Saddle Mountain fault deformation zone, Olympic Peninsula, Washington: Western boundary of the Seattle uplift

Geophysical, LIDAR (light detection and ranging), and paleoseismic studies are uncovering a rich history of Holocene earthquake activity in the Puget Lowland, northwestern United States. Paleomagnetic studies, global positioning system (GPS) measurements, and geologic arguments indicate that the Oregon forearc is rotating clockwise and moving northward with respect to cratonic North America at rates of ~1°/m.y. and 6–8 mm/yr, respectively (Wells et al., 1998; Mazzotti et al., 2002; McCaffrey et al., 2007). In the Puget Lowland, the resulting north-south compression causes 4.4 ± 0.3 mm/yr of permanent shortening (Mazzotti et al., 2002; McCaffrey et al., 2007), accommodated in part by a series of east- and southeast-striking faults that cross the lowland (Fig. 1). Although recurrence intervals and earthquake magnitudes are uncertain in most cases, recent surface- rupturing earthquakes have occurred on many of these faults, including the Utsalady Point fault (Johnson et al., 2004b), the southern Whidbey Island fault (Johnson, et al., 1996; Kelsey et al., 2004; Sherrod et al., 2008), the Seattle fault (Johnson et al., 1994; Pratt et al., 1997; Blakely et al., 2002; Nelson et al., 2003), the Tacoma fault (Johnson et al., 2004a; Sherrod et al., 2004), and the Olympia fault (Sherrod, 2001). The most recent large (M_W 7.5) crustal earthquake in the Puget Lowland occurred on the Seattle fault ~1100 yr ago, lifting the hanging wall of the Seattle fault 7 m, causing landslides, and generating a local tsunami (Bucknam et al., 1992; Atwater and Moore, 1992; Nelson et al., 2003; Sherrod, 2001; Karlin and Abella, 1996; ten Brink et al., 2006).

Active faults are also well known in the Olympic Peninsula west of the Puget Lowland (e.g., Carson, 1973; Nelson et al., 2007). Perhaps best known of these is the Saddle Mountain fault, with a surface trace initially reported to extend ~8 km (Figs. 2 and 3; Carson, 1973; Wilson, 1975; Hughes, 2005; Witter and Givler, 2008). Several issues warrant a closer look at the Saddle Mountain fault and surrounding tectonic framework. First, the Saddle Mountain fault generated an M 6.5–7.0 earthquake 1000–1300 yr ago (Hughes, 2005), within the same century or two as the M_W 7.5 Seattle fault earthquake. While radiometric dating techniques have insufficient resolution to determine if these two events were synchronous, it is important in assessing future earthquakes to understand any structural connections between the Saddle Mountain and Seattle faults. Second, although the surface expression of the Saddle Mountain fault was first reported to extend only ~8 km, geophysical evidence may help map the full extent of the fault in the subsurface, thus providing better estimates of its potential hazard. To help address these issues, we analyzed existing aeromagnetic and gravity data from the entire Olympic Peninsula and conducted ground-based magnetic surveys to characterize the Saddle Mountain fault in three dimensions. We also report on a paleoseismic trench investigation across one of the active faults along the southeast flank of the Olympic Mountains.

The Saddle Mountain fault is located on the southeast flank of the Olympic Mountains (Fig. 1), an accretionary complex consisting of two distinctive terranes. Highly deformed, pervasively sheared, and metamorphosed Eocene–Miocene sedimentary rocks form the core of the Olympic Peninsula. These core rocks are thrust under peripheral rocks of Eocene oceanic basalt and marine sediments along steeply dipping thrust faults (Figs. 4 and 5; Cady, 1975; Tabor and Cady, 1978a, 1978b). Severe disruption of the eastern part of the core and a general westward decrease in age provide strong evidence that all of these rocks were emplaced by subduction processes (Tabor and Cady, 1978a). Initial accretion of the complex began as early as late Oligocene, followed by exhumation that started ca. 18 Ma and continues today at approximately uniform rates (Brandon et al., 1998). Uplift of the Olympic Mountains is domal in shape, with highest rates (>1 mm/yr) near the center of the Olympic massif, tapering to less than 0.3 mm/yr in the Saddle Mountain area (Brandon et al., 1998).

