between calculated and observed gravity curves is as high as 50 mGal using the Nafe-Drake function [Nafe and Drake, 1957], is even larger using the Gardner function [Gardner et al., 1974], and occurs above the portion of the lower crust that we wish to constrain. Furthermore, the calculated gravity curve for the Mojave Desert region is consistently 20–30 mGal higher than the observed gravity data. The mismatch, based on its 40-km wavelength, is primarily caused by to midcrustal rocks.

Details are in the caption following the image
(a) Gravity data plotted against gravity calculated through two density models along line 1. Solid line is observed Bouguer gravity. Short-dashed line is the gravity calculated through a density model derived from the velocity model using the Nafe-Drake curve [Nafe and Drake, 1957] for velocities <5.5 km/s and the Christensen and Mooney [1995] velocity-density curve for velocities >5.5 km/s, and a density of 2.7 g/cm3 in the southern part of the crustal root, equivalent to a velocity of 6 km/s. Long-dashed line is the gravity calculated through a density model derived from the velocity model using the Gardner curve [Gardner et al., 1974] for velocities <5.5 km/s and the Christensen and Mooney [1995] velocity-density curve for velocities >5.5 km/s, and a density of 2.7 g/cm3 in the southern part of the crustal root, equivalent to a velocity of 6 km/s. (b) Density model (derived from the velocity model) for the curves shown in Figure 5a. Dashed lines show density contours and solid lines show density boundaries. Numbers are densities in g/cm3. (c) Gravity data plotted against gravity calculated through two density models along line 1. Solid line is observed Bouguer gravity. Short-dashed line is the gravity calculated through a density model derived from the velocity model using the conversions discussed in the text, and a density of 2.7 g/cm3 in the southern part of the crustal root (shaded region in Figure 5d), equivalent to a velocity of 6 km/s. Medium-dashed line is the gravity calculated through a density model derived from the velocity model using the conversions discussed in the text, and a density of 2.8 g/cm3 (equivalent to a velocity of 6.3 km/s) in the uppermost part of the crustal root and 2.9 g/cm3 (equivalent to a velocity of 6.5 km/s) in the lowermost part of the crustal root (shaded region in Figure 5d). Long-dashed line is the gravity calculated through a density model derived from the velocity model using the conversions discussed in the text, and a density of 2.9 g/cm3 (equivalent to a velocity of 6.5 km/s) in the crustal root (shaded region in Figure 5d). (d) Density model for the curves shown in Figure 5c. Dashed lines show density contours and solid lines show density boundaries. Numbers are densities in g/cm3. Shaded region is the region being tested in this model.

[29] We next evaluated whether the density-velocity function used is appropriate for lithologies exposed in the San Gabriel Mountains and Mojave Desert. McCaffree-Pellerin and Christensen [1998] found that the density of the Pelona schist (2.72 g/cm3) is higher than would be predicted from its seismic velocity (5.9 km/s; 2.68 g/cm3) using the function of Christensen and Mooney [1995]. We performed an additional 50 density measurements of Pelona Schist samples, and found an average density of 2.75 g/cm3. The crystalline rocks in the San Gabriel Mountains have an average density of 2.73 g/cm3 [Langenheim, 1999]. For the Mojave Desert, the average density of pre-Cenozoic rocks is 2.67 g/cm3. On the northern side of the San Andreas fault, the Christensen and Mooney [1995] function predicts higher densities than measured on the basis of velocity. Part of the discrepancy in the calculated and observed values near the San Andreas fault is resolved by recognizing that extensive fracturing most likely causes the low velocities in the velocity model. Fracturing does not affect density as severely as it does velocity. A study of granitic rocks in a well about 1 km from the San Andreas fault in the Gabilan Range (450 km to the northwest of line 1) found that saturated macrocracks produced 3–7% reduction in rock density, and a 30–50% reduction in seismic velocity [Stierman and Kovach, 1979].

[30] We used the above information to adjust the densities of the upper crust in the Mojave Desert and San Gabriel Mountains. We also adjusted (1) the densities in the offshore region to honor bathymetry, (2) the southwest boundary of the Los Angeles basin to honor borehole data, and (3) velocities in the mantle to reflect the high-velocity mantle root beneath the Transverse Ranges [Hadley and Kanamori, 1977; Kohler, 1999]. Modeling the seismic data only allowed us to determine that the velocity within the southern part of the crustal root (model coordinates 160-180 km) is between 6.0 and 6.5 km/s. Using the Christensen and Mooney [1995] velocity-density conversion, we tested densities of 2.7 g/cm3 (6.0 km/s) and 2.9 g/cm3 (6.5 km/s). The best fitting model has a density of 2.9 g/cm3 in the southern part of the crustal root, which corresponds to a velocity of 6.5 km/s.

