Insights into the geometry and evolution of the southern San Andreas fault from geophysical data, southern California

The southern San Andreas fault has not experienced a large earthquake in historical time and thus is considered to pose significant hazard to southern California (USA). The subsurface geometry of the fault is important for modeling how energy produced by a rupture propagates through the crust. Despite the San Andreas fault being one of the most studied fault zones in the world, its geometry and evolution are not fully understood. In particular, the dip of the San Andreas fault southeast of the restraining bend in San Gorgonio Pass (Fig. 1) has been the subject of debate, where the fault is depicted either as a vertical fault or as a steeply to moderately northeast-dipping fault.

Here we provide new constraints on the subsurface geometry of the San Andreas fault zone from modeling of potential-field anomalies along two profiles in northern Coachella Valley that build on earlier studies that incorporated seismic-reflection and velocity profiling (Fuis et al., 2012, 2017; Langenheim et al., 2014). Although potential-field anomalies provide mathematically nonunique solutions, these anomalies combined with mapped geologic relations provide important constraints on fault geometry (Saltus and Blakely, 2011). Analysis of potential-field anomalies along the southern San Andreas fault support earlier work based on seismicity, interferometric synthetic aperture radar (InSAR), and mechanical modeling that indicate a moderate to steep northeast dip of the fault in Coachella Valley (Fialko, 2006; Lin et al., 2007; Lindsey and Fialko, 2013; Fattaruso et al., 2014).

The present northeast dip of the southern San Andreas fault may not have been a constant throughout its history of displacement (Fattaruso et al., 2016; Langenheim et al., 2016). Although uplift northeast of the fault and transpression are predicted given the azimuth of present-day relative plate motion (Fattaruso et al., 2014), the motion of the Pacific plate relative to the North America plate during the lifetime of the San Andreas fault has rotated clockwise (Atwater and Stock, 1998, 2013; DeMets and Merkouriev, 2016). We combine geophysical constraints on key subsurface features such as basin geometry and basement terranes with tectonic reconstructions to evaluate the possible influences of these features on the history of San Andreas fault dip. We investigate the possibility that the dip of the San Andreas fault along the eastern margin of Coachella Valley has not been constant throughout its history of displacement and examine the possibility that a deep crustal projection of the Peninsular Ranges batholith caused the current dip of the fault. We use two tectonic reconstructions of San Andreas fault displacement (Bennett et al., 2016; Matti and Morton, 1993) to show the position of this deep crustal feature relative to the San Andreas fault during the past 6 m.y.

The southern San Andreas Fault System has a long and complex history that is the subject of considerable debate. One set of models posits that dextral displacement related to the Pacific–North America plate boundary prior to 5–6 Ma was concentrated in the continental borderlands west of Baja California, Mexico (Spencer and Normark, 1979; Stock and Hodges, 1989), and along proto–San Andreas faults in the Transverse Ranges northwest of Coachella Valley (Ingersoll and Rumelhart, 1999). At ca. 5–6 Ma, most of the dextral displacement shifted east of Baja California to the Gulf of California (Oskin and Stock, 2003; Bennett and Oskin, 2014) and extended into the Salton Trough and Coachella Valley via early San Andreas–like faults (e.g., Altar and Algodones faults in Fig.1). Another set of models argues that dextral displacement was accommodated in the Gulf of California between 12 and 6 Ma (Matti and Morton, 1993; Fletcher et al., 2007) or east of the Salton Trough considerably earlier, at ca. 17–12 Ma (Powell, 1993). More recently, detailed fault studies from the onshore margins of the northern Gulf of California and Salton Trough suggest that eastward shift of plate-boundary shear into the continent was a gradual process, with the onset of significant dextral displacement east of Baja California at ca. 9–7 Ma (Seiler et al., 2010, 2011; Dorsey et al., 2011; Bennett and Oskin, 2014; Bennett et al., 2017; Umhoefer et al., 2018).

Many tectonic reconstructions have been presented for the southern San Andreas fault, starting with early palinspastic reconstructions that invoked ~220–240 km and 40–60 km of slip on the San Andreas and San Gabriel faults, respectively (e.g., Crowell, 1962), that continue to be cited today. Subsequently, alternative cross-fault correlations based on regional geologic mapping indicated that the San Andreas fault has no more than 150–160 km of displacement at the latitude of San Gorgonio Pass (Weldon et al., 1993; Powell, 1993; Matti and Morton, 1993). Clockwise rotations in the eastern Transverse Ranges result in more dextral slip along the San Andreas fault south of the rotating blocks than to the north (Powell, 1993), a model that requires the two tectonic systems to be in part coeval. This solution was explored by Dickinson (1996) and Darin and Dorsey (2013). Most recently, Bennett et al. (2016) integrated what is known about translational and transrotational tectonics to develop an animated tectonic history for southern California during the past 11 m.y.

Although the timing of initiation of the San Andreas fault in Coachella Valley is debated, there is little question that strike-slip displacement in the Salton Trough and farther southeast was ongoing and at rates similar to modern-day rates by 5–6 Ma (Oskin and Stock, 2003; Bennett and Oskin, 2014). By 6 Ma, the southern San Andreas Fault System was kinematically linked to faults in the Gulf of California along oblique strike-slip faults such as the Altar and Algodones faults (Pacheco et al., 2006). Since at least 6 Ma, the southern San Andreas fault has formed the eastern margin of the basin underlying Coachella Valley north to the latitude of Indio, where the fault breaks up into multiple strands heading west into San Gorgonio Pass (Fig. 1), with a complicated history of fault reorganization (Matti and Morton, 1993).