Peripheral rocks consist primarily of early to middle Eocene Crescent Formation basalts and associated volcanic and sedimentary rocks, an overthickened volcanic assemblage of oceanic affinity and part of the Eocene Coast Range terrane extending from southern Oregon to Vancouver Island (Brown et al., 1960; Snavely and Wagner, 1963; Babcock et al., 1992). Tabor and Cady (1978a) considered the peripheral volcanic rocks to be a resistive bulwark against which subducting sediments were tilted, faulted, and sheared (Fig. 5). In the southeastern Olympic Peninsula, peripheral volcanic rocks are as much as 16 km thick, dip steeply eastward, and are overlain by Oligocene and younger sediments; farther east they are exposed in parts of the Seattle uplift and form part of the basement beneath the Seattle, Dewatto, and Tacoma basins (Fig. 1; Tabor and Cady, 1978b; Brocher et al., 2004; Johnson et al., 2004a).

Along the eastern and southeastern margin of the Olympic Peninsula, Crescent Formation basalts consist of two important units: an upper member of basalt flows and mudflow breccia and a lower member of massive flows, pillow basalts, breccia, and minor intrusive rocks (Tabor and Cady, 1978b). The lower member basalts are locally altered and metamorphosed to phrenite-pumpellyite and green-schist facies (Hirsch and Babcock, 2006) and commonly exhibit pillows and tube structures. Upper member basalts are characterized by closely spaced joints and local columnar jointing. On the basis of geochemical arguments, Glassley (1974) concluded that lower member basalts were created at a mid-ocean ridge in the early Eocene, subducted beneath similarly aged upper member basalts, and then faulted upward into their present juxtaposition in Miocene time. Cady (1975), however, described the lower and upper member contact as a simple upward gradation, from a deep-water marine origin to a shallow-water marine and terrestrial origin; these observations were later confirmed by Hirsch and Babcock (2006).

The Saddle Mountain fault exhibits surface traces easily seen in aerial photography and LIDAR images (Figs. 2 and 3). Carson (1973) first described scarps of the Saddle Mountain fault in detail, although there is anecdotal evidence that loggers recognized its topographic expression many years earlier. Wilson (1975) mapped the fault and recognized Pleistocene or younger deformation on three stands: the northeast-striking Saddle Mountain West and Saddle Mountain East faults and the northwest-striking Dow Mountain fault. Witter and Givler (2008) noted a fourth fault striking northeast along the flank of Dow Mountain (Figs. 2 and 3). Several other active faults with similar northeast strike are observed in the LIDAR data: the Frigid Creek fault ~2 km to the south (Fig. 2; Haugerud and Sherrod, 2007; Witter and Givler, 2008) and the Canyon River fault ~27 km to the southwest (Walsh and Logan, 2007).

LIDAR images and field examinations indicate that both the Saddle Mountain West and Saddle Mountain East faults exhibit southeast-side-up displacement, with scarps exceeding 8 m high in many places (Fig. 3). Early trench excavations (Wilson, 1975; Wilson et al., 1979) confirmed that the scarps were created by reverse faults offsetting late Pleistocene glacial deposits and early Eocene Crescent Formation basalt. A recent trench excavated across the Saddle Mountain West fault (Witter and Givler, 2008) exposed evidence for two earthquakes, one occurring between 17,000 and 8500 yr ago, and a second occurring after 1700 yr ago. Total vertical offset for both earthquakes at this most recent trench site was ~1 m, although the 1.7 m scarp height suggests that only part of the deformational history was exposed in the trench. Lateral slip also may have been important in these earthquakes. Southwest-plunging striations observed on the hanging wall of the Saddle Mountain East fault plane indicate left-lateral movement (Wilson et al., 1979), while a basalt cobble lodged in till and split by a secondary fault suggests dextral displacement along the Saddle Mountain West fault (Witter and Givler, 2008). The northeast-striking Canyon River fault, 27 km to the southwest, shows clear evidence of oblique left-lateral slip in Holocene time (Walsh and Logan, 2007), suggesting that right-lateral slip along the Saddle Mountain West fault seen by Witter and Givler (2008) was a local phenomenon rather than a persistent kinematic feature of the region.