7. Geological Interpretation of the Velocity Model

[31] For details in interpretation of the complex upper crust, the reader is referred to ten Brink et al. [2000] and Fuis et al. [2001a, 2001b].

[32] In our model, the middle and lower crust have a simpler velocity structure than the upper crust, although this may also result from the decrease in resolving power of seismic data with increasing depth. We suggest that beneath the Los Angeles basin, the mid and lower crust are most likely decoupled from the upper crust at the base of the seismogenic zone (about 15 km depth) [Hauksson, 2000]. Beneath the San Gabriel Valley and San Gabriel Mountains we follow the interpretation of Ryberg and Fuis [1998], that the lower crust is decoupled from the upper crust along a décollement, which dips northward from a depth of 15 km beneath the southern San Gabriel Valley to 20 km depth at the San Andreas fault, where it terminates. It is also possible that the lower crust is decoupled from the upper crust in the Mojave Desert at a zone of midcrustal reflectors at 20 km depth [Fuis et al., 2001a]. We include in “lower crust” all material beneath the interpreted decoupling zone (bold dashed line, Figure 3c), and we divide the lower crust into five regions (A–E) based on our velocity model (shaded regions, Figure 3c).

7.1. Region A, Continental Borderland

[33] The Continental Borderland can be divided into the Outer and Inner Continental Borderland (vertical dashed line at model coordinate 50, Figure 3c). The Outer Continental Borderland, south of model coordinate 50, has a low to intermediate velocity (6.0–6.6 km/s) midcrust (down to 18 km depth), and a relatively thin (5 km) intermediate-to high-velocity (6.6–6.8 km/s) lower crust. The Moho is at about 22 km depth (region A, Figure 3c). The Inner Continental Borderland, between model coordinates 50 and 100, includes a step in the thickness of both the middle and lower crust at about model coordinate 60 km, where the lower crust becomes thicker and the mid-crust becomes thinner. The Moho starts to deepen toward the north at model coordinate 50 km, from 22 to 27 km depth (region A, Figure 3c). In the interpretation of ten Brink et al. [2000], the entire Continental Borderland crust is composed of Catalina Schist. Our lower crustal velocities are too fast, however, to reasonably interpret as Catalina Schist (an equivalent to Pelona Schist) (Figure 6a). Our velocity structure for the Continental Borderland is consistent with Catalina Schist underlain by greenschist facies basaltic rocks at the base of the crust (Figure 6a). Such a structure is interpretable as a fossil subduction zone, where the basaltic rocks are tectonically underplated, similar to other such zones imaged along the coast of California (see Fuis and Mooney [1990]; Miller et al. [1992]; Howie et al. [1993]; Page and Brocher [1993]; Beaudoin et al. [1996]; Fuis et al. [1996]; Henstock et al. [1996]; Holbrook et al. [1996]; Godfrey et al. [1997], summarized by Fuis [1998]; and Ryberg and Fuis [1998]). Such structure is also interpretable as a core complex, similar to the hypothesis of ten Brink et al. [2000], but where the basaltic rocks are magmatically underplated.

Details are in the caption following the image
Velocity-depth profiles for our velocity model (shaded regions are envelopes enclosing these profiles) plotted with velocity-depth profiles for various lithologies [Christensen and Mooney, 1995] (solid black lines). The data from Christensen and Mooney [1995] are corrected for temperature with depth in an average heat flow regime. The range of values measured for the anisotropic Pelona Schist is shown by the patterned region. Pelona Schist velocity data is taken from Godfrey et al. [2000]. The long-dashed line shows average Pelona Schist values (two fast velocities and one slow velocity) [Godfrey et al., 2000]. The short-dashed line shows average Pelona Schist values taken from McCaffree-Pellerin and Christensen [1998]. (a) Continental Borderland. (b) Los Angeles and San Gabriel Valley basins and their basement. The light shaded region shows basin velocities. The darker shaded region shows basement velocities. (c) Mid and lower crust of the San Gabriel Mountains and the crustal root. Velocities taken from region C using the preferred velocity of 6.5 km/s in the crustal root (Figure 3c) are shown with medium shading. Velocities taken from region C using the other possible velocities of 6.0–6.5 km/s in the crustal root (Figure 3c) are shown with light shading. Velocities taken from region D (Figure 3c) are shown with dark shading (narrow block). (d) The Mojave Block.