Normal faulting in the Pacific–North America plate-boundary zone along the western margin of the Salton Trough was coeval with the development of dextral transtensional slip along the San Andreas Fault System (Axen and Fletcher, 1998). Initiation of the West Salton detachment fault in the Peninsular Ranges, the northernmost of these extensional structures, may have occurred as early as 12 Ma (Shirvell et al., 2009), but more recent thermochronologic and stratigraphic studies suggested that exhumation associated with the detachment began in earnest ca. 9–7 Ma (Dorsey et al., 2011; Mason et al., 2017). Total extension accommodated by the West Salton detachment is estimated to be a minimum of 8–10 km (Shirvell et al., 2009) and ended at ca. 1–1.2 Ma (Axen and Fletcher, 1998; Lutz et al., 2006).

An important factor to consider when discussing the southern San Andreas fault is that the fault consists of multiple strands, especially in the San Gorgonio Pass Region and across the northern part of Coachella Valley. From northeast to southwest, the Mission Creek, Banning, and Garnet Hill strands all have been active in the Quaternary. Hence, it is important to be specific about which San Andreas fault strand is active during any specific period of the fault zone's structural evolution. Here we apply the name “San Andreas fault” in Coachella Valley only to that segment southeast of the junction of the Mission Creek, Banning, and Garnet Hill fault strands where the San Andreas fault consists of a single major strand (SAF-CV in Fig.1). Elsewhere, for example, in northern Coachella Valley and San Gorgonio Pass, we will specifically name the strand of the San Andreas fault being discussed. We use the following interpretations put forth by Matti and Morton (1993) for the northern branches of the San Andreas fault. They include the Mission Creek, Mill Creek, Wilson Creek, and San Bernardino strands. The Mission Creek and Wilson Creek strands are responsible for most of the cumulative displacement of the southern San Andreas Fault System in San Gorgonio Pass (~134 km; Matti and Morton, 1993). These strands were active from ca. 5 Ma to ca. 1.5 Ma and together are considered here as the long-term strand of the fault in San Gorgonio Pass because they are almost everywhere coincident. For simplicity, we refer to this combined strand as the Mission Creek strand. The Mission Creek strand is coincident with the Mill Creek and San Bernardino strands southeast and northwest of San Gorgonio Pass, respectively.

The middle strand of the San Andreas fault, the active Banning strand in Coachella Valley, is collocated with an ancestral (>ca. 5 Ma) Banning fault as it enters and exits San Gorgonio Pass. These interpretations imply that the Banning fault separates Peninsular Ranges basement to the southwest from San Gabriel Mountains–type basement to the northeast and that the Mission Creek strand separates San Gabriel Mountains–type rocks to the southwest from Mojave Desert–type basement to the northeast (Fig. 1A). As discussed by Langenheim et al. (2005), these basement types can have characteristic gravity and magnetic signatures.

The bend in the plate boundary in the northern Coachella Valley was most likely developed (and required for continuity) when the plate boundary moved eastward from offshore Baja California to its current location in the Gulf of California. As the Pacific plate continued its northwestward movement, it converged with the North America plate and downwelled (underthrust) in the region of this bend, where its movement direction was oblique to the plate boundary (Fuis et al., 2012). Passive seismic imaging of the lithosphere in this region shows Pacific plate (Peninsular Ranges) lithosphere continuing northward beneath the North America plate and sinking to depths of as much as 200 km in the region beneath and north of the San Bernardino Mountains (Kohler et al., 2003; Fuis et al., 2012) whereas Peninsular Ranges lower crust appears to extend at least 30 km beneath the North America plate north of the traces of the San Andreas fault in San Gorgonio Pass (Barak et al., 2015). The presence of the Peninsular Ranges basement abutting the North America plate can be traced southward through Coachella Valley to about the middle of the Salton Sea (about the latitude of bb in Fig.1) (Han et al., 2016a, 2016b). Southward of this location on both sides of the plate boundary, metamorphosed and unmetamorphosed sediment forms the upper crust, and mafic intrusions form the lower crust (Fuis et al., 1984; Fuis and Kohler, 1984).

Many studies have examined the dip of the San Andreas fault southeast of San Gorgonio Pass using various approaches, including seismicity, geodesy, mechanical modeling, and various geophysical methods. Each by itself has advantages and limitations. Along much of the southern San Andreas fault, seismicity does not lie directly beneath the strands of the fault as expected for a vertical fault. Instead, in northern Coachella Valley, seismicity defines 45°–70° northeast dips for the faults that ruptured in the 1986 North Palm Springs and 1948 Desert Hot Springs earthquakes (Jones et al., 1986; Nicholson, 1996).

Farther south along the Coachella Valley segment of the San Andreas fault, seismicity patterns have been interpreted in two different ways. One interpretation attributes a vertical dip to the San Andreas fault based on a few earthquakes that lie beneath the mapped trace, and more abundant off-fault seismicity is relegated to other faults (Nicholson et al., 2009). According to this paradigm, the Mecca Hills stretch of the San Andreas fault is vertical, and seismicity that defines a northeast dip below depths of ~6 km is attributed to the Mecca Hills–Hidden Springs fault system, given the slight obliquity between the trend of the deep seismicity and that of the San Andreas fault surface trace and that seismicity defines a vertical San Andreas fault farther south at the Brawley seismic zone (Nicholson et al., 2015). The other interpretation attributes the northeast-dipping alignment of microearthquakes to the San Andreas fault at depth (Lin et al., 2007). Geodetic and InSAR measurements indicate asymmetric strain rates across the San Andreas fault are best fit by a northeast-dipping San Andreas fault (Lindsey and Fialko, 2013). Modeling of creep triggered by the 2017 M_w 8.2 Chiapas (Mexico) earthquake along the San Andreas fault from Bombay Beach to Mecca Hills along the eastern shore of the Salton Sea (Fig. 1) supports a steep northeast dip of the fault to a depth of 2–3 km (Tymofyeyeva et al., 2019). Mechanical modeling indicates that the pattern of subsidence and uplift is best fit by a northeast-dipping San Andreas fault (Fattaruso et al., 2014) compared to either a single vertical San Andreas fault or combination of a vertical San Andreas fault and a blind northeast-dipping Mecca Hills fault. These various data sets together indicate that active deformation is on a northeast-dipping San Andreas fault. Both models of fault geometry are in any case a simplification of the structure in the Mecca Hills, which is characterized by multiple faults with a complex history during the past 3–5 m.y. (Bergh et al., 2019; McNabb et al., 2017). These faults are interpreted as an active flower structure by Fuis et al. (2017).

