Paleoseismic patterns of Quaternary tectonic and magmatic surface deformation in the eastern Basin and Range, USA

Continental extensional provinces, or rifts, are commonly partitioned into segments in which deformation is dominated by either magma-assisted or tectonic extension (e.g., Buck, 1991; Hayward and Ebinger, 1996; Scholz and Contreras, 1998; Wright et al., 2006; Rowland et al., 2010; Muirhead et al., 2016). The underlying reasons for rift segmentation comprise an interplay of modern strain rate, total amount of cumulative strain, presence of inherited structures, thermal and mechanical layering of the lithosphere, and melt availability (e.g., Buck, 1991; Brun, 1999; Corti et al., 2007; Nestola et al., 2015). Depending on these factors, the mode of deformation within segments can vary with time; for instance, a segment dominated by crustal or lithospheric extension on normal faults can become dominated by magma-assisted rifting as melt production and availability evolve (e.g., Bursik and Sieh, 1989; Parsons and Thompson, 1991; Ebinger and Casey, 2001; Muirhead et al., 2015, 2016). Similarly, spatial variations in magma supply and lithospheric structure can lead to along-strike segmentation of rifts, with adjacent segments dominated by either tectonic or magma-assisted extension (e.g., Hayward and Ebinger, 1996). Extension can also be partitioned between these processes within a single basin (e.g., Bilham et al., 1999; Ebinger and Casey, 2001; Ebinger, 2005; Keir et al., 2006; Athens et al., 2016; Muirhead et al., 2016), with border faults typically accommodating tectonic extension and intrabasin faults accommodating both tectonic faulting and extension above shallow intrusions (Rowland et al., 2007; Ibs-von Seht et al., 2001; Calais et al., 2008; Athens et al., 2016). In all cases, the characteristics of active faults at the surface provide clues that assist in evaluating modes of extension.

In magma-assisted rifting, extension in the upper crust is partly accommodated by dike injection and secondary faulting around the intrusion (e.g., Rubin and Pollard, 1988; Bursik and Sieh, 1989; Wright et al., 2006; Rowland et al., 2007; Villamor et al., 2011; Athens et al., 2016; Gómez-Vasconcelos et al., 2017). The thinner effective elastic thickness of the crust in magma-assisted rifts means that seismic strain release is localized on shallowly rooted (<5 km) faults associated with propagating dikes. These faults do not usually accumulate elastic strain through a seismic cycle and do not typically produce M_w ≥6 earthquakes (Parsons and Thompson, 1991; Smith et al., 1996; Rowland et al., 2007). In tectonic rift segments, or during episodes of limited melt supply, upper-crustal extension can be accommodated entirely by tectonic faults that span the seismogenic portion of the crust (Rowland et al., 2010; Medynski et al., 2016). Surface rupturing earthquakes of M_w 7 or greater occur regularly on tectonic normal faults, despite evidence for temporal variability in the length of the seismic cycle (e.g., Friedrich et al., 2003; Gómez-Vasconcelos and Wernicke, 2017). Given the disparity in maximum magnitudes on dike-induced versus tectonic faults, distinguishing between the two is important for the purposes of characterizing seismic hazard, especially if they are both present within the same rift segment (Smith et al., 1996; Villamor et al., 2007; Rowland et al., 2010; Villamor et al., 2011; Gómez-Vasconcelos et al., 2017).

Tectonic and magma-assisted rifting can be distinguished based on the geometry, recurrence, and expression of active faults at the surface (Smith et al., 1996; Rowland et al., 2007; Payne et al., 2009; Villamor et al., 2011; Gómez-Vasconcelos et al., 2017; Stahl and Niemi, 2017). If magma-assisted rifting is accommodated by dike intrusion, faults exhibit a spatial and temporal coincidence of faulting with volcanism, and are intermixed with mode I extensional fissures (cf. Smith et al., 1996; Payne et al., 2009). Compared to regions dominated by tectonic faulting, there may also be discrepancies between geodetic and geologic extension estimates if some strain is accommodated aseismically (e.g., Smith et al., 2004; Wright et al., 2006). In contrast, the surface expressions of tectonic faults have displacement, length, and recurrence characteristics that roughly follow empirical scaling laws developed from catalogues of historical surface-rupturing earthquakes (e.g., Wells and Coppersmith, 1994; Wesnousky, 2008), and geological extension or moment estimates may be expected to more closely align with geodetic and seismological estimates.

The Basin and Range of the western United States is one of the characteristic examples of a wide continental rift (terminology of Buck, 1991) (Fig. 1). The mode of extension has not been uniform through time or across this nearly 1000-km-wide intraplate extensional province (Axen et al., 1993; McQuarrie and Wernicke, 2005; McQuarrie and Oskin, 2010). Extension is currently dominated by tectonic faulting on range-bounding normal faults that accommodate far-field, plate boundary–driven extension (e.g., Personius et al., 2017), with some evidence at the western edge of the Basin and Range for crustal extension being accommodated via intrusions in the crust (e.g., Bursik and Sieh, 1989; Smith et al., 2004; Athens et al., 2016).

The highest modern strain rates in the Basin and Range are concentrated near its margins (Bennett et al., 2003; Hammond and Thatcher, 2004; Zeng and Shen, 2017). At the eastern margin of the Basin and Range in central Utah, ∼4 mm yr⁻¹ of east-west extension is accommodated between the western edge of the stable Colorado Plateau and the Nevada border (Fig. 1). North of ∼39°N, most of this extension is accommodated on the Wasatch fault and associated structures (Friedrich et al., 2003). The Wasatch fault, however, does not continue as a discrete structure south of ∼39°N. Instead, a diffuse north-south–trending boundary separates extended and non-extended domains to the south, highlighted by resistivity and tomographic surveys that show the contrasting lithospheric structure between the eastern Basin and Range, Colorado Plateau, and an intervening “transition zone” (Fig. 1) (Wannamaker et al., 2008; Schmandt and Lin, 2014; Liu and Hasterok, 2016; Long, 2018).

Near the transition zone in the easternmost Basin and Range, narrow zones of low resistivity within ∼30-km-thick continental crust of the eastern Sevier Desert (Fig. 1) coincide with regions of high heat flow, Pleistocene–Holocene volcanism, and Pleistocene–Holocene faulting (Wannamaker et al., 2008, 2013; Liu and Hasterok, 2016; Stahl and Niemi, 2017; Long, 2018). A thin lithospheric mantle (∼25 km) and a prominent low-velocity zone in the underlying asthenospheric mantle at ∼65 km depth (York and Helmberger, 1973; Levander and Miller, 2012; Schmandt and Lin 2014) are characteristic of other volcanic regions of the Basin and Range and indicate favorable conditions for a continued, albeit slow, supply of melt to the upper crust in this region (e.g., Valentine and Perry, 2007).

Magmatic contributions to finite extension have not been fully quantified in the Sevier Desert, and the spatial distribution of extension and the mechanisms by which it is accommodated, particularly in the eastern Sevier Desert, are still debated (Hammond and Thatcher, 2004; Niemi et al., 2004; WGUEP, 2016; Stahl and Niemi, 2017; Yuan et al., 2018) (Fig. 1). Competing hypotheses for strain accommodation across the Sevier Desert include slip on a buried low-angle normal fault, the Sevier Desert detachment (e.g., Niemi et al., 2004; Yuan et al., 2018); salt tectonics and resultant surface faulting (e.g., Wills et al., 2005); and magma-assisted rifting via episodic dike injection (Stahl and Niemi, 2017). Quantifying the relative contributions of magmatic and tectonic processes to extension in the Sevier Desert can help resolve this debate, and the wealth of existing data make the Sevier Desert an important natural laboratory for resolving competing models of strain accommodation.