7.2. Region B, Los Angeles and San Gabriel Valley Basins and Their Basement

[34] The lower crust (below 15 km depth) beneath the Los Angeles and San Gabriel Valley basins has intermediate to high velocities (6.6–6.8 km/s) similar to the offshore region, but the region of these intermediate to high velocities is greater in thickness (10–12 km versus 5 km) (compare regions A and B, Figure 3c). The crust is about 27 km thick in region B. The midcrust (above 15 km depth) beneath the Los Angeles Basin is interpretable as either Peninsular Ranges batholithic rocks (e.g., granites, diorites and diabase), or Catalina Schist (an equivalent to Pelona Schist) (Figure 6b). The lower crust may be either the mafic root of the Peninsular Ranges batholith or underplated basaltic rocks (Figure 6b). Aeromagnetic data are consistent with Peninsular Ranges rocks in the subsurface, which are magnetic, rather than Catalina Schist, which is not magnetic [Langenheim, 1999].

7.3. Regions C and D, Crustal Root

[35] The midcrust of the San Gabriel Mountains, above the reflective zone imaged by Ryberg and Fuis [1998] at 15–20 km depth, south of the San Andreas fault has a relatively low velocity (5.8–6.0 km/s). In contrast, the midcrust beneath the Mojave Desert, north of the San Andreas fault, has a low to intermediate velocity (6.2–6.3 km/s). Velocities within the lower crust (beneath the reflective zone (décollement) imaged by Ryberg and Fuis [1998]) may also change across the deep projection of the San Andreas fault, which extends into the center of the 35-km-deep crustal root. Velocities in the southern part of the crustal root, south of the projection of the San Andreas fault, (region C, Figure 3c) are not well constrained (between 6.0 and 6.5 km/s), although gravity data suggest densities consistent with a faster velocity, close to 6.5 km/s. Such a velocity is slower than the lower crust of region B. The lower crust in the northern part of the root, north of the projection of the San Andreas fault, (region D, Figure 3c), has a velocity of 6.3 km/s, possibly lower than region C, and similar to region E. The rock in region D is consistent with intermediate to granitic rocks, or possibly Pelona Schist (Figure 6c).

7.4. Region E, Mojave Block

[36] North of the crustal root, mid-and lower crustal velocities remain fairly low (6.3 km/s) down to the Moho, which rises to ∼28 km depth. The relatively low velocity of region E, like that of region D, is consistent with intermediate to granitic composition, and/or Pelona Schist (Figure 6d). It should be noted, however, that the mid-and upper crustal velocities in the Mojave Desert are not compatible with Pelona Schist (Figure 6d) [Lutter et al., 1999; Fuis et al., 2001a], and models of aeromagnetic data suggest that nonmagnetic Pelona Schist can only exist at depths below 8 km [Langenheim, 1999].

8. Discussion

[37] The crustal root in our model is centered beneath the surface trace of the San Andreas fault, and is therefore north of the topographic high along line 1, the San Gabriel Mountains. A similar crustal root is modeled by Kohler and Davis [1997] and Kohler [1999], although the roots in those models are deeper because those authors assumed lower crustal velocities that were significantly different from those presented here. Airy compensation is apparently not occurring beneath the San Gabriel Mountains, but the crustal velocity model is more complex than is assumed in Airy compensation. Sheffels and McNutt [1986] modeled southern California as consisting of two plates that partially support the crustal load. Their model invokes a rigid northern plate (Mojave block), and a weaker southern plate (Continental Borderland and Los Angeles Basin region). The boundary between these plates is along the southern side of the San Gabriel Mountains, south of the San Andreas fault. Sheffels and McNutt [1986] propose that the Transverse Ranges are regionally compensated by the stiff northern plate, which is underthrust by the weaker southern plate. If the southern plate is actually subducting, or sinking vertically into the asthenosphere, it may be related to the mantle anomalies modeled by various authors [Hadley and Kanamori, 1977; Humphreys et al., 1984; Humphreys and Clayton, 1990; Kohler, 1999]. In particular, Kohler [1999] modeled a narrow, vertical anomaly that extends ∼200 km into the lithosphere, centered directly beneath the surface trace of the San Andreas fault and crustal root beneath LARSE line 1. A gravitationally unstable sinking plate may enhance crustal contraction at the plate boundary [Fleitout and Froidevaux, 1982] and help form a crustal root by drawing lower crustal material into the mantle above the sinking Pacific plate (Figure 7). In section 9 we calculate how much crustal contraction is implied by the formation of the crustal root in our model.