Other geophysical data sets that pertain to crustal structure provide information on the dip of various strands of the San Andreas fault in Coachella Valley. A northwest-trending gravity low is located along the length of Coachella Valley (Fig. 1B), reflecting a basin that shoals to the northwest (Langenheim et al., 2005); this feature is corroborated by seismic tomography (10 × 10 × 3 km grid; Ajala et al., 2019) and detailed seismic profiling (lines 4 and 6 of Fuis et al., 2017) of the Salton Seismic Imaging Project (SSIP). The Coachella Valley segment and the Mission Creek strand of the San Andreas fault coincide with a pronounced gravity gradient that marks the northeastern margin of the gravity low. The fault dips steeply to the northeast based on the position of the steepest horizontal gravity gradient in the lower half of the overall gradient (Figs. 1 and 2). The continuity of this gradient suggests a fairly simple basin margin, despite the presence of mapped faults that indicate a wider zone of deformation near the surface (Lancaster et al., 2012; Jänecke et al., 2018).

Coincident with the steep gravity gradient are prominent aeromagnetic highs that extend along the San Andreas fault from the Mojave Desert to the latitude of Indio (Fuis et al., 2012; Fig. 1C). The prominent highs diverge to the east away from the San Andreas fault and follow the northeastern side of the Clemens Well fault (Griscom and Jachens, 1990), interpreted to be a much older strand of the San Andreas system (Powell, 1993). Another set of magnetic highs, albeit more subdued, continue along the northeastern side of the San Andreas fault southeast to Bombay Beach. Modeling of these anomalies jointly with gravity across lines 4, 6, and 7 of Fuis et al. (2017) and Langenheim et al. (2014) along with a profile near Indio (Fuis et al., 2012) indicates a moderate to steep northeast dip along the San Andreas fault (Mission Creek strand along line 6). Like the continuity of the gravity gradient marking the eastern edge of the Coachella Valley basin, the continuity of magnetic highs along the Mission Creek strand and Coachella segment is consistent with the presence of a long-term strand that has accommodated most of the cumulative lateral and vertical offset.

Interpretations from seismic tomography also support to first order a northeast dip of the San Andreas fault and Mission Creek strand. Ajala et al. (2019) compared the location of velocity contrasts at depths of 2–9 km with the surface trace of the San Andreas fault and Mission Creek strand. Lower velocities associated with Cenozoic basin fill extend northeastward in Coachella Valley to the San Andreas fault at depths of 2–4 km, whereas higher velocities interpreted as basement extend northeast of the fault. This geometry is consistent with a northeast-dipping San Andreas fault and Mission Creek strand. Share et al. (2019) also interpreted a northeast dip of the Banning–Garnet Hill strand along one of their Vs and Vp/Vs profiles (Vs is S-wave velocity, and Vp is P-wave velocity), although it is difficult to project that dip onto adjacent profiles, perhaps because the resolution value (r_s) was below their best resolution value as determined by checkerboard tests.

Ambient-noise tomography provides constraints on the dip of the fault and structure within the lower crust. Barak et al. (2015) showed vertical to northeast-dipping shear-wave velocity contrasts throughout the crust of the North America plate southeast of Cajon Pass (cp in Fig.1) that project to the surface trace(s) of the San Andreas fault. High-velocity lower crust extends north from the Peninsular Ranges beneath the various surface traces of the San Andreas fault and beneath the San Bernardino Mountains north of those traces. The top of this body appears to coincide with the base of seismicity in San Gorgonio Pass that dips gently northward, most notably along their profiles e and f (dashed black lines in Fig.1). Barak et al. (2015) suggested that this lower-crustal body may have influenced the moderate dip of the San Andreas fault according to a magnetic model by Fuis et al. (2012) that is ~10 km southeast of profile f (sb in Fig.1). The underthrusting of Peninsular Ranges lower crust modeled by Barak et al. (2015) is consistent with the downwelling (underthrusting) of the Peninsular Ranges–Pacific plate lithosphere beneath the San Bernardino Mountains interpreted by Fuis et al. (2012).

Seismic reflections have been interpreted as dipping fault planes along various strands of the San Andreas fault. Seismic reflections along SSIP line 6 suggest a steep dip to the Banning fault in the upper 10 km or so, below which the fault dips more moderately to the northeast (Fuis et al., 2017). Aftershocks of the 1986 North Palm Spring earthquake (Jones et al., 1986) suggest that this transition may lie at or above 6 km depth if the earthquake ruptured the Banning (or Garnet Hill) strand. Similarly, seismic reflections along SSIP line 4 indicate a steep northeast dip of the San Andreas fault, changing to a moderate dip below depths of 6 km (Fuis et al., 2017). Modeling of seismicity, reflection data, and potential-field data indicate that the San Andreas fault is not only nonvertical but also nonplanar. An exception to a northeast dip is found in the reflection study by Catchings et al. (2009) located between SSIP lines 5 and 6 at Desert Hot Springs, where steep (80°), southwest-dipping reflections in the upper 200–250 m and modeling of isostatic gravity anomalies lead to an interpretation of a steeply southwest-dipping Mission Creek strand.

