Salt-dissolution faults versus tectonic faults from the case study of salt collapse in Spanish Valley, SE Utah (USA)

Seismogenic tectonic faults differ from nonseismogenic tectonic faults in that the former ones are able to cause high-magnitude earthquakes due to sudden slip (episodic motion), while the latter ones are characterized by creep, small slip, and small to moderate earthquakes (McCalpin, 2009a). According to the Nuclear Regulatory Commission, salt-dissolution faults may have significant geomorphic expression, but they are unable to produce significant earthquakes because of their limited rupture area and vertical extent (Huntoon, 1999). Several authors have reported on shallow micro-earthquakes related to collapse phenomena induced by salt dissolution in different geological contexts (Styles, 2003; Trifu and Shumila, 2010; Dahm et al., 2011; Jousset and Rohmer, 2012; Lenti et al., 2012; Land, 2013).

Mistakenly interpreting nontectonic salt dissolution–induced faults as seismogenic tectonic structures may have important practical implications and lead to seismic hazard overestimates (e.g., Gutiérrez et al., 2012). Evidence of episodic displacement has been proposed as a criterion to differentiate between seismogenic tectonic faults and nonseismogenic salt-dissolution faults in areas underlain by evaporites (e.g., Huntoon, 1999). However, several recent trenching investigations have documented evidence of episodic displacement on normal and flexural-slip faults related to interstratal dissolution of evaporites (Gutiérrez et al., 2012, 2014; Carbonel et al., 2013). Those case studies strongly suggest that episodic faulting is not a reliable criterion for distinguishing between seismogenic tectonic faults and salt-dissolution faulting. Moreover, the work by Gutiérrez et al. (2014) on flexural-slip faults related to evaporite dissolution in the Southern Rocky Mountains of Colorado suggests that, in some cases, such gravitational faults might produce damaging earthquakes (M_w 6); the rupture area of those flexural-slip faults may be as large as 200 km². Sudden surface rupture associated with faults characterized by episodic slip is typically recorded by colluvial wedges, fissure fills, and sharp angular unconformities (McCalpin, 2009b). In contrast, cumulative wedge-outs (growth strata) are commonly found in subsidence areas controlled by creeping faults.

Episodic gravitational faulting related to subsurface karstification of soluble rocks may be controlled by a number of interacting factors, including (1) the rate at which evaporites are dissolved by groundwater flow; (2) the extent of the area affected by karstification; (3) the mechanical strength of the strata overlying the karstification zone; and (4) changes in the base level of the hydrogeological system, which may be related to processes like fluvial incision and/or tectonic uplift (Gutiérrez et al., 2012). The intensity of salt dissolution may vary considerably through time, with rates increasing during climatic wet periods, when both groundwater recharge and flow increase (e.g., Berner, 1978; Raines and Dewers, 1997; Jeschke et al., 2001). Other factors that control the solubility and dissolution kinetics of evaporite minerals include temperature (Alkattan et al., 1997; Freyer and Voigt, 2004; McMurry, 2004), effective lithostatic pressure (Blount and Dickson, 1973; Newton and Manning, 2005), pH (Dove and Czank, 1995), the chemistry of the groundwater (Ponsjack, 1940; Wigley, 1973; Ford and Williams, 1989; Alkattan et al., 1997), and the velocity and hydrodynamic regime (laminar vs. turbulent) of the underground flows (Berner, 1978; Raines and Dewers, 1997; Jeschke et al., 2001). Changes in those variables may result in increases or reductions in the dissolution rate, contributing to the episodic behavior of salt dissolution–induced faults. The evolution of karst systems is also sensitive to spatial changes in the subterranean drainage network, such as the opening of new conduits and the collapse of older ones (Ewers and Quinlan, 1981; Ewers, 1982; Birk et al., 2003). Karst systems are also influenced by the spatial and temporal evolution of the surface drainage network. A base-level drop related to fluvial incision or fluvial capture alters the gradient and patterns of the groundwater flow (Palmer, 2000; Lauritzen and Lundberg, 2000).

Even if one assumes steady dissolution under stable groundwater flow conditions, stresses within the overburden change through time and may lead to failure conditions (faulting episodes) when the voids created by karstification reach a critical size and become unstable (Gutiérrez et al., 2008).

Since nontectonic salt-dissolution faults may show episodic behavior like seismogenic tectonic faults, the key question from a diagnostic perspective is: Do salt-dissolution faults have similar scaling and temporal parameters as tectonic faults, such as the aspect ratio of the scarps (height/length), maximum displacement/length ratio (D_max/L), slip rate, displacement per event, average recurrence, and recurrence variability?

To our knowledge, this study conducted in Spanish Valley, Utah, is the fourth site where salt dissolution–induced faults have been investigated by trenching and the second one in the United States. The previous sites were located in the Teruel graben (Gutiérrez et al., 2012) and the Cameros Massif (Carbonel et al., 2013) in the Iberian Chain, and the Carbondale collapse center in the Southern Rocky Mountains of Colorado (Gutiérrez et al., 2014). All these studies document evidence of episodic displacement as revealed by colluvial wedges, angular unconformities, upward truncation of faults, and lacustrine deposits indicative of sudden obstruction of drainages related to the rejuvenation of fault scarps. Scaling parameters estimated for some of the evaporite collapse faults in Spain yield D_max/L values comparable with those reported for tectonic faults, but higher slip rates, greater displacement per event, and much shorter recurrence than those expected for tectonic faults of the same length and in the same geotectonic setting. Unfortunately, the calculation of parameters for the flexural-slip faults of the Carbondale collapse center, Colorado, was hindered by the poorly constrained chronology of the faulting events.

