Holocene earthquakes and right-lateral slip on the left-lateral Darrington–Devils Mountain fault zone, northern Puget Sound, Washington

The Cascadia subduction zone is the dominant source of seismic hazard in the Puget Sound region of northwestern Washington (e.g., Petersen et al., 2008), but additional seismic sources related to crustal (upper-plate) faults in the forearc of the subduction zone have been identified in the last decade (Fig. 1) through the acquisition and analysis of high-resolution light detection and ranging (LiDAR) data (e.g., Harding and Berghoff, 2000; Haugerud et al., 2003). The E-W to NW-SE strikes of most of these faults and the coincidence of some of these faults with the margins of similarly trending Tertiary basins and uplifts are thought to reflect the N-S compression caused by northward transport of the Oregon block against a backstop of quasi-stable crust in southern British Columbia (Fig. 1A; Wells et al., 1998, 2014; Wells and Simpson, 2001; McCaffrey et al., 2007). From south to north, paleoseismic and geologic mapping studies have identified surface deformation indicating Holocene activity on the Olympia and Tacoma faults in the Tacoma Basin (Bucknam et al., 1992; Sherrod, 2001; Sherrod et al., 2004a, 2004b; Nelson et al., 2008; Barnett et al., 2010; Sherrod and Gomberg, 2014), the Seattle fault in the Seattle Basin (Bucknam et al., 1999; Johnson et al., 1999; Sherrod et al., 2000; Nelson et al., 2003a, 2003b, 2014; Kelsey et al., 2008), the Southern Whidbey Island, Utsalady Point, and Darrington–Devils Mountain faults in the Everett Basin (Johnson et al., 1996, 2001, 2004; Dragovich and DeOme, 2006; Sherrod et al., 2008), and the Boulder Creek, Birch Bay, Sandy Point, and Drayton Harbor faults in the Bellingham Basin (Kelsey et al., 2012; Sherrod et al., 2013; Sherrod and Gomberg, 2014). Other Holocene faults that lie outside of the margins of the Tertiary basins include the Lake Creek–Boundary Creek (Little River) fault and the Saddle Mountains fault zone, which form parts of the north and east flanks, respectively, of the uplifted Olympic Mountains (Wilson et al., 1979; Nelson et al., 2007; Walsh and Logan, 2007; Witter et al., 2008; Blakely et al., 2009; Sherrod and Gomberg, 2014). Shallow seismicity is associated with some of these structures but not others (Fig. 1B).

We focused our investigation on one of these crustal faults, the Darrington–Devils Mountain fault zone, because despite prior work indicating the likelihood of Holocene displacement, the late Quaternary paleoseismic history of this prominent structure was heretofore unknown. As currently mapped, the Darrington–Devils Mountain fault zone extends >125 km roughly E-W across the northern Puget Sound region, from near Victoria, British Columbia, eastward to south of Darrington, Washington, in the foothills of the North Cascades (Fig. 1B). The Darrington–Devils Mountain fault zone is thought to have formed in the early Eocene, and it juxtaposes Mesozoic accreted terranes (Trafton sequence, and Western and Eastern mélange belts) to the south against Upper Mesozoic terranes (Helena-Haystack mélange and Easton metamorphic suite) to the north (Fig. 2; Whetten et al., 1988; Tabor, 1994; Johnson et al., 1996; Tabor et al., 2002; Dragovich and DeOme, 2006). In our area of investigation, the Darrington–Devils Mountain fault zone is coincident with the northern margin of the Tertiary Everett Basin, and it juxtaposes a thick section of Eocene to Oligocene continental sedimentary rocks (Rocks of Bulson Creek) to the south against similar Eocene rocks (Chuckanut Formation) to the north. Less-extensive outcrops of Eocene intrusive and extrusive igneous rocks are localized along the Darrington–Devils Mountain fault zone and other structures in the area (Dragovich and DeOme, 2006). The Darrington–Devils Mountain fault zone has had a long (∼50 m.y.) and complicated slip history, which may have included an early period of right-lateral oblique slip (Tabor, 1994). Numerous NW-striking en echelon structures associated with the Darrington–Devils Mountain fault zone suggest left-lateral transpression in the later part of its history (Loveseth, 1975; Johnson et al., 2001; Hayward et al., 2006; Dragovich and DeOme, 2006), perhaps due to collision of the Siletzia block in the Eocene (Wells et al., 2014). Historic seismicity is sparse along the mapped trace of the Darrington–Devils Mountain fault zone (Fig. 1B), but a limited number of published focal mechanisms of nearby earthquakes are consistent with the N-S compressive stress regime in the region (Fig. 3).

The first major study of the seismic potential of the Darrington–Devils Mountain fault zone was conducted as part of seismic hazard investigations related to the proposed siting of a nuclear power plant in the Skagit River valley east of Mount Vernon, Washington (Puget Sound Power and Light, 1974; hereafter “Puget Power, 1974”). Puget Power used geologic, air photo, and side-looking airborne radar mapping to locate potentially active fault traces and excavated a total of 11 trenches and eight drill holes at four sites to determine the age of the most recent movement along the fault zone. Puget Power concluded that the Darrington–Devils Mountain fault zone had not been active in the last 35 k.y. and thus was not a capable seismic source (Puget Power, 1974; Bechtel, Inc., 1979; Crosby et al., 1986; Adair et al., 1989).

Extensive studies on Whidbey Island and surrounding marine waters provide abundant evidence of Quaternary deformation along the western portion of the Darrington–Devils Mountain fault zone (Johnson et al., 1996, 2001; Hayward et al., 2006). Johnson et al. (1996, 2001) used water well data, shoreline bluff exposures, and high-resolution seismic profiles to document displacements in last interglacial (oxygen isotope stage [OIS] 5, 80–125 ka) and younger deposits. Although the Darrington–Devils Mountain fault zone has no topographic expression where it projects across Whidbey Island, Johnson et al. (2001) described warping and minor offsets of postglacial (latest Pleistocene to Holocene) sediments in several marine seismic profiles west of Whidbey Island. In subsequent mapping in the area, Dragovich et al. (2005) delineated several strands of the Darrington–Devils Mountain fault zone across the island and inferred that deposits of the latest Pleistocene final glacial recession (Everson interstade) were offset by the Darrington–Devils Mountain fault zone in several cross sections. Hayward et al. (2006) reinterpreted the same seismic data used by Johnson et al. and concluded that the youngest deformation in the eastern Strait of Juan de Fuca makes a right step from their “primary” fault strand of the Devils Mountain fault onto the nearby Utsalady fault. Johnson et al. (2004) conducted a paleoseismic (trenching) investigation of the Utsalady fault and found evidence of multiple Holocene left-lateral transpressional earthquake ruptures. However, no such investigations of the timing of paleoearthquakes have been conducted on the main trace of the Darrington–Devils Mountain fault zone.