To further examine the dip of the Mission Creek strand and other faults, we constructed two potential-field models along the Desert Hot Springs profile of Catchings et al. (2009) and the SSIP line 5 profile. We used surface geology (Proctor, 1968; Erskine and Wenk, 1985) and fault locations (Lancaster et al., 2012; U.S. Geological Survey and California Geological Survey, 2012) as critical constraints in our models. We developed these models independently of seismic studies, but comparison with the velocity structure of Ajala et al. (2019) and Persaud (2016) indicates first-order agreement. Sedimentary rocks were chosen to have density contrasts of 600 kg/m³ to 400 kg/m³ with respect to basement, based on other modeling studies in southern California (e.g., Mabey, 1960; Anderson et al., 2004; Dorsey and Langenheim, 2015). These density contrasts are consistent with the density contrasts predicted by the velocity structure within the basin using the relationship of Gardner et al. (1974) to convert velocity to density. Note that basin depth depends on the assumed density contrast; the lower the density contrast, the deeper the basin. Densities and magnetic susceptibilities assigned to basement rocks are guided by measurements of hand samples in the region (Anderson et al., 2004; Langenheim and Powell, 2009).

We collected new gravity data (Table S1¹) along the Desert Hot Springs profile (Fig. 2A) because the gravity variations modeled by Catchings et al. (2009) were derived from contours from an isostatic gravity map (Ponce and Langenheim, 1992). The contours were constrained by only a couple of measurements along the profile. In this study, the model profile was also lengthened at both ends to extend to basement outcrops of the Peninsular Ranges and Little San Bernardino Mountains (Fig. 2). A geothermal well (Eichelberger No. 1, American Petroleum Institute [API] 06590214; California Division of Oil, Gas, and Geothermal Resources, 2020) near the profile provides a minimum depth to basement (152 m) along the profile (Fig. 3) northeast of the Mission Creek strand. We also collected new gravity data along the SSIP line 5 profile (Fig. 4A). To place these two-dimensional profiles in a broader context, we also show maximum horizontal gradients derived from the gravity and magnetic fields (Figs. 2 and 4) using the method of Blakely and Simpson (1986). Gradient maxima occur approximately over steeply dipping contacts that separate rocks of contrasting densities or magnetizations.

Modeling of the gravity anomalies along the Desert Hot Springs profile (dh in Fig. 1; Fig.3) indicates that the basin floor, defined by the contact between Cenozoic basin fill and various basement types, slopes gently northeast from the Peninsular Ranges basement contact to a depth of 2 km, then deepens into a depression just southwest of the Garnet Hill strand. A basement high of Peninsular Ranges type lies between the Garnet Hill and Banning strands at depths of <1 km. Between the Banning and Mission Creek strands, San Gabriel Mountains–type basement is interpreted to form a gentle asymmetric arch, overlain by Cenozoic deposits as much as 2.1–2.2 km and ~1.5 km thick northeast and southwest of the Banning and Mission Creek strands, respectively. The arch is also apparent in the magnetic data (yellow and green areas southwest of the Mission Creek strand in Fig. 2C), which we interpret as arising from a magnetic body within the upper-plate San Gabriel Mountains–type basement (blue region with relatively higher magnetic susceptibility [S] in Fig.3). The Mission Creek strand lies at the base of the gravity gradient (Fig. 2B), consistent with a northeast dip, although the presence of low-density basin-fill materials at the surface on either side of the fault makes determination of dip more difficult. Nonetheless, a vertical dip at depth would result in a gentler gravity gradient than observed (Fig. 3), with an even gentler gradient that deviates even more from the observed data if the fault dips to the southwest (not shown in Fig.3). A fairly gentle southwestward decrease in magnetic values from the Little San Bernardino Mountains to the Banning strand marks to first order the presence of magnetic rock types northeast of the Banning strand, particularly northeast of the Mission Creek strand. Superposed on the decrease are local, short-wavelength (<2 km wide) lows that coincide with the Garnet Hill and Mission Creek strands. Similar-wavelength highs coincide with the basement arch between the Banning and Mission Creek strands and the deepening of the basin southwest of the Garnet Hill strand. The wavelength of these anomalies indicates depth to the top of sources at or above the basement surface. As done for SSIP line 6 (Fuis et al., 2017) located a few kilometers to the northwest (Fig. 2), the short-wavelength high and low southwest of and beneath the Garnet Hill strand are modeled as Miocene(?) volcanic sources having strong remanent magnetization. The magnetic high associated with the basement arch at 12–13 km along-profile distance is clearly a localized body between the Mission Creek and Banning strands (Fig. 2C). A vertical dip of the Mission Creek fault would result in overestimation of the magnetic anomaly amplitude (Fig. 3), but this can be resolved by small modifications to the magnetizations of the basement bodies.

The joint gravity and magnetic model is consistent with a steep northeast dip of the Mission Creek fault in the upper few kilometers. The seismic profile of Catchings et al. (2009; red box in Fig.3) provides seismic velocity information to ~200 m depth, and although they showed multiple reflections interpreted as part of a flower structure to depths of 600 m, it is difficult to pick out offset reflections below ~200 m. This depth of observation may not be representative of the subsurface geometry below 200 m. It is common for high-angle strike-slip faults to exhibit multiple anastomosing fault surfaces, changes in dip direction, and flower structures near Earth's surface (Wilcox et al., 1973; McClay and Bonora, 2001; Jin and Fialko, 2020). Furthermore, it is also clear that the model profile crosses a section of the Mission Creek strand where the gravity and magnetic expression differs from that of the fault to the northwest and southeast in that the gravity gradient is wider (Fig. 2B) and there is a magnetic body immediately to the southeast of the fault (Fig. 2C). The model profile crosses a localized zone of uplifted basin fill (hills) that may reflect local geometric complexities in the fault zone, coincident with a short restraining bend (Fig. 2A), whereas to the northwest of the hills, the Mission Creek strand then forms a slight 500-m-wide releasing bend over a distance of ~5 km (circled in Fig.2). Such structural complication suggests a more complex fault system with strike-slip fault strands that can dip steeply in either direction. Thus, the steep southwest dip of one trace of the Mission Creek strand imaged by Catchings et al. (2009) may not be representative of the dip elsewhere along the strand or of other traces, or of the dip below ~200 m depth.

