Abstract

The 1983 Mw 6.9 Borah Peak earthquake generated ∼36 km of surface rupture along the Thousand Springs and Warm Springs sections of the Lost River fault zone (LRFZ, Idaho, USA). Although the rupture is a well-studied example of multisegment surface faulting, ambiguity remains regarding the degree to which a bedrock ridge and branch fault at the Willow Creek Hills influenced rupture progress. To explore the 1983 rupture in the context of the structural complexity, we reconstruct the spatial distribution of surface displacements for the northern 16 km of the 1983 rupture and prehistoric ruptures in the same reach of the LRFZ using 252 vertical-separation measurements made from high-resolution (5–10-cm-pixel) digital surface models. Our results suggest the 1983 Warm Springs rupture had an average vertical displacement of ∼0.3–0.4 m and released ∼6% of the seismic moment estimated for the Borah Peak earthquake and <12% of the moment accumulated on the Warm Springs section since its last prehistoric earthquake. The 1983 Warm Springs rupture is best described as the moderate-displacement continuation of primary rupture from the Thousand Springs section into and through a zone of structural complexity. Historical and prehistoric displacements show that the Willow Creek Hills have impeded some, but not all ruptures. We speculate that rupture termination or penetration is controlled by the history of LRFZ moment release, displacement, and rupture direction. Our results inform the interpretation of paleoseismic data from near zones of normal-fault structural complexity and demonstrate that these zones may modulate rather than impede rupture displacement.

INTRODUCTION

The 1983 Mw 6.9 Borah Peak earthquake ruptured ∼36 km of the ∼130-km-long Lost River fault zone (LRFZ, Idaho, USA) (Crone et al., 1987), one of several normal faults that accommodate dominantly SW-NE extension in the Centennial Tectonic Belt of the northern Basin and Range Province (Scott et al., 1985; Stickney and Bartholomew, 1987; Payne et al., 2013) (Fig. 1). Surface rupture occurred along two structural fault sections, including all of the Thousand Springs section, and the southern half of the Warm Springs section, north of the Willow Creek Hills structure, a prominent hanging-wall bedrock ridge where the LRFZ splits into multiple strands with differing strikes (Crone et al., 1987) (Fig. 1). As one of the largest intraplate normal-faulting earthquakes recorded historically and an example of the complex rupture of a multisegment normal fault system (Haller and Crone, 2004), the Borah Peak earthquake rupture offers an important opportunity to relate spatial and temporal patterns of surface displacement to fault-rupture processes (e.g., Wesnousky, 2008; Nissen et al., 2014; Haddon et al., 2016; Delano et al., 2017; Personius et al., 2017; Johnson et al., 2018).

Although the subsurface rupture geometry (Boatwright, 1985; Doser and Smith, 1985; Smith et al., 1985; Richins et al., 1987) and slip (e.g., Ward and Barrientos, 1986), far-field crustal deformation (Stein and Barrientos, 1985; Barrientos et al., 1987), fault-zone structure (Janecke, 1993; Susong et al., 1990; Bruhn et al., 1991), and surface rupture extent and displacement (Crone et al., 1987) of the Borah Peak earthquake are well documented, uncertainty remains regarding the role the Willow Creek Hills structure played in controlling the length of the rupture (Crone et al., 1985; Bruhn et al., 1991). That is, did the structure impede the lateral propagation of the 1983 rupture, where surface faulting to the north along the Warm Springs section is secondary (nonseismogenic) in nature (Crone et al., 1987)? Or is the 1983 earthquake an example of multisegment rupture in which the Willow Creek hills modulated, but did not fully stop slip propagation? Further, to what degree has the structure impeded the propagation of previous LRFZ ruptures? These questions are important in recognizing how conditional probabilities of rupture through structural barriers (e.g., Oskin et al., 2015) may help explain evidence of multi-modal fault behavior (e.g., single-segment and multi-segment rupture; e.g., DuRoss et al., 2016), and ultimately help improve earthquake-rupture forecasts (e.g., Field et al., 2014).

Here, we use high-resolution (5–10-cm-pixel) digital surface models (DSMs) to improve our understanding of the 1983 rupture in the context of slip propagation through the Willow Creek Hills structurally complex zone (Fig. 2). DSMs generated from low-altitude aerial photography derived from unmanned aircraft systems (UAS) allow us to map the geometry and extent of deformation in the 1983 rupture, estimate the vertical displacement of geomorphic surfaces faulted by the LRFZ in both the 1983 and prehistoric earthquakes (Fig. 3), and quantify trends in displacement along fault strike. We focus on the northernmost 16 km of rupture, north and south of the Willow Creek Hills, using 252 displacement observations (Supplemental Fig. S11). Measurements include geomorphic surfaces displaced by both the 1983 rupture (n = 196) as well as prehistoric ruptures (n = 56). These observations highlight the complex surface-rupture geometry near the Willow Creek Hills and clear and continuous rupture of the southernmost Warm Springs section, and provide evidence that prehistoric ruptures of the Warm Springs section had displacements and along-strike extents distinctly different from the 1983 rupture.

TECTONIC SETTING

Lost River Fault Zone

The LRFZ is one of several NW-striking normal faults that accommodate dominantly SW-NE extension and terminate to the south near the northern margin of the eastern Snake River plain (Baldwin, 1951; Scott et al., 1985; Payne et al., 2008) (Fig. 1). Six sections along the LRFZ have been proposed, including (north to south) the Challis, Warm Springs, Thousand Springs, Mackay, Pass Creek, and Arco sections (Scott et al., 1985; Crone et al., 1987). These sections range from 15 km (Arco and Warm Springs sections) to 26 km (Pass Creek section) in length (U.S. Geological Survey, 2018) and are defined using along-strike changes in fault geometry and geomorphology, structural relief of the range front (including the presence of hanging-wall bedrock salients), and differences in the timing of most recent fault movement (Scott et al., 1985).

Movement on the SW-dipping LRFZ since ca. 4–7 Ma has generated the prominent SW-facing Lost River range front and ∼2.7 km of maximum structural relief, accounting for basin fill and range front topography (Scott et al., 1985). This translates to a late Neogene average slip rate of ∼0.4–0.7 mm/yr. Scott et al. (1985) calculated a latest Pleistocene to present geologic slip rate of ∼0.3 mm/yr for the Thousand Springs section, based on 3.5–4.5 m of vertical offset (including displacement from the 1983 earthquake) measured across a ca. 15 ka fan surface (Pierce and Scott, 1982). Using the 1983 displacement (∼1.5–2 m) and time since the previous surface-rupturing earthquake (∼8 k.y.; Scott et al., 1985), Hanks and Schwartz (1987) reported a single-event closed-interval slip rate of ∼0.2 mm/yr. Slip rates for fault sections north and south of the Thousand Springs section are not well known but are possibly <0.2 mm/yr based on geologic mapping and trench investigations (Scott et al., 1985; Olig et al., 1995; U.S. Geological Survey, 2018).

1983 Borah Peak Rupture

The 1983 Mw 6.9 Borah Peak rupture initiated at a depth of ∼16 km (Doser and Smith, 1985) near the structural boundary between the Thousand Springs and Mackay sections (Susong et al., 1990) and propagated unilaterally NW (Boatwright, 1985; Doser and Smith, 1985; Richins et al., 1987), resulting in 36 ± 3 km of surface rupture (Crone et al., 1987) (Fig. 1). Finite fault models suggest 1.5–2 m of dominantly dip-slip displacement occurred on a 40°–50° SW-dipping sub-planar normal fault (Doser and Smith, 1985; Stein and Barrientos, 1985; Barrientos et al., 1987). Surface faulting in the 1983 earthquake includes rupture of (1) the entire 24-km-long Thousand Springs section, (2) the southern 8 km of the ∼15-km-long Warm Springs section to the north, and (3) a fault that branches WNW from the range-front trace of the LRFZ at the northern end of the Thousand Springs section (herein, the Arentson Gulch fault after Bruhn et al., 1991) and continues into the Willow Creek Hills (Crone et al., 1987) (Fig. 2). The Willow Creek Hills are a prominent bedrock ridge on the hanging wall of the LRFZ that form an apparent long-lived structural complexity between the Thousand Springs and Warm Springs sections (Figs. 1 and 2). Bruhn et al. (1991) suggest that the Willow Creek Hills are structurally perched (hung-up) on a SSW-plunging subsurface bedrock ridge in the LRFZ that increases in width with depth. Only a 9° change in the strike of the range-front trace of the LRFZ occurs across the Willow Creek Hills (Crone et al., 1987) (Fig. 1). Aftershocks following the 1983 earthquake were most abundant near the Willow Creek Hills, eventually migrating to near the Warm Springs and Challis sections (Boatwright, 1985; Zollweg and Richins, 1985; Richins et al., 1987). More recent (2010–2018) moderate-magnitude seismicity distributed along the Challis section (Fig. S2 [footnote 1]) is likely part of a long-lived aftershock sequence (Pang et al., 2018) and highlights the area of increased Coulomb stress along the Challis section following the Borah Peak earthquake (Payne et al., 2004).