Details are in the caption following the image
Schematic diagram showing the weak Pacific lithosphere sinking vertically into the asthenosphere at the Pacific–North America plate boundary [Fleitout and Froidevaux, 1982]. The lower crust has thickened into a crustal root (black) as a result of ductile deformation in response to horizontal compression across the plate boundary, enhanced by the Pacific plate sinking into the lithosphere. SAF is the San Andreas fault.

9. Implications of the Crustal Root

[38] It has been suggested that the upper crust is deforming by rigid block rotations that are decoupled from lower crustal and mantle deformation [Nicholson et al., 1994; Shen et al., 1996; Ryberg and Fuis, 1998; Kohler, 1999]. We hypothesize that lower crustal deformation has been occurring in a ductile manner based on the thickening of the lower crust in the region of the crustal root. We perform a simple mass balance to determine the amount of crustal shortening required to form this root. We assume the depth of decoupling is at the base of the seismogenic layer, at about 15 km depth, beneath the Los Angeles basin. We then assume it follows the décollement interpreted by Ryberg and Fuis [1998] and Fuis et al. [2001a] beneath the southern and central San Gabriel Mountains (Figure 3c). This décollement dips north from ∼15 km depth beneath the southern San Gabriel Valley to 20 km depth beneath the central San Gabriel Mountains. We also assume the midcrustal reflections at ∼20 km depth beneath the northern San Gabriel Mountains and Mojave Desert form a décollement in which the relative sense of motion is opposite, and symmetrical, to motion on the décollement beneath the southern and central San Gabriel Mountains [Fuis et al., 2001a]. We assume the depth of decoupling did not alter over the time of convergence. We assume for simplicity that all the thickening of the lower crust in regions C and D (Figure 3c) is due to north-south compression between the Pacific and North American plates, and that none of the thickening has resulted from out-of-plane motion, which makes our convergence estimate a maximum. Our calculation assumes that the lower crust of the Mojave block (region E, Figure 3c) and the mid and lower crust beneath the Los Angeles and San Gabriel Valley basins (region B, Figure 3c) are undeformed. We further assume that the San Andreas fault is a vertical plate boundary, although Fuis et al. [2001a] have modeled it with a steep (∼83°) northward dip. Our simple calculation estimates how much horizontal motion (contraction) is required to thicken the bounding blocks (Figure 8a) into the root seen in our model (Figure 8b).

Details are in the caption following the image
Schematic diagram of regions B, C, D, and E from Figure 3c showing how the mass balance calculation was performed. Not to scale. LAB is the Los Angeles basin, SAF is the San Andreas fault, SGM are the San Gabriel Mountains, and SGVB is the San Gabriel Valley basin. (a) Prior to compression. Mid and lower crust of the blocks before compression (shaded) showing the geometry used in the calculation. The upper boundary of these blocks is the decoupling zone that separates deformation by rigid block rotation in the upper crust (white) from ductile deformation in the lower crust (shaded). (b) Present-day. Mid and lower crust of undeformed blocks (I and V) with additional material gained by the lower crustal root as a result of kinking (II) and compression (III and IV). The upper boundary of these blocks is the decoupling zone that separates deformation by rigid block rotation in the upper crust (white) from ductile deformation (shaded) in the lower crust.

[39] The additional rock added to the base of region C (III, Figure 8b) amounts to an excess of 62 km2, which we assume was derived from region B (Peninsular Ranges and other lower crust) (I, Figure 8b), which has an average thickness of 12 km (Figure 8b). An excess of 62 km2 derived from a starting thickness of 12 km implies 5.2 km of shortening from the south (Figure 8). The material added to the base of region D (IV, Figure 8b) amounts to an excess of 238 km2, which we assume was derived from region E (Mojave block) (V, Figure 8b). We assume the Mojave block (region E, Figure 3c) is decoupled from the mid and upper crust at 20 km depth, and has a 28-km-deep Moho. It therefore has a thickness of 8 km. An excess of 238 km2 derived from starting thickness of 8 km implies 29.8 km of shortening from the north (Figure 8). There is also an additional amount of shortening due to the kinking of the lower crust in region C (II, Figure 8b). The kink produces an additional 0.6 km of shortening (Figure 8). Our simple calculation suggests that a total of about 35.6 km (0.6 km plus 5.2 km from the south and 29.8 km from the north) of shortening has taken place across the San Gabriel Mountains and San Andreas fault.

