The Salton Trough (southeastern California, USA) is the northernmost transtensional stepover of the Gulf of California oblique-divergent plate boundary and is also where the southern terminus of the San Andreas fault occurs. Until recently, the distribution of active faults in and around the Salton Sea and their displacement histories were largely unknown. Subbottom CHIRP (compressed high-intensity radar pulse) surveys in the Salton Sea are used to develop a seismic facies model for ancient Lake Cahuilla deposits, a detailed map of submerged active faults, and reconstructed fault displacement histories during the late Holocene. We observe as many as fourteen Lake Cahuilla sequences in the Salton Sea (last ~3 k.y.) and develop a chronostratigraphic framework for the last six sequences (last ~1200 yr) by integrating CHIRP data and cone penetrometer logs with radiocarbon-dated stratigraphy at an onshore paleoseismic site. The Salton Sea contains northern and southern subbasins that appear to be separated by a tectonic hinge zone, and a subsidence signal across hinge-zone faults of 6–9 mm/yr (since ca. A.D. 940) increases toward the south to >15 mm/yr. The faults mapped to the south of the hinge zone appear to accommodate transtension within the San Andreas–Imperial fault stepover. We identify 8–15 distinct growth events across hinge-zone faults, meaning growth occurred at least once every 100 yr since Lake Cahuilla sedimentation began. Several faults offset the top of the most recent Lake Cahuilla highstand deposits, and at least two faults have offset the Salton Sea flood deposits. Active faults and folds were also mapped to a limited extent within the northern subbasin and display growth, but their kinematics and rupture histories require further study. The broad distribution of active faulting suggests that strain between the San Andreas, San Jacinto, and Imperial faults is highly distributed, thus discrepancies between geologic and geodetic slip-rate estimates from these major fault systems are to be expected.
The Salton Sea, California (USA), is a large (~970 km2), shallow lake that covers the southernmost extension of the San Andreas fault (Fig. 1) and the northernmost transtensional stepover associated with Gulf of California oblique rifting (Elders et al., 1972; Lonsdale, 1989; Brothers et al., 2009). More than 80% of the contemporary plate motion between the Pacific and North America plates is accommodated along active faults within the Salton Trough (Lindsey and Fialko, 2013), thus numerous studies have focused on understanding the tectonic processes and earthquake hazards of the region (Axen and Fletcher, 1998; Brothers et al., 2009, 2011; Crowell et al., 2013; Dorsett et al., 2019; Dorsey et al., 2007; Elders et al., 1972; Fialko, 2006; Fuis et al., 1984; Hudnut et al., 1989b; Lohman and McGuire, 2007; Lonsdale, 1989; Meltzner et al., 2006; Nicholson et al., 1986; Philibosian et al., 2011; Rockwell et al., 2018). The first-order geologic evolution of the Salton Trough is characterized by a series of major reorganizational events, including a late Miocene phase of orthogonal rifting, lithospheric thinning, and subsidence (Axen and Fletcher, 1998; Stock and Hodges, 1989) followed by an eastward jump of the translational boundary between the North America and Pacific plates from the former continental slope to the Gulf of California–Salton Trough corridor (Lonsdale, 1989). The Quaternary evolution is marked by the development of several right-stepping dextral transform faults between the Salton Trough and the mouth of the Gulf of California (Dorsey et al., 2007). Lithospheric thinning and subsidence within the northern gulf and Salton Trough created a vast depocenter for terrigenous sediment sourced primarily from the Colorado River (Dorsey and Lazear, 2013) but also from local mountain catchments that debouche onto the basin floor. More than 5400 km2 of the trough is presently located below sea level. Rugged topography sur-rounds the trough on all sides except to the south, where aggradation of the Colorado River delta has constructed a broad topographic sill ~13 m above sea level, separating the Salton Trough from the Gulf of California (Fig. 1).
The central trough is presently occupied by the Salton Sea, a large terminal lake that formed in A.D. 1905 by an accidental, human-made diversion of Colorado River floodwaters (e.g., Waters, 1983), and ever since, the lake has been sustained by agri-cultural runoff. Despite its unnatural origins, the Salton Sea is located within the footprint of prehistoric “Lake Cahuilla”, a name assigned to a series of massive inland lakes that filled the valley during episodic diversions of the Colorado River into the Salton Trough (Philibosian et al., 2011; Waters, 1983). Lake Cahuilla (LC) inundated the surface exposures of several major fault systems, including the confluence of the San Andreas, San Jacinto, and Imperial faults (Fig. 1), and deposited thick sedimentary sequences that have recorded paleoseismic and hydroclimatic events in the region (Brothers et al., 2011; Meltzner et al., 2006; Philibosian et al., 2011; Rockwell et al., 2018). Because LC deposits span the Salton Sea's shoreline, we are provided a rare opportunity to combine the strengths of offshore geophysical approaches with onshore geological studies within a spatially and temporally continuous depositional system (Brothers et al., 2009, 2011; Philibosian et al., 2011). In this paper, we develop a detailed stratigraphic framework for the Salton Sea by integrating offshore seismic reflection profiles and sediment core data with onshore studies of Lake Cahuilla deposits. We also expand on previous efforts (e.g., Brothers et al., 2009, 2011) to characterize the timing and style of paleoseismic deformation on active faults in the Salton Sea and discuss the regional implications presented by the observed deformation patterns.