Folding and faulting of latest Pleistocene and Holocene lake sediments beneath Lake Cavanaugh (Fig. 3) imaged in high-resolution seismic profiles (Naugler et al., 1996) extend evidence of Quaternary displacement on the Darrington–Devils Mountain fault zone 40–50 km east of Whidbey Island. Although the seismic data are unpublished, Naugler et al. (1996) described a central graben with vertical offset of lake bottom sediments of 3–5 m and used deformation patterns in postglacial sediments to infer a predominant strike-slip component. In later mapping, Dragovich et al. (2004) inferred a similar amount of latest Quaternary vertical offset across their “main strand” of the Darrington–Devils Mountain fault zone under Lake Cavanaugh.

Recent 1:24,000 scale quadrangle geologic mapping, from Whidbey Island on the west to near Darrington on the east, documented abundant evidence of Quaternary displacement along the Darrington–Devils Mountain fault zone (see summary in Dragovich and DeOme, 2006). Such evidence includes possible uplifted alluvial deposits, apparent thinning of valley-fill sediments across the fault zone, observations of sheared Pleistocene sediments, and an anomalous increase in the gradient of Pilchuck Creek (Fig. 3) upstream of the mapped main trace of the Darrington–Devils Mountain fault zone. Dragovich and DeOme (2006) also reinterpreted the trenching study of Puget Power (1974) and concluded that the offset features mapped in some of the Puget Power trenches as slump scarps were instead likely caused by Holocene tectonic movement along the Darrington–Devils Mountain fault zone.

We began our study of the Darrington–Devils Mountain fault zone with a trench investigation on a remnant of glacial outwash (herein informally named the Whitman bench after nearby Whitman Road) perched ∼200 m above and north of the modern floodplain of the North Fork Stillaguamish River ∼20 km west of Darrington, Washington (Fig. 3). Dragovich et al. (2003a, 2003b) mapped the abandoned outwash plain as recessional outwash of the Everson interstade of the Fraser glaciation. We excavated two trenches across a 0.75-km-long NW-trending scarp that was first identified by Ralph Haugerud of the U.S. Geological Survey (USGS) on digital elevation models (DEMs) derived from a detailed LiDAR survey of the region (R.A. Haugerud, 2007, written commun.). We found no evidence of tectonic faulting or folding in either trench, and we conclude that the trenched scarp is a glacial-fluvial channel margin rather than a fault scarp. Large-scale logs, unit descriptions, a detailed geomorphic map, and a brief summary of the results of this investigation are included in Supplemental Figure S1¹.

The second phase of our investigation consisted of a trenching study of topographic lineaments in the Lake Creek area (Fig. 4) that were originally identified during the Puget Power (1974) investigation of the Darrington–Devils Mountain fault zone. Our evaluation of the Puget Power study in the Lake Creek area (see Supplemental Fig. S2²) reached conclusions similar to those of Dragovich and DeOme (2006): The lineaments identified by Puget Power are fault scarps marking the trace of the Darrington–Devils Mountain fault zone rather than slump headscarps as originally interpreted by Puget Power. We used the subsurface investigations of Puget Power (1974), detailed geologic mapping of Dragovich and DeOme (2006), and our own field and LiDAR investigations to site our trenches in the Lake Creek area. Our mapping of fault scarps on 2006 LiDAR DEMs (pixel size 1.8 m) resulted in a pattern of late Quaternary faulting in the area (Fig. 4A) similar to that of Dragovich and DeOme (2006). We both map a pair of E-W–trending fault scarps, an ∼1-km-long, south-facing northern strand, and an ∼0.5-km-long, predominantly north-facing southern strand (graben faults B and A, respectively, of Dragovich and DeOme, 2006). The major difference is that we were unable to confirm the existence of Dragovich and DeOme’s (2006) “main strand” that they mapped cutting Vashon-aged glacial deposits 20–50 m north of our northern strand. We found no topographic expression of their main strand either in the LiDAR data or in limited field searches, and none is shown in cross-section B–B′ of Dragovich and DeOme (2006). Puget Power (1974) also showed no indication of faulting, such as offset or truncated bedding or fracturing of the glacial sequence or sheared bedrock within 30–40 m of the projected location of this strand in their log of trench 1A (Supplemental Fig. S2 [see ^{footnote 2}]), which extended across the mapped trace of Dragovich and DeOme’s (2006) main strand. These lines of evidence and the fact that most of the evidence for significant Quaternary displacement is restricted to the northern strand, support our contention that the northern strand is the surface expression of the steeply dipping active trace of the Darrington–Devils Mountain fault zone.

We excavated three trenches on the northernmost scarp at the Lake Creek site (Fig. 4B) for the following reasons: (1) the northern scarp is longer and more continuously expressed in the landscape than the southern scarp; (2) the Puget Power trenches and drill holes across the northern scarp show significant (as much as 10 m or more) north-side-up vertical separation of the bedrock surface (Supplemental Fig. S2 [see ^{footnote 2}]), consistent with the sense of slip inferred along the northern scarp; and (3) vertical separation of the bedrock surface is much smaller (<4 m) and evidence of faulting is more ambiguous where Puget Power (1974) trenches 1A and 1D cross the southern scarp (Supplemental Fig. S2 [see ^{footnote 2}]). We consider the southern scarp as a minor splay of the main (northern) strand that likely has a similar history of Quaternary faulting.

We excavated the easternmost “Aplodontia” and “Springboard” trenches ∼25 m apart across an ∼1-m-high south-facing scarp (Fig. 5; Supplemental Fig. S3³) and exposed a sequence of glacial sediments consisting of a basal unit of lacustrine or fluvial sand and silt (unit 2), which overlies and is faulted against a fault-bounded remnant of fluvial gravel (unit 1). These basal units are overlain by a 30- to 90-cm-thick deposit of fluvial (outwash) sand and gravel (unit 3). We use the abrupt transition from gravel (unit 3b) to sand (unit 3a) in the Aplodontia trench at horizontal trench m 13.25 as a piercing point in a later analysis of lateral slip across the fault zone (see “Fault Geometry and Sense of Slip”). A massive to moderately bedded lacustrine and fluvial deposit of sand and silty sand (unit 4) overlies unit 3. An up to 2-m-thick deposit (unit 6a) of massive silt and silty sand (till) caps the glacial sequence. An extensively root-stirred soil (unit 8) is formed in the upper part of unit 6. The trenched deposits are mapped as part of the Vashon Stade advance of the Fraser glaciation (Dragovich and DeOme, 2006). Dragovich and DeOme (2006) used a radiocarbon age (14,725 ± 470 radiocarbon yr B.P.; 17.9 ± 1.0 ka [thousands of calibrated years ago, ±2σ]) from a depth of 3 m in Puget Power (1974) trench 1B (70 m west of the Aplodontia trench) to date the glacial sequence at the Lake Creek site. This age is in agreement with the timing of the advance and retreat of the Puget Lobe of the Cordilleran ice sheet during the Fraser glaciation (Porter and Swanson, 1998). Two radiocarbon samples (Table 1) from a depth of 3.2 m in unit 2 (45.3 ± 1.1 ka) and 2.8 m in unit 3b (26.2 ± 1.1 ka) in the Aplodontia trench may indicate a pre-Vashon age for the lower part of the sequence. However, because both of these samples were on detrital charcoal, we consider them to be maximum-limiting ages for the glacial sediments exposed in the Lake Creek trenches.