This paper analyzes the dissolution-induced subsidence structures developed in Spanish Valley on the basis of detailed mapping and characterizes the kinematic behavior of dissolution-induced faults in the area using the full displacement history inferred from a trench dug across the master normal fault on the northeastern margin of the graben.

Spanish Valley is a NW-SE–trending graben, 25 km long and 3 km wide, related to the collapse of the crest of a salt anticline in the Canyonlands section of the Colorado Plateau, Utah. The anticline is truncated by the Tertiary diorite laccolith of the La Sal Mountains and crossed transversely by the Colorado River (Fig. 1). The valley forms part of a suite of collapsed salt-cored anticlines in the Paradox fold-and-fault belt of Utah and Colorado. From NE to SW, these dissolution-induced grabens are Fisher Valley, Sinbad Valley, Salt-Cache Valley, Castle Valley, Paradox Valley, Spanish Valley, Gypsum Valley, and Lisbon Valley (Fig. 1). Spanish Valley is situated in the central and deepest part of the ancestral evaporitic Paradox Basin, which developed along the southwestern flank of the Uncompahgre uplift during middle Pennsylvanian to Middle Permian time in Utah and Colorado (Stevenson and Baars, 1986). The ancestral Uncompahgre uplift was a 50-km-wide, NW-SE–trending, basement-involved arch, bounded on both margins by 200–300-km-long fault zones (Barbeau, 2003). The Paradox Basin was interpreted as a pull-apart basin by Stevenson and Baars (1986), but Barbeau (2003) concluded that it was an intraforeland flexural basin adjacent to the Uncompahgre uplift. The Pennsylvanian-age evaporitic Paradox Formation records 29 distinct cycles of halite (Hite, 1960). During the Permian, the salt underwent lateral and upward flow, presumably favored by faulting and differential loading related to thickness variations in the overburden. This halokinetic activity led to the development of NW-SE salt-cored anticlines above Proterozoic basement fault lineaments, and intervening subsidence troughs (Cater, 1970; Stevenson and Baars, 1986; Doelling, 1988, 2000). The salt-bearing Paradox Formation was originally 1200 m thick (Cater, 1970), but it reaches up to 3500 m in thickness along the salt walls that form the core of the anticlines (Baars and Doelling, 1987). Structural and stratigraphic relationships (e.g., thickness variations of stratigraphic units across the anticlines) reveal that salt flowage continued over ∼100 m.y. during the Triassic, Jurassic, and Early Cretaceous, when the region became an inland sea with deposition of the Cretaceous Mancos Shale (Baars and Doelling, 1987; Doelling 1988, 2000). During the Late Cretaceous–Early Tertiary Laramide orogeny, the region underwent west-to-east tectonic shortening that reactivated the preexisting structures (Baars and Stevenson, 1981; Pevear et al., 1997). The Oligocene diorite laccolith of the La Sal Mountains intruded into the terrain at around 29 Ma (Doelling, 2000). Fluvial incision associated with the regional uplift of the Colorado Plateau started around 5–6 Ma (Hoffman et al., 2011), and it is still continuing (Pederson et al., 2013a, b). The consequent continuous base-level drop enhanced the circulation of groundwater through deeper strata, favoring interstratal karstification of the Paradox Formation (Cater, 1970; Doelling, 2000; Gutiérrez, 2004).

The sedimentary succession exposed in the studied sector of Spanish Valley ranges from leached evaporites of the middle Pennsylvanian Paradox Formation (cap rock) to the Cretaceous Mancos Shale. Drilling data indicate that the Paradox Formation has a minimum thickness of 1500 m in Spanish Valley (Doelling, 2001). The salt-rich beds grade upward into a karstic residuum, more than 250 m thick, devoid of salt and mainly consisting of highly distorted calcium sulfates and shales. This cap rock crops out along the valley margins as jumbled masses of contorted and brecciated gypsum, black shale, and dolomite beds affected by halokinetic flow and dissolution-induced collapse (Fig. 2A; Doelling, 2001).

Triassic (Moenkopi and Chinle Formations) and Lower Jurassic rocks (Wingate, Kayenta, and Navajo Formations) overlie the Paradox Formation along both flanks of the valley. Middle and Upper Jurassic and Cretaceous formations are only exposed at the southeastern end of the valley. The Mesozoic formations dip 15°–30° away from the collapsed core of the anticline, attaining a subhorizontal attitude several hundred meters away.

The Spanish Valley graben changes abruptly at its NW edge into a half graben controlled by the 45-km-long, NE-dipping Moab fault, located on the SW margin of the depression (Fig. 2A; Foxford et al., 1996). Doelling (2001, 2004) proposed that this fault extends along the axis of Spanish Valley and could be related to both salt dissolution and extensional tectonics (Berg and Skar, 2005). The Moab fault offsets a 5000-m-thick sequence of Carboniferous- to Cretaceous-age rocks with a maximum dip-slip displacement of 950 m (Foxford et al., 1996). Displacement along this listric fault formed a SW-verging, asymmetric roll-over anticline (Moab anticline) in the hanging wall (Garden et al., 2001). No evidence of Quaternary activity has been found associated with the Moab fault (Olig et al., 1996; Wong et al., 1996). From K-Ar analysis of clay gouge along the northern section of the Moab fault, Pevear et al. (1997) reported that the last faulting period occurred ca. 50–60 Ma. Garden et al. (2001) indicated that hydrocarbon migration in the Moab anticline associated with motion on the fault occurred at 65 Ma. Dating of neoformed illite and illite-smectite samples by ⁴⁰Ar/³⁹Ar in fault gouge and adjacent rocks yielded ages between 60 and 63 Ma (Solum et al., 2005). Thus, the last period of major activity on the Moab fault presumably dates back to the Laramide orogeny (Pevear et al., 1997; Garden et al., 2001; Solum et al., 2005). Furthermore, there is no microseismicity associated with the Moab fault. Most seismicity occurs along a NE-trending Precambrian basement fault zone along a 35-km-long stretch of the Colorado River southwest of Spanish Valley (Wong and Simon, 1981; Wong et al., 1996).