The Banning and Garnet Hill strands of the San Andreas fault zone have steep dips, based on the gravity along the Desert Hot Springs profile. The magnetic data do not inform on the dip of the faults in the basement because Peninsular Ranges basement and lower-plate San Gabriel Mountains–type basement (Pelona Schist) are not magnetic (Langenheim et al., 2005). The gravity model indicates a steep northeast dip of the Banning strand in the upper 2 km, with thick basin fill to the northeast. The Garnet Hill strand, whose location is concealed, coincides with a steep southwest dip of the basement surface. Gravity gradients indicate that the basement high between the Garnet Hill and Banning strands is ~9 km long; this feature is absent along SSIP line 6 (Fuis et al., 2017). The basement high may represent an upper-plate block in the hanging wall of the West Salton detachment that has been modified by subsequent transpression.

We modeled a profile along SSIP line 5, ~20 km to the southeast of the Desert Hot Springs profile, close to the junction of the Banning and Mission Creek strands where their surface traces are ~2 km apart (Fig. 4). Steep reflections imaged along line 5 nearest the Banning fault have been attributed to the Banning fault (Persaud, 2016) and change to a moderate northeast dip at a depth of ~5 km.

Gravity data show a gradient parallel to the Banning fault that slopes upward toward higher values in the Little San Bernardino Mountains (Fig. 4B). The lowest values are immediately southwest of the Banning fault, with a modeled maximum thickness of basin fill of ~3 km (Fig. 5). The basin floor slopes gently downward to the northeast from exposures of Peninsular Ranges basement with a couple of subtle inflections that may mark concealed basement topography related to hanging-wall blocks of the completely concealed West Salton detachment fault. Northeast of the gravity minimum near the Banning fault, the gravity gradient is subtly stepped, with inflections at the Banning and Mission Creek strands. These inflections along the profile are not limited to the profile; maximum horizontal gradients coincide with the Banning strand and the Mission Creek strand to the southeast and northwest of the profile, respectively (Fig. 4B). Weak to moderate gravity gradients are aligned with the southeastward projection of the Garnet Hill strand and extend across the profile location. More prominent gradients trend more southerly east of Point Happy and are most likely related to the western basin margin. As in the Desert Hot Springs profile, no density contrasts were assumed within the basement southwest of the Mission Creek strand, and thus fault dip is not constrained beneath the basin fill.

A magnetic high with an amplitude of ~300 nT is present over exposures of Mojave basement in the Little San Bernardino Mountains along the profile (Figs. 4C and 5). The position of this steep southwest-facing gradient is best matched by a northeast dip of the Mission Creek strand (Fig. 5); a vertical dip results in a shallower gradient that is located ~2 km to the southwest of where it is observed (Fig. 5). The basement between the Banning and Mission Creek strands either is non-magnetic or includes a very thin magnetic layer. The Mission Creek strand appears to be the southwestern boundary of magnetic rocks of Mojave basement in the Little San Bernardino Mountains.

The gravity variations along SSIP line 5 are to first order consistent with what is predicted by the seismic velocity models of Persaud (2016; Fig. 5B) and Ajala et al. (2019; Fig. 5C). Important differences between the gravity and the two velocity models are (1) high velocities that exceed 6.4 km/s at depths between 2 and 4 km at 12–20 km along-profile distance in the Ajala et al (2019) model (PRH in Fig. 5C) and (2) the geometry of the eastern margin of the basin. The high-velocity body PRH within the Peninsular Ranges basement would have a density of ~2850 kg/m³ using the density-velocity relationship of Christensen and Mooney (1995), similar to the upper end of densities measured for Peninsular Ranges basement rocks (Langenheim and Powell, 2009). Because of its location within the basement, it is difficult to incorporate such a dense body and fit the observed gravity variations unless it is thinner or its rock type has a density lower than that predicted by the density-velocity relationship. Ajala et al. (2019) suggested that this body reflects the Eastern Peninsular Ranges mylonite zone. Limited seismic-velocity anisotropy measurements of samples of the mylonite zone indicate velocities of <6.4 km/s, with similar average velocities of 6.1–6.2 km/s of both mylonite and surrounding protolith (Kern and Wenk, 1990). Because gravity anomalies are not sensitive to anisotropic variations, we suggest that the body is not only anisotropic but also contains marble to account for the disproportionately high velocity. Marble has a velocity of 6.7 km/s but a density of ~2750 kg/m³ (Christensen and Mooney, 1995) and is present in and near the mylonite zone (Erskine and Wenk, 1985; Fig. 4A).