Surface rupture in the Borah Peak earthquake produced mostly continuous, SW-facing normal-fault scarps with a minor (∼17%) left-lateral component (Crone et al., 1987). Vertical displacements are typically ∼0.5–2 m and reach a maximum of 2.5–2.7 m near the center of the Thousand Springs section (near Doublespring Pass Road; Crone et al., 1987). Measurements of laterally offset cultural features (e.g., fences and roads) indicate 0.4–1.0 m of left lateral strike-slip displacement along the Thousand Springs section (Crone et al., 1987). Along the northernmost Thousand Springs section near the Arentson Gulch fault, complex faulting includes discontinuous normal and thrust faults (Crone et al., 1987) (Fig. 2). Displacement along the Arentson Gulch fault decreases to the northwest: the southeastern part of the fault has 0.8–1.6 m of vertical and 0.3–1.2 m of left lateral strike-slip displacement, compared to 0.3–0.6 m of vertical and 0.3–0.7 m of right lateral strike-slip displacement to the north (Crone et al., 1987). At the northern end of the Arentson Gulch fault (northern flank of the Willow Creek Hills), Crone et al. (1987) mapped zones of surface cracking and a single 0.1 m vertical displacement scarp (Fig. 2). About 2–3 km west, two isolated and <0.05-m-displacement scarps correspond with the southern termination of the Lone Pine fault (Fig. 1), a Quaternary active, NE-dipping normal fault west of the LRFZ (U.S. Geological Survey, 2018). Crone et al. (1987) suggest a possible nontectonic (shaking-related) origin for the 1983 Lone Pine scarps on account of the minor and isolated nature of the slip.

An important aspect of the 1983 rupture is the occurrence of prominent gaps in surface displacement observed along the range-front-bounding LRFZ (Crone et al., 1987). Northeast of the Arentson Gulch fault, the Willow Creek Hills abut the Lost River range front, coinciding with a 5-km-long portion of the range-front trace of the LRFZ that did not rupture in 1983 (kilometer marks 8–13; Fig. 2). Older, discontinuous scarps in this area (Crone et al., 1987) indicate previous ruptures of the LRFZ; however, the late Quaternary surfaces that are faulted have not been precisely dated. Northwest of the Willow Creek Hills, 1983 surface rupture resumes on the southeastern half of the Warm Springs section (kilometers 0–8; Fig. 2). Displacements along this section decrease from ∼0.5–1 m in the south to <0.2 m in the north, and only a few, small (<0.2 m) lateral offsets were measured (Crone et al., 1987). A nearly 2 km gap in the 1983 surface rupture occurs at the Goosebury graben in a zone of prominent prehistoric scarps (kilometers 3–5, Fig. 2).

Paleoseismic Data

Paleoseismic data for the northern LRFZ are from four paleoseismic trench sites and a natural exposure.

Trenches excavated near Doublespring Pass Road (DP; Fig. 1) exposed evidence of at least one paleoearthquake on the Thousand Springs section. An initial trench excavated in 1976 by Hait and Scott (1978) found evidence for a single paleoearthquake expressed in alluvial-fan deposits postdating the most recent (Pinedale) glaciation (ca. 15 ka; see summaries by Crone, 1985; Schwartz and Crone, 1985; and Haller and Crone, 2004). A second trench, excavated immediately adjacent to the Hait and Scott (1978) trench following the 1983 earthquake, showed that 1983 rupture displacement at the site (1.7–2.0 m) had a similar magnitude compared to prehistoric displacement at the site (1.3–1.5 m) (Schwartz and Crone, 1985). Neither trench yielded age constraints for the paleoearthquake.

Stratigraphic evidence and geochronological results from an additional trench and natural exposure suggest that the Thousand Springs paleoearthquake may have occurred in the early Holocene. About 3 km northwest of Doublespring Pass Road, the Poison Spring (PS; Fig. 1) trench exposed faulted bedrock, hillslope colluvium, and scarp-derived colluvium from a prehistoric earthquake (Vincent, 1995). Charcoal from the uppermost part of the scarp-colluvial unit yielded an age of 9045 ± 100 14C yr B.P. that provides a minimum constraint on the timing of the prehistoric rupture. A natural exposure at Elkhorn Creek (EC; Fig. 1) provides additional evidence of the prehistoric rupture of the Thousand Springs section (Vincent, 1995). At this site, a soil A horizon formed in alluvial-fan sediments predates a prehistoric fault scarp; charcoal from the soil dated to 9710 ± 240 14C yr B.P. provides a maximum limit on the timing of the prehistoric rupture. Vincent (1995) used a similar pattern of surface displacement decreasing from the center (Doublespring Pass and Poison Spring) to southern (Elkhorn Creek) parts of the Thousand Spring section to infer that the Poison Spring and Elkhorn Creek sites record the same paleoearthquake. As a result, Vincent (1995) used the Poison Springs and Elkhorn Creek radiocarbon ages to bracket the timing of the prehistoric Thousand Springs rupture to between 9045 and 9710 14C yr B.P., or 10.2–11.1 ka, calendar calibrated here using OxCal (version 4.3; Bronk Ramsey, 2009).

Two trenches across the Warm Springs section (Fig. 1) provide evidence of a ∼mid-Holocene earthquake. Schwartz and Crone (1988) excavated trenches across the section at two locations: an unnamed drainage ∼2 km SE of McGowan Creek (the Rattlesnake Canyon trench, RC; Fig. 2) and an unnamed drainage within the northern part of the Willow Creek Hills near Sheep Creek (herein, the Sheep Creek trench, SC; Fig. 2). Although the RC and SC trench observations are unpublished, a summary by Schwartz and Crone (1988) indicates similar stratigraphic relations and ages. Charcoal derived from burn horizons within basal scarp-derived colluvium at the sites yielded radiocarbon ages of 4940 ± 200 14C yr B.P. (RC trench) and 5280 ± 240 14C yr B.P. (SC trench). Schwartz and Crone (1988) used the distance between the sites (∼7.5 km) and degree of soil development to infer that the trenches record the same paleoearthquake on the Warm Springs section, calendar calibrated here to shortly before 5.1–6.6 ka. Vertical displacement estimates are 0.75 m at the Rattlesnake Canyon trench and ∼2.2 m at the Sheep Creek trench (Schwartz, written communication, 2016).

Although constraints on event timing are sparse, we consider it most likely that earthquakes occurred at ca. 10–11 ka on the Thousand Springs section and shortly before ca. 5–7 ka on the Warm Springs section.

METHODS

To investigate surface displacement along the LRFZ, we acquired low-altitude aerial photographs using UAS and a tethered balloon in May 2015 (Warm Springs section; Fig. 2) and May 2016 (Willow Creek Hills area, including the northern Thousand Springs section and Arentson Gulch fault; Fig. 2). Using image-based (structure-from-motion) modeling, we generated 5–10-cm-pixel DSMs for the northern 16 km of the 1983 rupture (Fig. 4), georeferenced using ground control from kinetic differential global navigation satellite system (GNSS) measurements of survey targets. We measured the vertical separation (VS; Fig. 3) of geomorphic surfaces displaced by the LRFZ using topographic profiles extracted from the DSMs (e.g., Fig. 5; Fig. S1 [footnote 1]).

Digital Surface Models

We generated DSMs from photographs acquired from low altitudes (mostly <100 m) using UAS and a Helikite tethered balloon during field campaigns in 2015 and 2016. UAS platforms and sensors included (1) a DJI Phantom that we customized to carry a Sony a5100 camera and 16mm prime lens, (2) a Falcon fixed wing with a Sony a5100 camera and 20mm prime lens, and (3) a 3DR Solo with GoPro Hero4 Black (∼3mm fisheye) camera. A Canon SX230 compact camera set at a focal length of 28 mm was mounted on the Helikite. We used the Phantom and Helikite during the 2015 survey and the Phantom, Falcon, and Solo during the 2016 survey.

For geographic control, we placed ground-control targets throughout the survey area. Targets consisted of ∼0.3-m-diameter bucket lids, ∼0.5-m-wide mesh construction flags, and, most commonly, ∼1.5-m-square black and white (iron-cross pattern) vinyl targets. For our ∼11.9 km2 acquisition area, we placed 186 control points, including ∼28 control points/km2 along the Warm Springs section (∼4.6 km2) and ∼8 control points/km2 near the Willow Creek Hills (∼7.6 km2). Target locations were measured in the field using kinetic differential GNSS systems and local geodetic-quality base stations (Trimble 5700 receiver with Zephyr Geodetic I antenna) that were postprocessed in the NAD83 (2011) reference frame using the National Geodetic Survey’s Online Positioning User Service (OPUS; National Oceanic and Atmospheric Administration, 2018). Control-point positions were transformed into NAD83 (2011) UTM zone 12 and orthometric heights using GEOID12A.

We used Agisoft Photoscan image-based photogrammetric modeling software (versions 1.2.6–1.4.3) to generate two separate point clouds and DSMs (available at Open Topography, http://opentopo.sdsc.edu/datasets; Bunds et al., 2019), one for the Warm Springs section (2015 data) and one for the Willow Creek Hills area (2016 data). Image-based modeling uses feature recognition and photo alignment to generate a point cloud for a 3-D surface (Matthews, 2008; Harwin and Lucieer, 2012; Bemis et al., 2014; Johnson et al., 2014; Reitman et al., 2015). Our DSM workflow (Supplemental Text S12) consisted of the following steps. The surveyed areas were initially subdivided into 18 subregions. In Photoscan, for each subregion we (1) added photographs then generated a sparse point cloud and aligned cameras using Photoscan’s highest accuracy setting, (2) manually placed markers on ground-control target centers that were recognized in individual photographs, (3) used the OPUS-corrected differential GNSS data to provide spatial control for the markers, (4) for 2016 data, removed sparse cloud points with large uncertainty via Photoscan’s gradual selection tools, and (5) optimized cameras (least squares bundle adjustment). We merged 12 subregions that form the surveyed portion of the Warm Springs section in Photoscan, re-optimized the camera models, and built a dense point cloud using high quality and aggressive depth filtering. Six subregions of the Willow Creek Hills area were combined and processed in the same fashion as the Warm Springs section data except the merged sparse point cloud was cleaned of high uncertainty points. We did not automatically classify the dense point cloud because in some cases steep topography and locally heavy vegetation prevented imaging of the ground surface. The DSM of the Warm Springs section was generated in Photoscan from a mesh built in Photoscan. The DSM of the Willow Creek Hills area was generated in Photoscan using its DEM generation tool (inverse-distance weighting method). The merged DSMs (Bunds et al., 2019) are 10-cm-pixel resolutions. We used GIS software to visualize the DSMs, create slopeshade and hillshade maps, and map fault traces and surficial geology.