Carson (1973) first suggested that slip on the Saddle Mountain East fault dammed a tributary of Lilliwaup Creek ~1100 yr ago, creating Price Lake and drowning a forest that existed at the time (Figs. 2 and 3). Hughes (2005) analyzed stumps beneath Price Lake and concluded that both the Saddle Mountain West and Saddle Mountain East faults ruptured between 1000 and 1300 yr ago, possibly during the same earthquake. This earthquake may have caused as much as 4 m of vertical offset on the Saddle Mountain East fault and 2 m of offset on the Saddle Mountain West fault (Hughes, 2005), and probably generated the 1.7 m slip observed in the Saddle Mountain West fault trench (Witter and Givler, 2008). These estimates agree with trench excavations made in the 1970s across the Saddle Mountain East and West scarps (Wilson, 1975; Wilson et al., 1979) that revealed 3.5 m and 1.8 m of reverse slip, respectively. The bracketed time of this earthquake (1000–1300 yr ago) includes the time (1020–1050 yr ago; Atwater, 1999) of the M_W 7.5 earthquake on the Seattle fault and corresponds temporally with a wide variety of other fault and landslide activity over a wide region of the Olympic Peninsula and Puget Lowland (Schuster et al., 1992; Logan et al., 1998; Haugerud et al., 2003; Walsh and Logan, 2007).

Figure 6 shows gravity and magnetic anomalies of the eastern Olympic Peninsula and adjacent Puget Lowland. Magnetic anomalies are based on an aeromagnetic survey flown in 1997 by the U.S. Geological Survey (Blakely et al., 1999). Flight altitude was nominally 300 m over flat to moderate terrain, but significantly higher altitudes were necessary over river valleys and along the eastern margin of the Olympic Mountains. Most of the study area was flown along north-south flight lines spaced 400 m apart and along east-west tie lines spaced 8 km apart. Flight-line and tie-line spacings were doubled in the northwestern part of the study area, where the magnetic field is characterized by longer wavelengths. Gravity anomalies shown in Figure 6B are based on point measurements from the Pan-American Center for Earth and Environmental Studies (PACES) repository (http://paces.geo.utep.edu/home.shtml), supplemented with unpublished data (Thomas Wiley, 2008, written commun.). Gravity anomalies have been reduced to isostatic residual gravity values in order to emphasize middle and upper crustal sources (Simpson et al., 1986). Gravity measurements are sparse in some areas, especially over Hood Canal and within inaccessible parts of the Olympic Mountains, but in most areas station density is adequate (at least one station per 25 km²) to define regional-scale structures.

Low densities and magnetizations of sedimentary and metamorphic rocks of the core terrane produce subdued, low-amplitude gravity and magnetic anomalies. In contrast, relatively high densities and magnetizations of Crescent Formation rocks produce high-amplitude anomalies over the peripheral terrane. Moreover, the upper and lower members of the Crescent Formation appear geophysically distinct. This is particularly evident south of lat 48°45′N., where gravity anomalies (Fig. 6B) are broadly associated with the entire exposure of Crescent Formation basalts, but high-amplitude magnetic anomalies (Fig. 6A) primarily reflect only the upper member basalts.