In both trenches, the glacial sequence is bisected by a near-vertical, 2-m-wide fault zone with 45–70 cm of up-to-the-north vertical separation (Fig. 5; Supplemental Fig. S3 [see ^{footnote 3}]). Several near-vertical fractures and faults with millimeter-scale vertical displacements extend 4–6 m north and as much a 4 m south of the main fault zone, but they are more numerous and more continuous north of the fault zone. Many of the faults and fractures continue upward through the glacial sequence but were difficult to map in the root-stirred soil horizons that cap the sequence. Remnants of a charcoal-rich buried soil on the glacial deposits (unit 6Bb) and an overlying 10- to 15-cm-thick deposit of fault-scarp colluvium are preserved at the upward terminations of two fault splays in the Springboard trench. In addition, a cast of a large burned root offset by a single fault splay is preserved at a similar stratigraphic level in the upper part of the glacial sequence. Five radiocarbon ages on charcoal from the buried soil and burned root yield similar ages of 2.2–2.4 ka. We also observed a zone of subhorizontal shears and flame-like injections of glacial units 3, 4, and 6 north of the main fault zone in the Springboard trench. Their strata-bound and rootless geometry leads us to attribute these features to internal glacial deformation, rather than tectonic processes.

We excavated a third (“Boundary”) trench ∼125 m west of the Springboard trench across a 1.5-m-high scarp on the northern strand to examine along-strike variations in faulting and preservation of near-surface faulting relations. The trench exposed highly sheared bedrock eroded into and overlain by a sequence of glacial-fluvial, glacial-lacustrine, and till sediments deposited at the basal glacier-bedrock contact (Fig. 6; Supplemental Fig. S3 [see ^{footnote 3}]). We suspect that the differences in depth to bedrock between the Aplodontia/Springboard and Boundary trenches reflect hydraulic or topographic irregularities in the bed of the overriding ice sheet, rather than differences in vertical displacement between the two sites. The base of the glacial sequence consists of gravelly till (unit 3) and glacial-lacustrine sand and silt (unit 4) deposits locally preserved in subglacial channels or tunnels eroded into the underlying bedrock. The surficial and bedrock deposits were subsequently planed off by later glacial erosion. The unconformity created by this erosion event is marked throughout most of the trench by a thin (5–10 cm) deposit of lacustrine or fluvial silt to fine sand. This marker bed is in turn overlain by glacio-fluvial and till deposits (unit 6) of gravelly silt and sand. The sequence is capped by a gravelly glacio-fluvial deposit (unit 7) and a thick root-stirred soil (unit 8) formed in unit 7. Bedrock exposed in the bottom of the trench includes highly weathered sandstone and shale (unit 1) north of the fault zone, and pervasively sheared and convolutedly bedded sandstone, shale, and coal (unit 2) to the south. As mapped in the immediate area, these rocks probably correspond to the Eocene Chuckanut Formation and Oligocene to Eocene Rocks of Bulson Creek, respectively (Dragovich and DeOme, 2006).

The Boundary trench exposed a complex, steeply dipping, 1- to 1.5-m-wide zone consisting of two primary synthetic faults similar to those exposed in the Aplodontia and Springboard trenches. The numerous strands that comprise the fault zone continue upward through the glacial sequence until they are obscured by soil formation in unit 8. The thin marker bed at the bedrock-glacial contact (unit 5) is vertically separated 65 cm (up-to-the-north) across the fault zone. Abrupt changes in geometry and thickness of units 4 and 6 are apparent across the fault zone. Abundant evidence of preglacial deformation in the bedrock units includes two shear zones in the steeply dipping, highly weathered sandstone and shale in the footwall, and pervasive shearing and convoluted bedding in unit 2 in the 15-m-wide exposure of these rocks in the hanging wall. The coincidence of Quaternary faulting with such an abrupt change in bedrock properties is evidence that this strand of the Darrington–Devils Mountain fault zone has likely been recurrently active for millions of years.

We documented two sets of slickensides on the two primary postglacial fault traces in the Boundary trench. In the east wall, two slickenlines in highly sheared bedrock (unit 2) plunge 40° and 58° west on the western primary fault strand (Fig. 7), and in the west wall, two slickenlines in fine-grained glacial sediments (units 4 and 6) plunge 02° and 07° east on the eastern primary fault strand. The implications of these features are discussed in the “Fault Geometry and Sense of Slip” section later herein.

After opening the Lake Creek trenches, we recognized the probability of significant strike-slip displacement based on the very steep fault dips, flowerlike fault geometries, slickenside orientations, and major changes in the thickness and texture stratigraphic units across faults. To further explore the question of slip direction, we dug additional excavations adjacent to the Aplodontia and Boundary trenches near the end of our field investigation to quantify the sense and amount of postglacial lateral slip. The logs from these excavations are shown in Supplemental Figures S4⁴ and S5⁵, and the implications of these results are discussed in detail in Figures 8 and 9 and the “Fault Geometry and Sense of Slip” section.

The third phase of our investigation focused on refining the earthquake chronology on the Darrington–Devils Mountain fault zone. The three trench exposures at the Lake Creek site showed an almost complete lack of preservation of fault-related stratigraphy (e.g., scarp-derived colluvium) in the upper 0.5–1 m due to soil formation and extensive churning of the glacial sequence by root growth, tree-throw cratering, and burrowing animals. We only found evidence of a single earthquake-related deposit in one trench (scarp-colluvial unit 7 in Springboard trench; Fig. 5; Supplemental Fig. S3 [see ^{footnote 3}]), despite the presence in all three trenches of numerous fault splays that likely ruptured the ground surface. In order to help refine the age of this earthquake and to look for evidence of other earthquakes that may have been obscured by soil formation, we conducted a coring investigation of a small wetland (herein informally referred to as “Boomer marsh”) directly adjacent to the eastern end of the Lake Creek fault scarps (Fig. 4A).

In our refinement of Dragovich and DeOme’s (2006) mapping of the Lake Creek fault scarps, we mapped the eastern end of the northern LiDAR scarp to within a few tens of meters of the outlet of Boomer marsh (Fig. 4B). Dragovich and DeOme (2006) mapped the fault to the southeast as concealed beneath the center of the marsh, but the similar trend of the steep scarp that forms the northern margin of the marsh suggested to us that the Lake Creek fault scarp might instead project across the outlet of the Boomer marsh and continue eastward along the northern margin. If our mapping is correct, then a record of scarp formation during earthquakes might be manifested in the marsh stratigraphy by: (1) temporary deepening of the marsh by partial damming of the outlet, and (2) incursion of slope colluvium onto the nearshore marsh surface from exposure of glacial sediments by surface faulting along the northern margin of the marsh.