The timing of salt dissolution and collapse of the anticline crest in Spanish Valley is poorly constrained because of: (1) the scarcity of stratigraphic and geomorphic markers of Cenozoic age; (2) the limitations of the correlations of deposits based on the morphological stage of petrocalcic horizons; and (3) the limited optically stimulated luminescence (OSL), radiocarbon, and paleomagnetic dating. The oldest preserved alluvial deposit, T1 (Fig. 1), probably of glaciofluvial origin (Woodward-Clyde Consultants, 1980; Harden et al., 1985), occurs as a perched unit on the eastern flank of Spanish Valley, situated at 240 m above the depression floor (Fig. 2B). The deposit has reversed paleomagnetic polarity (Biggar et al., 1981), suggesting it is early Pleistocene–late Pliocene in age. This evidence suggests that the collapse of the valley started during or after the deposition of the alluvial unit T1.

Spanish Valley is bounded by 300–450-m-high vertical cliffs that are underlain by massive sandstone of the Jurassic Wingate and Navajo Formations (Fig. 1). Rock slides, rockfalls, and topples are responsible for cliff retreat, and the resulting deposits accumulate in the lower debris-covered slopes developed on less resistant Triassic bedrock. At the foot of the slopes, there is a sequence of mantled pediments that grade into axial terraces along the valley bottom, where the cover deposits mantle residual cap rock and foundered bedrock blocks. We mapped a stepped sequence of six terrace levels (T1 to T6; from oldest to youngest) belonging to Mill Creek and Pack Creek that drain the valley longitudinally, and four levels of mantled pediment correlative to some of the terraces (P2, P3, P4, P5; Fig. 1). The alluvium of terraces T1 to T4 (older terraces) consists of 7–10 m of poorly sorted, crudely stratified, boulder and cobble gravel with sand matrix. The pedogenic calcic horizons developed on these terrace deposits display continuous carbonate (stage III) or laminated carbonate (stage IV) depending on the locality (Biggar et al., 1981). Terraces T5 and T6 (younger terraces) are mainly composed by 2.5–4 m of sand and silt with some lenticular gravel beds.

There are three lines of geomorphic evidence indicating that dissolution-induced subsidence has been active in Spanish Valley since the Pleistocene:

• (1) According to Harden et al. (1985), the longitudinal profiles of the older terraces display a continuous downstream gradient decrease and disappear at the southeastern reach of the graben. Thus, older terraces converge in the graben to form a thickened alluvial deposit made up of superimposed terrace fills. The presence in the middle sector of Spanish Valley of a complex pattern of buried soils (Fig. 2C), and stacked alluvial units beneath terrace T4 support Harden’s et al. hypothesis that thickened deposits correlative to older terrace levels lie beneath younger terraces. This stratigraphic relationship should be related to dissolution-induced synsedimentary subsidence, as reported in other fluvial systems developed on halite-bearing evaporates (Guerrero et al., 2008, and references therein).

• (2) According to Smith and Goodknight (2005), although gravel strath terraces exist along the canyon carved by the Colorado River both upstream and downstream of Spanish Valley, the absence of Colorado River terraces in the graben is due to the superposition of fluvial fills caused by salt-dissolution subsidence. The log of a borehole drilled at the northern end of the graben (just outside the study area) reveals more than 120 m of fluvial gravels and sands resting on cap rock, recording an intense and sustained synsedimentary subsidence.

• (3) The presence of ponded subsidence depressions up to 1.3 km² in the Colorado River floodplain, where the water table intersects the topographic surface (“palustrine karstic areas” in Fig. 1), indicates a karstic subsidence rate higher than river aggradation.

The investigation started with the production of a preliminary map of the study area based on the interpretation of color stereoscopic aerial photographs printed at 1:24,000 scale, and the review of published geological information. The geological maps of Spanish Valley produced by Doelling (2001, 2004) and Doelling et al. (2002) constituted highly useful documents for the characterization of bedrock geology. Special attention was paid to the distribution of fault scarps and faulted Quaternary deposits as targets for the trenching investigation. In a subsequent phase, the map was refined in the field using color 1:10,000 scale orthophotographs.

The trenches were oriented perpendicular to the strike of the faults. We followed the procedure for logging paleoseismological trenches as described by McCalpin (2009a). After cleaning the trench walls, a reference grid with horizontal and vertical strings spaced 1 m apart was placed on the shaded side of the trench, which was then logged on graph paper at a scale of 1:50. Material datable by the radiocarbon method (peat, charcoal, and pieces of wood within silt or sand deposits) was collected from the excavated deposits, preferably in the units situated at the base of the graben fill and in those located just beneath and above the stratigraphic discontinuities recording faulting events (event horizons). The retrodeformational analysis of the trenches was conducted using Autocad 2008. The samples were dated at Beta Analytic, Inc., in Miami, Florida, using accelerated mass spectrometry (AMS), and the ages were calibrated using the INTCAL 04 calibration curve of Reimer et al. (2004) (Table 1).