Magnetic susceptibility measurements made at Crescent Formation outcrops confirm this overall pattern (Figs. 6A and 7; Table 1). In general, high susceptibilities are observed in regions with high-amplitude magnetic anomalies, which coincide with the mapped extent of upper member basalts. Low susceptibilities and low-amplitude magnetic anomalies characterize the lower member basalts. This relationship is particularly clear in the Saddle Mountain area (Fig. 7), where all mean susceptibility values in the Price Lake and Dow Mountain area (upper member basalts) exceed 18 × 10⁻³ SI (Système International) units, whereas all mean values along Lake Cushman (lower member basalts) fall below that value.

Figure 8 shows a cross-sectional model through the Saddle Mountain area based on forward simultaneous modeling of gravity and magnetic anomalies, constrained by geologic mapping (Tabor and Cady, 1978b). The model is consistent with low-density, weakly magnetic core rocks thrust beneath steeply dipping, sometimes overturned peripheral volcanic rocks. Peripheral volcanic rocks are modeled in Figure 8 with nearly uniform densities, whereas upper member basalts required a magnetization four times greater than lower member basalts, consistent with our susceptibility measurements (Table 1). The model assumes that magnetizations are entirely induced, a reasonable assumption based on poor paleomagnetic results from most of the main Crescent outcrops (C.S. Gromme, Myrl Beck, and David Engebretson, 1975–1982, personal communs.).

The Saddle Mountain East and West faults offset highly magnetic upper member basalt, and both faults are included in our cross- sectional model through Price Lake (Fig. 8). Price Lake afforded an excellent opportunity to investigate the magnetic characteristics of the Saddle Mountain West fault at outcrop scale (Fig. 9). The alpine setting of this small lake is devoid of artificial objects that deleteriously affect ground-based magnetic surveys, and the obvious lack of tree cover facilitated accurate GPS navigation. Measurements were made with a nonmagnetic canoe, GPS navigation, and a cesium-vapor magnetometer system. A stationary proton-precession magnetometer was maintained at the lakeshore to subsequently correct for time-varying fields. We acquired 26 track lines over the entire lake at an average track-line spacing of 40 m and a total line distance of 9 km; 15 track lines crossed the lakeward projection of the Saddle Mountain West LIDAR scarp, but the Saddle Mountain East scarp was unreachable by canoe. The canoe-magnetic survey illuminated a linear, north-northeast– trending magnetic trough, ~150 m wide and 1000 nT in amplitude, on strike with LIDAR scarps northeast and southwest of the lake (Fig. 9).

A cross-sectional model of the Saddle Mountain West fault (Fig. 10) based on the canoe magnetic survey is consistent with two distinct strands of the Saddle Mountain West fault. Both strands are modeled as southeast-side-up reverse faults that offset Pleistocene and Holocene deposits at the surface and Crescent Formation at >30 m depth. Although our modeled offsets of Crescent Formation project upward to LIDAR scarps observed at the surface, the offsets at depth have significantly larger displacements (>20 m) than implied by either topography (scarps as high as 8 m) or trench excavations (dip slip as much as 3.5 m; Wilson et al., 1979), suggesting a long history of Quaternary deformation.

The canoe-based magnetic survey of Price Lake demonstrates that individual strands of the Saddle Mountain fault are reflected in magnetic anomalies. We now step back and examine aeromagnetic anomalies at a regional scale. Our interpretation of the eastern and southeastern Olympic Peninsula (Fig. 11) employed a variety of both quantitative and qualitative analyses of magnetic and gravity data, but we relied primarily on a method recently described by Phillips et al. (2007): magnetic anomalies (Fig. 6A) were reduced to the pole, transformed to maximum horizontal gradients, and analyzed for mathematical curvature. This three-step methodology provides the locations of abrupt lateral variations in crustal magnetization, shown as sinuous alignments of small black dots in Figure 11B.