Following from a 2017 study (Stahl and Niemi, 2017), we hypothesize that late Quaternary extension in central Utah (Fig. 1) is partitioned into magma-assisted and tectonic segments. To test this hypothesis, we present new data on the Late Pleistocene to Holocene activity of active faults in a transect across the Sevier Desert (Fig. 1). We begin with a review of the late Quaternary geologic history of the Sevier Desert, including previous work that constrains the ages of Lake Bonneville highstands and volcanic landforms, which are essential geomorphic markers in this region, and an overview of new geochronology data from this study that bear on the ages of these features. The results from our fault studies across the Sevier Desert, including mapping, surveying, and geochronology, are then presented by region. Time-integrated slip and extension rates are calculated for the major range-bounding and intrabasin faults of the western and eastern Sevier Desert and compared to GPS extension rates. The timing of most-recent events (MREs) are estimated for select faults from ¹⁰Be and ³He exposure age dating of alluvial fans and tension fissures in basalt, respectively. We conclude with a new model of active tectonics in the Sevier Desert in which deformation is partitioned into domains that are currently dominated by either tectonic or magma-assisted extension.

The evolution of Lake Bonneville underpins the Late Pleistocene geomorphology of the Sevier Desert. Exhaustive reviews by Oviatt (2015) and Oviatt and Shroder (2016, and references therein) provide summaries of the >100 years of research on Lake Bonneville. Here, we provide an overview of the information that is relevant for assessing fault activity in the Sevier Desert area.

Lake Bonneville was a Late Pleistocene pluvial lake that covered >50,000 km² of western Utah at its maximum extent. While Lake Bonneville was only the most recent major lake cycle in the basin over the last 2 m.y., no evidence exists of an equivalent lake for tens of thousands of years prior to its rise (Thompson et al., 2016). There were three major phases of Lake Bonneville: (1) transgressive, (2) overflowing, and (3) regressive (Currey, 1990; Oviatt, 2015) (Fig. 2). The transgressive phase (phase 1) began during cooler and wetter climatic conditions in Utah at ca. 30 ka. During this time, all water entering the basin exited via evaporation only. Lake level rose to the elevation of the Bonneville shoreline (Fig. 1), which was maintained by the elevation of an alluvial fan dam near Red Rock Pass, Idaho. Eventual breaching of the alluvial fan caused catastrophic flooding that took place rapidly, with lake level lowering ∼120 m in an estimated ∼500 yr (O’Connor, 1993; Janecke and Oaks, 2011; Miller et al., 2013). Following lake-level lowering, there was sustained outflow of Lake Bonneville through the Snake River Plain for ∼3 k.y. during the overflowing stage (phase 2) (Oviatt, 2015). During this stage, the prominent Provo shoreline formed around the Great Salt Lake and Sevier sub-basins of Lake Bonneville (Figs. 1 and 2).

Lake level fell during the regressive phase (phase 3) after overflow ceased and the basin once again became hydrographically closed (Fig. 2). During this stage, Lake Bonneville segregated, and Lake Gunnison formed as a hydrologically separate lake in the present Sevier Desert. Overflow from Lake Gunnison reached the lower Great Salt Lake Basin to the north until Lake Gunnison dropped below the Old River Bed threshold and the two basins became hydrologically disconnected (Fig. 1). Continued regression of Lake Gunnison likely occurred less rapidly than in the Great Salt Lake Basin, as inflow to Lake Gunnison from the Sevier and Beaver Rivers moderated losses due to evaporation from ca. 13 to 11.5 ka (Hintze and Davis, 2003) (Fig. 2). During this time, Lake Tule (in the Tule Valley adjacent to the House Range; Fig. 1) became isolated from the rest of Lake Gunnison and subsequently regressed rapidly due to limited inflow (Sack, 1990; Hintze and Davis, 2003).

The most important aspect of Lake Bonneville chronology for this study is the age of the Provo shoreline. The Provo shoreline is well preserved in many locations across the Sevier Desert, represented by erosion into, or deposition on, lava flows and eruptive vents, alluvial fan surfaces, and bedrock promontories. The shoreline formed over an ∼3 k.y. interval from ca. 15 to 18 ka (Godsey et al., 2011; Oviatt, 2015; Miller, 2016). Miller et al. (2013) documented two Provo-aged shorelines at localities throughout the Bonneville basin. Subsequent work refined the age of the Provo shoreline to forming between 14.8 and 18.2 ka (Miller, 2016). We adopt a simplified age range of 15–18 ka for the Provo shoreline in this paper.

Bimodal volcanism in the Sevier Desert began in the late Miocene, but has been most active since ca. 2.5 Ma (Johnsen et al., 2010). Here, we focus our overview on the geochronology of the Black Rock Desert volcanic field (e.g., Hoover, 1974; Oviatt, 1991) in the eastern Sevier Desert (Figs. 1 and 3). Most volcanic units mapped in the eastern Sevier Desert are basalt and basaltic andesite lava flows (Johnsen et al., 2010). In some cases, dacites and rhyolites are exposed at the surface in close proximity to larger basalt flows. There is no clear spatial or temporal progression in composition of the volcanic units, although the youngest (Late Pleistocene to Holocene) ashes and flows are basaltic. A more detailed review of previous work, including ages, petrology, and chemistry of volcanic rocks in central Utah, is provided by Johnsen et al. (2010).

The youngest volcanic unit in the Sevier Desert is the Ice Springs basalt (Fig. 3). The Ice Springs basalt is commonly reported as being 660 ± 170 ¹⁴C yr B.P. or younger (Valastro et al. 1972; Oviatt, 1991; McBride et al., 2015). This age, however, is not considered robust, as it is based on dating a single root fragment within a soil on the edge of one of the four flow lobes (Valastro et al., 1972; Hoover, 1974; Patzkowsky et al., 2017). Recent exposure age and varnish microlamination dating suggests that the Ice Springs flows could be significantly older—between 9 and 12 ka (Schantz, 2016; Patzkowsky et al., 2017) (Fig. 2).

The Tabernacle Hill flow is a basalt flow ∼1 km south of the southern edge of the Ice Springs lava flow (Fig. 3). Original dating of the flow was conducted using tufa that was encrusted onto the outer margin of the flow (Oviatt and Nash, 1989). This sample yielded a radiocarbon age of 14.3 ± 0.9 ka (ca. 17.3 ± 0.3 cal. kyr B.P). Based on this minimum age for the flow, and on the presence of lava pillows and shoreline elevations around the flow, Oviatt (1991) interpreted the Tabernacle Hill flow as having erupted during the Provo highstand of Lake Bonneville. Subsequent sampling and dating of this tufa with procedures to avoid atmospheric contamination and using accelerator mass spectrometry (AMS) revealed a slightly older age of 18.2 ± 0.3 cal. kyr B.P. (Fig. 2), uncorrected for a near-negligible (∼200 yr) reservoir effect in Lake Bonneville at this stage (McGee et al., 2012; Lifton et al., 2015; Oviatt, 2015).

The largest flows (by exposed area) are those of the Pavant field (Fig. 3). Two sub-flows within this field were mapped by Hoover (1974) (Fig. 3). Reported ages for both basalt flows range from 30 ± 69 ka to 220 ± 260 ka based on K-Ar dating (Condie and Barsky, 1972; Hoover, 1974; Best et al., 1980; Johnsen et al., 2010) (Fig. 3). The Pavant I and II flows are spatially, but not temporally, coincident with a volcanic edifice called Pavant Butte (Fig. 2). Pavant Butte was constructed later than the surrounding flows and was the source of the widespread Pavant ash that can be found throughout the eastern Sevier Desert (Oviatt and Nash, 1989). The Pavant ash is relatively well constrained as having erupted into the late transgressive phase of Lake Bonneville at 15.3–16 ka (Oviatt and Nash, 1989) (ca. 19 cal. kyr B.P., using IntCal13 curve and CALIB 7.10 sotfware; Stuiver et al., 2019).