Our initial series of exploratory 1- to 6-m-deep gouge cores along the northern edge of the marsh revealed a laterally extensive stratigraphic sequence consisting of a basinward-thinning package of sandy and pebbly silts, over- and underlain by thick peat deposits. In contrast, reconnaissance coring along the southern edge of the marsh, which is marked by higher, steeper slopes in thin glacial deposits and sandstone bedrock (Dragovich and DeOme, 2006), revealed highly variable sequences of peat, lacustrine silt, and rocky debris; we infer the latter sediments were deposited by localized slope failures. We found no evidence of a basinward-thinning silty sequence similar to the northern margin in exploratory cores at numerous sites distributed along the southern margin (Fig. 4B).

We documented the geometry of the silty sequence present along the northern margin with six, 2.5-cm-wide gouge cores along a 40-m-long transect from the wetland edge to near the center of the marsh (Fig. 4B). We determined the gross stratigraphy with small gouge cores (Fig. 10; Supplemental Fig. S6⁶), and then we used a “Russian” coring device to obtain a larger-diameter (5-cm-wide), 70-cm-long core of the middle part of gouge core BO-01 for sedimentologic analysis and sampling (Fig. 11A; Supplemental Fig. S7⁷). In the laboratory, we made detailed sediment descriptions using the methods of Troels-Smith (1955), sampled and described diatom assemblages (Supplemental Fig. S8⁸), and carefully sampled for macroscopic samples of plant fragments and charcoal for radiocarbon dating (Table 1; Fig. 11A; Supplemental Fig. S7 [see ^{footnote 7}]).

The gouge and Russian cores revealed a sequence of well-bedded silts interbedded with sandy, pebbly silts, over- and underlain by 70- to >100-cm-thick peat deposits. The base of the Russian core consists of 15 cm of sandy decomposed peat, which is overlain in sharp contact by an 11-cm-thick sandy silt interval marked by a 2-cm-thick pebbly sand at its base, which in turn is overlain by a 1-cm-thick bed of mixed organic materials (Fig. 11A; Supplemental Fig. S7 [see ^{footnote 7}]). The organic bed grades upward to finely laminated sandy silt and a diatom assemblage indicative of deeper, quiet-water deposition. A peaty interval overlying the sandy silt suggests a temporary shallowing of water depths, but this interval transitions gradually back to deeper water depths with the deposition of ∼13 cm of finely laminated silt. The silt interval contains a prominent 2- to 3-cm-thick orange tephra bed chemically correlated with the Mazama ash, dated elsewhere (Zdanowicz et al., 1999) at 7630 ± 150 cal. yr B.P. (Supplemental Table S1⁹). The same distinctively colored bed was identified in several reconnaissance cores along the northern margin of the marsh west of our transect (Fig. 4B). At a depth of 99 cm, the silt interval is overlain in sharp contact by another sandy silt deposit with a 2-cm-thick basal layer of mixed sand, pebbles, and detrital wood similar to the lower sandy-silt deposit. Above the mixed basal layer, the upper sandy silt interval is discontinuously laminated, but it also contains numerous fragments of detrital wood, charcoal, and pebbles. Diatom assemblages suggest somewhat shallower water depths than the lower sandy silt (Fig. 11A). Subsequent deposition of a 5-cm-thick interval of diatomite- and organic-rich silt suggests water depths gradually deepened and then gradually shallowed with a return to peat deposition that extends to the modern-day surface of the marsh.

The marsh deposits described from the BO-1 core contain two sandy, pebbly silt intervals that we interpret as the sedimentologic response to two periods (colluvial events “CE1” and “CE2”) of colluvial progradation of hillslope sediment and organic debris into the marsh. We base this interpretation on the following evidence: (1) Both intervals are wedge-shaped and thin and become finer grained away from the wetland margin. (2) The sharp lower boundaries of both intervals suggest abrupt progradation of hillslope sediment and organic debris into the marsh rather than soil formation on an exposed wetland surface. (3) The presence of sand and pebbles in the basal sediments of both intervals indicates that they likely were derived from exposure of the glacial sediments currently covered by heavy vegetation and thick forest duff in the hillslopes adjacent to the marsh; the lack of any similar deposits in similar geomorphic positions in the rest of our transect or reconnaissance cores likely precludes a fluvial or flood origin for these coarse sediments.

We examined a number of possible origins for the sediments we interpret as colluvium along the northern margin of the Boomer marsh (see “Alternative Explanations” section) before concluding that these deposits were likely the result of one and possibly two surface-rupturing earthquakes on the Darrington–Devils Mountain fault zone. Evidence supporting this conclusion is circumstantial and stronger for CE1 than CE2: (1) Two laterally persistent stratigraphic packages of colluvial deposits are only found adjacent to the steep south-facing scarp that controls the northern margin of the marsh. The approximately E-W trend of the northern margin is parallel to the strike of the LiDAR-defined Lake Creek fault scarps and is anomalous because most nearby steep slopes have NW to WNW strikes controlled by the underlying bedrock (Dragovich and DeOme, 2006). The similar trend and near continuity with mapped fault scarps across the 25-m-wide outlet indicate to us that the surface trace of the Darrington–Devils Mountain fault zone is coincident with the steep slope bordering the northern margin of the marsh (Fig. 4B). This slope has been modified by road construction, but its height (6–8 m) suggested to us that it is probably a composite feature formed from glacial erosion along a preexisting fault zone and subsequent postglacial faulting. (2) Our analysis of the minimum constraints on the age of the CE1 colluvium in the marsh and the maximum constraints on the age of fault-scarp colluvium in the Springboard trench (LC1) indicates that both deposits are the same age and thus likely are related to a single surface-rupturing earthquake that occurred ca. 2 ka on this part of the Darrington–Devils Mountain fault zone (Table 1). (3) By analogy, we infer that similar sedimentologic characteristics of the younger (CE1) and older (CE2) colluvial deposits may indicate that both are related to surface-rupturing earthquakes. (4) Deposition of laminated silt beds and changes in diatom assemblages indicate that CE2 was followed by an abrupt deepening of the water level that may have been caused by partial damming of the outlet by formation of a north-side-up fault scarp. Significant deepening apparently did not accompany CE1, either because the earthquake was predominantly strike slip, or because of folding of the northern margin of the marsh. The former is supported by the set of near-horizontal slickensides observed in the Boundary trench, and the latter is supported by the unconformity defined by abrupt changes in sedimentation rates between deposition of the Mazama ash and the time of CE1. We infer that the dearth of evidence for CE2 in our Lake Creek trenches is attributable to near-complete disruption by burrowing animals and root stirring of the upper parts of the glacial deposits exposed in the walls of all three trenches.