Linear alignments of black dots in Figure 11B indicate the map projection of magnetic contacts, which have diverse explanations in our large study area. Some contacts represent faults, notably the contact between highly deformed core rocks thrust beneath the Crescent Formation in peripheral rocks (Fig. 11C, label CPF, core-peripheral fault). Other contacts are caused by lateral variations in lithology or geochemistry, such as the marked contrast in anomaly amplitudes over upper and lower members of the Crescent Formation south of lat 48°45′N. Still others may reflect folding within Crescent Formation rocks, although none of our calculated magnetic contacts (Fig. 11B) directly correlates with folds mapped by Tabor and Cady (1978b). Within the Crescent Formation, magnetic contacts may reflect lateral changes from normal to reverse magnetization, although metamorphism and alteration of these volcanic rocks have probably reduced significantly their primary remanent magnetization.

The contrast in anomaly amplitude between the upper and lower member Crescent Formation south of lat 48°45′N was an expected result in view of magnetic susceptibilities exhibited by these rocks (Figs. 6A and 7; Table 1). The contrast in magnetic properties reflects differences in lithologic characteristics and levels of tectonic deformation. Lower member basalts are highly sheared and deformed and contain abundant pillows, whereas upper member basalts are characterized by massive flows, local columnar jointing, sparse pillows, and less deformation. The contrast in magnetic properties may reflect primary formation of magnetic minerals, possibly due to contrasting local environments at the time of formation, submarine environments for lower member basalts and subaerial for upper member basalts. Alternatively, the weak magnetization of lower member basalts may reflect secondary alteration of magnetic minerals, either caused directly by the deformational events responsible for the pervasive shearing seen in these rocks, or promoted subsequently by migration of low-temperature water through loosely consolidated pillow structures (e.g., Marshall and Cox, 1972). The contrast in magnetic properties ends rather abruptly at lat 48°45′N; north of this latitude lower member basalts produce anomalies similar in amplitude to the upper member basalts. Subdued anomalies south of lat 48°45′N may reflect local metamorphism to greenschist facies.

A pronounced northeast-striking magnetic anomaly directly overlies Saddle and Dow Mountains (Figs. 2, 11, and 12), possibly caused by combined southeast-side uplift along both the Saddle Mountain East and West faults. The northwest-facing gradient of this magnetic anomaly yields a well-defined magnetic contact corresponding closely with the swath of scarps seen in LIDAR data (Fig. 12, label SMF). The magnetic contact extends southwestward 4 km from Price Lake to Lake Cushman (Fig. 12B), where it coincides with the northeast-striking Cushman Valley fault (Carson and Wilson, 1974; see Figs. 2B and 3B for location of Cushman Valley). Witter and Givler (2008) suggested that the Saddle Mountain and Cushman Valley faults are surface expressions of the same structure, and our observations support their interpretation. Moreover, the magnetic contact and LIDAR scarps suggest that the Saddle Mountain fault extends even farther to the southwest, to at least 6 km west of Lake Cushman (Fig. 12). Southwest of this point, the magnetic contact steps southward before continuing southwestward. Thus, the Saddle Mountain fault, as expressed by LIDAR scarps, extends a minimum of 15 km, from northeast of Price Lake to southwest of Lake Cushman, and magnetic anomalies suggest that the fault extends an additional 5 km southwestward (Fig. 12). We describe herein evidence for extending the Saddle Mountain fault northeastward, well beyond mapped LIDAR scarps.

The Frigid Creek fault (Figs. 2 and 13) is parallel to and 4 km south of the Saddle Mountain fault and exhibits a well-defined northwest-side-up scarp 2.7 m in height. A single trench excavated across the Frigid Creek scarp (Fig. 14) revealed conformable strata consisting of oxidized sandy gravels, sandy loams, and sandy silts. Radiocarbon analyses by Lawrence Livermore National Laboratory showed that charcoal clasts collected from units 2 and 3 ranged in age from 5657 to 5476 calendar yr B.P. (4850 ± 40 ¹⁴C yr B.P.) to 4513–4220 cal yr B.P. (3925 ± 40 ¹⁴C yr B.P.). A single clast of charcoal from unit 6 (part of the surface soil profile on the scarp) yielded an age of 526–319 cal yr B.P. (415 ± 40 ¹⁴C yr B.P.). These strata resemble deposits observed throughout the southeast Olympic Mountains that consist of intercalated Holocene debris flows and soils.