Among the oldest volcanic units in the study area (Fig. 3) is the tholeiitic basalt of the Deseret volcano (Johnsen et al., 2010). This flow has previously been dated by Best et al. (1980), yielding a K-Ar age of 400 ± 400 ka. Despite the large age uncertainty, the Deseret volcano is known to be pre–Lake Bonneville in age based on shorelines on the edge of the flow (Oviatt, 1989) (Figs. 1 and 2).

Several other volcanic vents and small flows are known in the eastern Sevier Desert, but are not radiometrically dated. In addition to the volcanic rocks exposed at the surface, extensive subsurface basalt flows, some as old as 6.9 Ma, have been interpreted in seismic lines across the eastern Sevier Desert and correlated to adjacent well data (Lindsey et al., 1981; Planke and Smith, 1991). There are no known Pliocene to recent igneous rocks west of the Cricket Mountains in the western Sevier Desert (Fig. 1).

Our goal in data collection and analysis is to measure cumulative displacements across faults, fissures, and fold scarps in the Sevier Desert, determine single-event displacements (SEDs), and, where possible, estimate fault slip rates and timing of most-recent events (MREs). Because fault-related deformation in the Sevier Desert region transects a range of geologic materials and time scales, such an analysis requires the application of a variety of survey and geochronologic approaches. Because our goal is to extract fault displacement information from both of these types of information, we first summarize the geochronologic methods we have used and the results we have obtained. We then summarize the methods we employed for obtaining high-resolution topography across the faults, and the assumptions and methods used to derive fault displacement information from these profiles. The introduction of all of these methods is followed by the integration of fault slip and age data pertinent to specific study areas across the Sevier Desert.

The methods and results of all of the age data collected in this study are presented in Table 1. In three locations, we used ⁴⁰Ar/³⁹Ar to date lava flows (e.g., the Deseret and Pavant I and II flows). For the Tabernacle fault, we used ³He exposure-age dating in olivine to establish timing of fissure openings along one major trace of the fault (e.g., Mackey et al., 2014). For the House Range fault, we used ¹⁰Be exposure-age dating to constrain the age of an unfaulted alluvial fan that postdates the MRE on the fault. Full methodological details, including preparation, analysis, and modeling, are included in the Supplemental Information¹. Further discussion of the results are provided for each of the faults separately in subsequent sections.

We conducted detailed geomorphic mapping at select locations along faults using aerial imagery and shaded relief derived from the 5 m digital elevation models (DEMs) available from the Utah Automated Geographic Reference Center (AGRC; https://gis.utah.gov/data/). At the House Range, Cricket Mountains, and Tabernacle faults, we created high-resolution digital surface models (DSMs) generated from terrestrial lidar scanning (TLS) and unmanned aerial system (UAS)–based structure-from-motion (SfM) acquisitions (details of acquisition and processing are included in the Supplemental Information [^{footnote 1}]). Topographic profiles of faulted shorelines, fan surfaces, and lava flows were generated using field survey data from real-time kinematic (RTK) Global Navigation Satellite System (GNSS), the 5 m DEM, and TLS- and SfM-derived DSM products.

Topographic profiles were used to calculate dip slip, considered to be the net slip, for each mapped fault trace. Net slip was calculated using estimates of fault dip, fault position along the scarp, and regressions fit to the hanging wall, footwall, and scarp, following the Monte Carlo simulation methodology developed by Thompson et al. (2002). From these slip estimates, we calculated cumulative extension and throw across each fault zone. In our calculations, we assumed that all motion is dip slip, as we observed no evidence of strike-slip displacement in the field. We used input parameters and uncertainties for fault parameters deemed appropriate for each fault based on our mapping, subsurface data from other studies (e.g., Crone and Harding, 1984; Greene 2014; McBride et al., 2015), and exposures (Table 2). Most near-surface geophysical investigations and our observations show that all faults are steeply dipping (70°–85°) within ∼1 km of the surface, but, where available, we prefer moderate dips with conservatively large ranges to reflect average dip at depth from seismic reflection and interpreted geologic cross sections (e.g., Allmendinger et al., 1983; Planke and Smith, 1991; Wills et al., 2005; Greene 2014). Where we have limited or no data on where a fault projects to the surface, we assume that the fault intersects near the midpoint of the scarp (Table 2). Based on unpublished trench exposures in alluvial fans, faults tend to project to the surface 50%–75% of the way up the scarp from the hanging wall in these deposits, and we use that range instead (Table 2). Monoclines were treated as fold scarps and have slip resolved on faults at depth, some of which have been identified in seismic lines and ground penetrating radar surveys (McBride et al., 2015). We measured extension across fissures where not associated with fault traces, and added these values to the cumulative extension across studied fault zones. Fissures in lava flows are not likely to span the depth of the crust, and, given no local vertical separation, likely represent the surface expression of distributed dike-induced dilation at the surface.

In places where the ages of displaced landforms are known (Table 1), we also calculated slip and extension rates using the same Monte Carlo approach (Thompson et al., 2002), with ages and uncertainties constrained by age data in Table 1. SEDs are presented wherever they are evident. Time-integrated slip rates are presented only for faults where reliable SEDs can be estimated across the entire fault zone (i.e., the House Range and Cricket Mountains faults). As these faults have long recurrence intervals relative to the age of displaced landforms, the slip rates we present depend on the time elapsed since the last event. To account for this effect, we assigned an additional slip rate uncertainty of 0.1 mm yr⁻¹ to these faults following the approach of Personius et al. (2017).

In order to investigate extension rate variability, we compared our paleoseismic-time-scale slip rate data to GPS and geologically derived extension rates. We used the horizontal component of maximum net slip calculated in this study for paleoseismic extension estimates to account for surface slip deficits (e.g., Wells and Coppersmith, 1994; Dolan and Haravitch, 2014; Personius et al., 2017). Geologic extension rates from the Clear Lake to the Pavant faults in the eastern Sevier Desert are presented based on estimates from Planke and Smith (1991), calculated by summing fault heave across a 6.4 Ma basalt flow (Wills et al., 2005) imaged in seismic sections, and assuming that the displacement is not resolved on the Sevier Desert detachment.

We calculated GPS velocity profiles across the Sevier Desert using easting velocities from Basin and Range Geodetic Network (BARGEN) and UNAVCO Plate Boundary Observatory stations in the NAM08 reference frame (available at https://www.unavco.org/data/gps-gnss/derived-products/derived-products.html). GPS velocity solutions are current as of December 2017. GPS velocity vectors are approximately perpendicular to average fault strike (∼185°–190°) across this region, so we do not take into account northing velocities. Stations and the differential velocities between them are listed in Table 3.

We focused on four fault zones in the eastern Sevier Desert that are at roughly the same latitude as existing geodetic velocities and paleoseismic studies (e.g., Niemi et al., 2004). Below, we describe our new mapping, surveying, and sampling strategies for each of these fault zones. Extension rates are presented at the end of each section. It is important to note that due to sedimentation through several Pleistocene lake cycles, Holocene dune deposition, and the limited extent of radiometrically dated lava flows exposed at the surface, it is difficult to account for all of the extension in the eastern Sevier Desert. Numerous small fissures, such as those across Clear Lake playa and at Tabernacle Hill, accommodate a portion of extensional strain but are poorly preserved due to lack of significant vertical separation and the small scale of the features. Not all extension may be expressed as brittle deformation at the surface (Wright et al., 2006)—in analog experiments and geodetic inversions of dike-induced deformation, extension accommodated by faults and fissures at the surface is usually less than dike thickness at depth (e.g., Mastin and Pollard, 1988; Hollingsworth et al., 2013). Additionally, we only consider one-dimensional extension (i.e., extension measured at the surface in one direction) in adjacent transects. In two- and three-dimensional strain, faults and fissures with <10 km length, as is typical in the eastern Sevier Desert, may only accommodate ∼20% of strain over geologic time scales (e.g., Cowie et al., 1993).