We acknowledge that our interpretation of the stratigraphy observed in the Boomer marsh is not unique. Other processes that could produce some of the characteristics we observed in our core transect include (1) water-level fluctuations related to outlet damming by mass wasting or animal activity (beavers), (2), erosion of adjacent hillslopes during and after forest fires or large storm events, and (3) changes in hillslope sedimentation and water level due to rapid changes in climate and vegetation. We dismiss the first process because we see no evidence of landslide modification or old beaver dams near the outlet, and because it seems unlikely that outlet damming would induce the simultaneous deposition of colluvial sediments documented in our coring campaign. Hillslope erosion from nontectonic processes is less easy to dismiss, but our best evidence of a tectonic origin is the localization of the colluvial wedge deposits only along the anomalously E-W–trending northern margin of the marsh. We interpret the restriction of sand and pebbles to the bases of both wedge-shaped colluvial deposits as evidence that these sediments were shed from the adjacent hillsides rather than from flooding or high stream flow, which would likely distribute pulses of coarser sediment throughout the marsh. We interpret two other silty sand deposits in the lower parts of cores BO-1 and BO-04 (Fig. 10; Supplemental Fig. S6 [see ^{footnote 6}]) as probable examples of such fluvial processes, because both deposits thin toward the margin of the marsh. Both CE1and CE2 (ca. 2 ka and ca. 8 ka, respectively; see discussion in “Earthquake Chronology” section) are roughly correlative with changes in regional climate (e.g., Galloway et al., 2009, their fig. 6), but we suspect that even the most rapid changes in climate would induce gradual changes in vegetation that would not cause the abrupt and localized changes in hillslope erosion we infer from our coring data. However, shorter-term climate fluctuations, such as periods of extended drought, could cause nondeposition and soil formation near the margins of the marsh. We consider this origin unlikely, because desiccation should cause a “bathtub ring” with similar sedimentary characteristics around the perimeter of the wetland. Our reconnaissance coring shows that the southern margin exhibits highly variable sedimentation patterns that are strikingly different from those along the northern margin. In addition, the sharp lower contacts of both colluvial packages are inconsistent with soil formation and weathering of an emergent marsh surface.

The very steep fault dips exposed in the Lake Creek trenches are a significant source of uncertainty in our determination of the sense of slip on this reach of the Darrington–Devils Mountain fault zone. Depending on dip direction, the consistent north-side-up vertical separations apparent in all three trenches indicate either a component of normal displacement on a steeply south-dipping fault, or north-side-up reverse faulting on a steeply north-dipping fault. The latter scenario is supported by the apparent gentle anticlinal folding of the glacial sequence in the upthrown wall in the Aplodontia trench (Fig. 5), and it is consistent with north-south compression in the region (Mazzotti et al., 2002; Hyndman et al., 2003; Lewis et al., 2003). However, this scenario is at odds with Dragovich and DeOme’s (2006, their cross-section B) interpretation of the Lake Creek scarps as graben structures in the hanging wall of their south-dipping “main strand.” Given the modern stress regime and our doubt about the existence of their main strand, we suspect the fault zone is vertical or dips steeply to the north at depth. However, the possibility of triggered normal slip on upper-plate structures in Cascadia cannot be ruled out (Sherrod and Gomberg, 2014), because examples of large aftershocks with normal faulting mechanisms have occurred on favorably oriented upper-plate structures following recent subduction zone earthquakes in Chile (M8.8, 2010; Ryder et al., 2012) and Japan (M9.0, 2011; Kato et al., 2011). The angular relationship between the Darrington–Devils Mountain fault zone and the convergence direction of the Cascadia subduction zone is similar to some of the normal-slip aftershocks in the Chile and Japan sequences, so the Darrington–Devils Mountain fault zone may be favorably oriented for extensional reactivation during great Cascadia earthquakes.

Three-dimensional trenching at the Aplodontia and Boundary excavations support our inference of a significant strike-skip component of faulting exposed in the original excavations (Figs. 8 and 9; Supplemental Figs. S4 and S5 [see ^{footnotes 4} and ⁵]). This inference is consistent with previous studies that documented a component of left-lateral slip on parts of the fault zone (Johnson et al., 2001; Dragovich et al., 2004; Dragovich and DeOme, 2006). We also recognized that our main fault trace at the site makes a subtle right-stepping bend (Fig. 4B) that might accommodate localized extensional slip, but only if the strike-slip component is right-lateral.

At the Aplodontia trench, we excavated short slot trenches on both sides of and parallel to the main fault zone in the west wall of our original excavation (Fig. 8; Supplemental Fig. S4 [see ^{footnote 4}]). This wall was chosen because we recognized a linear piercing point in the east wall of the trench—the lateral edge of an apparent channel margin (unit 3b) ∼1.5 m north of the main fault zone that was not exposed in the original west wall. Fortuitously, the southern slot trench exposed an identical channel margin in both walls of the slot trench. This line feature is subhorizontal and trends ∼030°.

Projection of the channel margin observed in the east wall of the main trench at the same orientation to its intersection with the projection of the fault zone between the east and west walls yields a best estimate of 2.2 (+0.6, –1.0) m of right-lateral displacement of this piercing point. The listed uncertainties are likely maximum and minimum estimates based on two different projections of the channel between the east and west walls of the main trench.

Combined with the vertical separation of 45 cm of the unit 4–unit 3 contact across the fault zone in the east wall of the main trench (Fig. 5), and the 50 cm difference in elevation between the channel margins in the slot and main trenches (Fig. 5; Supplemental Fig. S4 [see ^{footnote 4}]), our lateral displacement measurement yields a horizontal-to-vertical ratio of ∼5:1.

We also excavated a slot trench on the south side of and parallel to the main fault zone in the west wall of the Boundary trench. The only piercing point found in the original trench was a distinctive, ∼30-cm-wide fault zone that juxtaposes weathered bedrock (unit 1) against highly sheared bedrock (unit 2) in the east wall of the trench (Fig. 9; Supplemental Fig. S5 [see ^{footnote 5}]). The bedrock fault zone projects across the floor of the main trench and intersects the west wall, where it is overprinted by shearing along the main fault zone. We exposed a similar bedrock fault zone juxtaposing units 1 and 2 in the north and west walls of the slot trench that also indicates right-lateral slip. However, the bedrock fault zone in the slot trench consists of several strands that require at least two projections of possible piercing points. These projections yield an estimate of right-lateral slip (2.3 +0.5, –0.9 m) that is nearly identical to our estimate from the Aplodontia trench.

A second line of evidence for right-lateral oblique slip in the Aplodontia and Springboard trenches is the presence of numerous en echelon, subvertical fractures and secondary faults in glacial deposits in a 4- to 6-m-wide zone surrounding the main fault zones (Fig. 5). A similar set of fractures was mapped in a similar position in trench 1A by Puget Power (1974; Supplemental Fig. S2 [see ^{footnote 2}]). Most of these features have been disturbed by infiltration of small roots and groundwater. We mapped these features either as secondary faults where tiny (millimeter-scale) vertical offsets were visible in finely stratified glacial sediments, or as fractures where evidence of displacement could not be confirmed. For the former structures, the steep (∼90°) dips and lack of consistent vertical offset directions suggest primarily strike-slip displacement. Orientations of all structures in the trenches were measured where the features were observed in the ∼2-m-wide floors and both walls of the trenches (Supplemental Table S2A¹⁰).