Beneath the scarp, the strata are offset by a normal master fault and several smaller antithetic normal faults, forming a small graben along the scarp. We interpret the offset strata as the result of movement along the normal master fault during an earthquake between 5657 and 319 cal yr B.P. The deformation is best interpreted as downward movement of two hanging-wall blocks (blocks 2 and 3) relative to the footwall block (block 1; Fig. 14B). Downward movement and clockwise rotation of block 2 formed a small graben adjacent to the scarp. Piercing points observed in the excavation show that 2.5 m of vertical separation can be accounted for in one event (93% of total 2.7 m scarp height). The unaccounted 0.2 m of scarp height suggests either an earlier and much smaller earthquake, or differential erosion and deposition along the scarp after the earthquake.

We interpret the Frigid Creek fault as a bending-moment fault in the hanging wall of a large thrust sheet, or as a normal fault associated with a bend or stepover in a lateral fault system. The Frigid Creek fault is astride a high- amplitude, sinuous magnetic anomaly (Fig. 12) that we interpret as a fold in Crescent Formation. A magnetic contact is not observed directly along the Frigid Creek fault scarp, suggesting that the fault is entirely above or roots into underlying Crescent Formation basalts. The Frigid Creek fault, with southeast side down, may be responding to subsidence of the Dewatto and Tacoma basins directly to the east and midway between the Olympia and Seattle uplifts (Fig. 1B).

A narrow but pronounced magnetic anomaly (Fig. 12, label OF, Olympia fault) and coincident gravity anomaly (Fig. 6B) extend south-eastward from Crescent Formation exposures in the Olympic Mountains to the southern edge of our study area, where the source of the anomaly is entirely concealed beneath Pleistocene glacial deposits. South of our study area, anomaly OF merges with the Olympia structure (Magsino et al., 2003), which juxtaposes near-surface Crescent Formation to the southwest against the Tacoma basin to the northeast (Fig. 1). Coseismic subsidence occurred along the Olympia structure ~1100 yr ago (Sherrod, 2001), and the linear nature of associated gravity and magnetic anomalies is highly suggestive of a near-surface fault. However, detailed models based on gravity and magnetic data were unable to determine whether the anomalies reflect faults or folds (Magsino et al., 2003).

A magnetic contact (Fig. 12, label CRF) coincides approximately with the Canyon River fault (Walsh and Logan, 2007) and extends 5 km northeastward beyond its mapped surface expression (Schuster, 2005). High-resolution aeromagnetic data do not cover the western end of the Canyon River fault. The Canyon River fault is expressed topographically as a 3-m-high northwest-facing scarp. A trench excavated across the scarp revealed evidence for oblique reverse-left-lateral slip during an M 7–7.5 earthquake in late Holocene time (Walsh and Logan, 2007). The sense of the magnetic contact is consistent with southeast-side-up slip seen in the trench. Although the Canyon River fault is essentially on strike with the Saddle Mountain fault, magnetic anomalies provide no obvious way to connect them as a single, continuous structure. If the Canyon River and Saddle Mountain faults are linked, it is apparently accomplished through complex en echelon relationships.

The Saddle Mountain fault is expressed topographically over a length of 15 km, and magnetic anomalies suggest that the fault extends an additional 5 km to the southwest (Fig. 12) and 15 km to the northeast (Fig. 15), a total span of 35 km. Moreover, the alignment of the Saddle Mountain, Frigid Creek, and Canyon River faults (Figs. 11 and 12) may reflect a zone of faulting extending more than 45 km. The opposing sense of slip on these three faults (southeast-side-up reverse slip on the Saddle Mountain and Canyon River faults, northwest-side-up normal slip on the Frigid Creek fault) may reflect their positions relative to deformational patterns in the Puget Lowland immediately to the east (Fig. 1). The Saddle Mountain and Canyon River faults are adjacent to the Seattle and Olympia uplifts, respectively, whereas the Frigid Creek fault is adjacent to the subsiding Dewatto and Tacoma basins. Alternatively, the Frigid Creek fault could represent either a releasing bend fault or a bending moment normal fault in the hanging wall of the Saddle Mountain fault zone.