Error in our DSMs results from a complex interplay of factors such as camera exposure settings, lens specifications, camera calibrations applied, flight design (e.g., flight-line geometry and altitude), photograph overlap, sensor dimensions and ground sample distance, ground-control target placement and GNSS measurement, and effects from vegetation and topography. We applied a large number of strategies to reduce error, most important of which were (1) a dense network of ground-control targets, (2) placement of targets on relatively flat ground away from vegetation to minimize any vertical and/or horizontal movement during use, (3) use of high-resolution cameras, fast shutter speeds (>1/1000 s), and camera mounts that did not maintain perfect nadir camera direction, (4) varied flight altitudes, (5) GNSS base station occupations exceeding 8 h, and (6) application of best practices for processing in Agisoft Photoscan (Matthews, 2008; James and Robson, 2012; Bunds et al., 2015). We assessed DSM errors using checkpoints and methods similar to those specified for U.S. Geological Survey Q1 airborne lidar (Heidemann, 2018). Checkpoints are differential GNSS location measurements taken on bare, relatively flat ground across the surveyed area. We measured 82 checkpoints on the DSM of the Warm Springs section, and 28 on the Arentson Gulch fault and northern Thousand Springs section. The misfit in elevation of the checkpoints relative to the DSMs was determined using GIS software, and root-mean-square-error (RMSE) was calculated for the checkpoints. RMSE for the Warm Springs section DSM is 6.4 cm and 5.8 cm for the Arentson Gulch fault and northern Thousand Springs section. DSM precision across distances of meters to tens of meters (i.e., common topographic profile lengths) is likely better than the overall RMSE for the models as the fine topography of features such as cobbles, boulders, and small plants are well resolved in the DSMs.

Vertical Separation Measurements

We used topographic profiles extracted from the DSMs to estimate the vertical separation (VS) of geomorphic surfaces displaced by the Warm Springs section (n = 133) and northern Thousand Springs section (n = 119), including the Arentson Gulch fault (Fig. S1 [footnote 1] and Text S2 [footnote 2]). Profile locations were manually chosen and drawn on hillshade and slopeshade maps using GIS software and extracted from the DSMs using Quick Terrain Modeler. The profile lengths are mostly <25 m long for recent and steep 1983 scarps and <300 m long for the larger and more eroded, prehistoric compound scarps. For some 1983 scarps, short (<10-m-long) profile lengths reflect the limited extent of correlative surfaces across the rupture (e.g., Figs. 3 and 5). We found that randomly generated profiles or strictly imposed sample-distance intervals resulted in poorly sited measurements and introduced unnecessarily inflated error estimates. With these limitations in mind, we subjectively chose profile locations across correlative surfaces with the best geologic preservation and sparsest vegetation to ensure the highest-quality measurements.

For each profile, we measured VS using a Matlab script (Supplemental Code S13 and Text S3 [footnote 2]) and the following process. For each topographic profile, linear approximations of the footwall and hanging-wall geomorphic surfaces were projected to the midpoint of the scarp face; VS is the difference in elevation between the surfaces, measured at the scarp midpoint (Caskey, 1995) (Fig. 5). For profiles with heavy vegetation, we manually selected and removed profile points above the ground surface prior to defining footwall and hanging-wall surface projections. Because footwall fault exposures (preserved free faces) along the 1983 rupture are rare, we calculate VS rather than fault throw, which requires assumptions about near-surface fault dip (e.g., Johnson et al., 2018). Although slope steepness and fault dip can enhance the discrepancy between actual fault throw and measured VS (e.g., Mackenzie and Elliott, 2017), our VS measurements are likely ∼80%–90% or greater of throw because steep (60°–90°; Crone et al., 1987) near-surface faults dip in the same direction as moderately (<15°) sloping geomorphic surfaces (Caskey, 1995). Differences between VS and throw along strike are small relative to along-strike changes in the VS and throw values themselves, making our VS measurements valid for evaluating along-strike surface-rupture behavior. Our VS measurements do not account for lateral slip and the possibility of two-dimensional profiles sampling non-correlative surfaces because of laterally offset geomorphic features (e.g., Mackenzie and Elliott, 2017). However, as a whole, we expect the influence of lateral slip on VS to be minimal because of the minor (∼17%) strike-slip component, steep fault dips, and moderate and laterally continuous surface slopes.

For complex fault rupture patterns (e.g., multiple synthetic and/or antithetic traces), we adopted two approaches to measuring VS. For narrow zones of faulting (less than a few tens of meters wide), we accounted for complex faulting (e.g., graben formation) by selecting upper and lower far-field points outside of the fault zone(s). For complex scarps, such as those consisting of multiple synthetic scarps and/or antithetic scarps several tens of meters or more from the primary scarp, we subdivided the profile into shorter segments and summed the VS results. For compound scarps consisting of both 1983 and prehistoric rupture, we used short profiles (generally <10 m in length) to reconstruct 1983 VS, and longer profiles (tens of meters in length or more) to extract the cumulative (prehistoric and 1983) VS.

To account for uncertainties in our measurements, we conducted multiple (typically five to seven) VS-measurement iterations for each profile. Each iteration used a different subset of the original ground-elevation points to define alternative upper and lower far-field surface projections, thereby standardizing and facilitating multiple geomorphic interpretations, both factors that control measurement error (e.g., Salisbury et al., 2015). This method is especially useful for non-bare-earth DSMs, where geomorphic surfaces adjacent to the rupture can be partially obscured by vegetation. We use the VS results from these iterations to define minimum and maximum VS as well as a subjective best-fit preferred VS selected by assessing the linear fits to the geomorphic surfaces and the results of the multiple measurement iterations. We assume that the minimum and maximum VS bounds capture the range of possible VS measurements (e.g., similar to Scharer et al., 2014 and Gold et al., 2015).

The precision of our VS measurements is limited by DSM resolution, surface texture, and vegetation height and density. For most of the 1983 rupture, we were able to resolve displacements as small as ∼0.2 m. Relatively smooth ground surfaces and a predominance of shortgrass vegetation allowed us to resolve ∼0.1 m displacements along the Warm Springs section (northernmost extent of the 1983 rupture). VS values for compound scarps generally exceed these lower limits.

Along-Strike Displacement Curves

We constructed along-strike displacement curves using our VS data (Figs. 6 and 7). We projected all VS data to a simplified fault trace with kilometer tick marks shown on Figure 2. For the Arentson Gulch fault, we also project VS measurements to an 8-km-long simplified trace, oriented WNW-ESE (Fig. 2). The mean along-strike displacement curve represents a 0.5 km moving average applied to our preferred estimates of VS, linearly interpolated at 50 m intervals. These parameters yield displacement curves that strike a balance between short-wavelength, hectometer-scale variability and longer-wavelength, kilometer-scale trends. To define the uncertainty bounds in the curves, we applied the same linear interpolation and moving average methods to the minimum and maximum VS values from each scarp profile. Thus, our uncertainty ranges approximate the uncertainty in the moving average, rather than the full range of possible VS values. Mean VS values (and min-max values in parentheses) reported below for different ruptures or fault sections are based on the displacement curves, rather than the individual observations, which are unequally distributed along the fault trace.

DISPLACEMENT ALONG THE 1983 BORAH PEAK RUPTURE

1983 Warm Springs Rupture

The Borah Peak earthquake ruptured ∼8 km of the Warm Springs section north of the Willow Creek Hills (kilometers 0–8; Fig. 4A). We document rupture both north and south of the Goosebury graben (kilometers 3–5; Fig. 4A), an ∼2-km-long by ∼0.1–0.3-km-wide zone of synthetic and antithetic faulting. Crone et al. (1987) reported only minor 1983 rupture of the northernmost part of the graben, including 0.03 m of synthetic displacement and 0.05 m of antithetic displacement. Mean VS for the Warm Springs rupture is 0.3 (0.2–0.4) m including the displacement gap along the Goosebury graben (single, 8-km-long fault trace), or 0.4 (0.3–0.5) m, excluding the graben (two fault traces totaling 6.5 km in length).

The 1983 rupture is most prominent south of the Goosebury graben (kilometers 5–8; Fig. 4A). Here, the rupture has a continuous, sinuous trace for ∼3 km––a geometry that follows prehistoric fault scarps along the Warm Springs section and thus likely mimics previous ruptures (discussed below) (Fig. 4A). VS measurements have considerable scatter (∼0.2–1 m) but exhibit a clear displacement peak (∼0.7–1.0 m) ∼1–2 km north of the Willow Creek Hills (kilometers 7.0–7.5; Fig. 6A). VS decreases relatively steeply to the south, toward the northern flank of the Willow Creek Hills, and more gently to the north, toward the Goosebury graben. Our VS measurements broadly agree with those made by Crone et al. (1987) (Fig. 6A). Only a single displacement recorded by Crone et al. (1987) is outside the range of our nearby VS data. At kilometer ∼6.2 (Fig. 6A), Crone et al. (1987) measured ∼1 m of 1983 displacement on a compound scarp, whereas we report ∼0.5–0.7 m of VS for six profiles within ∼0.1 km of this original measurement. It is possible that our estimates are minima or that the Crone et al. (1987) measurement includes a minor component of prehistoric displacement.