Salton Trough Tectonic Overview
The San Andreas (SAF), San Jacinto (SJF), and Imperial (IF) faults are the major, throughgoing, northwest-striking dextral strike-slip faults that accommodate at least 80% of the total relative Pacific–North America plate motion. Estimated slip rates for all three faults vary depending on the type of study (geologic versus geodetic) and the time scale of measurement. Geodetic studies place the combined rate of the SAF and SJF between 35 and 40 mm/yr (Fay and Humphreys, 2005; Fialko, 2006; Lindsey and Fialko, 2013; Meade and Hager, 2005), but rates based on geologic data are generally lower than geodetically derived rates and are more variable between studies (Behr et al., 2010; Brothers et al., 2009; Fletcher et al., 2007; Kendrick et al., 2002; Rockwell et al., 1990). Regardless, both faults appear to transfer their entire slip budgets southward onto the IF, which carries 35 ± 2 mm/yr of the plate motion southward across the United States–Mexico border (Fig. 1; Bennett et al., 1996). The IF has ruptured twice during the past 82 yr (Mw 6.9 in 1940 and Mw 6.4 in 1979), whereas the SAF and SJF are both considered late in the interseismic phase of their earthquake cycles (e.g., Field et al., 2015). The southernmost 200 km of the SAF has not produced a major rupture in 300+ yr despite evidence that its average recurrence interval is between ~150 and 220 yr and it continues to accumulate significant strain (Behr et al., 2010; Blanton et al., 2020; Dingler et al., 2016; Fumal et al., 2002; Lindsey and Fialko, 2013; Philibosian et al., 2011; Seitz and Williams, 2007; Spinler et al., 2010). The potential for a Mw ≥7.0 rupture along the SAF to nucleate at the Salton Sea and propagate northward is a major concern for the Los Angeles metropolitan region (Brothers et al., 2011; Dingler et al., 2016; Olsen et al., 2006).
Near the town of Bombay Beach, the SAF takes a releasing step across the Salton Sea and toward the IF (Fig. 1; Brothers et al., 2009). Sub-aqueous geophysical surveys have produced the first fault maps in the Salton Sea and provided first-order constraints on kinematic history of faults as recorded in Holocene LC deposits (Fig. 2; Brothers et al., 2009, 2011). The lake is split into two depositional subbasins along an array of en echelon, southeast-dipping faults that define a tectonic hinge zone and display down-to-the-southeast vertical displacement in seismic reflection profiles. These faults are identified in the upper 50 m of the substrate and trend ~N15°E. Ongoing and rapid subsidence in the southern Salton Sea is thought to be in response to transtension induced by the SAF-IF stepover (Brothers et al., 2009; Crowell et al., 2013) as well as anthropogenic-induced poroelastic contraction associated with geothermal fluid production (e.g., Barbour et al., 2016). Southward of the tectonic hinge zone, LC strata diverge from each other and progressively thicken, suggesting that maximum subsidence occurs near the southeastern shoreline, perhaps at rates as high as 20 mm/yr over the last ~1 k.y. (Brothers et al., 2009). The projected zone of maximum subsidence is approximately coincident with a major geothermal anomaly, including a northeast-oriented string of active volcanoes (Fig. 2C), which implies that lithospheric extension within the SAF-IF stepover is at least partially compensated by magmatic injection (Fuis et al., 1984; Schmitt and Vazquez, 2006).
Numerous other secondary fault structures have been mapped to the south and west of the Salton Sea, including several short (<15 km), northeast-trending sinistral faults (Figs. 1 and 2C). Although minor amounts of late Holocene offset are observed along the Elmore Ranch fault and the Extra fault (Fig. 2C), the regional kinematic significance of these structures is not clear (Brothers et al., 2009; Hudnut et al., 1989b; Dorsett et al., 2019; Nicholson et al., 1986). Brothers et al. (2009) initially referred to the collection of northeast-trending faults in the central Salton Sea as the Extra fault hinge zone, but their Holocene deformation history and their relationship to the Extra fault proper, as originally documented by Hudnut et al. (1989b), remain uncertain. To the southeast of the Salton Sea, slip is transferred between the IF and a series of north-south–oriented faults that define the southern boundary of an active pull-apart basin (Mesquite Basin); the faults appear to accommodate a relatively high rate of regional dilation and extensional subsidence and have experienced numerous late Holocene earthquakes (Crowell et al., 2013; Mann et al., 1983; Meltzner et al., 2006). Of significant concern in regional hazard assessments is the potential for ruptures along secondary fault structures in and around the Salton Sea to trigger major earthquakes along the larger SAF, SJF, or IF systems (e.g., Hudnut et al., 1989b; Brothers et al., 2011).
The Brawley seismic zone defines a region of high seismicity (Fig. 2A) that spans the SAF-IF step-over and is inferred to result from a complicated interplay among tectonics, magmatic injection, and geothermal fluid migration. Several moderate and large historical earthquakes have occurred in and around the Brawley seismic zone, some associated with mapped surface faults and others not (Barbour et al., 2016; Crowell et al., 2013; Hauksson et al., 2013; Hudnut et al., 1989b; Johnson and Hadley, 1976; Lin et al., 2007; Lohman and McGuire, 2007; Nicholson et al., 1986). Most notable within the Brawley seismic zone is the relatively frequent occurrence of earthquake swarms typically defined by a handful of small and/or moderate earthquakes, a mixture of predominantly right-lateral and left-lateral focal mechanisms, and hundreds of microseismic events that spatially and temporally align along short (<10 km long) lineaments (Fig. 2A). The orientation of the lineaments and focal mechanisms of larger events (i.e., Mw >3.5; Hauksson et al., 2011, 2013, 2017) generally do not align with the trend and sense of displacement observed along mapped fault traces in the Salton Sea (Brothers et al., 2009). Some swarms appear to be linked with perturbations to the local stress field due to both geothermal fluid migration (Brodsky and Lajoie, 2013; Chen and Shearer, 2011; Roland and McGuire, 2009; Taira et al., 2018) and aseismic creep along active faults (Lohman and McGuire, 2007; Jiang and Lohman, 2021). Earthquakes with only moderate moment release (Mw >5.0) within the Brawley seismic zone have produced unusually large surface displacements, possibly due to a combination of brittle deformation, fluid migration, and aseismic creep (Barbour et al., 2016; Hauksson et al., 2013; Lohman and McGuire, 2007; Taira et al., 2018). Interseismic creep and triggered slip along several faults, including the SAF and IF, have been observed throughout the Salton Trough with rates varying between 2 and 20 mm/yr (Blanton et al., 2020; Hudnut et al., 1989b; Lohman and McGuire, 2007; Lyons and Sandwell, 2003; Lyons et al., 2002; Sieh and Williams, 1990; Sylvester et al., 1993; Wei et al., 2011). Although the significance of aseismic and/or triggered displacement in longer-term fault behavior remains poorly understood, it poses implications for paleoseismic reconstructions along many of the faults in the region (e.g., Sieh and Williams, 1990; Meltzner et al., 2006).