Analysis of the secondary structures in the Aplodontia and Springboard trenches suggested to us that these features are Riedel shears for which orientations could be used to determine the lateral-slip direction of the main fault zone (Fig. 12). R shears form left-stepping, en echelon arrays along right-lateral shear zones, with orientations clockwise to the main shear zone. In contrast, R shears in left-lateral shear zones form right-stepping arrays with orientations counterclockwise to the main shear zone. In addition, Riedel theory predicts that R fractures should have synthetic shear and an angular difference from the main shear direction equal to ϕ/2, where ϕ is the angle of internal friction of the faulted material. In contrast, R′ shears have antithetic shear and a much higher angular difference (90° – ϕ/2) from the main shear direction (Riedel, 1929; Tchalenko, 1968, 1970). If our secondary structures are R shears rather than R′ shears, then their orientation indicates right-lateral slip on the Lake Creek portion of the Darrington–Devils Mountain fault zone. We used eight values of ϕ of Vashon-aged glacial deposits from a compilation of values for Quaternary sediments in the Seattle area (Harp et al., 2008, their table 1) to estimate ϕ for the faulted glacial sediments at the Lake Creek site (Supplemental Table S2B [see ^{footnote 10}]). These eight values range from 24° to 38° and yield average values of 32° ± 5° (mean), 33° (median), or 34° (mode). Thus, the theoretical angular difference for R shears in average Vashon glacial deposits is 32°–34°/2 = 16°–17° ± 5°. The attitudes of mapped fractures and shears in the Aplodontia and Springboard trenches (mean strike of 303°, n = 10), and the average strike of the main fault zone (286°, n = 2) yield an actual angular difference in strike (17° ± 5°) that is nearly identical to the theoretical values calculated from the estimated internal angle of friction of the faulted sediments in the trenches (Fig. 12). In contrast, if the fractures and faults were R′ shears, then their predicted strike difference would be 73° ± 5°. No fractures or shears with such orientations were observed in any of our Lake Creek trenches, so we conclude that the secondary structures are Riedel R shears. Although the mapped lengths of the R shears in our excavations are too short (∼2 m) to discern clear stepping patterns, these shears are clearly oriented clockwise to the main shear direction (Fig. 12) and thus are indicative of right-lateral displacement.

The orientation of slickensides observed in the Boundary trench constitutes a third line of evidence for a significant component of lateral displacement on the Darrington–Devils Mountain fault zone in the late Quaternary. We interpret the western plunge of slickenlines in sheared bedrock along the western primary fault and clear evidence of north-side-up vertical separation of trench units and fault scarps as evidence of right-lateral oblique slip during at least one surface-rupturing earthquake (Fig. 7). In addition, the two slickenside measurements in glacial sediments on the eastern primary fault indicate nearly pure strike-slip displacement. The significance of the differences in orientation of these two sets of slickensides is unknown, but the indicators of nearly pure strike slip may be evidence of a second post–last-glacial earthquake with significantly less vertical displacement. The lack of evidence for a significant increase in water depth associated with the youngest colluvial event (CE1) in the BO-1 core (Fig. 11A) indicates that CE1 may have been accompanied by less vertical offset than event CE2. Given the evidence for both pure strike-slip and oblique-slip postglacial surface faulting, we conclude that the most robust method of calculating a net post–last-glacial slip vector is to use the combined slip and fault orientation data from the Lake Creek trench site. We used the averages of the displacement measurements from the Aplodontia and Boundary trenches (2.25 ± 1.1 m right-lateral and 0.6 ± 0.1 m north-side-up vertical) and fault orientation data (strike 286°, dip 90°) to determine a net slip vector of 2.3 ± 1.1 m plunging 14° west.

Our preferred interpretation of right-lateral, north-side-up oblique slip on a vertical to steeply north-dipping fault zone is exactly opposite the sense of slip determined for the Lake Creek site by Dragovich and DeOme (2006). The reversal of shear direction is difficult to explain (see discussion in “Regional Implications” section), but we suspect that the differences in sense of vertical slip are likely due to minor variations in dip angle along strike of the near-vertical Darrington–Devils Mountain fault zone. Although generally mapped as south-side-up, recent quadrangle mapping in the region shows evidence of both north-side-up and south-side-up senses of Quaternary slip along the central Darrington–Devils Mountain fault zone. For example, 10 km east of the Lake Creek site, an anomalous increase in gradient (Dragovich and DeOme, 2006, their fig. 3) and the presence of uplifted(?) Holocene terraces (Dragovich et al., 2004) on Pilchuck Creek north of the mapped trace of the Darrington–Devils Mountain fault zone indicate north-side-up slip, but 8–10 km further east, Dragovich et al. (2004) showed their “main strand” offsetting the lake bottom south-side-up beneath Lake Cavanaugh. Still further east, between Deer Creek and Stillaguamish River, cross sections based on water well data (Dragovich et al., 2003a, 2003b) suggest north-side-up offsets of Olympia-aged (>20–60 ka) interglacial and older (pre-Vashon) glacial sediments along the Darrington–Devils Mountain fault zone. Such changes in apparent vertical slip direction along strike are a classic characteristic of strike-slip fault systems (e.g., Sylvester, 1988).

We used accelerator mass spectrometry (AMS) radiocarbon dating of charcoal, wood, and macrofossil plant and animal remains from our trench and coring investigations (Table 1) to constrain the ages of two possible surface-rupturing earthquakes on the Darrington–Devils Mountain fault zone at the Lake Creek site. To calibrate the laboratory-reported ages and put them in stratigraphic context, we constructed a combined depth model (Fig. 11B) in OxCal (v. 4.2; Bronk Ramsey, 1995, 2008, 2009), using the IntCal09 calibration curve of Reimer et al. (2009). OxCal uses Bayesian analysis of prior information such as laboratory uncertainties and stratigraphic relationships, along with the likelihood (represented by probability distribution functions or PDFs) of each age determination, to estimate the ages of undated events, in our case, prehistoric earthquakes.

We used a combination of ages from our trench and coring investigations to estimate the time of the most recent earthquake (MRE) on this part of the Darrington–Devils Mountain fault zone. The mean of four nearly identical charcoal ages from the Springboard trench provides a close maximum age of the MRE (LC1), based on the location of samples from a buried soil directly beneath fault scarp colluvial deposits and from a burned root faulted during the earthquake (Fig. 5; Supplemental Fig. S3 [see ^{footnote 3}]). No minimum ages for the trench MRE were obtained, so we used our correlation of earthquake LC1 with colluvial event CE1 (see previous discussion) to justify the addition of the charcoal ages from the trench to an OxCal depth plot of ages from core BO-1 to better constrain the time of the LC1/CE1 earthquake. The minimum time of this event is constrained by two somewhat conflicting ages, charcoal sample BO-97-R5 (2330 ± 80 cal. yr B.P.) from a core depth of 97 cm, and wood sample BO-98-R13 (1480 ± 80 cal. yr B.P.) from a core depth of 98 cm (directly above the CE1 unconformity). We have reason to suspect both of these samples: The older charcoal sample likely has some inherited age (e.g., Gavin, 2001), as suggested by the ∼200–1100 yr inheritance for the two other charcoal samples from the core (Table 1), and the younger wood age, which is younger than (and does not overlap with at 2σ uncertainties) the likely more accurate age on seed sample BO-78-R11 (1670 ± 80 cal. yr), which lies 20 cm higher in the core. Although the charcoal age (2330 ± 80 cal. yr B.P.) may be a few hundred years too old, we prefer this age because it yields a post-CE1 sedimentation rate (∼0.35 mm/yr) that is nearly identical to the pre-Mazama-ash rate (∼0.3 mm/yr), rather than the anomalously fast post-CE1 rate (∼12 mm/yr) inferred from the stratigraphically inverted wood age (Fig. 11B). Our OxCal models using the charcoal and wood minimum ages yielded modeled timing estimates of LC1/CE1 of 2.2 ± 0.1 ka and 1.9 ± 0.4 ka, respectively (Fig 11B). Given the concerns we have about both of the minimum constraints, we consider 2.3–1.5 ka (1.9 ± 0.4 ka) to be a reasonable estimate of the 2σ time range for this earthquake.