The 1983 rupture south of the graben includes a previously unrecognized, ∼1-km-long zone of distributed splay faulting that continues 0.3–0.5 km into the hanging wall of the range-front trace of the LRFZ (kilometer ∼5; Fig. 4A). The northern limit of this zone of distributed scarps is unknown as it falls within a gap in DSM coverage, but it is likely <0.4 km northwest of scarps mapped in this study as we did not observe similar faulting in DSMs southwest of the Goosebury graben (Fig. 4A). Displacement along the splays corresponds with a section of the LRFZ south of the Goosebury graben that did not rupture in 1983. VS measurements for the splay faults decrease from south (∼0.6–1.0 m) to north (∼0.2–0.3 m). These data form a cluster of VS observations at the southern margin of the Goosebury graben (kilometer ∼5.0–5.2; Fig. 6A), 0.2–0.4 km north of a previous observation made by Crone et al. (1987) (kilometer 5.4; Fig. 6A).

North of the Goosebury graben, the 1983 rupture has a nearly continuous and sublinear trace for at least 3 km (kilometers 0–3; Fig. 4A). Here, our preferred VS measurements yield a VS curve that is consistently ∼0.2–0.3 m (Fig. 6A); however, large uncertainties in the individual VS measurements allow a range of values from ∼0.1 to ∼0.6 m. Although the scatter in these data reflect the difficulty in using a non-bare-earth DSM to capture the small-displacement rupture tail, we have confidence in the results because the coherent along-strike VS signal is generally consistent with small, less than ∼0.3-m-high scarps observed in the field and measured by Crone et al. (1987).

Our VS measurements for the rupture of the Warm Springs section north of the Goosebury graben generally exceed by as much as ∼0.2 m those made by Crone et al. (1987) (Fig. 6A). Although unrecognized prehistoric scarps along this part of the Warm Springs section could account for the VS disparity, we suspect that differences in scarp-measurement methods explain our greater VS values. For some Crone et al. (1987) measurements, a tape measure was used to measure the vertical height of the scarp free face (J. Lienkaemper, written communication, 2018). Our profiles, which ranged from meters to tens of meters long, may have captured additional off-fault, but near-field distributed deformation and yielded larger VS values.

1983 Thousand Springs Rupture

Our analysis of the Thousand Springs section is focused on 7 km of the 1983 surface rupture along the section near the Willow Creek Hills (Fig. 2). The 1983 surface rupture in this area is complex, including three zones of faulting: (1) displacement along the range-front trace of the LRFZ at the base of the Lost River range, (2) the Arentson Gulch fault, which deviates to the WNW from the range-front fault at an ∼35° angle (Fig. 4B), and (3) a complex zone of E-W to ESE-WNW–trending faults between the range-front LRFZ and Arentson Gulch faults.

We imaged ∼1 km of the 1983 rupture along the range-front trace of the Thousand Springs section (kilometers ∼13.5–14.5; Fig. 4B). Here, 1983 scarps (∼1 m VS) are superimposed on large (∼5–8 m VS) scarps within latest Pleistocene glacial sediment (Pierce and Scott, 1982). For most of this 1 km section of rupture, VS is ∼0.8–1.0 m (∼0.5–1.5 m with scatter and uncertainties; Fig. S1 [footnote 1]), consistent with four previous observations of ∼0.8–1.2 m (Crone et al., 1987). VS decreases abruptly to the north to ∼0.2–0.3 m, ∼0.25 km north of where the E-W transfer faults and NW-SE range front faults intersect (Fig. 4B). We did not construct displacement curves for this part of the rupture because of the complex deformation and limited along-strike extent of our VS measurements.

The Arentson Gulch fault comprises an 8-km-long and <1-km-wide zone of en echelon, mostly south-facing fault scarps that nearly cross the Willow Creek Hills ∼3.5 km west of the range-front trace of the LRFZ (Fig. 2). The rupture trace is complex near its SE end, forming a discontinuous zone of faults that terminate ∼0.5 km west of the range-front LRFZ trace (Fig. 4B). The NW end is likely near the northern flank of the Willow Creek Hills (3 km NW of the northernmost scarps measured in this study), based on small (∼0.1–0.6 m VS) scarps measured by Crone et al. (1987) (Fig. 2). Although additional 1983 scarps roughly on trend with the Arentson Gulch fault are present ∼5 km NW of the Willow Creek Hills (Crone et al., 1987), we exclude these disconnected and small (mostly ∼0.05 m) displacement scarps in our 8 km length estimate as we cannot rule out a secondary (i.e., shaking-related) origin. VS along the Arentson Gulch fault increases toward the center of the fault, from ∼0.2–0.4 m near the SE end and ∼0.4–0.8 m near the NW end to ∼2.0 m near the center (Fig. 7A). These measurements are similar to, and in some cases, slightly (<0.5 m) larger than, previous estimates (Crone et al., 1987). Mean displacement for the Arentson Gulch fault is 0.7 (0.6–0.8) m.

Near the northern end of the 1983 rupture of the Thousand Springs section, mostly E-W–oriented fault scarps form a zone of faults that connect the Arentson Gulch and range-front-trace faults (kilometer ∼13.5; Fig. 4B). This ∼1-km-long zone includes three parallel traces, each having ∼0.1–0.4 m of VS (Fig. S1 [footnote 1]). A single fault trace that connects to the range-front trace of the Thousand Springs section has ∼0.5–1.0 m of displacement. Based on the sum of the VS measurements for the separate traces and the maximum VS for the single strand, these faults accommodate ∼1 m of down-to-the-south displacement. Although right-lateral faulting is possible given the orientation of the faults, we did not find evidence of laterally offset geomorphic features.

1983 Borah Peak Earthquake Mean Displacement

The mean vertical displacement for the entire Borah Peak rupture is important for estimating seismic moment release (discussed below) and placing our observations in the context of global regressions relating surface displacement to earthquake magnitude (e.g., Wells and Coppersmith, 1994). We estimate a mean vertical displacement for the rupture of 0.93 (0.76–1.10) m, based on displacement curves (0.5 km moving average) fit to linearly interpolated (50 m spacing) VS data reported in Crone et al. (1987) for the Thousand Springs section and our VS measurements for the Warms Springs section and Arentson Gulch fault, which are generally denser and slightly larger than those reported by Crone et al. (1987). Although Crone et al. (1987) did not report a mean displacement, we estimate a mean of 0.84 (0.67–1.01) m using displacement curves fit to their original VS data set for the entire rupture. Wesnousky (2008) reported 0.94 m based on linearly interpolated (>100 m spacing) data from Crone et al. (1987). We report a total length of 34.2 km, which reflects surface displacements measured at the northern extent of our study area near McGowan Creek to the southern rupture terminus as mapped by Crone et al. (1987) (Fig. 1). This length excludes an ∼1-km-long zone of surface cracks and a single 0.04 m displacement scarp ∼3–4 km northwest of McGowan Creek (Crone et al., 1987) as we cannot rule out a shaking related origin for these features.

PREHISTORIC RUPTURES OF THE LRFZ

Warm Springs Section

Our high-resolution DSMs allow detailed inspection of 1983 and prehistoric fault scarps along the Warm Springs section. With the exception of the Goosebury graben and the southernmost terminus of the Warm Springs section, the scarps are compound in nature with evidence of both the 1983 and at least two prehistoric ruptures (e.g., Figs. 3 and 5). In total, VS measurements for 34 prehistoric scarps range from 1.4 to 4.5 m (Fig. 6B). VS ranges are similar north of (1.4–3.2 m), south of (1.5–4.5 m), and within (1.4–4.1 m) of the Goosebury graben (Fig. 6B). Displacements along the Warm Springs section generally cluster within two groups: those ≤2 m (n = 8) and those >2 m, which are concentrated between ∼2.5 and 4.0 m (n = 26). Based on these displacement populations as well as our geomorphic mapping discussed below, we infer that at least two earthquakes prior to the 1983 earthquake ruptured the Warm Springs section since the latest Pleistocene (time of the last glacial maximum): PE2 (VS >2 m) and PE1 (VS ≤2 m).

At three areas along the Warm Springs section (kilometers 3, ∼3.5, and 8; Fig. 2), prehistoric scarps cut multiple geomorphic surfaces and support our interpretation of two previous earthquake ruptures (Fig. 4A). At these sites, the largest VS measurements (∼3–4 m) correspond with the stratigraphically oldest (highest elevation) alluvial-fan surfaces and likely include displacement from the 1983, PE2, and PE1 ruptures. Inset surfaces exhibit less displacement (generally, ≤2 m), and likely only include faulting from the 1983 and PE1 ruptures. For example, within the Goosebury graben (kilometers ∼3–3.5; Fig. 4A), the youngest faulted alluvial-fan surfaces have ∼1.4–2.1 m of VS, whereas the highest (oldest) surfaces are displaced ∼2.8–3.8 m (Figs. 8 and S3 [footnote 1]). Beyond the southern extent of the 1983 rupture along the Warm Springs section (kilometer ∼8; Fig. 4A), an inset fan surface has 2.0 m of VS compared to older surfaces that are displaced 2.7–3.5 m (Fig. S4).