Late Holocene Sedimentation
The upper 1 km of sediment deposited in the central Salton Trough is sourced primarily from the Colorado River. Smaller lobes of coarse-grained alluvium are locally sourced from steep drainages that debouch onto the basin floor. During the late Holocene, the Colorado River episodically breached its banks, flowing northward into the Salton Trough and filling LC (Van de Kamp, 1973). When full, LC was ~100 m deep and its shoreline was located ~13 m above sea level (Fig. 1), the height of Colorado River delta (Waters, 1983). Colorado River floodwaters deposited a series of relatively coarse-grained, stacked delta sequences near the southern shoreline of the Salton Sea (Van de Kamp, 1973; Waters, 1983). During flooding and shoreline transgression, delta foresets back-stepped from the deepest parts of LC and lower-energy, deep-water lacustrine conditions prevailed until the Colorado River reverted southward back to the Gulf of California. Without significant inflow, the lake gradually desiccated and the basin became sub-aerially exposed, allowing delta foresets sourced from local catchments to prograde lakeward, culminating in the emplacement of alluvial sediments mantling the former lake bed. Repeated flooding and desiccation cycles associated with LC created distinctive depositional facies that span the modern shoreline of the Salton Sea. Onshore paleoseismic studies (Meltzner et al., 2006; Philibosian et al., 2011; Rockwell et al., 2018) have constrained the age and lithologic character of LC sediment packages over the last five to six flooding-desiccation cycles (since ca. A.D. 800), which are critical for developing a chronostratigraphic framework for LC strata observed in the Salton Sea. Philibosian et al. (2011) described lacustrine deposits at the Coachella paleoseismic site (Fig. 1) as laminated fine sands, silts, and clays, whereas the subaerial deposits were notably coarser grained and derived from both aeolian and fluvial sources. Evidence for as many as seven major earthquakes was documented at the Coachella site.
DATA AND METHODOLOGY
Seismic Reflection and Cone Penetrometer Test Data
Between 2006 and 2008, >1000 line km of high-resolution seismic CHIRP (compressed high-intensity radar pulse) profiles were acquired in the Salton Sea using Scripps Institution of Ocean-ography's Edgetech subbottom profiler (Fig. 2B; Brothers et al., 2022). Profiles were collected using the following CHIRP pulses depending on subbottom characteristics: a 0.7–3.0 kHz, 50 ms pulse and a 1.0–15.0 kHz, 30 ms pulse, producing vertical resolution of ~60 cm and ~15 cm, respectively. All shots were time stamped with differential GPS navigation. Shot traces were processed using SIOSEIS software (https://sioseis.ucsd.edu) and interpreted using Kingdom Suite software. All depth measurements from seismic sections assumed a P-wave velocity of 1600 m/s.
Between 2003 and 2006, United Research Services (URS) Corporation subcontracted Gregg In Situ to collect offshore borings and cone penetration test (CPT) data in the Salton Sea (Fig. 2B) aboard a self-propelled jack-up barge (URS Corporation, 2004). A separate study by Schroeder et al. (2002) analyzed the chemical makeup of several short cores to delineate the thickness and base of the post–A.D. 1905 Salton Sea sediment. Combined, these studies were used to ground-truth acoustic character variations observed in the high-resolution CHIRP data and develop a seismic facies model for the Salton Sea. CPTs were incrementally advanced to subbottom depths between 10 and 20 m and measured cone resistance, sleeve friction, and dynamic pore-water pressure at 2.5 cm intervals during penetration. Empirical relationships between these parameters were used to assign soil behavior types at each penetration interval (URS Corporation, 2004; Robertson, 1990, 2009). CPT-derived interpretations of soil behavior type are superposed on CHIRP sections (e.g., Figs. 3 and 4) to examine the relationships between seismic impedance boundaries and the mechanical properties of the soil (e.g., shear strength). Depth estimates from CHIRP data appear to be accurate to within ~30 cm based on visual agreement between discrete reflections and coarse-grained units in CPT plots. Several material index properties, including water content and void ratio, were measured from bore samples, but these measurements were sparsely distributed and did not show the consistent trends that are needed to correct depth measurements for compaction. Clay layers, particularly near the seafloor, have water content between 50% and 80%. Water content decreases with depth, implying that significant post-depositional compaction may have occurred (URS Corporation, 2004). We did not apply a decompaction correction; therefore, all offset measurements should be considered minimum estimates.
Fault Growth and Displacement Reconstructions
Growth faults record the interaction between vertical fault displacement and on-fault depositional processes. If the sedimentation rate of material transported as bed load outpaces the rate of displacement, we can assume the relief created during an earthquake would be infilled prior to any subsequent earthquakes. In this case, to reconstruct the displacement history, one needs to be able to resolve stratigraphic horizons that were emplaced during interseismic time periods. Thus, vertical displacement along faults in the Salton Sea is rapidly infilled by Colorado River sediment, and paleo-displacement events can be reconstructed at decadal-scale resolution (Brothers et al., 2011). Nevertheless, this approach is not reliable for strike-slip faults that do not produce resolvable vertical offset and/or for displacement events that occur during depositional intervals that do not produce differential sediment thickness across the fault (i.e., mantled deposits). Therefore, fault growth needs to be examined in conjunction with seismic facies, namely, changes in acoustic character and stratal geometry.
Identification of displacement event horizons is difficult if growth occurred during highstand intervals and displacement resulted in only slight thickening of the fine-grained facies. Most accom-modation was infilled during subsequent floods and coarse-grained deposition. The degree of subjectivity in event identification is reduced by combining quantitative (throw plots) and qualitative (digital flattening) approaches. When the sedimentation rate keeps pace with or exceeds the fault displacement rate, throw plots can be useful for extracting the relative timing of paleo-displacement events from the stratigraphic record (Brothers et al., 2011; Cartwright et al., 1998; Castelltort et al., 2004). To generate throw plots, we digitize the tops and bottoms of every identifiable stratigraphic layer using shot-trace correlation and manual picking. We match layers offset by faulting by progressively flat-tening horizons down-section. Vertical throw across each fault is measured across every horizon and along five evenly spaced vertical profiles in order to compute an average throw for each horizon.