Evidence for an earthquake origin for colluvial event CE2 is much more equivocal, but in contrast to the uncertainties surrounding the timing of the LC1/CE1 earthquake, radiocarbon ages on seeds sampled directly above and below the CE2 unconformity (Fig. 11A; Supplemental Fig. S7 [see ^{footnote 7}]) yield a tightly constrained model time estimate of 8.1 ± 0.1 ka (2σ) for this possible penultimate earthquake.

Calculations of prehistoric earthquake magnitudes are commonly based on estimates of surface-rupture length and displacement (e.g., Wells and Coppersmith, 1994). Estimating surface-rupture length is especially difficult in the heavily vegetated Pacific Northwest. We follow Dragovich and DeOme (2006) in mapping fault scarps in glacial deposits in the Lake Creek area, but these scarps are only ∼1 km long and likely were part of a much longer surface rupture or ruptures that are now obscured by vegetation and steep slopes in the foothills of the Cascade Range. We estimate a minimum surface-rupture length of ∼30 km, from the prominent bedrock lineaments on the south flank of Devils Mountain west of Lake Creek, through Lake Cavanaugh, where Holocene lake bottom sediments are apparently offset in reflection profiles, to the vicinity of Deer Creek, where Dragovich et al. (2003b) noted the “probable displacement of Quaternary strata” along a subsidiary fault (Fig. 3). This minimum surface-rupture length yields estimates of Mw of >6.8 (Wells and Coppersmith, 1994; all faults data set), >6.8 (Wesnousky, 2008; all faults data set), and >7.0 (Stirling et al., 2002; average of instrumental and preinstrumental data sets). Calculations of prehistoric earthquake magnitudes based on estimates of surface displacement are complicated by uncertainties about whether the displacement estimates are average or maximum values, as well as the number of earthquakes responsible for our displacement values. In our preferred scenario (two earthquakes), one half of our preferred displacement value of 2.3 ± 1.1 m (1.15 m) yields an estimated Mw of 7.0; if our preferred displacement value is a maximum, then it yields an estimated Mw of 6.7 (Wells and Coppersmith, 1994; all faults data sets). Alternatively, if our preferred displacement value is the result of a single earthquake, then the same empirical relations yield magnitude estimates of Mw of 7.2 (average slip) and Mw 7.0 (maximum slip).

We can also use our average displacement and surface-rupture length values, an estimated seismogenic depth of 15 km, and a dip of 90° to calculate moment magnitude using the area/magnitude relation of Hanks and Kanamori (1979). This relation yields estimates of Mw 6.8 (two earthquakes) and Mw 7.0 (one earthquake). We conclude that the past surface-rupturing earthquakes were likely Mw 6.7–7.0, and they produced strong ground shaking in the region.

Our trench and coring studies produced a record of one and possibly two paleoearthquakes on the central Darrington–Devils Mountain fault zone since deglaciation of the Puget Lowland in the latest Pleistocene. These two possible earthquakes, dated at 1.9 ± 0.4 ka and 8.1 ± 0.1 ka, yield a single recurrence interval of ∼6 k.y., and an elapsed time (open interval) of ∼2 k.y. since the MRE. We measured 2.3 ± 1.1 m of right-lateral oblique slip postdating deglaciation of the Lake Creek site, which we infer was ice-free by ca. 16 ± 0.5 ka (e.g., Dethier et al., 1995; Porter and Swanson, 1998). These data yield an average postglacial slip rate of 0.14 ± 0.1 mm/yr. We consider this a minimum rate because of the unknown (but likely small) contribution of slip from the untrenched southern strand and possibly from other unmapped fault traces, and uncertainties related to our reliance on slip values from a single paleoseismic site.

Despite its length and geologic longevity, our results indicate lower rates of paleoseismic activity on the Darrington–Devils Mountain fault zone than several less prominent Holocene-active faults in the northern Puget Lowland. For instance, the Sandy Point fault in the Bellingham Basin (Fig. 1B) has generated three earthquakes resulting in coseismic uplift of ∼4 m in the last 6 k.y. (Kelsey et al., 2012). In the northeast corner of the Bellingham Basin, the Boulder Creek fault has experienced at least three surface-rupturing earthquakes in the last 8 k.y. and has a postglacial vertical slip rate of 0.3–0.6 mm/yr (Sherrod et al., 2013). The Utsalady Point fault in the northwest corner of the Everett Basin (Fig. 1B) has probably experienced two postglacial surface-rupturing earthquakes, but both occurred in the last 2.2 k.y. and were accompanied by larger surface displacements: One trench showed evidence of two earthquakes that totaled as much as 4.5 m of vertical separation, and a second trench showed evidence of one earthquake that totaled 1–1.5 m of vertical and ∼2 m of lateral slip (Johnson et al., 2004). Although the apparently low rate of Holocene paleoseismicity on the Darrington–Devils Mountain fault zone may be attributable to an incomplete paleoseismic record, we believe our recurrence and slip rates from the Lake Creek site are consistent with the dispersed and subdued expression of postglacial rupture along the mapped trace of the Darrington–Devils Mountain fault zone.

Comparisons of regional paleoearthquake histories also raise questions about the possibility of clustering of Darrington–Devils Mountain fault zone paleoearthquakes with earthquakes on other crustal faults in the region, as well as the potential for triggering crustal earthquakes by great earthquakes on the Cascadia subduction zone (e.g., Sherrod and Gomberg, 2014). Several regional structures have dated earthquakes with ages similar to the youngest earthquake on the Darrington–Devils Mountain fault zone, such as the Sandy Point fault (ca. 2.1 ka; Kelsey et al., 2012), the Utsalady fault (2.1–1.2 ka; Johnson et al., 2004), the Seattle fault (2.6–1.9 ka; Nelson et al., 2003a, 2014), and the Lake Creek–Boundary Creek fault (2.0–0.6 ka; Nelson et al., 2007). However, neither the Darrington–Devils Mountain fault zone nor the regional earthquake time estimates are precise enough for us to confidently correlate these earthquakes with one another or with a great earthquake on the Cascadia subduction zone.