We computed along-strike displacement curves for scarps having VS values of ≤2 m, including the 2.1 m VS measurement for the inset Goosebury graben surface (herein, ≤2 m curve) and >2 m (herein, >2 m curve) (Fig. 6B). The curves are cumulative and include the 1983 rupture displacement. Both curves have similar, broadly uniform along-strike shapes and barely overlap within their uncertainties. Although these curves reasonably approximate prehistoric displacement along the Warm Springs section, because we lack VS data for the northernmost part of the section, they may not capture the total variability or along-strike trend in prehistoric VS. The ≤2 m curve suggests a pattern of approximately uniform along-strike displacement and a mean VS of 1.7 m (1.3–1.9 m). The >2 m curve has a mean VS of 3.3 m (2.8–3.8 m), with peak displacement at the Goosebury graben.

Northern Thousand Springs Section

Along the northern Thousand Springs section, prehistoric scarps provide evidence of multiple prehistoric ruptures of the range-front trace and at least one previous rupture of the Arentson Gulch fault.

Displacement along the range-front trace of the Thousand Springs section decreases to the north near its northern intersection with the Willow Creek Hills (kilometer ∼13.2; Fig. 4B). Late Pleistocene(?) surfaces along the western flank of Dickey Peak (Fig. 2) record as much as 4.5–8.0 m of VS (this study; Fig. 4B). Displacement decreases to the north to ∼2.0–3.5 m, where a prominent prehistoric scarp lacks evidence of 1983 rupture displacement (kilometer ∼12.5; Fig. 4B) (Crone et al., 1987). This pattern of prehistoric displacement decreasing steeply toward the Willow Creek Hills is consistent with our VS measurements for the 1983 rupture. Discontinuous scarps to the northwest (east of the Willow Creek Hills; kilometers 8–13 km; Fig. 2) suggest prehistoric faulting without reactivation in 1983. These scarps are outside of our DSM data set and thus we were unable to measure their VS.

Along the Arentson Gulch fault, compound (including 1983) and single-event (prehistoric) scarps record at least one previous (prehistoric) rupture (Figs. 7B and 9; Fig. S5). Near Arentson Gulch (kilometer ∼5.5; Fig. 4B), single-event (1983) scarps that cross the lowermost (youngest) faulted alluvial-fan surface (T1) have 1.5–1.8 m of VS (Fig. 9). Compound scarps on older fan surfaces record VS of ∼1.9–3.5 m (n = 5), consistent with at least one previous rupture of the Arentson Gulch fault. Less than 0.1 km to the southwest, a subparallel compound scarp has 1.5–1.6 m of VS, compared to 1983 VS measurements that are ∼0.3–0.6 m. Near the northern extent of the Arentson Gulch fault within the Willow Creek Hills (kilometer ∼3; Fig. 4B), single-event (prehistoric) scarps have ∼0.2–0.5 m of VS (Fig. S5). A single-event (prehistoric) scarp near the southern terminus of the Arentson Gulch fault has 0.6 m of VS (Fig. 4B). We suspect that additional compound scarps northwest of Arentson Gulch may account for our slightly larger VS measurements than those of Crone et al. (1987). However, we were unable to resolve prehistoric scarps along this section of the Arentson Gulch fault because evidence of older, eroded prehistoric scarp crests are not preserved in the steeply sloping geomorphic surfaces.

We computed a cumulative along-strike displacement curve for the Arentson Gulch fault. The curve suggests peak VS of ∼3 m near Arentson Gulch, with displacement tapering to the north and south (Fig. 7B). The mean (cumulative) VS is 1.5 m (1.2–1.6 m).

Displacement Difference Curves

Prehistoric displacement curves for the Warm Springs (Fig. 6B) and northern Thousand Springs (Fig. 7B) sections show the cumulative displacement expressed along these faults. To evaluate plausible per-event displacement along the Warm Springs section, we differenced displacement curves for the 1983, PE1, and PE2 ruptures (Fig. 10). For each rupture, we differenced the mean VS curves, as well as the min-max curves separately to determine uncertainty bounds.

For the Warm Springs section, we compute two difference curves that represent the displacement in at least two prehistoric earthquakes (Fig. 10A). The per-event displacement profile for PE2 is the >2 m displacement curve minus the ≤2 m curve. The displacement for PE1 is the ≤2 m curve minus the 1983 displacement curve. Although more than two paleoearthquakes could be recorded in the prehistoric scarps, we interpret the PE2 and PE1 curves as each representing discrete events because the curves have similar along-strike shapes, with mean per-event displacement varying between ∼1 and 2 m (Fig. 10A), which is broadly similar to the 1983 rupture of the entire Thousand Springs section south of Willow Creek Hills. The PE2 curve has broader uncertainty because of more scatter in the >2 m data. The PE2 curve has an approximately uniform along-strike shape, in contrast with the PE1 mean curve that peaks at the Goosebury graben, where no net displacement was observed in 1983. Because we cannot rule out the possibility that graben faults reactivated in 1983 with distributed faulting that was not expressed or detectable at the surface, it is possible that the PE1 peak includes decimeter-scale displacement from the 1983 rupture. Mean displacement (VS) for PE2 is 1.6 m (0.8–2.5 m), compared to 1.4 m (0.9–1.7 m) for PE1.

To estimate the VS for the prehistoric rupture of the Arentson Gulch fault, we use single-event scarps at the northern and southern ends that record only a prehistoric event as well as compound scarps near Arentson Gulch (Fig. 7B). For the compound scarps, we subtracted the VS measurements for the closest 1–2 single-event (1983) scarps from the individual VS measurements for the compound scarps, yielding per-event VS values that range from 0.8 to 2.0 m (Fig. 10B). We then fit displacement curves to the corrected data, yielding a paleoearthquake curve with a mean VS of 0.9 m (0.6–1.0 m). This is similar to the 1983 displacement curve that yields a mean VS of 0.7 m (0.6–0.8 m). Our Arentson Gulch fault prehistoric displacement curve is constrained by 14 VS measurements and is similar in shape, length, and magnitude to that for the 1983 rupture.

MOMENT CALCULATIONS

We calculate seismic moment (M0) and moment accumulation rates (forumla) for the northern LRFZ (Table 1) to evaluate Warm Springs and Thousand Springs section ruptures in the context of M0 accumulation and release. The main questions we seek to address with these simple M0 calculations are (1) what percentage of the total 1983 M0 release occurred on the Warm Springs section and (2) how does M0 in the 1983 rupture of the Warm Springs section compare to the M0 of prehistoric ruptures of the Warm Springs section? We treat M0 calculations in the standard way (Aki, 1966), where M0 is estimated as the product of rupture area (based on rupture length and down-dip width), average coseismic fault-parallel displacement, and the shear modulus of the crust (3.3x1010 Pa; Doser and Smith, 1985). forumla is the product of rupture area, shear modulus, and fault-parallel slip rate (Brune, 1968). We use a fault dip of 45° (Doser and Smith, 1985; Richins et al., 1987) and, in most cases, a maximum seismogenic depth of 15.5 km (midpoint of 15–16 km maximum depth of mainshock and aftershocks reported in Doser and Smith, 1985 and Smith et al., 1985) to define down-dip rupture width (Table 1). Because our M0 estimates are generally used for comparative purposes, we do not account for complex uncertainties related to minor oblique slip, VS versus fault throw, and the assumption that surface displacements can be used to approximate average displacement over the rupture at seismogenic depths. M0 estimates are converted to earthquake magnitude (Mw) using the equation of Hanks and Kanamori (1979): 
graphic

Our observations yield a geologic M0 estimate for the total Borah Peak rupture that is similar to previous seismic and geodetic M0 estimates. Using a surface-rupture length of 34.2 km and an average displacement of 0.93 m extending from the surface to 15.5 km depth (although see Discussion for uncertainties regarding 1983 coseismic slip along the Warm Springs section), we estimate a geologic M0 of 3.3 × 1019 Nm (Mw 6.9) for the Borah Peak earthquake (Table 1). Seismic M0 (including main shock and 3.5 weeks of aftershocks; Smith et al., 1985) and geodetic M0 (based on a 70-km-long leveling line surveyed in 1933, 1948, and 1984; Stein and Barrientos, 1985) estimates are 2.9 × 1019 Nm (Mw 6.9) and 3.2 × 1019 Nm (Mw 7.0), respectively (Table 1). By comparison, the Thousand Springs section alone, excluding the Arentson Gulch fault, yields a M0 of 2.4 × 1019 Nm (Mw 6.9), assuming a length of 21.4 km and average displacement of 1.1 m. While moment calculations from average surface slip are only a rough approximation of slip at depth, these calculations underscore the simple observation that the bulk of moment in 1983 occurred south of the Willow Creek Hills.