The eastern margin of the basin according to the seismic velocity model of Ajala et al. (2019) dips gently to the southwest, reaching a maximum depth of ~2.2 km based on the 4.5 km/s contour, interpreted by Ajala et al. (2019) as the top of basement, whereas the depth of the basin in Persaud (2016) is ~2.7 km. The gravity model suggests that the deepest part of the basin is ~3 km and is bounded on the northeast by the steeply northeast-dipping Banning strand. The seismic velocity models can be reconciled with the gravity model by recognizing that (1) velocity of basement (granite or gneiss or schist) in general increases rapidly with depth in the upper 2–3 km from 3.5 km/s to >5.5 km/s (McCaffree Pellerin and Christensen, 1998; Brocher, 2008) and (2) fracturing decreases velocity more significantly than density (Stierman and Kovach, 1979). Assuming that the top of basement can be defined by a single velocity value of 4.5 km/s or higher along SSIP line 5 is problematic, resulting in the interpretation that basement is nowhere (Ajala et al., 2019) or sporadically (Persaud, 2016) exposed at the surface along the profile, which is not supported by the geologic map (Fig. 4A). Furthermore, the deepest part of the basin may be 3–3.5 km if a velocity of 5.5 km/s is assumed to be the top of basement. Fracturing may result in lower velocities in and above the basin-basement contact between the Banning and Mission Creek strands. Given that the two strands are only ~2 km apart, smoothing in the final tomography models may also obscure abrupt changes in velocity. Likely the true geometry of the eastern basin margin lies within the range of the potential-field and tomographic geophysical models.

The magnetic anomaly bounded by the Mission Creek strand and the Coachella Valley segment of the San Andreas fault from San Gorgonio Pass to Bombay Beach indicates a northeast dip between magnetic rocks to the northeast and less magnetic rocks to the southwest where these rocks are modeled. Nicholson et al. (2009) noted that the Mission Creek strand dips to the northeast based on gravity and velocity modeling but argued that this contrast likely reflected an older ancestral basin-bounding Mission Creek fault, not an active vertical Mission Creek fault. A northeast-dipping fault in transtension would produce a basin northeast of such a fault; instead, the basin fill is thicker to the southwest of this proposed older fault. Furthermore, the magnetic contrast extends to or near the surface trace of the Mission Creek strand with magnetic rocks to the northeast, where a subsequent vertical fault at that same location at the surface would displace weakly magnetic rocks southwest of the fault and thus one would not expect to find offset magnetic equivalent bodies on the southwestern side of the fault. However, magnetic equivalent rocks are present where predicted after reconstructing offset on the San Gabriel fault and ~160 km of right-lateral offset across all strands of the San Andreas fault (A1–A4 in Fig. 1C; Griscom and Jachens, 1990). If the magnetic interface reflects only an older Mission Creek strand, it must be recently that a vertical neotectonic Mission Creek fault became active. Although the Mission Creek fault has had a slip rate of ~20 mm/yr during the past 100,000 yr in the Indio Hills near its intersection with the Banning fault (Blisniuk et al., 2021), its geomorphic expression north of Desert Hot Springs and the absence of offset Holocene deposits north of the Indio Hills (Matti and Morton, 1993; Yule et al., 2021) argue for slip transfer onto faults in the eastern Transverse Ranges, the Banning fault, and/or the Garnet Hill fault (Gold et al., 2015), and thus the Mission Creek strand may not accommodate significant offset outside of northern Coachella Valley during the past 100,000 years.

A northeast dip of the long-term San Andreas fault (Mission Creek strand and Coachella segment) southeast of San Gorgonio Pass is supported by multiple lines of geophysical data and modeling (Fuis et al., 2012, 2017). Potential-field models have the best resolution in the upper 1–5 km and indicate steep northeast dips (>65°–70°) southeast of San Gorgonio Pass. Magnetic modeling fits the observed gradients if the fault is planar to depths of 10 km or changes to more moderate dips below 5–6 km, as indicated by seismic reflections and seismicity (Fuis et al., 2017).

A nonvertical San Andreas fault has implications for vertical deformation patterns (Dair and Cooke, 2009), particularly if the fault is oriented obliquely to the direction of plate motion. Boundary-element modeling (Fattaruso et al., 2014) shows that uplift occurs on the northeastern side of a northeast-dipping San Andreas fault if the relative plate motion is rotated clockwise with respect to the strike of the fault. The present-day plate motion is oriented 321°–322° (DeMets and Merkouriev, 2016), which is rotated 7°–8° clockwise from the average trend of the Coachella Valley segment of the San Andreas fault (314°), resulting in minor transpression (Fattaruso et al., 2014). If a northeast-dipping San Andreas fault is oriented parallel (Fattaruso et al., 2016) or counterclockwise to plate motion, or transtension is superposed on a northeast-dipping San Andreas fault, subsidence is predicted on the northeastern side of fault (Fattaruso et al., 2014, 2016). During most of the time frame that the Mission Creek strand, ancestral Banning fault, and West Salton detachment were active, from at least 7–9 Ma to 1–1.5 Ma, the azimuth of Pacific–North America plate motion has rotated clockwise (Atwater and Stock, 1998, 2013; DeMets and Merkouriev, 2016), with transtension predicted across the long-term strands of the San Andreas fault in Coachella Valley until ca. 3–5 Ma. A long, large gravity low along the length of Coachella Valley lies to the southwest of the long-term San Andreas fault strands (Figs. 1B, 6B, and 7A). Although the Coachella Valley gravity low dissipates toward the southeast within the Salton Sea because of hydrothermally altered sediments and intrusions (Fig. 1B), seismic-refraction studies south of the Salton Sea (Fuis and Kohler, 1984; Persaud et al., 2016) indicate thickening of basin fill on the southwestern side of the Algodones fault, interpreted as a long-term proto–San Andreas fault (Pacheco et al., 2006), further highlighting the absence of significant basins developed to the northeast of the San Andreas fault. This disparity could be compounded because the blocks on either side of the San Andreas fault have been displaced at least 160 km during the past 5–6 m.y. (Haxel and Dillon, 1973; Matti and Morton, 1993; Powell, 1993).