Wilson et al. (1979) noted that the Saddle Mountain fault is on strike with a narrow zone of fracturing mapped by Glassley (1974) immediately to the northeast. Glassley (1974) viewed this fracture zone as a remnant of a major tectonic event that brought lower and upper member basalts into contact in Miocene time, and Wilson et al. (1979) suggested that Holocene displacement on the Saddle Mountain fault may reflect a reactivation of this Miocene structure. Hirsch and Babcock (2006) showed, however, that the upper member–lower member contact is more likely an abrupt change in metamorphic grade, lower member basalt having been altered to greenschist facies. Moreover, Figure 12 shows that the Saddle Mountain fault, as displayed in both LIDAR and aeromagnetic data, is not coincident with the upper member–lower member contact, but rather is 2–3 km to the southeast.

The circular shaped anomaly at the eastern edge of our study area (Fig. 11, label GM, Green Mountain) is caused by highly magnetic rocks in the hanging walls of the Seattle and Tacoma faults. The southern margin of anomaly GM is along the Tacoma fault (Johnson et al., 2004a; Sherrod et al., 2004); the northern margin is along the Seattle fault (Blakely et al., 2002). Crescent Formation basalt and Tertiary intrusive rocks exposed at Green and Gold Mountain near the northern edge of anomaly GM probably represent the source of the entire anomaly; likewise the broad limits of anomaly GM predict where these magnetic rocks are located at relatively shallow depths within the Seattle uplift.

The Crescent Formation, exposed at the surface in the hanging wall of the Seattle fault, is 9–10 km deep beneath the Seattle basin immediately to the north (Brocher et al., 2001). Widely divergent models have been proposed to explain this large vertical offset. Johnson et al. (1994), Pratt et al. (1997), and ten Brink et al. (2002) envisioned two or three south-dipping thrust faults that extend to depths of 10–20 km, bringing the hanging wall of the Seattle fault northward over the Seattle basin. Brocher et al. (2004) and Kelsey et al. (2008), however, found evidence for a roof thrust that merges at shallow (<5 km) depth with a south-dipping floor thrust well north of Crescent Formation exposures, thus forming a northward-advancing crustal wedge. In their model, large vertical offset in the Crescent Formation is accomplished along one or more south-dipping imbricate thrust faults within the wedge (see Fig. 8; Brocher et al., 2004).

Both models require a means to accommodate strain beyond the westward limit of the Seattle fault. Johnson et al. (1994) proposed the existence of a north- to northeast-trending strike-slip fault beneath Hood Canal that transfers strain on the Seattle fault northward to other faults, possibly the southern Whidbey Island fault (Fig. 1, dotted line). Although the linear nature of Hood Canal lends credence to this interpretation, no seismic evidence has been found to support the existence of a fault beneath and parallel to Hood Canal (Haug, 1998). The magnetic analysis shown in Figure 11 also shows no evidence for a Hood Canal fault, although the altitude of the aeromagnetic survey in this area exceeded 1400 m above ground because of the proximity of the Olympic Mountains range front. A prominent magnetic gradient is west and parallel to part of Hood Canal (Fig. 11C), but the broad gradient is inconsistent with strike-slip faulting. It more likely reflects the eastward-dipping contact between the Eocene Crescent Formation of the Olympic Peninsula and younger overlying sediments of the Seattle basin (Fig. 8). This contact presumably forms a broad syncline beneath Hood Canal, shallowing eastward within the Seattle uplift, where it produces anomaly GM (Fig. 11). As an alternative to faulting along Hood Canal, we identified several prominent magnetic anomalies in the northern part of our study area (Fig. 11C, label DBF, Dabob Bay fault) with abrupt, linear, northwest-striking gradients. They may reflect right-lateral strike-slip faults that pass through Dabob Bay and transfer strain from the Seattle fault northward.