DEVELOPMENT OF A LATE HOLOCENE STRATIGRAPHIC FRAMEWORK
In order to understand the timing and nature of fault displacement in the Salton Sea, we must first establish a high-resolution chronostratigraphic framework for the Lake Cahuilla (LC) sedimentary section by correlating onshore and offshore data. A seismic profile co-located with a mid-sea alignment of CPT soundings provides an opportunity to ground-truth seismic stratigraphy (Figs. 2C and 3). The high-amplitude reflections correspond to facies containing greater silt and/or sand components, whereas the low-amplitude and transparent reflections correspond to fine-grained clay facies. Variations in stratal geometry, acoustic character, and lithology represent cycling between rapidly deposited Colorado River flood deposits, highstand lacustrine deposits, and, in some areas, thin, reworked subaerial deposits (Figs. 3–5). Colorado River floods are associated with emplacement of relatively thick wedges of silt- and sand-rich flood deposits having maximum thickness in the southern Salton Sea (Brothers et al., 2009, 2011). As shoreline transgression occurs during flooding, deltas back-step toward the southeast and deposition in the distal, deeper waters is dominated by low-energy lacustrine conditions, resulting in deposition of pre-dominantly fine-grained silts and clays. When the Colorado River inflow resumes its path southward into the Gulf of California, LC begins to evaporate and experience shoreline regression. During periods of desiccation and relative lowstand, episodic fluvial discharge from the basin margins transports coarse-grained material onto the former lake bed, forming localized fan-delta lobes that that are apparent in topographic maps and in the subsurface data (Figs. 2C and 3B). These flood-desiccation cycles generate distinctive marker beds that are exposed in onshore paleoseismic excavations (e.g., Philibosian et al., 2011) and can be identified in CPT and boring data and in seismic reflection profiles based on acoustic character and lithology (Figs. 3–5).
Acoustic horizons representing key marker beds were extrapolated from the mid-sea type section (Fig. 3) onto crossing profiles and correlated throughout the sea. Marker beds were assigned a lithology and number that correspond to a particular lake sequence (following the convention developed by Philibosian  and Philibosian et al. ). For example, each sequence contains a fine-grained highstand lacustrine facies (“L”), a coarsening-upward lowstand (“S”) facies, and/or a fining-upward flood deposit (“F”) (Fig. 5). In most CHIRP profiles, it is difficult to identify the surface that separates the subaerial portion of the previous lake from the flooding stage of the incipient lake. Therefore, the boundaries between flood and lowstand facies are indistinguishable in the seismic stratigraphy and are labeled “S/F”. Following these criteria, offshore imaging reveals at least six additional stratigraphic sequences that have not been logged at onshore sites (Philibosian, 2007; Philibosian et al., 2011). In total, at least 14 stratigraphic sequences have been identified along a mid-sea type section and in the southern subbasin; however, the sequences are condensed in the northern subbasin and along the lake margins, making it difficult to identify variations in acoustic character below layer 9S/8F (Fig. 6).
A set of relationships need to be established between onshore and offshore LC deposits to build a chronostratigraphic framework for the seismic facies observed beneath the Salton Sea. Although we do not have direct age control for the offshore LC sediments, Philibosian et al. (2011) developed age estimates for the LC flood-desiccation cycles based on historical accounts from early explorers passing through the region and extensive radiocarbon dating of the subaerial and lacustrine units exposed at the Coachella paleoseismic site. Philibosian et al. (2011) approximated the durations of lacustrine intervals at the Coachella site using the full 95% confidence range of the modeled radio carbon ages and the date-constrained sedimentation rates for each of the intervals. We use their model (spanning the last six LC sequences) to assign approximate ages to the corresponding LC facies observed offshore.
The A.D. 1905 flood deposits and subsequent Salton Sea sediments are easily identified as a low-amplitude transparent layer overlying a southward-thickening package of high-amplitude material; this interpretation is supported by chemical profiles extracted from shallow sediment cores (Schroeder et al., 2002) and provides a historical marker bed. For the underlying units, we assume that highstand LC formed when the entire flow of the Colorado River was diverted to the Salton Trough, requiring ~20 yr to completely fill the basin to the sill elevation (Waters, 1983). As the shoreline transgressed, the transition from coarse-grained flood deposits to fine-grained, deep-water lacustrine facies was probably coincident with the Colorado River reverting southward to the Gulf of California. Given that radiocarbon age dating uncertainties are greater than the short amount of time required to fill LC, the base of each correlative lacustrine unit onshore and offshore is assumed to be synchronous. Waters (1983) estimated that once the Colorado River diverted back to the Gulf of California, LC evaporated at ~1.8 m/yr, causing subaerial exposure beneath the Salton Sea to occur nearly 60 yr later than at the highstand shorelines. The top of unit 6S/5F (or base of unit 5L), dated to ca. A.D. 940 (Table 1; Philibosian et al., 2011), represents a key marker horizon mapped in the subbottom data. The total thickness of sediment deposited above this horizon can be used to estimate average sedimentation rates throughout the basin (e.g., Figs. 3, 6, and 7) and suggests the entire LC section was emplaced during the last 3000 yr.
The Salton Sea contains two primary depocenters separated by the tectonic hinge zone described by Brothers et al. (2009). CHIRP profiles image the upper 40–70 m throughout most of the northern basin, but penetration was limited in the southern basin and in all nearshore areas by gas trapped in near-surface sediments (see shaded region in Fig. 2C). To the northwest of the hinge zone, tilted beds of the Brawley Formation are truncated along an angular unconformity that underlies the LC section for several kilometers (Brothers et al., 2009; Figs. 3, 6, and 7). Beneath the northern subbasin, a transparent wedge of sediment thickens to the northwest and separates the Brawley Formation from the overlying LC section. The majority of the LC section near the depocenters consists of fine-grained lacustrine layers (L); coarse-grained subaerial and/or flood (S/F) layers downlap basin-ward, thickening toward ephemeral streams (Fig. 3B inset; Fig. 7B) and commonly display chaotic internal bedding and irregular bounding surfaces.