To evaluate geophysical correlations across the San Andreas fault and assess the influence of long-term motion of key geophysical features on the evolution of fault geometry, we track the past position of geophysical features using published tectonic reconstructions of southern California during the past 6 m.y. We present two alternative reconstructions. First is the animated reconstruction of Bennett et al. (2016; Fig.6), which utilizes the magnitude and direction of relative Pacific–North America plate motions through time provided by Atwater and Stock (1998, 2013) as fundamental constraints. Although the Bennett et al. (2016) tectonic reconstruction shows past positions of fault-bounded crustal blocks through time, it does not explicitly track the past positions of faults. Furthermore, their model is based on relative plate motions of Atwater and Stock (1998, 2013), which differ from other reconstructions, such as that of DeMets and Merkouriev (2016), in that the Atwater and Stock plate motions vary more in azimuth and rate. Note that these variations, however, are within the uncertainties of the DeMets and Merkouriev model for the past 6 m.y. The Bennett et al. reconstruction also appears to overshoot a best-fit reconstruction of basement terranes by as much as 50 km (Powell, 1993; Matti and Morton, 1993). Thus, as an alternative, we also show the reconstruction of Matti and Morton (1993) focused in the San Gorgonio Pass restraining bend, which assumes plate motion parallel to the strike of the San Andreas fault south of the bend (Fig. 7).

For both reconstructions, we outline and track the Coachella Valley gravity low (see elongate white polygon in Figs. 6 and 7) to better understand the evolving position of this basin and to evaluate which geophysical features are juxtaposed across the San Andreas fault back in time. We also outline and track the high-velocity lower-crustal body that currently lies beneath the San Bernardino Mountains in San Gorgonio Pass (see dashed blue line in Figs. 6 and 7), interpreted by Barak et al. (2015) as an extension of Peninsular Ranges lower-crustal basement beneath the surface trace of the San Andreas fault. Tracking this body back in time provides insight into how and when the Peninsular Ranges lower crust may have interacted with crustal blocks northeast of the San Andreas fault.

Both reconstructions highlight the fact that crystalline basement likely has been at or near the surface on the northeastern side of the San Andreas fault for at least the last ~6 m.y., with no indications for widespread and deeper basin formation northeast of the fault, as might be predicted given a transtensional regime from 6 to ca. 1–1.5 Ma, a northeast San Andreas dip, and the assumption of no strain partitioning (where only pure strike-slip is allowed on the San Andreas fault and any vertical slip is partitioned onto the West Salton detachment fault). Furthermore, the Matti and Morton (1993) reconstruction (Fig. 7) predicts uplift northeast of the San Andreas fault resulting from the formation of a restraining bend in the Mission Creek fault at ca. 2.5 Ma due to movement of the Pinto Mountain fault. A northeast-dipping Mission Creek fault would have resulted in uplift of the northeastern block (San Bernardino and Little San Bernardino Mountains) as the blocks passed through the bend. This uplifted block northeast of the fault would then have been translated to its current location, juxtaposed against basin fill deposited on the southwestern block developed in the hanging wall of the West Salton detachment fault. This scenario explains the juxtaposition of basin fill on the southwestern side and uplifted basement on the northeastern side of the fault in northern Coachella Valley, although thermochronologic data suggest that the timing of uplift in the Little San Bernardino Mountains was earlier at 5 Ma, when the block was interpreted to be next to the San Gabriel Mountains by Spotila et al. (2020; see also Figs. 6, 7E, and 7F).

The absence of basins in the hanging wall of a northeast-dipping San Andreas fault, as predicted under transtension, could be explained if the fault did not initiate with a northeast dip but was rotated into that orientation from an initial near-vertical attitude (>80°). Fattaruso et al. (2016) suggested that extension on the West Salton detachment may have tilted the San Andreas fault from a vertical attitude to a northeast dip. Another possible driver for this change in dip is the oblique collision with North America of the high-velocity lower-crustal body that currently lies beneath the San Bernardino Mountains in San Gorgonio Pass (Barak et al., 2015). Barak et al. (2015) interpreted this block as underthrust Peninsular Ranges lower-crustal basement that extends northward beneath the surface trace of the San Andreas fault zone to account for the gentler dip of the fault there. Alternatively, the high-velocity body may be a remnant of the Farallon plate (Barak et al., 2015). A Farallon plate fragment beneath the San Bernardino Mountains would have been emplaced before initiation of the San Andreas Fault System. Given that the San Andreas fault appears to cut across the full thickness of the crust and possibly further into the mantle (Fuis et al., 2012), this remnant of Farallon plate presumably would have extended across the San Andreas Fault System and thus be offset by the San Andreas fault at least 160 km. Such an equivalent high-velocity feature is absent on the southwestern side of the San Andreas fault and now ~160 km northwest of this high-velocity body (see Barak et al., 2015, their figure 7b), thus supporting the interpretation that the high-velocity lower crustal body is of local Peninsular Ranges affinity and thus is attached to and related to the block southwest of the San Jacinto and San Andreas fault strands.

Rotation of the San Andreas fault from near vertical to a northeast dip predicts southwest tilts on either side of the fault, assuming the crust is rigid. Such southwest tilts on either side of Coachella Valley are not documented by available thermochronologic data or geomorphic analyses, except in the San Jacinto Mountains (Wolf et al., 1997; Spotila et al., 2020). Northeast tilting of the Little San Bernardino and Santa Rosa Mountains (Spotila et al., 2020; Dorsey and Langenheim, 2015) is opposite of what might be expected from rigid block tilting from rotating the San Andreas fault from near vertical to a northeast dip. However, the crust is likely not rigid on either side of the San Andreas fault. Northeast of the fault along Coachella Valley is the eastern Transverse Ranges, a domain of sinistral, east-striking faults that accommodated clockwise rotation. The absence of basins at the intersection of these faults with the San Andreas fault indicates nonrigid behavior (Schelle and Grünthal, 1996; Langenheim and Powell, 2009). Southwest of the San Andreas fault, the West Salton detachment fault accommodated transtension by faulting the crust of the Peninsular Ranges. Hanging-wall normal faults, although rarely exposed along the western margin of the Salton Trough, are interpreted on the SSIP line 5 and the Desert Hot Springs profiles as well as imaged in a seismic-reflection profile west of the Salton Sea (Severson, 1987). Such faults can have a large impact on patterns of uplift and subsidence (Fattaruso et al., 2016). Thus, off-fault deformation may have permitted the San Andreas fault to rotate to a northeast dip.