The Tabernacle faults (Fig. 3) comprise a zone of faults, monoclines, and fissures developed in basalt across the Tabernacle Hill flow (Oviatt, 1991). Some lineaments included in the U.S. Geological Survey Quaternary fault database (U.S. Geological Survey and Utah Geological Survey, 2006) are actually fissures that have a close spatial association with the central tuff cone within the flow (Stahl and Niemi, 2017). Faults have maximum vertical separations of 10 m and transition within 1–2 km along strike to tension cracks with no vertical separation. One of the fault traces is mapped across the Tabernacle Hill flow (18.2 ka) and into the older Beaver Ridge flow (<1 Ma), potentially indicating continued displacement or reactivation from mid-Pleistocene to recent time (Fig. 3). At the northern end of the Tabernacle Hill flow, McBride et al. (2015) conducted active-source P- and SH-wave surveys across a horst and interpreted it to be bounded by a set of steeply dipping (∼70°) faults. They also conducted augering and found ash units (interpreted to be from Tabernacle Hill) displaced vertically across the prominent west-facing scarp by 10.4 m (McBride et al., 2015).

We focused our paleoseismic study on the northern end of the Tabernacle Hill flow along a prominent horst that projects northward to, and is buried by, the younger Ice Springs flow (McBride et al., 2015). This location was selected because it has the clearest expression of Provo shoreline development around the flow edge (Fig. 4). The most continuous of the fault traces at this site is west facing and forms a ∼3-km-long monocline. Maximum vertical separation along the trace is ∼10 m; this value decreases to zero ∼2 km to the south, where the monocline splays into a series of fissures around the Tabernacle cone (Stahl and Niemi, 2017).

Mapping in the field and on a 1 m TLS DSM reveals two paleoshorelines preserved on the hanging wall of the west-facing fault. The lower shoreline is marked by a zone of rounded boulders (P1, Fig. 4), and the upper is a wave-smoothed platform (P2, Fig. 4). A single wave-smoothed platform with rounded, CaCO₃-encrusted boulders, and an inferred wave-cut “inner-edge” carved into a tephra ridge, are present on the footwall of the fault (Fig. 4B). These correlative features are interpreted to be Provo highstand shorelines (15–18 ka) based on their elevation and age of the Tabernacle Hill flow. The lower shoreline on the hanging wall, P1, is displaced from the footwall shoreline by 10–12 m, which matches the ∼10 m offset of basaltic ash found to the north in the subsurface (McBride et al., 2015). The upper shoreline, P2, is displaced by 5–6 m relative to the footwall shoreline (Fig. 4). Based on mapping and survey results at this site, two displacement events are interpreted on the fault trace at this location. Event 1 (E1) occurred during initial development of the Provo shoreline (P1) on the flow edge, with 5–6 m of vertical separation; E2 occurred after reoccupation of the Provo shoreline (P2) across the fault, and accumulated another 5–6 m vertical separation across the scarp.

We followed the sampling protocol and procedure of Mackey et al. (2014) for dating the timing of these faulting events using cosmogenic exposure ages to date fissure walls. The fissure selected is developed within a monocline and therefore has an opening history linked to near-surface faulting (Fig. 4B) (Mackey et al., 2014). There was no evidence of block toppling at the sites, and we sampled from locations on the fissure wall that could be visually reconstructed to the opposite fissure wall. We collected four samples from the walls of the fissure along the northern edge of the Tabernacle Hills flow: two each from different fissure depths on opposing walls where there was no evidence of block toppling, other erosion, or burial by surface deposits. Samples were collected at >2.5 m depths to avoid inheritance from the flow surface. Topographic and self-shielding was corrected for by taking field measurements of the skyline geometry within the fissure and calculating a shielding correction factor in CRONUS-Earth (Balco et al., 2008). Analytical data are listed in Tables S1 and S2 (^{footnote 1}).

Exposure ages of the fissure yield evidence of either one or two fissure-opening events caused by faulting (Fig. 5). In the single-event model, three samples (TH200, TH201, TH101) are used to define a single E1 event, and sample TH100 is considered a younger outlier (see individual ages in the Supplemental Information [^{footnote 1}]). This chronology yields an error-weighted mean age of E1 at 15.9 ± 1.8 ka (2σ). It is possible that the fissure was progressively opened in more than one event, with the top two samples having been exposed first. In this second scenario, in which all ages are accepted and sample ages from opposite walls at the same elevation are averaged, events E1 and E2 have error-weighted mean ages of 16.3 ± 1.8 and 13.6 ± 2.5 ka, respectively (Fig. 5). The ages for these two events overlap within 2σ uncertainty, but are distinguishable in a Kolmogorov-Smirnov test at the 5% significance level.

A two-event displacement history is consistent with the observed offset of the Provo shoreline at this site (Fig. 4), and we therefore favor the interpretation in which a minimum of two events have occurred on this trace of the Tabernacle fault since the flow formed at ca. 18 ka, with 5–6 m of maximum vertical separation per event. Our data do not permit discrimination of several, closely spaced smaller events that could have produced the observed ages, nor can we rule out entirely the possibility that lake-level regression from the Provo shoreline at 16 ka influenced the observed exposure ages. We do, however, consider the latter unlikely given the 3 k.y. age difference between the upper and lower samples.

We calculate cumulative horizontal extension across the 5-km-wide Tabernacle Hill flow, from a combination of all faults and fissures, to be m. Using an age of 18.2 ka for the flow (Lifton et al., 2015), we calculate a minimum extension rate of mm yr⁻¹.

Traces of the Pavant fault run from west of the Ice Springs flow to ∼3 km south of Pavant Butte (Fig. 3). The fault traces trend NE to NW, including a ∼60° bend (Fig. 3). The trace of the fault north of this bend comprises a west-facing scarp and/or flexural monocline in columnar basalt. Vertical separation generally increases northwards along the scarp from ∼2 m in the south to a maximum vertical separation of >20 m in the Pavant II lavas, then decreases again or is buried closer to Pavant Butte. There are two locations where faulting is observed off of this main fault zone. A fault trace is identified 1 km to the west of the main trace within the Pavant II lava and subparallel to the main zone, and another is identified 10–20 km to the northeast, trending northeast (Fig. 3).

Field work along the Pavant fault was focused on constraining the timing of the MRE and assessing long-term extension rates by using ⁴⁰Ar/³⁹Ar to date the offset flows (Table 1). To determine the timing of the MRE, we excavated a pit in fissure fill located within a monocline forelimb along the fault to evaluate evidence of syn- or post-Provo-aged displacement. For the assessment of deformation rates, we obtained ⁴⁰Ar/³⁹Ar ages from the Pavant I and II lavas (after Hoover, 1974; Johnsen et al., 2010) and conducted scarp surveys with RTK GPS supplemented with profiles extracted from a 5 m DEM.