To more closely examine the spatial distribution of M0 release north of the Willow Creek Hills in 1983, we find that M0 for the 1983 rupture of the Warm Springs section is 1.9 × 1018 Nm (Mw 6.1) based on our 1983 scarp mapping and VS data and assuming a seismogenic depth of the rupture of 10 km, consistent with the slip-inversion models of Ward and Barrientos (1986), Mendoza and Hartzell (1988), and Du et al. (1992). M0 estimates for the Arentson Gulch fault are poorly constrained because of uncertainty in its down-dip extent and intersection with the Thousand Springs section. Assuming that the Arentson Gulch fault lies entirely within the hanging wall of the LRFZ, and using the orientation of the two fault planes, we use two-thirds of the rupture area calculated using the 8 km length and 15.5 km seismogenic depth. Using an average displacement of 0.7 m (1983 rupture) results in a M0 of 3.7 × 1018 (Mw 6.3) for the 1983 rupture of the Arentson Gulch fault. In summary, we estimate that the Warm Springs section and the Arentson Gulch fault only contributed 9% and 11% of 1983 M0, respectively (Table 1).

To gain insight on prehistoric ruptures centered on and north of the Willow Creek Hills, we also estimate M0 for the Arentson Gulch fault and PE1 and PE2 along the Warm Springs section. For the Arentson Gulch fault, we use an average displacement of 0.9 m for the prehistoric rupture to obtain a M0 of 4.8 × 1018 (Mw 6.4). To estimate M0 for PE1 and PE2 along the 15-km-long Warm Springs section, we use along-strike displacement curves for PE1 and PE2. For PE1, we speculate that the rupture includes the range-front trace of the LRFZ through the Willow Creek Hills (discussed below) and increase the length by 2 km (total of 17 km). Because the displacement curves for PE1 and PE2 do not cover the entire length of the Warm Springs section, likely excluding a decrease in displacement toward the rupture ends, we apply a tapered-slip model for the Warm Springs section which arbitrarily reduces the mean displacement by 50% for the northern and southern 2 km of the ruptures, consistent with the displacement taper along the northern and southern ends of the Thousand Springs section. As a result, overall mean displacement is reduced from 1.4 m to 1.2 m for PE1 and 1.6 m to 1.4 m for PE2. M0 for PE1 and PE2 is 2.1 × 1019 Nm (Mw 6.8) and 2.2 × 1019 Nm (Mw 6.8), respectively.

forumla calculations for the Warm Springs section allow us to calculate the total M0 accumulated on the section since the most recent earthquake (PE1). We calculate a vertical slip rate for the section of ∼0.2 mm/yr using the average VS of PE2 scarps (∼3.3 m) and the approximate age of the displaced geomorphic surface (ca. 15 ka; Pierce and Scott, 1982). This rate is similar to vertical slip rates of 0.2–0.3 mm/yr calculated for the Thousand Springs section and yields a Warm Springs forumla of 3.1 × 1015 Nm/yr. The Warm Springs section M0 rate, as well as the M0 estimates for PE1 and PE2, suggest that the Warm Springs section is capable of generating large (Mw ∼6.8) earthquakes approximately every 7 k.y. Assuming that at least 5.1–6.6 k.y. have elapsed since earthquake PE1 and subtracting our M0 estimate for the 1983 rupture of the section, the net M0 accumulated on the Warm Springs section since PE1 is 1.4–1.8 × 1019 Nm, enough to generate a Mw 6.7–6.8 earthquake.

DISCUSSION

1983 Rupture of the Northern LRFZ

Surface rupture in the 1983 Borah Peak earthquake propagated northward from the southernmost Thousand Springs section toward the Willow Creek Hills (Crone et al., 1987; Richins et al., 1987). Our observations, based primarily on remapping of 1983 surface displacement, allow us to closely examine how slip was modulated by the Willow Creek Hills structural complexity in 1983.

Within ∼1 km of the Willow Creek Hills, displacement on the Thousand Springs section in the 1983 rupture decreased abruptly to the north and a complex pattern of surface faults transferred slip to the 8-km-long Arentson Gulch fault (Fig. 4B), which continues into and nearly through the Willow Creek Hills (Fig. 2). The rupture did not break the surface in the ∼5-km-long portion of the range-front trace of the LRFZ along the eastern edge of the Willow Creek Hills (Fig. 2). Surface rupture occurred along the southernmost Warm Springs section, with less displacement and a more discontinuous trace than along the Thousand Springs section (Fig. 4A). Although the direction of the Warm Springs rupture propagation is uncertain, it likely continued south to north, based on the unilateral rupture propagation inferred for the Borah Peak earthquake (e.g., Doser and Smith, 1985) and the northward-decreasing displacement tail observed north of the Goosebury graben (consistent with Ward, 1997).

Prehistoric Rupture of the Northern LRFZ

Our displacement measurements fall into two clear populations that allow us to infer that at least two prehistoric earthquakes (PE1 and PE2; Fig. 11A) have ruptured the Warm Springs section. PE2 likely occurred after last glacial maximum deglaciation (ca. 15 ka) as scarps are formed in extensive surfaces related to this period (Pierce and Scott, 1982). PE1 occurred shortly before ca. 5.1–6.6 ka based on radiocarbon ages from the Rattlesnake Canyon and Sheep Creek trenches (Schwartz and Crone, 1988). Although per-event displacement patterns for PE1 and PE2 are similar (Figs. 10A and 11B), peak displacement in PE1 at the Goosebury graben could explain the 1983 displacement minimum within this 2-km-long structure. Alternatively, the PE1 displacement peak could signal and include decimeter-scale displacement across the graben in 1983 that was not observable at the surface.

Although we document evidence for pre-1983 rupture of the Arentson Gulch fault (Fig. 10B), we lack information on displaced surface ages and thus cannot constrain the timing of the rupture. However, it is possible that the rupture is contemporaneous with the previous rupture of the Thousand Springs section, constrained to ca. 10–11 ka from paleoseismic trenches (Crone, 1985; Vincent, 1995).

Primary versus Secondary Origin of 1983 Scarps along the Warm Springs Section

A central motivation for this study is to help interpret the significance of the northern extent of the 1983 rupture pattern, addressing two related questions. First, is the Warm Springs rupture primary or secondary in nature? We define primary rupture as surface slip related to coseismic rupture propagation at seismogenic depths, whereas secondary surface slip may reflect a range of processes such as nonseismogenic gravitational failure, shallow triggered coseismic slip, or postseismic slip (Fig. 12). Second, how is normal-fault rupture arrested or modulated as it approaches a major trans-basin structural complexity such as the Willow Creek Hills?

Several lines of evidence point toward coseismic, tectonic slip as the driver for surficial displacement along the Warm Springs section. Continuous, moderate-displacement (∼0.2–1.0 m VS) scarps are locally superimposed on larger (∼1.5–4.5 m VS) prehistoric scarps (Fig. 4A; Fig. S1 [footnote 1]). Although the 1983 mean displacement is small (∼0.35 m VS), peak displacement near the Willow Creek Hills is significant (∼0.7–1.0 m VS), similar to coseismic displacement observed along the Thousand Springs section and Arentson Gulch fault (Fig. 11). A shallow, shaking-related origin of displacements along the Warm Springs section is unlikely as there is a lack of complex, isolated, arcuate, and/or compressional scarps along the main fault trace and especially within the hanging wall, that might be expected if only shaking-induced gravitational failure or consolidation of near-surface deposits had occurred (e.g., Caskey et al., 1996). Instead, the 1983 rupture appears to accurately track the prehistoric rupture trace, just with less displacement and possibly in more discontinuous fashion than prehistoric ruptures.

As a result of the geometrically simple fault pattern and moderate displacements, Crone et al. (1987) and Crone and Haller (1991) concluded that the rupture of the Warm Springs section is not directly related to the primary rupture on the Thousand Springs section, but may be the result of insignificant, shallow faulting triggered by strong shaking and the directivity of the primary rupture (Fig. 12A). Slip inversions using teleseismic waveforms and geodetic observations have poor resolution along the Warm Springs section, but these models suggest that slip in the 1983 Warm Springs rupture occurred at least to depths of ∼5 km (Ward and Barrientos, 1986) or ∼8–10 km (Mendoza and Hartzell, 1988; Du et al., 1992). Predictions of deep slip in the finite fault models and decimeter-scale surface displacements observed along the Warm Springs section make it unlikely that surface displacement along the Warm Springs section reflects secondary, dynamically triggered shallow slip. Triggered slip would likely produce shallow, sub-centimeter displacements (e.g., Rymer, 2000; Wei et al., 2011) that would likely be undetectable at the surface.

Although the 1983 surface displacement of the Warm Springs section is likely deeply (at least 5–10 km) rooted (Fig. 12B), it is possible that some portion of the observed displacement represents postseismic slip adjacent to coseismic slip patches on the Thousand Springs section and Arentson Gulch fault (Fig. 12C). Postseismic slip is commonly observed following large earthquakes, including normal fault ruptures (e.g., D’Agostino et al., 2012). Unfortunately, we lack the field and remote sensing observations necessary to evaluate a postseismic slip signal, especially since the 1983 Warm Springs section scarps were not visited in the field on the day of the earthquake (Ostenaa et al., 1984). Postseismic slip that continued after the Crone et al. (1987) measurements along 1983 Warm Springs section could explain some of the greater values we obtain for VS measurements for the northernmost part of the rupture as the Crone et al. (1987) measurements were made between a day to several years following the earthquake (A.J. Crone, written communication, 2018). However, this explanation for the VS difference is unlikely for those field measurements obtained years after the rupture, as rates of postseismic slip typically decay within days to months following the main shock (e.g., Cheloni et al., 2014; Gualandi et al., 2014; DeLong et al., 2016). Average and maximum displacements for the 1983 rupture of the Warm Springs section are both <40% of those for the Thousand Springs section rupture, but global postseismic-coseismic slip relations show that the ratio of coseismic to postseismic slip can vary substantially: postseismic slip (or M0) is most commonly ∼10%–30% or less of coseismic slip (Wdowinski et al., 1997, Reilinger et al., 2000; Ozawa et al., 2012; Lin et al., 2013), but in some earthquakes exceeds 30% (Wang et al., 2015) or could be as much as 100% (DeLong et al., 2016) of coseismic slip. However, the magnitude of 1983 Warm Springs section displacement (maximum of ∼1 m) is less consistent with postseismic slip observations, which are mostly 0.1–0.5 m (Wdowinski et al., 1997; Reilinger et al., 2000; Cheloni et al., 2014; DeLong et al., 2016). In summary, we cannot rule out the possibility that postseismic slip contributed to the observed surface displacement on the Warm Springs section.