Dissolution-induced subsidence in the crest of the salt anticline has been accommodated by both passive bending (folding) and collapse (normal faulting). The downdropped supra-evaporitic sediments are buried by more than 100 m of alluvium in the bottom of the Spanish Valley, concealing the subsidence structures (Smith and Goodknight, 2005). However, exposures along the flanks of the valley allow us to observe the deformation structures and infer the mechanism involved in the subsidence. Differential sagging in the Mesozoic formations caused by intrastratal karstification of the Paradox Formation produced a sequence of NW-SE–trending anticlines and synclines exposed on both margins of the valley. The folds have wavelengths from 50 to 400 m and lengths in excess of 17 km. Swarms of synthetic and antithetic normal faults within the folds accommodate most of the vertical displacement. Synthetic faults have throws of more than 80 m, while vertical separation on antithetic faults is mostly between 5 m and 20 m and never exceeds 30 m. Locally, the dissolution-induced deformation has generated complex structures in which the strike and dip of the strata change significantly over short distances. According to Cater (1970), the initial radius of curvature of the folds controlled deformation within the salt anticlines. Open anticlines favored ductile deformation (passive bending), whereas subsidence was mainly accommodated by brittle collapse (normal faulting) in tighter anticlines with shorter radius of curvature.

The subsidence magnitude attenuates toward the southeastern end of the valley, where the number of normal faults, their throw, and the dip of the fold limbs gradually decrease, to form single monoclines with dips in the limbs of 20°–30° on both flanks of the valley (Fig. 2D). These paired monoclines terminate at a WNW-ESE–trending syncline (Pack Creek syncline) and a NE-trending fault system developed in Jurassic and Cretaceous rocks that are partially covered by alluvial deposits (Doelling, 2004).

When considered in detail, the amount and structural style of subsidence differ slightly along both margins of Spanish Valley. On the NE flank, the Mesozoic strata exhibit several gentle to open anticlines and synclines with limb dips ranging from 5° to 40° (Fig. 3A). Locally, the folds are affected by graben structures along their hinges up to 1 km long and 100 m wide, controlled by subvertical synthetic and antithetic normal faults. Two nearly vertical normal faults, 5 km and 1.6 km long, accommodate most of the deformation where Mill Creek enters Spanish Valley (Fig. 1). Both faults juxtapose the Kayenta and Navajo Formations. We have estimated a minimum vertical separation of 30 m based on outcrop thickness of Navajo Sandstone in the hanging wall. The easternmost fault is associated with the collapse of the hinge of a syncline, while the fault closest to the valley is related to collapse of the SW limb of a SW-verging anticline (Fig. 3B). In the downthrown block, terrace deposits of Pack Creek and Mill Creek reach more than 30 m thickness and are offset by multiple normal faults (see section 2 of Data Repository data¹).

Subsidence is apparently greater on the SW flank of Spanish Valley. Ductile downwarping and brittle collapse of the competent Mesozoic sandstone beds created a sinuous monocline, 17.6 km long, at the upper part of the valley wall. The crest of the monocline is cut by a keystone graben that is 15.5 km long and 150 to 350 m wide with conspicuous geomorphic expression (Fig. 4A). The keystone graben is bounded by a master, NE-dipping synthetic fault and a swarm of secondary antithetic, SW-dipping faults with en-echelon arrangement that are connected by relay ramps in the step-over zones (Fig. 4B). Throw on the synthetic fault is between 90 m and 150 m in the central sector of the valley and decreases progressively toward both ends. In contrast, the antithetic faults are 150 m to more than 1300 m in length, i.e., considerably shorter than the master synthetic fault. The throw on these secondary faults is 5–45 m and is expressed by sackung-like uphill-facing scarps up to 30 m high. On the upslope side of the fresh-looking antislope scarps, there are elongated enclosed depressions, 150–950 m long and 45–115 m wide, that behave as sediment traps filled by colluvium, slope wash, and eolian deposits (Figs. 4B and 4D). Some of these fault-controlled depressions have been breached by the drainage network, creating deep gullies that expose a Quaternary graben fill up to 20 m thick. Extension has also produced linear fissures up to 90 m long and 3 m wide along the master synthetic fault scarp (Fig. 4A). Cover suffosion and cover collapse sinkholes up to 7 m and 2 m in diameter, respectively, are also relatively common along the trace of the fissures. Their formation is attributed to the downward migration (raveling and piping) of trough-filling deposits through open fissures in the bedrock. Trench 2 was dug across the scarp associated with a 350-m-long antithetic fault that offset Quaternary trough-filling deposits (section 4 of Data Repository data [see ^{footnote 1}]).

In the northeaster limb of this monocline, a shorter and shallower linear graben, more than 6 km long and up to 145 m wide, caused the collapse of Kayenta beds into Wingate strata, with a minimum vertical displacement of 30 m. Despite the intense deformation, the graben is strongly dissected by the drainage network, and fault scarps are degraded, indicating limited or no current activity. The fact that faults are more active and show more recent activity at the marginal sectors of the valley suggests that gravitational deformation migrates away from the valley axis, with a consequent increase in valley width.