The occurrence of the devastating 22 March 2014 landslide along the south flank of the Whitman bench has drawn attention to the numerous older landslides mapped in the immediate area (Dragovich et al., 2003a, 2003b; Haugerud, 2014). The 2014 landslide was not seismically induced (B.W. Presgrave, U.S. Geological Survey, 2014, written commun.), but given the inherent instability of steep slopes along the bench, we speculate that large earthquakes such as our ca. 2 ka LC1/CE1 event on the Darrington–Devils Mountain fault zone might have triggered some of the older landslides in the area. Confirmation of a seismic origin must await future studies on the ages of the older landslide deposits, but the proximity of the Darrington–Devils Mountain fault zone to unstable slopes along the Whitman bench should be taken into account in future assessments of landslide hazards in the region.

One of the more curious aspects of our results is the evidence for right-lateral oblique displacement on what has previously been described as primarily a left-lateral reverse structure (Johnson et al., 2001; Dragovich and DeOme, 2006). Because the modern stress field is dominated by the rapid (>40 mm/yr) oblique subduction of the Juan de Fuca plate beneath the region, current geodetic techniques may be of limited use in determining slip sense and rates on crustal faults with much lower (<1 mm/yr) rates of slip (McCaffrey et al., 2013). However, most seismicity studies indicate N-S–directed compression in the Puget Lowland (Ma et al., 1996; Mazzotti et al., 2002; Van Wagoner et al., 2002; Hyndman et al., 2003; Lewis et al., 2003). Seismicity along the Darrington–Devils Mountain fault zone is sparse (Figs. 1B and 3), and only a handful of focal mechanisms from small (Mw <4) earthquakes near the Darrington–Devils Mountain fault zone have been published (Zollweg and Johnson, 1989; Ma et al., 1996; Van Wagoner et al., 2002). We used the structural grain visible on published geologic maps (Dragovich et al., 2002a, 2002b, 2002c, 2003a, 2004; Dragovich and DeOme, 2006) near the projected surface locations of the two possible fault planes to select preferred fault plane orientations from five of these mechanisms (Fig. 3). Although we do not know if these earthquakes occurred on the Darrington–Devils Mountain fault zone, their likely orientations primarily indicate reverse oblique displacement on near-vertical to steeply north-dipping, roughly E-W–striking fault planes. Such orientations are consistent with regional N-S compression. Two of the focal mechanisms from Van Wagoner et al. (2002) also indicate a small component of right-lateral oblique slip, and, as noted by Van Wagoner et al. (2002), these mechanisms are not consistent with a left-lateral slip component on the Darrington–Devils Mountain fault zone as proposed in previous studies (e.g., Johnson et al., 2001; Dragovich and DeOme, 2006). We reiterate that the relationship between these small-magnitude earthquakes and the Darrington–Devils Mountain fault zone is unknown, but the very limited focal mechanism data are at least supportive of our interpretation of Holocene right-lateral slip on the central part of the Darrington–Devils Mountain fault zone.

The cause of the apparent slip reversal is unknown, but aspects of local and regional tectonic controls provide some potential explanations. One possible explanation is that the Lake Creek scarps are located on one of several right-lateral, northwest-striking faults that intersect the Darrington–Devils Mountain fault zone in the vicinity of our Lake Creek site (Dragovich and DeOme, 2006), rather than being located on the Darrington–Devils Mountain fault zone itself. One such structure is mapped as intersecting the Darrington–Devils Mountain fault zone in the western end of the Boomer marsh (Fig. 4). This alternative does not require any change in slip direction, but it is not supported by (1) the abundant evidence of late Quaternary slip elsewhere along the main trace of the Darrington–Devils Mountain fault zone, (2) our evidence for surface faulting along the northern margin of Boomer marsh, and (3) the lack of any evidence for Quaternary displacement elsewhere along the NW-striking bedrock faults. Another possible, and perhaps more likely, explanation is that clockwise rotation of the North Cascadia block (Wells et al., 1998; McCaffrey et al., 2007) has reoriented the Darrington–Devils Mountain fault zone to a more northwesterly strike that is slightly more favorable to a right-lateral component of slip in the current N-S compressive stress field. A recent comparison of paleomagnetic and geodetic rotation rates in the northwestern United States indicates that the Puget Lowland is currently rotating clockwise at a rate of ∼0.5 mm/yr, and it has rotated as much as 6°–7° in the last 12 m.y. (McCaffrey et al., 2013; Wells and McCaffrey, 2013). However, more detailed study will be required to document if this progressive change in orientation was large enough to change the sense of slip of the Darrington–Devils Mountain fault zone, and when such a change might have occurred.

We used the results of three-dimensional trenching of a prominent fault scarp and coring of an adjacent wetland to compile a composite Holocene history of surface-rupturing earthquakes on the Darrington–Devils Mountain fault zone at the Lake Creek site southeast of Mount Vernon, Washington. Trenching revealed evidence of extensive biogenic disruption of the upper 50–100 cm of our exposures, but radiocarbon dating of charcoal from a single preserved deposit of fault-scarp colluvium indicates at least one large earthquake ruptured through the site ca. 2 ka. In the adjacent wetland, we interpret the presence of two basinward-thinning silty colluvial deposits along our inferred trace of the Darrington–Devils Mountain fault zone as evidence of two possible earthquakes, one coincident with the timing of the earthquake documented in the trenches ca. 2 ka, and an earlier possible earthquake ca. 8 ka. Slip components determined from faulted glacial deposits (2.25 ± 1.1 m right-lateral and 0.6 ± 0.1 m north-side-up vertical) yield a right-lateral oblique net slip of 2.3 ± 1.1 m plunging 14° west on a vertical fault striking 286°, and a postglacial minimum slip rate of 0.14 ± 0.1 mm/yr. Individual displacements could not be determined, but if each earthquake accounted for about half of the net slip, then empirical relations based on historical surface faulting indicate magnitudes of Mw 6.7–7.0. The timing of these two possible earthquakes yields a single recurrence interval of ∼6 k.y., and an elapsed time (open interval) of ∼2 k.y. since the most recent earthquake. A single recurrence interval is insufficient to quantify the probability of future seismic activity on the fault zone, but the occurrence of at least one and possibly two surface-rupturing earthquakes in the middle and late Holocene indicate the Darrington–Devils Mountain fault zone is a potential source of future large (Mw 6.7–7.0) earthquakes in northwestern Washington.

This research was funded by the Earthquake Hazards Reduction Program of the U.S. Geological Survey (USGS). We wish to thank the landowners and managers of the Whitman bench (Grandy Lake Forest Associates, Ken Osborne, Arbor Pacific Forest Products; Washington Department of Natural Resources [WADNR]) and Lake Creek (Keith Sorenson; Washington Department of Natural Resources) sites for permission to conduct our investigations, and Gary Wright and the crew of Gary’s Grader Service for excavation services. We also acknowledge the excellent field help of Kelsay Davis and Monte Jones, logistics assistance of John Coyle (WADNR), insights into the local geology of Joe Dragovich, Tim Walsh, and Ray Cakir (WADNR), and Puget Sound Energy for granting permission to reproduce parts of a 1970’s-era report on seismic hazards investigations for the proposed Skagit Nuclear Power Plant. The manuscript was improved by comments from USGS reviewer Scott Bennett, Geosphere reviewer Harvey Kelsey, an anonymous reviewer, and Editor Raymond Russo.