The excavation within fissure fill was conducted at ∼1500 m above sea level (i.e., ∼50 m below the Bonneville and ∼50 m above the Provo shorelines; Fig. 2) within an ∼20-m-wide fissure (Fig. 6). The 1.7-m-deep pit revealed basaltic lapilli at the base overlain by in situ lacustrine marl, silt, and sands that were in turn overlain by aeolian sand (Fig. 6). The basal basaltic ash–lacustrine sequence is interpreted to represent the end of the transgressive phase of Lake Bonneville, with the lapilli likely being sourced from the Pavant ash erupted from nearby Pavant Butte (Figs. 2 and 3) (Oviatt and Nash, 1989). In this interpretation, the overlying marl and lacustrine sediments must date to the Bonneville highstand at ca. 18.5 ka, and would imply that this fissure was open well before the Provo shoreline had formed. While we cannot rule out post- or syn-Provo events, no evidence has been found (e.g., a buried layer of basalt colluvium from scarp collapse) for such events at this site. There is, however, strong evidence from the stratigraphy of the fissure fill that this fissure was open pre-Provo and probably pre-Bonneville highstand.

Further, new ⁴⁰Ar/³⁹Ar ages for the Pavant lavas (this study) demand reevaluation of previously proposed relative age relationships and K-Ar ages (Table 1; Fig. 3). Previous studies designated the Pavant I lava flow as being older than the inset and topographically higher Pavant II flow (Hoover, 1974; Johnsen et al., 2010). Our analysis indicates that samples from both flows have extremely low radiogenic yields (⁴⁰Ar*), and so a trapped atmospheric initial ⁴⁰Ar/³⁶Ar value was assumed in age calculations determined from plateau analysis. The plateau age (n steps = 8) for the Pavant I lava from sample ARAR5 is 66 ± 13 ka (1σ) (Fig. 7B). Sample DKF1 from the Pavant II lava yields a plateau age of 758 ± 21 ka (Fig. 7C). Sample ARAR5 yielded a disturbed age spectrum and has a high value of mean square weighted deviates (MSWD) at 10.6 (Fig. 7B). Sample DKF1 has a MSWD of 1.7, below the commonly accepted maximum value of 2.5, and a relatively straightforward spectrum with ages plateauing at ca. 0.75 Ma (n steps = 6) in higher-temperature steps.

We consider both of the new ⁴⁰Ar/³⁹Ar ages to be improvements upon existing K-Ar ages, which had ∼100% uncertainties. As such, we recommend adopting these new ages for the Pavant flows. These new results require that the Pavant II lava, previously considered to be younger than the Pavant I lava, is more than an order of magnitude older. In a relative sense, the morphologies of the Pavant I and Pavant II flows do suggest that this is accurate—the Pavant II flow surface appears more modified, wavy, and oxidized than that of Pavant I. The map pattern that led to the original interpretation (i.e., Pavant II flows “inset” within Pavant I) could be explained by the Pavant I lava flowing around and filling depressions adjacent to the much older, and locally thicker, Pavant II lava flow (Fig. 3).

Despite the ∼600 k.y. age difference, there is not a significant difference in the cumulative vertical separation across the Pavant fault on these two flows. The maximum vertical separation of the Pavant I lava is 18 m. Maximum vertical separation on the same fault trace where it displaces the Pavant II lava is on the order of ∼20 m. Thus no, or at most very little, displacement occurred on the Pavant fault prior to extrusion of the Pavant I flow, which suggests late-stage development of the fault trace through both flows.

Using the cumulative extension across the Pavant II flow surface, which has greater potential for preservation owing to its higher position in the landscape and better age control, we calculate 29.1 ± 4.9 m of extension in the last 758 ± 21 k.y. This leads to a minimum extension rate estimate of ∼0.04 ± 0.01 mm yr⁻¹. Using the more uncertain age of the Pavant I lava in the calculation yields an extension rate estimate of 0.46 ± 0.13 mm yr⁻¹. In our summary of extension rates (below), we use the minimum rate of 0.04 mm yr⁻¹.

Numerous west- and east-dipping faults displace the Deseret flow into a series of horsts and grabens. Between 12 and 20 fault traces, depending on location along the flow, displace the Deseret basalt flow by up to ∼20 m (Stahl and Niemi, 2017) (Fig. 3). The displacement histories on these faults were interpreted by Oviatt (1989) to predate Lake Bonneville, based on the observation that Bonneville sediments overlie the Deseret fault scarps. However, there are no established MRE timings on any of the fault traces. Faults that offset the flow surface do not disrupt the surrounding playa floor, although the adjacent Clear Lake fault scarps and fissures do displace the modern playa surface, and the westernmost Clear Lake scarp forms the edge of the Deseret flow (McBride et al., 2015) (Fig. 3).

The Deseret flow has previously been K-Ar dated to 0.4 ± 0.4 Ma. We sampled the flow interior for ⁴⁰Ar/³⁹Ar geochronology to revise this age estimate. Sample DesAR01 had a low radiogenic yield, although high-temperature steps reached 15% yield. During analysis, an excess trapped argon component was suggested by a high ⁴⁰Ar/³⁶Ar value of 299.7 ± 0.2, and a reliable plateau age could not be reached (Fig. 7A). The isochron (“errorchron”) age is therefore the preferred age for sample DesAR01 at 668 ± 9 ka (1σ) with a MSWD of 9 (Fig. 7D). This MSWD value is significantly larger than commonly accepted values (<2.5), but is considered a useful improvement upon previously reported ages for the flow.

Net extension from normal faulting across the 668 ± 9 ka flow is m. We therefore obtain extension rates across the Deseret flow of 0.12 ± 0.03 mm yr⁻¹.

The Clear Lake faults comprise a 5–10-km-wide, ∼35-km-long zone of faults and fissures between the Deseret and Pavant lava flows (Fig. 3). Cumulative vertical separation across the zone is negligible (Stahl and Niemi, 2017). There is little subsurface information on most of this complex zone of fissures and faults, although the westernmost scarp, with 4 m maximum vertical separation, has been imaged via seismic reflection at various depths, with progressively larger displacements on older strata (Crone and Harding, 1984; Planke and Smith, 1991; McBride et al., 2015). McBride et al. (2015) inferred ∼10 m of post–Lake Bonneville vertical separation on this fault trace. Some have noted that this fault can be traced to a prominent reflector at depth, and may therefore be kinematically linked to slip on the Sevier Desert detachment (Crone and Harding, 1984; Oviatt, 1989; Planke and Smith, 1991). Others have questioned this interpretation (e.g., Wills et al., 2005). Currey (1982) considered that the Clear Lake fault zone may be related to subsidence into a magma chamber at depth, potentially related to adjacent Pleistocene volcanism. Because the Clear Lake faults do not displace any lava flows, and subsurface information is limited, we cannot fully address the recent activity of these faults in this study.

We calculated net extension of 7.7 ± 1.6 m across the Clear Lake playa, which cuts surficial strata with an estimated age of 10 ± 5 ka (McBride et al., 2015; Stahl and Niemi, 2017). This leads to a minimum extension rate estimate of mm yr⁻¹.

We focus on two major, range-bounding faults in the western Sevier Desert, the House Range and Cricket Mountains faults (locations in Fig. 1). Mapping and surveying (Bunds et al., 2019; this study) were conducted to constrain the displacement and timing of surface-rupturing earthquakes. In the House Range, we used ¹⁰Be exposure age dating coupled with a new modeling procedure to constrain the age of a Late Pleistocene–Holocene fan surface, the formation of which postdates the last surface-rupturing event. Extension rates are estimated for both faults in order to compare with rates across the eastern Sevier Desert faults.

The House Range fault is an ∼37–50-km-long west-dipping normal fault located on the western side of the House Range (Figs. 1 and 8). For most of its length, the surface trace is concealed and the fault’s surface projection is mapped as an inferred trace along the gravity anomaly at the boundary between the structurally controlled Tule Valley basin and the steep western escarpment of the House Range (Hintze and Davis, 2003). At depth, the fault is well imaged in seismic-reflection profiles and is shown to be high-angle through the upper 5 km of the crust and listric below that depth, soling into the Sevier-aged Canyon Range thrust (Allmendinger et al., 1983; Greene, 2014). It is unclear if the House Range fault extends to the south between the Black Hills and southern Confusion Range, where the eastern edge of Tule Valley is still structurally controlled by a Cenozoic listric normal fault (Greene, 2014).