Two trenches were excavated across faults that offset Quaternary deposits on both margins of Spanish Valley (see location in Fig. 1). Trench 1 explores the main fault on the NE margin of the valley where it offsets Pack Creek terrace T5 deposits. Trench 2 was dug across an antislope scarp related to a secondary antithetic fault in a marginal graben depression on the SW margin of the valley (Fig. 4C). The log of trench 1 is presented in this section because it records a protracted faulting history constrained by AMS radiocarbon dates. The information related to trench 2, from which no numerical dates were obtained, is included in section 4 of the Data Repository (see ^{footnote 1}).

Trench 1 was a 27-m-long, 4.5-m-deep excavation located at 38°32′20.11N, 109°29′42.78W that exposed the longest fault mapped on the NE margin of Spanish Valley (Fig. 1). This trench corresponds to a preexisting artificial excavation from the 1950s that was improved using hand tools. The fault has a cartographic length of 5 km and a minimum throw of 30 m in its central sector, where the Navajo Sandstone is juxtaposed against Kayenta Formation (Fig. 5A). The trench was oriented N060E and located close to the SE edge of the mapped trace of the fault. At the trench site, the deposits of Pack Creek T5 terrace are deformed and in fault contact with Navajo Sandstone (Fig. 5B). This Quaternary fault is expressed in the landscape as a 5-m-high fault scarp (Fig. 5C). The fault splays into three smaller branches 100 m to the southeast of the trench site and then disappears under Pack Creek floodplain deposits. Several springs occur at the foot of the fault scarp (Fig. 5C), supporting the growth of bushes and trees. Calcite, quartz, goethite, and manganese oxides cement the fault zone. Finally, the trench location near the tip of fault implies that the measured displacements and slip rates have to be taken as minimum values.

The trench exposed 18 sedimentary units and a horst and graben structure affecting Holocene deposits in the hanging wall of the mapped fault. The mapped structure comprises an intermediate bedrock horst flanked by an upper graben, 9 m wide, and a downthrown block to the west (Fig. 6). Section 3 of the Data Repository (see ^{footnote 1}) includes a detailed description of the stratigraphic units and the deformation structures. In total, nine samples collected from key stratigraphic units were dated by the AMS radiocarbon method (Table 1). The upper graben includes a tapering-downward fissure (unit 17) that is 42 cm wide at the top (Figs. 7A and 7B) and eight subvertical faults (F10 to F17). F10 shows a 2.30-m-wide shear zone. The downthrown block west of the horst is bounded by the synthetic master fault F7, dipping 50°SW (Fig. 6). The subvertical faults F1 to F6 constitute a splay of secondary failure planes emanating from fault F7. Displacement on faults F7 and F5 is responsible for the folding of units 4–13, the deposition and dragging of two colluvial wedges with thicknesses of 77 cm and 48 cm thick (Figs. 7C and 7D), respectively, and complex soft sediment deformation in the sandy unit 4.

The structural and stratigraphic relationships observed in the trench, together with the available numerical dates, allow us to infer a minimum of nine faulting events involving the development of multiple surface ruptures, locally accompanied by folding in the unconsolidated Holocene deposits. The oldest recorded surface-rupturing event (event R), for which timing can be bracketed at 4290–4500 cal yr B.P., deformed colluvial unit 2 and created a scarp associated with fault F5 that led to the deposition of a 48-cm-thick colluvial wedge (unit 3). Unit 4, consisting of horizontally laminated fluvial sand with peat beds, suggests that subsidence at the foot of the scarp generated a poorly drained depression affected by periodic flooding. Considering that fault scarps tend to be about twice as high as the associated colluvial wedges (McCalpin, 2009b), we can roughly estimate that the fault scarp was around 1.0 m high.

The second event (event S) is recorded by deposition of the 77-cm-thick upper colluvial wedge between 3330 and 4340 cal yr B.P. This surface-rupturing event rejuvenated the scarp, leading to the colluvial deposition that locally and temporally interrupted the sedimentation of the fine-grained unit 4. Considering the thickness of the wedge, the displacement on F5 would be around 1.5 m.

The third event (event T) is recorded by the formation of the depression associated with the upper graben, in which the fine-grained unit 6 was deposited. The age of the base of unit 6 (2680–2640 cal yr B.P.), deposition of which started immediately after the creation of the sediment trap, can be considered as a good approximation for the timing of this event (ca. 2.7 ka). Probably, displacement on fault F10 generated an antislope scarp that confined deposition of unit 6 in the upper graben, as suggested by the absence of this unit in the downthrown block to the west of the horst. The thickness of unit 6 (81.5 cm), the top of which corresponds to an erosional contact, provides a minimum estimate for the vertical displacement achieved in this faulting event.

The fourth event (event U) occurred between 2340 and 2680 cal yr B.P. It is inferred from deformed units in the upper graben and the western downthrown block. In the upper graben, extension associated with fault F11 led to the formation of a fissure that is 42 cm wide. This fissure cuts through units 2 and 6 and is truncated by unit 7. The throw on fault F11 is just 17 cm. Moreover, on the eastern margin of the upper graben, unit 6 is offset by fault F12 with a minimum vertical displacement of 2.5 cm; unit 6 was eroded from the footwall. In the western downthrown block, this event is recorded by offset of units 2 and 4 on fault F1, which is truncated by unit 7. From the exposed thickness of unit 2 in the footwall, we can estimate a minimum throw of 45.1 cm on fault F1. The sum of the displacements yields a minimum total offset of around 65 cm for event U.