All units appear to thicken toward the south-eastern shoreline where the sedimentary section is dominated by intermittent high-amplitude packages that onlap the hinge zone, including the A.D. 1905 flood deposits (Figs. 6A and 7). The estimated time required to fill LC suggests that the thick, coarse-grained fan-delta deposits accumulated in 10–20 yr, a reasonable time frame when compared with contemporary Colorado River sedimentation rates in Lake Powell upriver ~1300 km (e.g., Ferrari, 1988). Layers 2S/1F, 4S/3F, 6S/5F, and 9S/8F contain distinctive internal beds that fluctuate between transparent and high-amplitude layers (e.g., Figs. 3 and 7) that show marked divergence and thickening to the south. Brothers et al. (2011) interpreted layers 4S/3F, 6S/5F, and 9S/8F to have been emplaced during Col-orado River flooding into a dry basin or nearly dry basin. In contrast, the time lapse between highstand layers 3L and 2L (Philibosian et al., 2011; Fig. 5; Table 1) and the relatively limited spatial distribution of layer 2S/1F in the southern sub-basin suggests unit 2S/1F was emplaced during a Colorado River diversion into an existing partially filled lake. Like-wise, intervals 3S/2F, 5S/4F, 7S/6F, 10S/9F, 11S/10F, 12S/11F, and 13S/12F are relatively thin, implying they were deposited when the Colorado River was diverted northward into an existing lake (Brothers et al., 2011). Figure 5B shows a simple schematic of the interplay between shoreline position and stratigraphic sequence, where the relative thickness of lowstand and/or flood deposits depends on the position of the shoreline during the onset of Colorado River floods. Floods into a dry basin are expected to emplace thicker deposits in the southern subbasins than floods into an existing body of water. Fan-delta deposits are not present at the Coachella site due to its distal location in the basin. Instead, lacustrine deposits are separated by coarse-grained subaerial and fluvial material (e.g., Fig. 5; Philibosian, 2007; Philibosian et al., 2011).
CHARACTER AND TIMING OF TECTONIC DEFORMATION
Most of the tectonic activity appears to be concentrated in the southern subbasin, within the footprint of the San Andreas fault–Imperial fault (SAF-IF) stepover (Brothers et al., 2009). Unfortunately, the offshore extension of the SAF is obscured by gas wipeout and the total length of most faults is unknown. Penetration generally improves to the west of the inferred SAF (Fig. 2C), but in places it is difficult to distinguish between faulted layers and soft-sediment deformation likely caused by substrate fluid expulsion. These imaging limitations imply that several active faults, particularly in the southern basin, have yet to be mapped. We aim to examine the recent activity along faults that are reasonably well resolved. For discussion purposes, hinge-zone faults (“H”) are differentiated using numbers, and other faults are named according to their proximity to nearby landmarks.
The hinge zone described by Brothers et al. (2009) contains a series of relatively short, en echelon faults and fault propagation folds (Fig. 2C) that strike ~N15°E and bound the northern extent of active extension and subsidence in the SAF-IF step-over basin. Figures 6A, 7, and 8 illustrate the stratal divergence and differential subsidence across the hinge zone that is inferred to be bounded to the south by a northwest-dipping fault system (Brothers et al., 2009). Deformation across faults within the hinge zone is almost exclusively down-to-the-south-east, yet the nature of the offset and the character of the stratigraphy vary along strike. Near the western shoreline, deformation along faults H1–H6 is expressed as down-to-the-southeast, monocline growth folds and subtle syncline folds (Figs. 7B and 8). Lake Cahuilla (LC) strata do not contain strong evidence for punctuated displacement (onlap and increased throw) across these faults, which limits our ability to constrain the timing and magnitude of displacement events. Instead, the hanging-wall accommodation appears to be gradually infilled over multiple depositional episodes. The apparent timing and magnitude of vertical displacement varies from profile to profile (e.g., Figs. 6, 7, and 8).
Each fault is associated with a break in seabed gradient along at least one CHIRP profile (Fig. 8); however, the up-section extent of vertically deformed strata is highly variable, and we cannot determine whether this is due to along-strike variation in vertical offset, minor components of strike-slip displacement, or the way the LC deposits infill coseismic accommodation. Each of the faults in the southwestern hinge (H1–H6) has roughly equal cumulative offset across the base of unit 9S/8F of ~1.5–2 m. Assuming layers were originally deposited horizontally, the net subsidence since ca. A.D. 940 (top of unit 6S/5F) across faults H1–H5 is ~7.5 m (Fig. 7B), yielding a subsidence rate of ~7 mm/yr. Strata continue to diverge southward of fault H5 (Figs. 7 and 8), but acoustic penetration was limited by gas.
Closer to Bombay Beach, the wavelength of hinge-zone roll-over remains relatively constant (Figs. 7 and 8), though subsidence appears to be more concentrated along discrete faults showing larger amounts of cumulative throw (e.g., fault H7 displays ~5 m offset across horizon 9S/8F). The structural relationship between faults H1–H6 and faults H7–H9 is not entirely clear. Faults H7–H9 are closer to the basin depocenter and proximal to the projected trace of the SAF; they also have sharply defined vertical offset and display subtle seafloor scarps (Figs. 7–10). To assess the timing of fault displacement, we apply two operations to characteristic seismic profiles. First, we generate throw-plots to quantify fault growth across individual stratigraphic horizons. Faults H7, H8 and H9 each display evidence for punctuated growth and onlap, which are prerequisites for accurate reconstruction of the displacement history. Correlations between the footwall and hanging-wall strata provided >60 vertical piercing points across each fault (Fig. 11). Second, we visually corroborate growth events in the throw plots with variations in stratal geometry by digitally flattening seismic horizons immediately up- and down-section of event horizons (Fig. 12; see also Brothers et al., 2011, their figure 3).
