Ongoing oblique slip at the Pacific–North America plate boundary in the Salton Trough produced the Imperial Valley (California, USA), a seismically active area with deformation distributed across a complex network of exposed and buried faults. To better understand the shallow crustal structure in this region and the connectivity of faults and seismicity lineaments, we used data primarily from the Salton Seismic Imaging Project to construct a three-dimensional P-wave velocity model down to 8 km depth and a velocity profile to 15 km depth, both at 1 km grid spacing. A VP = 5.65–5.85 km/s layer of possibly metamorphosed sediments within, and crystalline basement outside, the valley is locally as thick as 5 km, but is thickest and deepest in fault zones and near seismicity lineaments, suggesting a causative relationship between the low velocities and faulting. Both seismicity lineaments and surface faults control the structural architecture of the western part of the larger wedge-shaped basin, where two deep subbasins are located. We estimate basement depths, and show that high velocities at shallow depths and possible basement highs characterize the geothermal areas.
The Imperial Valley, located south of the Salton Sea in southern California (USA; Fig. 1), is a transtensional region involving (1) a complex connection between the main plate boundary Cerro Prieto fault in the south and the San Jacinto fault in the west, and (2) a major right step between the Cerro Prieto fault and the southern San Andreas fault (e.g., Fuis and Mooney, 1990, and references therein). The valley-fill sediments were deposited by the Colorado River from 5.3 Ma, and rapidly buried, heated, and intermingled with mafic intrusions to form metasedimentary crust (Fuis et al., 1984; Dorsey, 2010). Valley boundaries are the West Mesa and East Mesa and adjacent mountains (Fig. 1), which are mostly underlain by pre-Cenozoic crystalline rocks (Jennings, 2010). Thinned crust (∼18–20 km thick; Han, 2016) is found along the valley axis, and relatively high-velocity lower crust and low-velocity upper mantle underlie the region and the Chocolate Mountains to the east (Parsons and McCarthy, 1996). The upper and mid-crustal structure was previously studied in several seismic profiles, although with limited horizontal resolution (2 shots per profile despite the 0.5–1 km instrument spacing) by Fuis et al. (1984) who defined a sedimentary layer, a low-velocity metasedimentary basement, and a mafic subbasement. Lin (2013) combined these data with earthquake data and presented a three-dimensional (3-D) VP upper and mid-crustal model. The model resolution is determined by the earthquake data and is thus relatively coarse. In addition, the VS crustal-upper mantle structure has been studied using ambient noise (Barak et al., 2015).
Despite clear seismicity lineaments beneath the valley, surface traces of the Superstition Mountain fault and the Superstition Hills fault are not visible in the valley, possibly due to recent sedimentation by Lake Cahuilla followed by extensive agricultural activity. Similarly, there is no clear surface trace of the Brawley fault in the Brawley seismic zone, a speculated step-over fault between the San Andreas fault and the Imperial fault. In this region, earthquakes tend to occur in swarms (e.g., the August 2012 Brawley swarms), often related to the geothermal systems (Hill et al., 1975). Increased recent seismicity in the Salton geothermal area has been attributed to fluid extraction and injection during geothermal energy production (Brodsky and Lajoie, 2013). Studies suggest that the swarms are generally <10 km deep (Johnson and Hadley, 1976; Doser and Kanamori, 1986; Chen and Shearer, 2011; Hauksson et al., 2013); therefore, a more accurate shallow velocity model is of particular importance for studying the source properties of swarms and refining event locations (Chu and Helmberger, 2013; Wei et al., 2013).
As part of the 2011 Salton Seismic Imaging Project (SSIP; Rose et al., 2013), a 2 km spaced seismic array, along with three 0.1 km receiver-spacing seismic lines were deployed in the Imperial Valley (Fig. 1). Here we invert first-arrival traveltimes to determine the upper and mid-crustal structure along a profile from the Peninsular Ranges to the Chocolate Mountains, and a 3-D VP upper crustal model. Our results reveal significant variations in the seismic structure, and provide improved and more detailed information for earthquake hazard studies.
DATA AND METHOD
We used a subset of SSIP shots and receivers (Fig. 1). Because of zigzags along the profile (SSIP line 2, green line in Fig. 1), we performed a 3-D inversion, but strongly smoothed the third dimension in which structural variations are assumed to be small. For the 2-D model, we manually picked 8636 first arrivals from 24 shots recorded at 773 SSIP receivers. For the 3-D model (blue outline in Fig. 1), we picked an additional 25,565 arrivals from a total of 47 shots recorded at 1602 SSIP receivers. To increase the 3-D ray coverage, we added 538 picks from SSIP shots recorded by the Southern California Seismic Network (SCSN), 875 picks from the 1979 active source experiment (IV1979; Fuis et al., 1982), and 214,463 picks from SCSN earthquake recordings processed by the Southern California Earthquake Data Center (Fig. 1). The initial earthquake hypocenters are from the full time span (1981 to June 2011) of the relocated catalog of Hauksson et al. (2012) that were recorded by at least four SCSN stations and had focal depths ≤10 km, which is approximately the maximum depth constrained by the explosion data. Inversion results without earthquake data are also shown for comparison. Following Magistrale (2002), we used earthquakes over a larger region to define seismicity lineaments L1–L4 (Fig. DR1 in the GSA Data Repository1; discussed herein).