The structural and stratigraphic evidence of the fifth event (event V), which occurred between 2460 and 2340 cal yr B.P., includes in the downthrown block: (1) offset of unit 7 and older units by fault F2 with a vertical displacement of 50 cm, where fault F2 is truncated by unit 8; (2) faulting of units 4 and 7 and the two colluvial wedges (units 3 and 5) by fault F3 and F4 with a total vertical displacement of 3.5 cm; (3) development of a drag fold on both colluvial wedges (units 3 and 5) and units 4 and 7 due to slip on fault F5 (using the base of the upper colluvial wedge [unit 5], and assuming a horizontal original attitude, we estimate a minimum vertical displacement of 56 cm due to dragging along F5); and (4) disharmonic folding of units 4 and 7 related to local compression between faults F2 and F3. A minimum vertical cumulative displacement of 1.09 m is required to explain the deformational structures affecting these units, which are all truncated by unit 8.

The next three events (event W, X, and Y) took place in a short time span between 2140 and 2360 cal yr B.P. Unfortunately, it was impossible to better constrain their chronology. The sixth event (event W) is recorded by: (1) the offset of units 8 and 9 by faults F5 and F6 in the western downthrown block, with a minimum cumulative vertical throw of 32 cm since unit 8 has been eroded from the footwall of fault F6, and truncation of faults F5 and F6 by unit 10, which unconformably overlies the faulted unit 9; and (2) rotation of unit 9 by 7° to the west according to our retrodeformational analysis. Considering the dip of 2° in the western part of the half graben as the original angle of deposition, the vertical displacement required to restore unit 9 to its original orientation is 35 cm. Thus, the total displacement for this faulting event was over 67 cm.

The seventh event (event X) caused significant vertical displacement on fault F10 in the upper graben. The fault postdates unit 10 and precedes unit 11, which was deposited across the shear zone associated with fault F10. Deposition of units 7–13 in the upper graben and western downthrown block suggests that the vertical displacement on fault F10 was not enough to restrict sedimentation in the upper graben. Therefore, the vertical distance between the base of unit 7 in the hanging wall and the top of the shear zone (unit 17) provides a minimum throw of at least 96 cm, since unit 7 has been eroded in the footwall of F10.

The eighth event (event Y) displaced units 7–11 in the upper graben by motion on faults F13, F15, and F16, all of which predate unit 12. These structures record a minimum cumulative throw of 1.13 m.

The most recent event, or MRE (event Z), occurred after 2330 cal yr B.P. and affected the youngest dated unit in both grabens (unit 12). In the upper graben, movement along faults F10, F16, and F17 was accommodated by ductile deformation and the development of monoclinal drape folds at both margins of the graben. In the western block, dragging along F7 tilted unit 13 and the underlying ones. Assuming a syndepositional dip of 2° for unit 12, which is the gradient of the lowest present-day slope, we measure a minimum net vertical displacement due to tilting of 2.36 m. Because 2.36 m is significantly greater than that of all previous events, it is possible that cumulative ductile deformation was caused by two or more events instead of just one faulting event.

The slip history diagram (Fig. 8) shows seven closed and two open displacement cycles. The diagram has been constructed considering (1) the value of the age ranges assigned to single events; (2) uniform recurrence for the last three cycles, constrained by the same bracketing ages; and (3) the available data on displacement, either rough estimates based on colluvial wedge thickness (events R and S) or minimum vertical displacement values. The mean slip rates and recurrence intervals computed for the seven cycles are, respectively, from the oldest to the youngest, as follows: ∼2.67 mm/yr and 560 yr, >1 mm/yr and 815 yr, >1.22 mm/yr and 530 yr, >12.1 mm/yr and 90 yr, >8.7 mm/yr and 73 yr, >13.1 mm/yr and 73 yr, and >15.5 mm/yr and 73 yr. Considering a minimum cumulative vertical displacement of 6.81 m and a time span of 2218 yr for the seven cycles, we obtain a minimum mean vertical slip rate of 3.07 mm/yr and an average recurrence of 316 yr. The apparent minimum slip rate and the apparent recurrence computed for all the open and closed cycles are 2.26 mm/yr and 500 yr, respectively. In this latter calculation, we consider the age of the oldest dated sedimentary unit and the minimum total displacement (10.17 m) recorded in the trench.

Unfortunately, the chronology of the last faulting event is poorly constrained by a maximum age (2330 cal yr B.P.), since the youngest sediments exposed in the trench largely correspond to anthropogenic deposits accumulated on a man-made excavation surface. Nevertheless, the slip history suggests that the inferred MRE, which shows an anomalously high vertical displacement (>2.36 m), may actually correspond to two or more surface-rupturing events, supporting recurrent fault motion during the last two millennia. According to the hypothesis of two faulting events younger than 2330 cal yr B.P., the apparent average recurrence would be 450 yr or less. Regardless of the actual age of the last faulting event that occurred at the trench site, which most probably yields an underrepresented late Holocene deformation history, the available data strongly suggest that instantaneous dissolution-induced surface faulting is not an exhausted phenomenon in Spanish Valley. The existence of large subsidence depressions in the Colorado River floodplain (Fig. 1) provides geomorphic evidence of active evaporite dissolution and indicates that the karstic subsidence rate is notably higher than the local aggradation rates of the Colorado River and tributaries; otherwise, they would be quickly filled and obliterated by alluvial deposition. Moreover, the stacking of the Colorado River terraces in Spanish Valley as thickened alluvium (Smith and Goodknight, 2005) suggests that the Colorado River has been unable to keep pace with subsidence over a long period in the Quaternary. The regional incision rate of around 0.45 mm/yr estimated in the central Colorado Plateau over the late Pleistocene (Pederson et al., 2013a, 2103b) indicates a regional entrenchment trend that is clearly exceeded by the local base-level drop induced by salt-dissolution subsidence in Spanish Valley. The minimum slip rate of 3.07 mm/yr estimated in trench 1 for the main fault of the northeastern margin of the graben suggests that salt-dissolution subsidence may be locally six times greater than regional incision rates.