Evidence of late Quaternary activity is only present along ∼10 km of the central section of the fault (Fig. 8). Here, the fault displaces pre–Lake Bonneville alluvial fans, pre–Bonneville highstand transgressive shorelines, and the Provo shoreline (Sack, 1990; Hintze and Davis, 2003). Piekarski (1980) performed scarp diffusion dating along this section of the fault from 23 fault scarp profiles and estimated that the fault scarp formed more than 12,000 yr ago, with an average vertical separation of 1.4 m. It was unclear from previous mapping (1) whether the fault deforms post-Provo alluvial fans or (2) if the Provo shoreline and older transgressive shorelines are displaced by the same or differing amounts (i.e., that the average vertical separation of Piekarski [1980] was calculated from a single or multiple events). Sack (1990) and Hintze and Davis (2003) proposed that regression from the Provo shoreline took place rapidly in the Tule Valley due to little inflow following the isolation of Lake Tule from the rest of the Lake Gunnison (Sevier) basin. If this is the case, the relatively rapid, local base-level drop may have caused concurrent progradation of post-Provo alluvial fans to the new Tule Valley floor over a short time period. The ages of these fans, if undeformed, would then tightly constrain the age of the MRE to between the ages of the Provo shoreline and the ages of the post-Provo alluvial fans (Fig. 8).

Geomorphic mapping was carried out to supplement past surficial mapping and to target sample sites for dating the MRE on the House Range fault. Mapping was conducted on a combination of National Agriculture Imagery Program (NAIP) aerial photography and a high-resolution DSM produced from photogrammetry (Bunds et al., 2019). The UAS-based photogrammetry encompasses ∼6 km² and images the House Range fault, multiple shorelines, and pre-Bonneville and post-Provo alluvial fans. These high-resolution aerial imagery and DSM products enhance the detail of geomorphic mapping over previous efforts.

The House Range fault displaces both the transgressive shoreline sequence (estimated to have formed at ca. 20 ka in Lake Tule; Sack, 1990) and the inset Provo shoreline at ∼1462 m (Fig. 8). Vertical separations range from ∼0.4 m to 1.5 m and can vary by up to 1 m over distances of ∼250 m on features of the same age (e.g., the Provo shoreline bench). The fault interacts with fan surfaces of three different relative ages, based on their morphology. The oldest fan unit comprises pre-Bonneville alluvial fans with a preserved mantle of lacustrine deposits (unit Qla₁), as represented by the suite of transgressive shorelines (Fig. 8). Unit Qla₁ surfaces have local resurfacing and bar-and-swale topography, but channel density is low and channels are typically spatially limited to elevations below a Provo-aged tufa zone (Fig. 8).The second-oldest fan surface is unit Qla₂, which is post–Provo highstand in age as indicated by poor preservation of lacustrine deposits and shorelines due to deposition of younger alluvium over the unit Qla₁ surface. In places, unit Qla₂ has remnant shorelines preserved due to the variable amounts of post-Provo aggradation that took place across the surface (Fig. 8). In a single location on unit Qla₂, there appears to be a window into the underlying unit Qla₁ surface, and a degraded fault scarp is preserved (Fig. 8B). Channel density is higher on unit Qla₂ than on unit Qla₁, and bar-and-swale relief is lower. The youngest fan surface is unit Qaf, which consists of post-Provo alluvial fans with the highest observed channel density, low bar-and-swale relief (<∼1.5 m), and no preservation of lacustrine deposits. Units Qaf and Qla₂ are unfaulted in the study region, and the distinction between the two is made solely on geomorphic grounds.

Net slip estimates at 18 locations along the House Range fault vary between ∼0.5 and 1.7 m (Fig. 9), with a mean of 0.95 m. Transgressive shorelines young with increased elevation, and there is no systematic increase in net slip with shoreline marker elevation. In fact, some of the smallest values of net slip are from a pre-Provo unit Qla₁ fan surface below the level of preserved shorelines. Because similarly small slips can be measured on both the Provo shoreline and transgressive shorelines above the Provo level, we attribute the range of net slip to along-strike variability in surface displacement over this short (∼3 km) section of fault. The best indication of a time-integrated slip rate is given by the displacement of the Provo shoreline, which yields a maximum slip rate since Provo time of 0.05 ± 0.12 mm yr⁻¹, including the additional epistemic uncertainty of ±0.1 mm yr⁻¹ (Personius et al., 2017).

The ages of undeformed, post–Provo highstand alluvial fans (units Qaf and Qla₂, Fig. 8) would constrain the minimum age of faulting on the House Range fault. We collected six samples from Prospect Mountain quartzite boulders embedded in unit Qaf and Qla₂ surfaces and four whole-clast samples from an active dry wash. Individual boulder ¹⁰Be surface-exposure ages vary between ca. 28 and 137 ka (Table S4 [^{footnote 1}]) with a mean of ca. 63 ka. Samples HRC1–HRC4 (whole-clast samples taken from the dry wash), which were originally collected to provide an estimate of inheritance (as theoretically recently exposed clasts; e.g., Owen et al., 2011), instead yielded a broad range of exposure ages between 43 and 85 ka, and were older than some clasts from the fan surface. Given that there is definitive geomorphic evidence that the unit Qaf fans are younger than the Provo shoreline (Sack, 1990; Hintze and Davis, 2003; Fig. 8), all of the surface exposure ages from the fan boulders and clasts must carry significant and variable amounts of inherited cosmogenic ¹⁰Be. Therefore, even taking the minimum clast age (28.01 ± 2.70 ka) is not tenable.

While we cannot tightly constrain the age of the unit Qaf surface, preliminary modeling of exposure ages (after Oskin and Prush, 2015) suggests that the resurfacing of unit Qla₁, and deposition of units Qla₂ and Qaf, likely occurred soon after final regression from the Provo level sometime between ca. 13 ka and 18 ka (Fig. S1 [^{footnote 1}]). This is consistent with a purely geomorphic interpretation that lake-level drop in the Tule Valley took place rapidly following regression from the Provo shoreline and isolation of Lake Tule from other sub-basins of Lake Bonneville.

Based on all of the new data, the House Range fault has experienced one surface-rupturing earthquake in the last 20–30 k.y., and this earthquake occurred soon after regression of then–Lake Tule from the Provo shoreline. It is considered likely that the event occurred soon after 18.2 ka, between the minimum age of formation of the unit Qaf surface (ca. 13 ka) and the maximum age of the Provo shoreline (ca. 18 ka).

We used the maximum net slip at the surface to calculate an estimate of extension. Thus, the House Range fault has accumulated ∼0.72 m of extension since ca. 15–18 ka (age of the Provo shoreline), which yields an extension rate estimate of ∼0.04 ± 0.02 mm yr⁻¹ since that time.

The Cricket Mountains fault is an ∼40–55-km-long west-dipping normal fault on the western side of the Cricket Mountains (Fig. 1). The observed surface expression of faulting is limited to 40 km of this length, while the full extent of the structure is inferred from both topographic relief and structural relief estimated from gravity surveys (Case and Cook, 1979; Hintze and Davis, 2003). The Cricket Mountains fault has been inferred to sole into the reactivated Canyon Range or Pavant thrust within ∼5 km below the surface (Allmendinger et al., 1983; DeCelles and Coogan, 2006).