We perform tomographic inversions at 1 km grid spacing using a back-projection method (Hole, 1992; Hole et al., 2006) in which traveltimes are calculated through a finite-difference solution of the eikonal equation, and the model is updated iteratively by back-projecting the traveltime residuals along the ray paths. In the case of the 3-D model, alternating inversions of the velocity structure and the earthquake parameters were performed. For each iteration, the model is smoothed using a moving average to ensure stability, with a gradual reduction in the size of the smoothing window after every third iteration. The final result was smoothed over 5 × 5 × 2 km.
For the 2-D model, we used a smoothed version of the Southern California Earthquake Center Community Velocity Model–Harvard (CVM-H; http://scec.usc.edu/scecpedia/CVM-H) as the starting model, modified with linearly increasing VP from 1 km/s at 5 km elevation to the top of the CVM-H model at 1 km depth. These results closely match those obtained with a 1-D starting model of linearly increasing speeds, indicating limited dependence on the starting model. For our 3-D model, we used a smoothed 1-D starting model from Kanamori and Hadley (1975), padded at the top with linearly increasing VP ranging from 1 km/s at 5 km elevation to 5 km/s at 1 km depth.
RESULTS AND DISCUSSION
Due to reduced ray coverage below 15 km in the profile, results to 15 km depth are shown (Fig. 2), along with the number of rays in each cell, and the 5 km smoothed initial CVM-H model. The result to 40 km depth is provided in Figure DR3. Depth slices of our 3-D model at 1, 3, 5, and 7 km are shown in Figures 3A–3D; the cells lacking ray penetration are shaded gray. Because the 3-D structure is obtained from both active source (SSIP and IV1979) and earthquake data, we outline the area with ray coverage from the explosive shots in cyan. Several prominent features are noteworthy and discussed in the following.
2-D Model from the Peninsular Ranges to the Chocolate Mountains
In the valley, we refer to the layer with VP = 2.0–5.65 km/s as the sedimentary layer; typically VP = 5.9–6.0 km/s represents the basement outside the valley (Fuis et al., 1982). Intense seismicity starts below the 5.65 km/s contour (purple line, Fig. 2A), extends down to ∼12 km depth, and is more depth restricted in the metasedimentary basement than in the basin flanks (also noted by Magistrale, 2002). The central basin is deepest at a profile distance of ∼140 km, just west of the Imperial fault. Here the sedimentary layer is ∼7 km thick. This coincides with the eastern end of a deep east-west–trending basin imaged in the 3-D model (Fig. 3D). At profile distances <80 km where the Peninsular Ranges basement is exposed at the surface (Jennings, 2010), the velocities are interpreted to indicate ∼1 km of weathered or cracked crystalline rocks above unaltered or intact ∼6 km/s crystalline rock.
A VP = 5.65–5.85 km/s layer (purple and green contours in Fig. 2A) is locally up to ∼5.0 km thick (across ∼10–15-km-wide zones) beneath the valley, Chocolate Mountains, and West Mesa. This thickened layer in West Mesa is likely low-velocity crystalline basement (LVZ1, Fig. 2A), and is also present in the CVM-H model (Fig. 2C), although its lower boundary varies more in our model. Previous workers assumed that in the valley, this layer represents new crust formed through the combination of sedimentary and magmatic processes (Fuis et al., 1984; Schmitt and Vazquez, 2006; Dorsey, 2010) occurring above mafic intrusions (VP ≥ 7 km/s) that underlie the valley and the Chocolate Mountains (Parsons and McCarthy, 1996). However, we note that regardless of the basement type, this layer is thickest near faults, thinning to <1 km elsewhere, and is part of a larger velocity depression, particularly near seismicity lineaments and also east of the along-strike projection of the inactive Algodones fault in East Mesa (Figs. 1 and 2). We therefore suggest that the increased thickness of this layer may in part be an indicator of active or inactive faulting. Furthermore, our model is ∼0.4–1.4 km/s slower than the CVM-H in the shallow crust above LVZ1 (0–4 km depth at 80–110 km distance). This region is associated with northeast trends in seismicity or possibly the Elsinore fault, suggesting a possible faulting-related mechanism for generating the low velocities and/or localization of faulting into low-VP damage zones.
The 6.6 km/s contour (white line, Fig. 2A) may indicate the transition from the metasedimentary basement to mafic subbasement (Fuis et al., 1982), and is shallowest at ∼11 km depth east of the Brawley fault, deepening to ∼12 km near the Imperial fault. To explain the deeper seismicity near the Imperial fault, Doser and Kanamori (1986) modeled the top of a subbasement dome at 10 km depth to produce a depression of the brittle-ductile transition zone to 12 km depth, with intense seismicity extending into the subbasement, a configuration slightly different than ours. At 11–15 km depths, similar to observations by Parsons and McCarthy (1996), the subbasement forms a broad dome underlying most of the valley, extending from 120 km profile distance near the western sea-level contour to 170 km distance near a series of short unnamed faults, and the eastern sea-level contour (Figs. 1 and 2). Approximately along-strike of the Algodones fault, at ∼17 km depth (Fig. DR3), there is an abrupt ∼5 km deepening of the 6.6 km/s contour, which is relatively flat-lying at ∼22 km depth farther east. We interpret this transition as the eastern limit of the mafic subbasement.