However, hydrochemical data from the Colorado River from 2003 to 2011 suggest limited salt dissolution in Spanish Valley at the present time (Shope and Gerner, 2014). According to Shope and Gerner, the salinity of the Colorado River between two measurement stations located 13 km upstream and 2 km downstream of Spanish Valley does not increase, showing a stable concentration of around 100 mg/L of chloride. According to Woodhouse et al. (2010) and U.S. Geological Survey (2004, 2011), the decrease in precipitation during the monitoring period might be responsible for a decline in groundwater recharge and the amount of evaporites dissolved. The significance of this observation is uncertain. That is, dissolution and hence subsidence and faulting in Spanish Valley may have ended permanently, but more likely, this hydrochemical evidence, based on a short record, may be of little significance to longer-term salt karstification processes, the episodic behavior of which might be closely related to climate variability.

The stratigraphic and structural relationships observed in trenches 1 and 2 reveal episodic displacement on faults that accommodate subsidence caused by intrastratal dissolution of evaporites. Similar episodic faulting has been found in collapse sinkholes (Gutiérrez et al., 2009, 2011; Carbonel et al., 2014) and faults related to interstratal evaporite dissolution in Spain (Gutiérrez et al., 2012; Carbonel et al., 2013) and Colorado (Gutiérrez et al., 2014). Given that episodic faulting is the expected kinematic style for evaporite collapse structures, what other criteria may be robust discriminators of dissolution versus tectonic faulting? Here, we consider three fault parameters: (1) fault displacement to length ratio, (2) slip rate, and (3) statistical properties of the recurrence interval. Common scaling relationships for tectonic faults include ratios of maximum cumulative displacement to total fault length (D_cum/L), maximum coseismic displacement to total length (MD/L), average coseismic slip to fault length (AD/L), and average coseismic slip to maximum displacement (AD/MD). D_cum/L for tectonic faults with lengths similar to those in Spanish Valley show a broad range between 10⁻³ and 10⁻¹ (Dawers et al., 1993; Schlische et al., 1996; Davis et al., 2005; Kim and Sanderson, 2005). We obtain a D_cum/L value of ∼6 × 10⁻³ for the primary boundary fault along the northeastern margin of Spanish Valley, decreasing to 2 × 10⁻³ where trench 1 is located. The D_cum/L value calculated for the master synthetic fault of the marginal keystone graben is 9.6 × 10⁻³. These ratios are comparable with those reported for extensional tectonic faults (Fig. 9).

In contrast, the MD/L and AD/L ratios obtained for the salt dissolution–induced faults of Spanish Valley are significantly higher than those reported for tectonic faults (for extended explanation of slip-length ratio, see Shaw, 2013). Displacement measurements from trench 1 indicate a MD and an AD of 1.5 m and 1 m, respectively, for a fault around 5 km long. It is important to note that: (1) the inferred MD value corresponds to event S (the 2.36 m of displacement recorded in the MRE [event Z] was rejected from the analysis since it may be attributed to more than one faulting event, as discussed already), and (2) the AD is mainly derived from minimum values. Scholz (1994) pointed out that the MD/L ratio is usually on the order of 10⁻⁴–10⁻⁵. That is, the MD of a 5-km-long tectonic fault is expected to range between 50 cm and 5 cm. Similarly, the empirical relationships of Wells and Coppersmith (1994) and Leonard (2010) predict a maximum displacement per event (MD) of 6.7 cm and 20 cm. These values are 3–30, 20, and 7.5 times lower than the maximum displacement inferred in trench 1, respectively, for 5-km-long tectonic faults. Also according to those authors, an AD of 1 m would be expected for 30- and 36-km-long, normal fault surface ruptures, respectively. Regarding trench 2 (see section 4 of the Data Repository [see ^{footnote 1}]), the minimum MD and AD values estimated for the 350-m-long antithetic secondary normal fault investigated in trench 2 are 49 and 37 cm, respectively. The MD/L ratio is 14–140 times higher than Scholz’s (1994) range. In addition, according to the relationships of Wells and Coppersmith (1994) and Leonard (2010), the obtained average slip would be expected from surface-rupture lengths of 17 km and 11 km, respectively. Liu-Zeng et al. (2005) suggested that the AD/L ratio fits a linear relationship given by the formula D = (2.5 × 10⁻⁵)L (Fig. 10). The AD values of trench 1 and trench 2 are 8 and 42 times higher than the expected values for tectonic faults using the formula of Liu-Zeng et al. (2005), respectively.

McCalpin (2009c), based on an extensive bibliographic review, indicated that the AD/MD ratio of tectonic faults averages around 0.5, with a range of 0.2–0.8. The AD/MD ratios calculated for the faults investigated in trenches 1 and 2 are 0.6 and 0.75, respectively, which seem to match with those reported for tectonic faults.