Given the unknown postseismic slip contribution to 1983 scarps along the Warm Springs section, we prefer the interpretation that primary coseismic slip is the main cause of these features (Fig. 12B). This is the simplest explanation and is consistent with (1) recognition of the scarps within about a day of the Thousand Springs section rupture, (2) the magnitude of the surface displacement, and (3) the northward-decreasing displacement profile, which is suggestive of a gradual (if discontinuous) taper in displacement at the primary rupture terminus (e.g., similar to rupture tails for the Dixie Valley, Nevada, USA [Caskey et al., 1996], and Edgecumbe, New Zealand [Beanland et al., 1989] earthquakes and models of [Ward 1997]). Further, aftershocks occurring within 24 h of the 1983 mainshock are distributed along the entire length of the Thousand Springs and Warm Springs section ruptures (Richins et al., 1987), suggesting the continuation of mainshock slip north of the Willow Creek Hills for at least 8 km. We note that none of these criteria are absolute and all allow some combination of primary coseismic slip plus secondary afterslip as the origin of the Warm Springs scarps.

Ruptures prior to 1983 along the Warm Springs Section

We combine our vertical separation observations with M0 and forumla estimates to compare the 1983 and prehistoric ruptures of the Warm Springs section and place the 1983 rupture in the context of prior ruptures. We find that the 1983 Warm Springs rupture is fundamentally different from prehistoric earthquakes PE1 and PE2 (Fig. 11B). Our M0 estimate for the 1983 Warm Springs rupture is 1.9 × 1018 Nm (equivalent to a Mw 6.1 earthquake), or ∼9% of the potential M0 released in PE1 and PE2 (2.1–2.2 × 1018 Nm; Mw ∼6.8 earthquakes) (Table 1). This implies that even though the 1983 earthquake ruptured ∼50% of the Warm Springs section, the moment release is minor compared to possible previous ruptures of the section. If we assume a simplistic earthquake cycle with similar strain accumulation and release, we can calculate the Warm Springs section forumla of 3.1 × 1015 Nm/yr and an elapsed time since PE1 of at least 5.1–6.6 k.y. to infer that the M0 accumulation on the Warm Springs section at the time of the Borah Peak earthquake was ∼1.6–2.0 × 1019 Nm, or about ∼74%–95% of the M0 release estimated in PE1 and PE2 (2.1–2.2 × 1019 Nm). This suggests that the 1983 Warm Springs rupture released only ∼9%–12% of this accumulated M0 (Table 1), delaying the approximate 7 k.y. Warm Springs clock (based on the section’s forumla and prehistoric M0 release) by ∼0.6 k.y.

Only a minor percentage of the total M0 release in the 1983 Borah Peak earthquake occurred along the Warm Springs section. Although the 1983 Warm Springs rupture contributed 24% to the total Borah Peak rupture length (8 of 34 km total), the M0 for the rupture is only 6% of our estimate for the entire Borah Peak earthquake (Mw 6.9) (Table 1). In contrast, M0 release on the Warm Springs section in PE1 and PE2 was ∼65% that of the Borah Peak earthquake. In relative terms, the 1983 rupture of the Thousand Springs section (Mw 6.9) released ∼13 times the M0 for the 1983 Warm Springs rupture, or ∼75% of the total Borah Peak M0. Taken together, the surface faulting pattern, displacement profile, and moment comparisons to prehistoric events suggest that the 1983 Warm Springs rupture is the moderate displacement continuation of a rupture whose propagation energy dissipated during complex, branch-fault rupture into the Willow Creek Hills structure.

Rupture Modulation at the Willow Creek Hills

The earliest post-1983 earthquake studies recognized that 1983 Borah Peak rupture propagation was profoundly affected by a structural discontinuity at the Willow Creek Hills (Crone et al., 1985), and this was interpreted as a rupture “barrier” in the emerging terminology of the day, which focused on fault segmentation and characteristic earthquake models (Schwartz and Coppersmith, 1984). The observations we present here show that per-event and cumulative displacement data for the northern LRFZ suggest a complex history of rupture at and through the Willow Creek Hills. Multisegment ruptures during historical earthquakes on strike slip faults (e.g., Sieh et al., 1993; Treiman et al., 2002; Klinger et al., 2005; Oskin et al., 2012; Hamling et al., 2017), complex prehistoric slip histories across historical rupture terminations in subduction settings (e.g., Shennan et al., 2009; Briggs et al., 2014), and prehistoric “spillover” ruptures along the Wasatch normal fault zone (Personius et al., 2012; DuRoss et al., 2016) are evidence of variable rupture response to along-strike structural complexities in several tectonic settings. Such features have been recently termed “rupture gates,” defined as fault complexities that can act as a conditional barrier to rupture propagation “as a result of proximal fault geometry, rupture direction, and prior earthquake history” (Oskin et al., 2015).

Neither the segment boundary nor earthquake gate analogies are perfect, because the 1983 rupture was not fully arrested by the Willow Creek Hills, but instead the rupture style and slip amount were significantly modulated (Fig. 13). That is, the Borah Peak earthquake did not pass through the Willow Creek Hills without penalty; rupture continued, but without sufficient energy to initiate a full (displacement and length) rupture of the Warm Springs section. There are several plausible reasons for modulation of rupture at a structural complexity. First, complex, interlocking faults within a structural barrier have been shown to impede rupture progress (King, 1983; King and Nabelek, 1985; Bruhn et al., 1987). Limited rupture north of the Willow Creek Hills in 1983 may also relate to the Warm Springs section’s prior stress conditions; it was further from failure than the Thousand Springs section because of its more recent paleoearthquake. Other possibilities are the interruption of dynamic rupture propagation by varying fault geometry (Ando et al., 2017), the reduction in rupture velocity approaching a fault branch (Templeton et al., 2010), and energy loss to off-fault deformation in a damaged zone (Andrews, 2005). Regardless, the Warm Springs rupture is best described as a spillover rupture (DuRoss et al., 2016), or rupture across a “leaky” boundary (Crone and Haller, 1991), which added significant length to the rupture, but only minor displacement and moment.

It is useful to place the 1983 rupture in the context of prehistoric ruptures to appreciate the range of slip behaviors observed across the Willow Creek Hills. Paleoearthquakes PE1 and PE2 resulted in a significantly different pattern of displacement compared to the 1983 rupture (Fig. 11). These earthquakes may have spanned the entire ∼15 km length of the Warm Springs section based on ∼2–3 m VS scarps identified at the northern edge of our study area. We infer that the PE1 rupture continued south into the Willow Creek Hills based on displacements measured at the Sheep Creek trench (Schwartz and Crone, 1988). If our along-strike correlation of PE1 with the >5–7 ka Warm Springs rupture (Schwartz and Crone, 1988) is correct, PE1 displacement may increase toward the structure, from ∼1–1.5 m to ∼2.2 m (Fig. 11B). Thus, we speculate that the PE1 ruptured the Willow Creek Hills and possibly the northernmost Thousand Springs section. The ∼2 m VS scarp on the northernmost Thousand Springs section, east of the Willow Creek Hills, that did not break in 1983 (kilometer ∼12.5; Figs. 2 and 4B) could be related to earthquake PE1. A mid-Holocene prehistoric rupture of the Willow Creek Hills may partly explain why the 1983 rupture did not continue northwest along the range-front trace of the fault, but instead took a complicated path along the Arentson Gulch fault. We suspect that PE2 terminated at or just north of the Willow Creek Hills (e.g., kilometer ∼8–9; Figs. 2 and 11B), based on an apparent decrease in PE2 displacement from ∼1.5–2 m in the north to <1 m toward the Willow Creek Hills and the lack of evidence for PE2 in the Sheep Creek trench. However, it is possible that the Sheep Creek trench did not expose strata predating and displaced by the PE2 event. Although ruptures PE2 and PE1 may have influenced the northward propagation of Thousand Springs ruptures (including 1983), an explanation for their southern extents remains unresolved.