Previous workers have reported a maximum vertical displacement of the Provo shoreline of between 1.3 m (Hecker, 1993) and 2–2.4 m (Oviatt, 1989; Ertec Western Inc., 1981) across the fault. On the southern section of the fault, Hintze and Davis (2002b) mapped a fault trace that apparently displaces a post-Provo alluvial fan. We found this trace to be present where mapped, but determined that it is preserved on a degraded remnant of pre-Bonneville fan where that remnant grades laterally into a younger fan surface (similar to unit Qla₂ in the House Range). Nowhere else are post-Provo fans obviously displaced by the Cricket Mountains fault, constraining the last event to have taken place between formation of the Provo shoreline and formation of post-Provo-aged fans.

Mapping of the Cricket Mountains fault was conducted on NAIP aerial photography and a 1 m TLS DSM (Fig. 10). The age of unit Qaf on the western slope of the Cricket Mountains is more difficult to constrain than in the House Range. Unlike in Tule Valley, inflow from the Beaver and Sevier Rivers in the eastern Sevier Desert kept Lake Gunnison levels relatively high throughout the Late Pleistocene and Holocene (Hintze and Davis, 2003). Therefore, there is no expectation that the formation of unit Qaf was related to rapid regression from the Provo shoreline; it more likely formed at various times along the range front, as Lake Gunnison slowly regressed from the Provo shoreline and receded to the extent of the modern-day Sevier Lake (Figs. 1 and 2).

Our TLS scan captures the displaced Provo shoreline clearly (Fig. 10), but the resulting DSM is of limited use in defining the transgressive shorelines due to locally low point density. Just outside the scan perimeter (Fig. 10), a unit Qaf surface was observed in the field to be undeformed by the fault. Net slip was calculated as 0.55 m on the upper unit Qla₁ surface and 0.73 m on the Provo bench, with the two being indistinguishable within uncertainty. Using an age range for the unit Qla₁ surface that spans from the youngest age of the Provo shoreline (15 ka) to the oldest potential ages of transgressive shorelines (30 ka) yields a time-integrated slip rate at this site of ∼0.03 ± 0.12 mm yr⁻¹, which includes additional epistemic uncertainty of ±0.1 mm yr⁻¹.

Based on mapping and surveying at this single location, the Cricket Mountains fault’s MRE is constrained to be post- or syn-Provo in age. Hecker (1993) considered the MRE to have occurred before ca. 8 ka, however this number is based on scarp diffusion modeling and subject to significant uncertainty. Similar to the House Range fault, there is no evidence at this location of more than one event occurring since the formation of the transgressive-stage shorelines.

Based on our data, we confirm that the Provo shoreline is displaced and that previously reported displacements across it are reasonable. Accepting the minimum age of the Provo shoreline as the last reliably dated strain marker, and the reported maximum vertical separation of 2.4 m, we calculate 1.0 m extension on the Cricket Mountains fault over last 15–18 k.y. This leads to an extension rate estimate of ∼0.06 ± 0.02 mm yr⁻¹.

Two faulting domains can be distinguished across the Sevier Desert on the basis of surface expression, extension rate, and paleoseismicity. The first domain is defined by characteristics associated with faults, folds, and fissures found in the Black Rock volcanic field, which we will refer to as the eastern domain. The western domain is defined by fault characteristics associated with the House Range and Cricket Mountains fault.

Intrabasin faults of the eastern domain (Deseret, Clear Lake, Pavant, and Tabernacle faults) are spatially coincident with Plio-Pleistocene volcanic edifices. Surface deformation is expressed as a combination of basalt escarpments, monoclines, tension cracks, and fault scarps similar to morphologies observed the Volcanic Tableland of California (Ferrill et al., 2016). Individual fault zones are ∼5 km wide; however, the zones together comprise a broader trend of surface deformation that is 15–20 km wide (Fig. 3). Within this domain, individual fault traces are short (commonly <5 km) despite having several meters of cumulative, on-fault displacement over 10³ to 10⁵ yr time scales. Faults generally exhibit greater cumulative extension and larger vertical separations on older volcanic units, indicating recurrent motion over the last ∼0.75–6.5 m.y. (e.g., Wills et al., 2005) on at least some fault traces.

Despite recurring activity and low sedimentation and erosion rates, most topographic relief in the eastern Sevier Desert is produced by the volcanic centers rather than faults. Offsets that arise from displacement on individual fault traces are either countered by displacement on antithetic faults or recovered over short fault-normal distances (Stahl and Niemi, 2017). Minimum on-fault extension rates across the eastern Sevier Desert faults are high (∼0.1–0.8 mm yr⁻¹) relative to faults in the western Sevier domain (< 0.1 mm yr⁻¹), and were particularly elevated in the Late Pleistocene and early Holocene compared to geodetic and longer-term geologic extension rates (Fig. 11).

There are a few possible explanations for the discrepancy between paleoseismic extension rates and those derived over other (e.g., geologic and modern) time scales in the eastern domain. One possibility is that the geologic extension rate (0.19–0.35 mm yr⁻¹; Fig. 11) represents a temporally averaged mixture of extension rates that vary on shorter time scales (e.g., Friedrich et al., 2003; Niemi et al., 2004; Pérouse and Wernicke, 2017). Elsewhere in the Basin and Range, where late Quaternary magmatism is not present or recognized, there is evidence for time-varying strain accumulation as well as release (Friedrich et al., 2003; Pérouse and Wernicke, 2017), with higher Pleistocene–Holocene slip rates reflecting short-term earthquake clustering. In the eastern Sevier Desert, slow (or negative) modern extension rates (−0.1 to 0.1 mm yr⁻¹) and late Miocene to recent geologic extension rates (0.19–0.35 mm yr⁻¹) are similar, but are significantly lower than Late Pleistocene to Holocene rates on the Tabernacle and Clear Lake faults (∼0.74 mm yr⁻¹) (Fig. 11).

Faster Late Pleistocene to Holocene extension rates on the Tabernacle and Clear Lake faults may be the result of a transient episode of strain release driven by recent volcanism and associated deformation (Smith et al., 2004; Rowland et al., 2010; Acocella and Trippanera, 2016). If this hypothesis is correct, faulting would be not only spatially coincident with volcanism, but temporally clustered during periods of dike intrusion into the upper crust. There is some paleoseismic evidence that supports this relationship: two events on the Tabernacle fault occurred during or soon after emplacement of the Tabernacle Hill flow, with no observed paleoseismic activity after emplacement of the undeformed Ice Springs flow at ca. 9–12 ka (Figs. 2 and 3). The Clear Lake fault zone does not directly coincide with volcanism at the surface, but is bounded on either side by the Deseret, Ice Springs, and Pavant lavas (Fig. 3).

Other models of active faulting in the eastern Sevier Desert have been proposed. Salt mobility and saline fluids may play a subordinate role in assisting faulting (e.g., Yuan et al., 2018), but there is no evidence of a direct influence from salt tectonics in our limited paleoseismic data set. Faulting due to salt dissolution is characterized by short and erratic recurrence intervals (Guerrero et al., 2015), and a strong temporal correlation between faulting and volcanism would be unexpected, as seems to be the case for at least the Tabernacle fault. Active detachment faulting underlying a thin, hot upper plate may lead to a similar surface expression of faulting as observed in the eastern Sevier Desert (Jänecke and Evans, 2017), but it is difficult to reconcile the variability in eastern Sevier Desert fault extension rates with them being hanging-wall splays of a common detachment at depth. Furthermore, neither modern strain observed in GPS nor fault displacements are adequately fit by creep or interseismic strain accumulation on an active Sevier Desert detachment (Stahl and Niemi, 2017).