We propose that the Saddle Mountain fault forms part of the western boundary of the Seattle uplift. As evidence, we note a subtle west-striking magnetic lineament (Fig. 15, label SF) that is on strike with the Seattle fault to the east, crosses Hood Canal, and extends westward 10 km into Crescent Formation volcanic rocks to the west. At its western end, within Crescent Formation exposures, anomaly SF passes through a pronounced 3–4 km right step in magnetic anomalies and mapped folds (Tabor and Cady, 1978b). The westward terminus of lineament SF is near the northern terminus of the Saddle Mountain fault (Fig. 15, label SMF). We suggest that lineament SF is the westward extension of the Seattle fault.

We suggest that the Saddle Mountain, Frigid Creek, and Canyon River faults are elements of a deformation zone that accommodates the northward shortening of Puget Lowland crust inboard of the Olympic massif (label SMDZ, Saddle Mountain deformation zone, Fig. 16). This model predicts sinistral slip on all elements of the Saddle Mountain deformation zone. We also expect vertical slip to be important, with sense determined by the proximity of large-scale deformation to the east. The Saddle Mountain and Canyon River faults (southeast side up) respond to the Seattle and Olympia uplifts, respectively, whereas the Frigid Creek fault (northwest side up) is influenced by subsidence of the Dewatto and Tacoma basins. Lineament SF (Fig. 15) may be the westward continuation of the Seattle fault, with the Saddle Mountain fault (Fig. 15, label SMF) marking the western edge of the Seattle uplift.

We propose that the Saddle Mountain, Frigid Creek, and Canyon River faults are elements of a 45-km-long zone of deformation that accommodates shortening of Puget Lowland crust inboard of the Olympic massif. We see subtle evidence in geophysical anomalies that the Seattle fault extends westward across Hood Canal and 10 km into Crescent Formation exposures on the Olympic Peninsula, ending near the northern terminus of the Saddle Mountain deformation zone. In this framework, the Saddle Mountain and Seattle faults are boundaries of the same crustal block, the Seattle uplift. Previous studies have shown that the Saddle Mountain fault produced a M_W 6.5–7.0 earthquake 1000–1300 yr ago, within the same century as the M_W 7.5 Seattle fault earthquake. The temporal coincidence of these two earthquakes suggests that the Saddle Mountain deformation zone and Seattle fault zone are kinematically linked, and our geophysical studies further suggest that the two fault zones are spatially linked as well. We have mapped a magnetic contact that coincides with the topographic expression of the Saddle Mountain fault and shows that it extends at least 35 km. A magnetic survey of Price Lake conducted from a nonmagnetic canoe has allowed us to model the Saddle Mountain West fault in detail. The model includes two east-dipping reverse faults, consistent with scarps identified in LIDAR data and in the field. The opposing sense of slip on the Saddle Mountain, Frigid Creek, and Canyon River faults (southeast-side-up reverse slip on the Saddle Mountain and Canyon River faults, southeast-side-down normal slip on the Frigid Creek fault) may be a reflection of deformational patterns in the Puget Lowland immediately to the east. The Saddle Mountain and Canyon River faults are adjacent to the Seattle and Olympia uplifts, whereas the Frigid Creek fault is near the subsiding Dewatto and Tacoma basins.

We are grateful to Jerry Kvale, Washington Department of Natural Resources, for providing access to Price Lake. Tom Wiley graciously provided unpublished gravity data from the southeast Olympic Mountains. We also thank Tom Brocher, Gerry Connard, Jonathan Glen, Darcy McPhee, Rowland Tabor, and Rob Witter for helpful discussions and critical readings of early versions of our manuscript. Very helpful reviews were provided by Harvey Kelsey and Cathy Snelson.