[40] If the Big Bend region of the San Andreas fault formed at 5 Ma, 35.6 km of shortening implies a north-south convergence rate of ∼7.1 mm/yr in this region. If this convergence instead began with the uplift of the Transverse Ranges at 3.4–3.9 Ma, then the rate is ∼9.2 to 10.6 mm/yr. The maximum estimate for the component of motion perpendicular to the San Andreas fault based on geometrical considerations for the Big Bend region is ∼15 mm/yr [Bird and Rosenstock, 1984]. Geodetically measured rates perpendicular to the San Andreas fault in the Transverse Ranges vary from “very little” to ∼10 mm/yr [Weldon and Humphreys, 1986; Cheng et al., 1987; Lisowski et al., 1991; Shen, 1991; Donnellan et al., 1993; Feigl et al., 1993; Shen et al., 1996]. Geodetic rates, of course, measure the amount of upper crustal shortening, and this amount might be quite different from that in the lower crust and mantle. Weldon and Humphreys [1986] have shown that the upper crust accommodates compression largely by rotations and “escape” tectonics. There is little actual convergence thickening and uplift between individual blocks. Whether the calculated convergence rate is 7.1 or 10.6 mm/yr, the rates we estimate are reasonable compared to modeled and/or measured rates. The crustal root could, therefore, result from convergence. In our calculation, most of the crustal root was formed from Mojave block material. This suggests that the lower crust of the Mojave block is weaker, and more easily able to flow into the root, than lower crustal material from the south [cf. Sheffels and McNutt, 1986]. The inference of a weaker lower crust beneath the Mojave block is consistent with its relatively low velocity. Quartz-rich (low velocity) rocks should be weaker than feldspar-and/or olivine-rich (higher velocity) rocks.

10. Conclusions

[41] Data from the 1994 active phase of the Los Angeles Region Seismic Experiment have been modeled to show an approximately 8-km-thick lower crustal root centered beneath the surface trace of the San Andreas fault. This root is about 80 km wide, and consists of material with different velocities and densities on either side of the deep extrapolation of the San Andreas fault. This may be additional evidence that the San Andreas fault is vertical or steeply dipping, and penetrates the entire crust. We assume that and lower crustal deformation associated with north-south compression across the San Andreas fault in the Big Bend region is decoupled in the midcrust. Beneath the offshore Continental Borderland and Los Angeles basin regions we assume decoupling occurs at the base of the seismogenic crust (∼15-km depth). Beneath the San Gabriel Valley and San Gabriel Mountains, we assume decoupling occurs along a décollement interpreted from reflection data [Ryberg and Fuis, 1998; Fuis et al., 2001a] at 15–20 km depth. Beneath the Mojave Desert we assume the décollement coincides with reflections at ∼20 km depth [Fuis et al., 2001a]. In our model, the lower crust deforms in a ductile manner by thickening at the plate boundary. A simple mass balance calculation suggests about 36 km of north-south shortening has occurred across the San Gabriel Mountains and San Andreas fault. If shortening began with either the uplift of the Transverse Ranges, or the formation of the Big Bend in the San Andreas fault at 3.4–5 Ma, shortening of 36 km implies a convergence rate of 7.1–10.6 mm/yr.

[42] Our model also shows three regions of dramatically different lower crustal velocities and thickness. North of the San Andreas fault, beneath the Mojave Desert, lower crustal velocities are relatively low (∼6.3 km/s), and the lower crust is about 8 km thick, although the lower crust is apparently indistinguishable from the midcrust in this region. South of the San Andreas fault, beneath the San Gabriel Mountains, San Gabriel Valley and Los Angeles basin, lower crustal velocities are much higher (6.6–6.8 km/s), and the lower crust is thicker (10–12 km). Further south, beneath the Continental Borderland, lower crustal velocities remain high (6.6–6.8 km/s), but the lower crust is thin (5 km). This has implications for other models and earthquake locations that currently assume a constant lower crustal velocity across the entire Los Angeles region.

Acknowledgments

[43] We would like to thank all those involved in collecting the 1994 LARSE data sets including the U.S. Geological Survey, the Southern California Earthquake Center (SCEC), the Geological Survey of Canada, IRIS/PASSCAL, and all the volunteers who helped in the field. These people are listed by Murphy et al. [1996]. We also thank R. Hawman, G. Spence and an Associate Editor of JGR for their thoughtful and helpful reviews. The data are archived at the IRIS Data Management Center and are available at http://www.iris.edu. Generic Mapping Tools (GMT) software [Wessel and Smith, 1991] was used to make some of the figures presented in this paper. This research was supported by the Southern California Earthquake Center (SCEC). SCEC is funded by NSF Cooperative Agreement EAR-8920136 and USGS Cooperative Agreements 14-08-0001-A0899 and 1434-HQ-97AG01718. SCEC contribution 536.