Slip rate and recurrence intervals may also be useful discriminators when classifying faults as tectonic versus nontectonic structures associated with evaporite dissolution. The fault exposed in trench 1 has a surprisingly high slip rate of >3.07 mm/yr calculated from seven closed slip cycles. These values are significantly greater than the slip rates for most tectonic normal faults. For example, the slip rate of the Wasatch fault along the eastern boundary of the Basin and Range Province in Utah ranges from 1.4 mm/yr to 0.12 mm/yr, depending on the segment and the time period over which slip is measured (Machette et al., 1992; McCalpin and Nishenko, 1996; Mattson and Bruhn, 2001; Friedrich et al., 2003; Lund, 2005; Chang et al., 2006; Mayo et al., 2009). The average slip rate at Spanish Valley is between 2 and 25 times greater than the range estimated for the Wasatch fault.

The slip history diagram derived from trench 1 shows stochastic behavior with a large variation in the closed cycle average slip rates and average recurrences, with ranges of 1–15 mm/yr and 815–73 yr, respectively (Fig. 8). The coefficient of variation (CV) or aperiodicity (standard deviation of the recurrence data/long-term average earthquake recurrence), which simply reflects the variability in the recurrence interval, is 1.04. In extensional tectonic settings, there is an increasing body of evidence indicating that slip rate and recurrence vary in space and time (Mitchell et al., 2001; Friedrich et al., 2003; Bull et al., 2006; Nicol et al., 2009). Values of CV quoted in the literature vary from 0.02 to 1.0 (Ellsworth et al., 1999; Cowie et al., 2012), approaching a more representative value of 0.3–0.5 as the number of dated paleoearthquake ruptures increases, to provide a more complete fault rupture history (Console et al., 2008; McCalpin, 2009c; Schlagenhauf et al., 2010). The obtained CV value of 1.04 is around double that expected for a tectonic fault.

The karstic subsidence of the crest of the Spanish Valley salt-intruded anticline has led to the folding and collapse of the overlying strata, which have been buried by a stepped sequence of six terrace levels and four covered pediment levels correlative to some of the terraces at the bottom of the valley. The anomalies of the longitudinal profile of the older terraces at the southern end of the valley together with the existence of a complex pattern of buried soils beneath the planar depositional surface of terrace T3 support Harden et al.’s (1985) studies that older terraces converge downstream to form a single thickened alluvial surface made up of superimposed terrace fills in the graben.

The extension related to the intrastratal dissolution of the Paradox Formation has produced a sequence of anticlines and synclines that are 50–400 m in wavelength and up to 17 km long, trending NW-SE parallel to the axis of the graben on both margins of the valley. The limb and hinge of the folds are often affected by swarms of synthetic and antithetic normal faults, giving rise to an intensely brecciated stratigraphic sequence with highly variable strikes and dips. The gravitational deformation attenuates toward the southeastern end of the valley, where the number of normal faults and the dip of the fold limbs gradually decrease to form single monoclines with dips in the limbs of 20°–30° on both flanks of the valley. This complex deformational context fits well with Gutiérrez’s (2004) interstratal karstification model of salt anticlines.

A trench, 27 m long and 4.5 m deep, was excavated across the hanging wall of the 5-km-long primary normal fault of the eastern flank. The trench shows a complex structure made up of an upper graben and a lower half graben separated by a horst structure. The structural and stratigraphic relationships observed in the trench may be explained by a minimum of nine faulting events involving the development of multiple surface ruptures, the flexure and ductile deformation of units, and the formation of a fissure fill and two colluvial wedges, which are interpreted as evidences of episodic displacement. The slip history diagram of the fault shows seven closed and two open seismic cycles that allow us to obtain the following parameters: (1) D_cum/Length ratio of 6 × 10⁻³ in the middle sector and 2 × 10⁻³ where trench 1 was located, close to the SE edge of the mapped trace of the fault, values which are similar to those expected from tectonic faults; (2) from 2 to 25 times higher slip rate than the Wasatch fault, with an average slip rate of 3.07 mm/yr and an average recurrence interval of 316 yr calculated from seven closed slip cycles; (3) a stochastic regime with a CV value of 1.04, which is double that of tectonic faults with a long paleoseismologic record; (4) a maximum displacement (MD) of 1.5 m, which is 3–30 times higher than tectonic faults of the same length; and (5) an average slip displacement (AD) of 1 m. This value is 5–15 times higher than the expected AD for a 5-km-long tectonic fault according to the empirical relationships of Wells and Coppersmith (1994) and Leonard (2010).

Salt dissolution–induced faults show an episodic behavior but do not follow the experimental relationships used to characterized normal seismogenic tectonic faults. Although both types of faults have similar aspect ratios, salt-dissolution faults will probably yield: (1) higher long-term slip rates than the surrounding tectonic faults because the rate of salt dissolution will probably be higher than the telluric movement rate, (2) a random behavior with very high CV values, probably due to changes in the dissolution rate and karstic hydrogeological system, and (3) a very high slip per event in relation to their length. These differences would allow the genesis of faults in complex areas affected by tectonic activity and salt dissolution to be distinguished and the possible seismic risk related to them to be determined.

The work has been partially financed by the projects CGL2010-16775 (Ministerio de Ciencia e Innovación and FEDER) and CGL2013-40867-P (Ministerio de Economía y Competitividad). Jesús Guerrero carried out a postdoctoral stay at the University of Utah funded by the Ministerio de Educación of Spain. The authors acknowledge the assistance provided by Helmut Doelling. Thanks go to the City Hall of Moab, for approving a permit to work on the artificial excavation corresponding to trench 1.