Range-bounding faults of the western domain (Cricket Mountains and House Range) are characterized by simple, continuous fault traces at the surface. These faults displace Late Pleistocene to Holocene alluvial fans with SEDs of ∼1 m over narrow zones (∼10 m) of deformation, but have accumulated significant topographic relief (hundreds of meters) since the late Cenozoic (ca. 15–20 Ma; Stockli et al., 2001). The Cricket Mountains and House Range faults exhibit slower extension rates than the eastern domain faults (Fig. 11). While it is likely that preservation is not an issue at the sites we considered, it is possible that off-fault, intrabasin deformation in the Tule Valley (House Range) and Sevier Lake (Cricket Mountains) (Fig. 1) is not preserved as well as on volcanic edifices in the eastern domain, and therefore could account for at least some of the rate discrepancy. However, the slip rates we report are similar to rates and recurrence intervals observed on other range-bounding (or intrabasin) normal faults in the interior of the Basin and Range (Pérouse and Wernicke, 2017; Personius et al., 2017), and we therefore consider that off-fault deformation cannot account for the large disparities we observe over the same time scales. Collectively, these observations lead to the interpretation of western-domain faults as being tectonic faults that extend through the seismogenic crust (e.g., Allmendinger et al., 1983; Greene, 2014), and therefore accumulate and release elastic strain over the course of a seismic cycle.

Integrating our observations with regional geophysical data, including (1) the presence of near-vertical zones of low resistivity in the upper crust, from the surface to >6 km depth, beneath the Clear Lake, Tabernacle, and Pavant faults (Wannamaker et al., 2013; Liu and Hasterok, 2016), (2) regionally high heat flow and geothermal gradients (Hardwick and Chapman, 2012; Gwynn et al., 2013), and (3) locally thin (∼25 km) lithospheric mantle and prominent low-velocity zones in the sublithospheric mantle (York and Helmberger, 1973; Nelson and Tingey, 1997; Schmandt and Lin 2014; Valentine et al., 2017), we propose a revised neotectonic model for the eastern Sevier Desert that builds on the framework of Wannamaker et al. (2008) (Fig. 12). In this model, far-field tectonic strain is accommodated in the eastern Sevier Desert by thinning of the lithospheric mantle (e.g., Wannamaker et al., 2008) and, depending on magma supply, eventually culminates in episodic dike intrusion and associated faulting in the upper crust. This differs from previously proposed modes of late Quaternary faulting, which include salt tectonics (Wills et al., 2005) and underlying detachment faulting on the Sevier Desert detachment (e.g., Niemi et al., 2004).

This model indicates episodes of transient, elevated strain rates during dike-fed volcanism, and can explain the elevated paleoseismic extension rates on the Clear Lake and Tabernacle faults, as compared to modern GPS rates (Fig. 11). There are examples of this process occurring elsewhere in the Basin and Range. Mackey et al. (2014) found evidence for synchronous dike-fed volcanism and faulting at ca. 14 ka in the Fort Rock Basin of Oregon, with minimal or no activity since. In 2003, a dike injection event beneath Lake Tahoe resulted in a local strain transient, as implied by GPS velocities, of at least an order of magnitude greater than observed over the previous seven years (Smith et al., 2004). Had the resulting extension been resolved on faults at the surface, apparent extension rates across those structures would far exceed modern GPS or longer-term geologic rates, similar to the temporal relationships that we observe on the Clear Lake and Tabernacle faults (Fig. 11).

This revised model for the eastern Sever basin requires that strain is currently partitioned into magma-assisted and tectonic-dominated segments across the Sevier Desert, akin to the segmentation observed within and across segments of narrow rifts (e.g., Ebinger, 2005; Rowland et al., 2010) (Fig. 12). However, our model also has intriguing consequences for along-strike variations in strain accommodation along the eastern margin of the Basin and Range. One implication of our model is that the eastern domain of the Sevier Desert also marks an along-strike transition in strain accommodation, from solely tectonic faulting along the Wasatch fault to the north, to the observed <1 mm yr⁻¹ component of localized magma-assisted rifting in the eastern Sevier Desert to the south (with most of the total ∼4 mm yr⁻¹ still being accommodated on or near the Wasatch fault). Such variability in modes of strain accommodation is perhaps not surprising given that both modes of extension and surface faulting are observed along the western Basin and Range margin (Bursik and Sieh, 1989; Smith et al., 2004), between western and eastern branches of the East African Rift, along the eastern branch of the East African Rift (Hayward and Ebinger, 1996; Muirhead et al., 2015, 2016), and in the Taupo Volcanic Zone of New Zealand (Rowland et al., 2010). However, such variability in strain accommodation mechanisms has only recently been proposed for slowly extending sub-basins within the Basin and Range (e.g., Athens et al., 2016), and our results may imply that dike-driven extension is more relevant than previously recognized. The wealth of geophysical data available within the Sevier Desert region, along with ease of access and exposure, may thus provide a natural laboratory to investigate spatially complex segmentation that can be used to understand the varying modes of continental extension elsewhere.

Our model for strain accommodation in the Sevier Desert region (Fig. 12) is consistent with geodetic and paleoseismic observations and applies over the duration of pervasive volcanic activity in the Black Rock volcanic field (minimum 2.5 m.y.; Johnsen et al., 2010). Prior to the Pleistocene, strain accommodation in the Sevier Desert was likely accommodated via slip on the Sevier Desert detachment (Fig. 12), for which there is sound geologic evidence (e.g., Stockli et al., 2001; DeCelles and Coogan, 2006). Although consideration of this earlier phase of faulting and the reasons for a temporal transition in strain accommodation are beyond the scope of this study, we consider it most likely that the Sevier Desert detachment accommodated relatively rapid extension in the Oligocene–Miocene (Stockli et al., 2001), and Plio-Pleistocene volcanism and associated faulting represents a late to post-extensional phase of basaltic volcanism common throughout the Basin and Range (Gans and Bohrson, 1998).

Using new mapping, geochronology, and survey data, we present a revised model of late Quaternary tectono-magmatic activity in the Sevier Desert of Utah. High-resolution topography and Quaternary geochronology helped to constrain the styles, rates, and timing of surface faulting. The results demonstrate that faults at the eastern margin of the Basin and Range in central Utah comprise two distinct fault domains. Time-integrated extension rates increase from west to east across the Sevier Desert, and extension rates across Late Pleistocene and Holocene markers in the eastern Sevier Desert are an order of magnitude higher than modern GPS-derived extension rates, which we attribute to punctuated strain release during dike-fed volcanism. The revised model of active tectonics presented here points toward across-strike segmentation of the eastern Basin and Range into magma-assisted and tectonic segments, making the Sevier Desert an important locale for understanding spatial and temporal controls on rift segmentation.

This work was supported by U.S. National Science Foundation (NSF) grants EAR 1451466 to Stahl and CAREER 1151247 to Niemi. ¹⁰Be dating was funded by a Purdue Rare Isotope Measurement (PRIME) Lab (Purdue University, West Lafayette, Indiana, USA) Seed grant to Niemi and Stahl. Amanda Maslyn assisted with quartz and olivine mineral separation. We thank Matt Heizler at New Mexico Geochronology Research Laboratory (Socorro, New Mexico, USA), Mark Kurz and Josh Curtice at Woods Hole Oceanographic Institution (Woods Hole, Massachusetts, USA), and Tom Woodruff and Tom Clifton at PRIME Lab for their help and support. We would like to thank the Utah Geological Survey (Salt Lake City, Utah), including Steve Bowman, Adam Hiscock, Gregg Beukelman, Greg McDonald, and Ben Erickson, for field assistance and discussion. Thorough reviews by Craig McGee, Gary Axen, and two anonymous reviewers significantly improved this manuscript. John Sandru and Chris Crosby at UNAVCO (Boulder, Colorado) are thanked for their generous support, for assistance in processing TLS data, and for access to Trimble Business Center software for processing survey data.