We identify between 8 and 15 distinct growth events across faults H7, H8, and H9 (Fig. 11). Digitally flattened seismic horizons illustrate how stratal geometry can be used to corroborate events identified in throw plots (e.g., Fig. 12). Strata above event horizons onlap the fault and appear to infill coseismically generated accommodation, whereas the pre-event strata show little to no thickness variability across the fault. Fault displacement events are assigned letters, and events that may have been coincident across multiple faults are assigned the same letter (Fig. 13). Growth appears to have occurred six separate times (events b, e, g, h, i, and l) across each of the three faults and within the same depositional units, which suggests that (at times) displacement was roughly coincident amongst the three faults (Fig. 13; Brothers et al., 2011). Further-more, faults H7, H8, and H9 each offset the top of layer 1L and layer 1S/SSF onlaps the footwall of these faults (Figs. 7–11), implying that displacement occurred within the past 300 yr. Fault H7, the most comprehensively mapped fault in the Salton Sea, shows evidence for displacement since the Salton Sea flooding occurred between A.D. 1905 and 1907 (Figs. 7–9 and 11). The total throw measured over the top of layer 14L for faults H7, H8, and H9 is ~7.1 m, 1.7 m, and 4.2 m, respectively (Figs. 9–11).
The average vertical displacement per event for faults H7 and H8 is 0.7 ± 0.2 m and 0.4 ± 0.2 m, respectively; maximum displacement per event is 1.4 ± 0.2 m and 0.6 ± 0.2 m, respectively. Displacement estimates are misleading for fault H9, which shows slip inversion in layers below 6S/F (Figs. 10 and 11C). For example, the average displacement per event for the six most recent events is 0.5 ± 0.3 m and for the three oldest events, −0.3 ± 0.2 m. The maximum offset per event measured on fault H9 is 1.0 m. At least one event along faults in the Salton Sea has occurred every 100 yr since ca. A.D. 840. The minimum average vertical slip rates since ca. A.D. 840 on faults H7, H8, and H9 are 3.4, 1.1, and 3.4 mm/yr, respectively, although these rates may change slightly with updated age control on LC sediment sequences (e.g., Rockwell et al., 2022). Due to the limited acoustic penetration throughout much of the basin, we did not attempt to quantify along-strike variation in fault offset and slip rates.
Niland Fault and Southern Basin Faults
The Niland fault, located ~5 km to the southeast of the hinge zone, strikes approximately ~N10°E and is characterized by two to three parallel splays (Fig. 2C). Each splay exhibits down-to-the-southeast vertical offset of strata emplaced above layer 2S/F; deeper fault structure is obscured by gas (Figs. 10 and 14). One of the splays produced a seafloor scarp during the most recent displacement event (Fig. 14). The scarp morphology is sharp with little to no sediment drape. The average scarp height (measured on 15 profiles) is 0.4 ± 0.1 m. Displacement during the penultimate event is observed across the top of layer 1L or bottom of layer 1S/SSF. Parallel splays are associated with growth folding during the same time intervals (e.g., Fig. 14). A few additional unnamed faults to the south of fault H9 (Figs. 2 and 10) appear to have had multiple displacement events since layer 1L was emplaced; their spatial extents and displacement histories remain unknown due to shallow gas wipeout.
Northern Basin Faults
Only a few faults are observed in the LC section to the north of the hinge zone. Numerous folds and faults, however, are observed in older strata beneath a regional unconformity surface (Figs. 3 and 6; Brothers et al., 2009). The absolute age of the deformed strata is unknown, but it is inferred to be part of the late Pleistocene Brawley Formation mapped onshore in the San Felipe Hills and at Durmid Hill (Fig. 2C; Bürgmann, 1991; Dibblee, 1984). Dipping beds within the Brawley Formation can be traced throughout most of the northern basin (Figs. 3, 6, and 15). The Durmid Hill fault (Figs. 2 and 3) is associated with significant deformation in the Brawley Formation but only subtle synclinal folding in the LC section. The Durmid Hill fault trends subparallel to the SAF for at least 16 km. The timing of displacement is difficult to constrain, but the displacement appears to extend into unit 6L deposits and possibly younger. Bürgmann (1991) and Sahakian et al. (2016) proposed that the existence of a fault along the eastern shoreline of the Salton Sea that trendure parallel to the SAF could potentially explain deformation patterns observed at Durmid Hill; the Durmid Hill fault may be associated with this structure.
Two north-trending fault splays, termed the Desert Shores fault, are observed along the north-western shoreline (Figs. 2C and 15) for only ~2.5 km (horizontal distance) before being obscured by gas. The fault zone trends parallel to topographic contours including the Salton Sea shoreline, but synclinal folding is most prominent in the Brawley Formation. Two episodes of subtle fold growth are observed in the LC section, suggesting the Desert Shores fault has produced at least two events in the last ~3 k.y. (Fig. 15).
Lake Cahuilla Stratigraphy
The seismic CHIRP profiles capture the interplay between tectonically generated accommodation in the southern Salton Sea and sediment supplied by the Colorado River. The tectonic model proposed by Brothers et al. (2009) predicts the Lake Cahuilla (LC) deposition patterns observed in the Salton Sea. The regions north and south of the hinge zone record different depositional processes due to variation in the basin physiography and proximity to Colorado River inflow during floods. The region north of the hinge is mostly devoid of thick, highly reflective packages that are observed in the southern basin (Figs. 5–7). The thin high-amplitude “S/F” packages in the northern basin are most likely sourced locally by arroyos. Given the extreme aridity of the Salton Trough, ephemeral streams that drain local catchments deliver coarse-grained sediment to the northern basin during episodic storms, where the deposition that occurs is largely governed by lake level at the time. When LC fills, fine-grained, distally sourced Colorado River sediment infills the northern basin because the hinge acts as a natural dam blocking the coarse, bed load–transported Colorado River sediment from reaching the northern basin. In contrast, the rapidly subsiding southern basin appears to be filled with coarser-grained Colorado River fan-delta deposits based on the stratal divergence and onlap patterns (e.g., Fig. 7A). Many of the layers deposited during Colorado River floods onlap the hinge zone and back-step toward the southern shoreline presumably as transgression commences.