Rupture into and through the Willow Creek Hills may depend on several factors, including rupture propagation direction, displacement magnitude, and the history of past strain release on the Thousand Springs and Warm Springs sections (Fig. 13). PE1 and PE2 could have ruptured into, and possibly through the structure, but PE2 was not observed at the Sheep Creek trench (southernmost Warm Springs section) and the southern terminus of PE1 is only constrained by its absence in the Doublespring Pass trench (no ruptures younger than ca. 10–11 ka on the central Thousand Springs section). However, peak displacement in PE1 at the Willow Creek Hills could be evidence of rupture penetration, and we speculate, north to south propagation, into the Willow Creek Hills. In contrast, prehistoric ruptures of the Thousand Springs section (e.g., TS PE; Fig. 13) may have stopped short of the Willow Creek Hills based on the clear decrease in scarp VS along the Thousand Springs section as it nears the structure (Fig. 11). Prehistoric scarps along the Arentson Gulch fault suggest at least one prehistoric rupture of the branch fault with an along-strike displacement pattern similar to that for the 1983 rupture of the fault. We speculate that reactivation of the Arentson Gulch fault is more likely during northward propagating ruptures on the Thousand Springs section based on its proximity to the Arentson Gulch fault trace and the geometric configuration more favorable for simple branching. However, this geometry does not preclude simultaneous rupture of the Arentson Gulch fault and Warm Springs section, which could depend on unresolved structural relationships between the two at depth. Using the 1983 and possible prehistoric ruptures, it appears that the Willow Creek Hills structural complexity is not a hard barrier to rupture, but effectively modulates or limits the displacement in ruptures breaching the structure (Fig. 13). This is consistent with the conclusion of Bruhn et al. (1991) that prehistoric scarps along the eastern margin of the Willow Creek Hills (along the range-front trace of the LRFZ) are evidence that the structure acts as a nonpersistent (after Wheeler and Krystinik, 1992) rupture barrier with only occasional full-displacement, throughgoing ruptures (Crone et al., 1987).

From our observations of 1983 and previous surface-rupturing earthquakes, we speculate that ruptures through the Willow Creek Hills contribute only minor moment and cumulative vertical separation in the context of large ruptures to the north and south (Table 1; Fig. 13). There is only limited evidence for throughgoing ruptures with significant displacement and moment. Additional paleoseismic data for the southern Warm Springs and northern Thousand Springs sections would serve to refine the ages and displacements of prehistoric ruptures of these sections and further test the models of rupture at and through the Willow Creek Hills suggested here.

Analogs to the Willow Creek Hills can be found along the Wasatch fault zone (WFZ), Utah, USA, where paleoseismic data provide evidence of ruptures modulated by long-lived normal-fault structural complexities. For example, near the center of the Holocene-active part of the WFZ, a prominent hanging-wall bedrock ridge and fault bend separate the Salt Lake City and Provo sections and form the Traverse Mountains structural boundary (Bruhn et al., 1992; Machette et al., 1992). Although a complex, distributed zone of fault scarps connect the segments (Toké et al., 2017), similar to faults at the eastern edge of the Willow Creek Hills, this structure has previously been considered a hard barrier to rupture (e.g., Schwartz and Coppersmith, 1984; Machette et al., 1992; Wheeler and Krystinik, 1992). Paleoseismic data from near this structure indicate that it has also impeded some but not all surface-faulting earthquakes and may also be a source of <30-km-long ruptures centered near the structure (Bennett et al., 2018; DuRoss et al., 2018). To the north along the WFZ, the Pleasant View salient consists of a hanging-wall bedrock ridge and fault step that have long been considered a barrier to ruptures on the Brigham City and Weber segments (Machette et al., 1992). However, Personius et al. (2012) used paleoseismic and structural data to infer a small-displacement rupture through the barrier, similar to that observed in the Borah Peak rupture. Spillover rupture across the barrier, possibly from the Weber to Brigham City segment, broke ∼8 km of the 36-km-long Brigham City segment and reduced M0 accumulated since the most recent rupture on the segment by ∼11%–13%.

Implications for Normal-Fault Paleoseismology

Our results should be taken into account when interpreting rupture length and M0 along complex multisegment normal faults from paleoseismic data. For example, for geometric fault sections separated by a trans-basin high such as the Willow Creek Hills, complex spillover and single-section ruptures may be more common than unimpeded rupture through the structural complexity with peak per-event displacement centered near the structure. Our results suggest that although structural barriers can be long lived, they may act as “rupture gates” and conditionally allow ruptures to penetrate, but not without modulating rupture length and displacement. Spillover ruptures suggest that structural barriers are not hard limits to rupture but may moderate rupture displacement and progress and be regions of frequent strong ground shaking. For example, based on our rupture model (Figs. 11 and 13), the Willow Creek Hills structure has experienced more earthquake ruptures than either the Warm Springs or Thousand Springs sections. Our conclusion is similar to that for the Traverse Mountains structure on the Wasatch fault zone, which has recorded at least six late Holocene ruptures compared to about four ruptures on the segments bordering the structure (Bennett et al., 2018; DuRoss et al., 2018). This lends support to higher probabilities assigned to single-segment and spillover ruptures than full two-segment ruptures in normal-fault rupture forecasts (e.g., Working Group on Utah Earthquake Probabilities, 2016).

The fortuitous 1983 spillover rupture of the Warm Springs section offers essential lessons on the interpretation of sparse paleoseismic data obtained near normal fault structural complexities. First, because of moderate displacements in spillover ruptures, their preservation potential is limited as they can easily be overprinted by large-displacement ruptures. That is, it is not clear that the small, decimeter-scale displacements in 1983 will be preserved at the century or millennial scale of paleoseismic observations. This may explain why such ruptures are rarely interpreted from normal-fault paleoseismic data sets. If spillover displacement onto a fault section adjacent to the primary rupture is significant enough to be observed in a typical trench setting (i.e., >0.25–0.5 m, depending on stratigraphy), the moderate M0 event could be misinterpreted as the complete and full-displacement rupture of the fault section hosting the spillover. The collection of high-resolution topographic data and dense paleoseismic data from multiple sites approaching structural barriers may help resolve the extent and timing of these types of ruptures compared to ruptures that terminate cleanly at structural barriers.

CONCLUSIONS

This study demonstrates the utility in using high-resolution, UAS-based DSMs and a high-density suite of VS measurements to reconstruct along-strike surface displacements for both historical and prehistoric normal-faulting earthquake ruptures. Our results for the northern 16 km of the 1983 Mw 6.9 Borah Peak earthquake rupture suggest that (1) complex surface faulting near and within the Willow Creek Hills (e.g., the Arentson Gulch fault) includes evidence of at least one prehistoric rupture; (2) although the 1983 rupture of the Warm Springs section is part of the primary Borah Peak rupture, it is a moderate-displacement spillover rupture (defined by significant rupture length but modest M0 release) across a structural complexity; (3) two prehistoric ruptures of the Warm Springs section (PE1 and PE2) had larger (∼1–2 m) displacements and local M0 release than the 1983 spillover rupture; and (4) a variable history of rupture termination and penetration through the Willow Creek Hills suggests that the structure impeded some, but not all earthquakes, possibly depending on the history of M0 release along the fault (prior stress conditions) as well as fault displacement and rupture direction. Ultimately, our VS data and displacement distributions help improve our understanding of the role structures such as the Willow Creek Hills play in influencing rupture length, displacement, and M0 release. Our results have broad implications for the interpretation of normal-fault paleoseismic data and can help inform the range and weights of rupture models in regional seismic-hazard assessments.

ACKNOWLEDGMENTS

We thank Kendra Johnson, Lia Lajoie, and Edwin Nissen for assistance completing Helikite surveys of the Warm Springs section and the U.S. Geological Survey (USGS) Unmanned Aircraft Systems (UAS) Project, including Jeff Sloan, Mark Bauer, Joe Adams, and Todd Burton, for discussions of UAS photography and for conducting a Falcon flight near the Willow Creek Hills. Brad Koeckeritz provided aerial images (3DR Solo and GoPro) of the northern Arentson Gulch area. The remaining UAS flights were conducted by M.P.B. and N.A.T. Digital surface models (DSMs) used in this study are available at USGS ScienceBase (https://doi.org/10.5066/P9CH0IQ4); DSMs, point clouds, and metadata are available at Open Topography (Bunds et al., 2019; https://doi.org/10.5069/G9222RWR). Thanks to David Schwartz for discussions of this work and for providing unpublished paleoseismic data for the Warm Springs section. We also thank Utah Valley University (UVU) students Jeremy Andreini, Bret Huffaker, Kenneth Larsen, Rick Lines, Ephram Matheson, Brittany Ungerman, and Alexandra Valenzuela for able assistance in the field. Finally, we thank the UVU College of Science Scholarly Activities program for financial support to M.P.B. and N.A.T., and Nvidia Corporation for computing support to UVU through their Graphics Processing Unit (GPU) Grant Program. Idaho State University provided accommodations at the Lost River field station. We thank Jaime Delano (USGS), Austin Elliot, and one anonymous reviewer for their constructive peer reviews. Any use of trade, product, or firm names is for descriptive purposes only and does not imply endorsement by the U.S. Government.

1Supplemental Figures. Lost River fault zone mapping, topographic profiles, and vertical separation data (Fig. S1), aftershocks of the 1983 Borah Peak earthquake (Fig. S2), vertical separation in the Goosebury graben (Fig. S3), vertical separation along the southern Warm Springs section (Fig. S4), vertical separation along the northern Arentson Gulch fault (Fig. S5). Please visit https://doi.org/10.1130/GES02096.S1 or access the full-text article on www.gsapubs.org to view the Supplemental Figures.
2Supplemental Text. Agisoft Photoscan digital surface model workflow (Text S1), vertical separation data (Text S2), tutorial for Matlab code Scarp_VS.m (Text S3). Please visit https://doi.org/10.1130/GES02096.S2 or access the full-text article on www.gsapubs.org to view the Supplemental Text.
3Supplemental Code. Matlab code Scarp_VS.m (Code S1). Please visit https://doi.org/10.1130/GES02096.S3 or access the full-text article on www.gsapubs.org to view the Supplemental Code.
Science Editor: David E. Fastovsky
Associate Editor: Jose M. Hurtado
Gold Open Access: This paper is published under the terms of the CC-BY-NC license.