3-D Structure of the Imperial Valley
The depth slices reveal a narrow north-northwest–trending velocity gradient at the western basin boundary that mimics the sea-level contour (orange line) at 1 km depth (Fig. 3A). To highlight the progressively better match of seismicity trends (L1–L4, dashed purple lines in Fig. 3E) and faults with the shape of deeper basin edges, we show depths to the 5.65 km/s surface (Fig. 3F), representing approximate basement depths within the valley, and a shallower surface outside the valley where basement VP is typically higher. A shoaling basement block (HVZ1) is between the Elsinore fault and Superstition Mountain fault. The Yuha Wells fault corresponds to the southeast boundary of this block. Relatively low velocities (LVZ1) to the southwest of HVZ1 (Fig. 2 and 3C) coincide with the along-strike extension of the Elsinore fault.
The Mesquite basin, bounded by the Imperial fault and Brawley fault, appears to be a low-velocity zone at 3 km depth (Fig. 3B). In contrast, at 5 and 7 km depth, a more prominent velocity low and possible basin (Figs. 2, 3C, and 3D; LVZ2) occupies the left step between the Superstition Hills fault and the Imperial fault, west of the Mesquite basin. This velocity depression (∼15 km long, 8 km deep; Fig. 3F) coincides with a gravity low (Biehler, 1971). Based on its location in a region of local compression (assuming right-lateral slip on the Superstition Hills fault and Imperial fault), if this feature is a basin, it may have rotated from where it was first formed, or may represent an inherited structure from an earlier extensional phase that predates the Imperial fault and the San Jacinto fault zone, possibly reflecting the overprinting or destruction of an older pull-apart basin.
Another velocity low and possible basin (LVZ3, Figs. 3C and 3D) northwest of the Heber geothermal area appears to be bounded by two subparallel northwest trends in seismicity, i.e., by L3, which extends along strike from the Cerro Prieto fault, and by the Imperial fault, or to a lesser extent L2 (Fig. 3F). This basin lies in a poorly imaged region, where the exact VP values may be less accurate; however, it coincides with a broad gravity low (Biehler, 1971) and unlike the larger wedge-shaped Imperial Valley basin, which is controlled by faulting on virtually all sides, it may be a rhombochasm in an interpreted 9-km-wide right step between the Cerro Prieto fault and the Superstition Mountain fault (Magistrale, 2002).
Two kinks are observed in the basement and VP gradient that bound the western basin edge: the first is along strike of L3 at the northwestern edge of LVZ3, and the second occurs in the overlap between L1 and L2 (Figs. 3C and 3F). In addition, L1, L3, and L4 bound the western basin edge. Thus, seismicity lineaments, although mostly lacking known surface faults, exert substantial influence on the structural architecture of the valley, and along with the mapped faults appear to control the western basin boundary. This supports the continuity of slip between the San Jacinto fault zone and the Cerro Prieto fault suggested by Magistrale (2002), along conrolling structures L1–L3, indicating a more complex pattern than is evident in geodetic data (Lindsey and Fialko, 2016).
The Salton and Brawley geothermal areas are both associated with local gravity maxima (Biehler, 1971), and overlie or are located at the edge of high-velocity zones HVZ2 and HVZ3 (Fig. 3C), which at shallow depths probably resulted from cementation, recrystallization, and thermal metamorphism of sediments by circulating hot brines (Elders et al., 1972). At 7 km depth, HVZ2 and HVZ3 appear to merge into an elongated high-velocity zone that forms the eastern boundary of the BSZ.
We image a complex wedge-shaped basin at the southern end of the San Andreas fault system and its structural connections to faults and seismicity trends. New findings in our profile include localized regions of low VP (thickening of a 5.65–5.85 km/s layer) near faults or seismicity lineaments interpreted as possibly faulting related. Our 3-D VP model and basement map reveal velocity highs associated with the geothermal areas in the eastern valley. Two deep subbasins (<5.65 km/s) are located in the western part of the larger Imperial Valley basin, where seismicity trends and active faults play a significant role in shaping the basin edge.
We thank P. Umhoefer, G. Axen, D. Scheirer, V. Langenheim, editor Bob Holdsworth, and an anonymous reviewer for their comments on the manuscript. The Salton Seismic Imaging Project (SSIP) was funded by the U.S. Geological Survey Multihazards Project, and the National Science Foundation Earthscope and Margins Programs through grants OCE-0742253 (to California Institute of Technology) and OCE-0742263 (to Virginia Polytechnic Institute and State University). Persaud was supported by U.S. Geological Survey grant G15AP00062.
- Received 26 April 2016.
- Revision received 1 July 2016.
- Accepted 4 July 2016.
- © 2016 Geological Society of America