Avulsion processes likely control migration of the Colorado River away from LC. Therefore, even during highstand conditions the Colorado River may continue to deliver sediment to the Salton Trough. Desiccation observed onshore at higher elevations does not always represent subaerial exposure at lower elevations nor does it necessarily mean that a component of Colorado River flow did not maintain a smaller lake body. Desiccation intervals observed near the LC high shoreline (e.g., Coachella site at 3 m below the high shoreline) may simply record brief shoreline regressions. For example, high-amplitude layers 2S/1F and 3S/2F appear to have been deposited when the LC shoreline temporarily dropped, resulting in basinward shift of coarse-grained material, but the primary depocenters of the basin remained subaqueous. Layer 2S/1F thickens southward of the hinge zone, implying it formed when the Colorado River flow returned and deposited a thick fan-delta sequence. Subunits within layers 2S/1F and 4S/3F fluctuate between high and low amplitude (Figs. 6–10). These fluctuations are interpreted as localized parase-quence sets formed during lobe shifting that results in along-shore variability between distal (low-amplitude) and proximal (high-amplitude) facies.
Fault Displacement History and Lake Cahuilla Inundation
Recurrence intervals for faults in the southern Salton Sea are relatively short (Table 1; Figs. 11 and 13) and are comparable to recurrence estimates of 150–220 yr on the southern San Andreas fault (SAF) (Philibosian et al., 2011). Faults H7, H8, and H9 and the Niland fault each had events between ca. A.D. 1720 and 1907 that may have occurred around the same time as a number of paleoearthquakes on faults in and around the Salton Sea (Hudnut et al., 1989c; Hudnut and Sieh, 1989; Meltzner et al., 2006; Philibosian, 2007; Williams et al., 1990). Based on the youthful nature of the seafloor scarp across the Niland fault (event a) and the subtle seafloor offset associated with fault H7, we propose these faults ruptured during the last ~40–60 yr, though we cannot determine whether it was during an earthquake or due to triggered slip. Correlating the onshore events on the SAF to offshore events along the hinge zone is tentative, and only a few events appear to be contemporaneous (Fig. 13; Brothers et al., 2011). Brothers et al. (2011) performed a detailed investigation of potential correlations, or synchronicity, between events offshore and events at the Coachella site (Fig. 13). Three of the five roughly coincident displacement events observed on faults H7, H8, and H9 (events e, h, and i) potentially correlate with the SAF events.
More paleoseismic data and more precise age control on LC sediment facies are required at different elevations in the basin in order to better assess the relative timing of events (e.g., Rockwell et al., 2018, 2022). Nevertheless, there is significant potential for stress interaction between these fault systems (Brothers et al., 2011; Dorsett et al., 2019), and stress triggering amongst nearby faults occurred in A.D. 1987 (Hudnut et al., 1989b). Coulomb models for normal displacement on fault H7 suggest that static shear stress along the SAF near Bombay Beach would increase by >1 MPa (Brothers et al., 2011). Despite uncertainties in the correlation between onshore events and offshore events, the displacement reconstructions offer robust correlation between events on the hinge-zone faults and Colorado River floods (Figs. 12 and 13; Brothers et al., 2011). The Colorado River appears to also play an important role in modulating the earth-quake cycle beneath the Salton Sea. Rapid loading by water and sediment may induce lithospheric flexure and changes in pore fluid pressure, which in turn would be expected to promote rupture of normal faults in the Salton Sea. Some displacement events on the hinge-zone faults H7–H9 (events b, g, i, and l) cluster around times when the Colorado River flood deposits are relatively thick, suggesting the river was spilling into a mostly dry basin and causing a rapid transition from a subaerial to a lacustrine environment. The correlation suggests a common process causes these faults to rupture in concert or release strain through clustered seismicity during flooding. Stress models suggest that normal faults are impacted more by the flooding of the basin than major strike-slip faults in the region. Flood-induced rupture of normal faults may provide the missing link between prehistoric ruptures of the SAF and major hydroclimatic events in the Salton Trough.
Regional Tectonic Significance
In our preferred tectonic model (Brothers et al., 2009), the hinge-zone faults are accommodating, in large part, transtensional deformation along the northern boundary of the Salton Sea pullapart basin. Displacement timing in the western hinge zone, which is closest to the set of onshore sinistral faults that were previously projected to extend across the Salton Sea (e.g., Hudnut et al., 1989a, 1989b, 1989c), is difficult to constrain due to the subtle nature of deformation and absence of definitive event horizons (Figs. 7 and 8). With the exception of fault H9, which appears to accommodate oblique slip as it nears the SAF (Fig. 10), the late Holocene–present deformation across hingezone faults appears to be predominantly dip slip in nature and represents the northern boundary of the San Andreas fault–Imperial fault (SAF-IF) pull-apart basin. Deeper structural geometry observed in seismic reflection imaging presented by Kell (2014) support this interpretation. These faults trend ~50° oblique to the SAF and show consistent down-to-the-southeast vertical displacement despite minor along-strike jogs and bends. The vertical offset per event and vertical slip rates are relatively large and thus accommodate a significant percentage of the strain transferred from the SAF to the IF (Brothers et al., 2009).
Assuming LC strata were originally deposited horizontally, it becomes difficult to attribute the observed structural relief across the hinge zone to a strike-slip fault regime. For example, for an array of strike-slip faults to generate >7 m of vertical mismatch of late Holocene LC strata would require an unrealistically large amount of horizontal offset. Nevertheless, synclinal folding does suggest there is some component of strike slip along these faults. In Figure 10, for example, fault H9 appears to have a component of strike slip in the region imaged by CHIRP profiles. Slip inversion and fold growth on fault H9 is indicative of either a change in kinematics through time or gradual fold growth due to oblique displacement. The confluence of faults within the pull-apart basin and the SAF is expected to produce complicated deformation as strain is transferred between fault systems. The proximity of fault H9 to the projected trace of the SAF suggests it may be a major splay of the SAF that is accommodating significant oblique slip. Constraining the relative amounts of strike-slip versus dip-slip offset on faults in the Salton Sea requires further studies.