We examine how the high-velocity lower crust may have affected transpression along the San Andreas fault. Transpression along the San Andreas fault that produced localized uplifts in the Mecca and Indio Hills can be explained by small clockwise changes in fault azimuth (Bilham and Williams, 1985), but these localized, young uplifts are superposed on a much broader pattern of relative uplift northeast of the San Andreas fault as suggested by the gravity gradient bounding the Coachella Valley gravity low. The location of high-velocity lower crust through time varies slightly between the tectonic reconstructions by Bennett et al. (2016; Fig.6) and Matti and Morton (1993; Fig.7). The Bennett et al. reconstruction places the body adjacent to the Mecca Hills between 2 and 3 Ma (Fig. 6), whereas the Matti and Morton reconstruction places the body beneath the Mecca Hills at that time (Fig. 7D). In the latter case, transpressional deformation at Mecca Hills may be expected if the lower-crustal body acted as a wedge as it was translated obliquely along and into the San Andreas fault and thereby tilting the fault from near vertical into a northeast dip. McNabb et al. (2017) interpreted a widespread time-transgressive unconformity between members of the Palm Spring Formation as evidence of transpression in the Mecca Hills at 3.0–2.2 Ma (preferred 2.6–2.3 Ma). A similar, but poorly dated, unconformity within the Palm Spring Formation is exposed 25 km to the northwest in the Indio Hills. In a very speculative model, we would predict that the unconformity is ~1 m.y. younger in the Indio Hills and is time-transgressive, getting younger toward the northwest from the Mecca Hills (ca. 3.0–2.2 Ma) to the Indio Hills (ca. 1 Ma). Alternatively, McNabb et al. (2017) suggested that the unconformity may be related to a poorly dated, but previously documented, reorganization of plate-boundary faults in the northern Gulf of California (Nagy and Stock, 2000; Aragón-Arreola and Martín-Barajas, 2007). Furthermore, the global plate-circuit model of Atwater and Stock (1998, 2013) indicates that the Pacific–North America relative plate motion became 5°–10° more northerly at ca. 3 Ma, potentially driving uplift in the Mecca Hills and elsewhere along the plate boundary. In these scenarios, the unconformity within the Palm Spring Formation would likely be the same age throughout Coachella Valley.

Although the tectonic reconstructions of Bennett et al. (2016) and Matti and Morton (1993) differ in detail, both place the high-velocity body in the lower crust adjacent to or beneath the strands of the southern San Andreas fault at ca. 2 Ma (Figs. 6 and 7). A major reorganization within the San Andreas Fault System led to initiation of slip on the San Jacinto fault at 1.1–1.3 Ma (Matti and Morton, 1993; Jänecke et al., 2010) and development of the San Gorgonio Pass thrust zone at ca. 1.2 Ma (Matti and Morton, 1993). We speculate that oblique collision of the Pacific plate, with its deep, high-velocity lower crust, with the North America plate may have facilitated this reorganization in San Gorgonio Pass and driven the Quaternary development of the San Jacinto fault, highlighting the influence of the Peninsular Ranges batholith as it translated northwestward within the San Andreas system (e.g., Langenheim et al., 2004). A quantitative analysis of how movement of the lower crust of the Peninsular Ranges beneath the North America plate would affect the stress regime is beyond the scope of this paper. Other factors, such as displacement on the Pinto Mountain fault (Matti and Morton, 1993), displacement on the West Salton detachment (Fattaruso et al., 2016), and changes in relative plate motion (Atwater and Stock, 1998, 2013), must have all played a role in the development of the San Gorgonio Pass restraining bend.

Modeling of potential-field anomalies along two profiles in northern Coachella Valley indicates steep to moderate northeast dips of strands of the San Andreas fault in Coachella Valley, consistent with previous studies based on seismicity, reflection data, geodesy, boundary-element modeling, and other geophysical methods. The northeasternmost strand of the San Andreas fault (Mission Creek fault and Coachella Valley segment) also forms the northeastern margin of a Cenozoic basin beneath Coachella Valley. Boundary-element modeling (Fattaruso et al., 2014) indicates subsidence to the southwest of the fault given the northeast fault dip and oblique relative plate motion, consistent with the location of the basin. We speculate that a prong of high-velocity lower crust of the Peninsular Ranges has obliquely collided with the North America plate and is associated with the northeast dip of the fault. This collision may have facilitated a major fault reorganization in San Gorgonio Pass during Quaternary time.

We thank the National Cooperative Geologic Mapping and Earthquake Hazards programs of the U.S. Geological Survey for support. Jon Matti and Bob Powell provided geologic reality checks on an early version of the manuscript. Scott Bennett provided very thoughtful and helpful comments and suggestions that significantly improved the manuscript. Susanne Jänecke's very thorough and rigorous review resulted in a more honest appraisal of the shortcomings of the manuscript. Joann Stock and Craig Nicholson helped clarify phrasing and pointed to additional pertinent references. Comments from an anonymous reviewer pointed to additional support for a northeast dip along the Salton Sea. Tait Earney and Noah Athens provided crucial field support for collecting gravity measurements in Coachella Valley.