The structural and kinematic relationships between offshore faults and onshore faults surrounding the Salton Sea remain poorly understood. Distinct seafloor scarps along the Niland fault zone suggest recent (post–A.D. 1907) displacement. These faults trend ~N10°E and have a consistent down-to-the-southeast sense of slip. The A.D. 1987 Elmore Ranch fault event recorded ~0.15 m of sinistral offset along an ~N40°E plane (Hudnut et al., 1989a; Sipkin, 1989). As the Elmore Ranch fault enters the SAF-IF stepover its azimuth may adjust to accommodate extension, or these faults may be entirely separate structures. Further mapping is required to understand the relationship of the southern basin faults to the Elmore Ranch fault system. Aseismic creep and triggered slip have been observed on faults surrounding the Salton Sea, including the Extra fault (Hudnut et al., 1989a; Lohman and McGuire, 2007; Lyons and Sandwell, 2003; Lyons et al., 2002; Meltzner et al., 2006). Some of the seafloor offset observed, particularly along faults in the western hinge zone, may be from triggered slip following large events in the region. Although evidence for fault displacement in the Salton Sea during discrete events is definitive, we cannot discount sudden aseismic slip as a possible displacement mechanism (Hauksson et al., 2017), and similar events have been observed in other active rift basins (Doubre and Peltzer, 2007).
Since ca. A.D. 940 (top of layer 6S/5F), the subsidence rate across the hinge-zone faults has ranged from 6 to 9 mm/yr (Figs. 6–8); 9.5 m of total subsidence is observed between the hinge zone and the Niland fault since ca. A.D. 1380 (top of layer 4S/3F; Fig. 7A), which yields a regional sub-sidence rate of ~15 mm/yr, which is slightly higher than the 9–12 mm/yr estimated by Brothers et al. (2009). Southward divergence and rotated beds continue to the south of the Niland fault, meaning subsidence rates may have been as high as 20 mm/yr during the last several hundred years. Assuming extension and subsidence is bound by a northwest-dipping boundary fault system near the southern shoreline (Brothers et al., 2009), the conservative regional subsidence estimate of 9–15 mm/yr can be converted to 5–9 mm/yr of horizontal extension (assuming a 60°NW-dipping fault plane). Similarly, the estimated subsidence rate across only hinge-zone faults suggests they collectively carry 3–5 mm/yr of horizontal extension. In rift basins, between 20% and 50% of the total extension is typically distributed across secondary faulting in the hinge zone (Morley, 1995), implying that the majority of the estimated 23 mm/yr SAF slip budget (Meade and Hager, 2005) is accommodated by extension and subsidence along secondary faults in the SAF-IF stepover. Despite inferences for profuse strike-slip faulting and block rotation in the Salton Sea (Nicholson et al., 1986), we propose that over recent geologic time scales (last few thousand years), only a minor component of strike-slip faulting has occurred within the SAF-IF stepover basin and the majority of deformation has been accommodated by normal faulting.
The regional significance of faults discovered to the north of the hinge zone is poorly understood and needs to be studied further. The Desert Shores fault has experienced displacement during the last 3 k.y. Syncline folds and distinct down-to-the-southeast vertical displacement implies this fault is normaloblique. Perhaps there is an active fault system that trends along the western edge of the valley and connects to the SAF north of the Salton Sea. As mentioned, the Durmid Hill fault (Figs. 2C and 3) is subparallel to the SAF and may be linked to the East Shoreline fault inferred to be located in shallow water and parallel to the shoreline north of Bombay Beach, thus helping to explain transpressional folding in Durmid Hill (Bürgmann, 1991; Sahakian et al., 2016). However, its slip rate appears to be relatively low based on the minor amounts of offset and folding observed in the LC section. The presence of such structures suggests that strain is distributed across a broad region, which may help reconcile differences between geologic and geodetic based slip-rate estimates for the SAF.
This study represents the most comprehensive investigation of late Holocene tectonic activity and Lake Cahuilla (LC) sedimentation history in the Salton Sea. Colorado River diversions into the Salton Trough result in rapid emplacement of fandelta facies that onlap the tectonic hinge zone and infill tectonically generated accommodation in the southern subbasin.
Correlations between high-amplitude reflections in CHIRP profiles and coarse-grained sediment layers in cone penetrometer test soundings allowed us to establish a facies model for the LC depositional history in the Salton Sea. At least 14 LC sedimentary sequences are observed in the Salton Sea. The last six LC sequences have been radiocarbon dated at the Coachella paleoseismic site (Philibosian et al., 2011), providing chronostratigraphic constraints back to ca. A.D. 840. Punctuated growth intervals suggest that faults in the southern Salton Sea accommodate primarily extensional deformation associated with the San Andreas fault–Imperial fault (SAF-IF) stepover and rupture every 100–200 yr, and that vertical slip rates along individual faults can exceed 3 mm/yr. The comprehensive analysis presented in this study supports results presented by Brothers et al. (2011), confirming that at least four displacement events on faults H7, H8, and H9 appear coincident with Colorado River floods into the basin. Of the five events that were roughly coincident on all three of these offshore faults, three potentially correlate with paleoearthquakes on the SAF. Given the proximity of these hinge-zone faults to the SAF, static and/or dynamic stress changes following rupture of offshore faults have the potential to promote a northward-propagating rupture of the SAF. However, the role of secondary faults within the SAF-IF stepover as stress modulators for any of the major strike-slip faults of the region needs to be examined further. Future studies aimed at direct sampling and improving the age control on LC sediments in and around the Salton Sea (e.g., Rockwell et al., 2022) and a reassessment of the correlations between onshore and offshore stratigraphy are necessary to refine the displacement history of subaqueous faults, particularly prior to emplacement of layer 6L. Deeper-penetrating seismic reflection data will provide important constraints on the fault geometry at depth and the longer-term evolution of these faults and how they interact.
We thank Janet Watt, Harvey Kelsey, and Chris Sorlien for constructive reviews of this manuscript. This project was funded by the State of California Department of Water Resources, Scripps Institution of Oceanography, University of California Academic Senate, U.S. National Science Foundation (grants OCE-0112058 and EAR-0545250), Southern California Earthquake Center (grant 2008-08127), and U.S. Geological Survey. We would like to thank the Sonny Bono Salton Sea National Wild-life Refuge and the Salton Sea State Recreation Area for their hospitality and providing access to their facilities during data collection. We would also like to thank E. Aaron, L. Johnstone, J. McCullaugh, J. Dingler, C. Takeuchi, and numerous others for field assistance. Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the U.S. Government.