Geologic Evidence of Earthquakes at the Snohomish Delta, Washington, in the Past 1200 Yr

Geologic evidence of earthquakes at the Snohomish delta, Washington, in the past 1200 yr

Joanne Bourgeois* Department of Geological Sciences, University of Washington, Seattle, Washington 98195, USA Samuel Y. Johnson² U.S. Geological Survey, M.S. 966, Box 25046, Denver Federal Center, Denver, Colorado 80225, USA

ABSTRACT suggest that one set of these dikes formed 1996, 1999a; Pratt et al., 1997). Only the ca. A.D. 910±990; radiocarbon ages on a downgoing plate has produced large historic Exposed channel banks along distribu- younger set indicate a limiting maximum earthquakes (e.g., 1949 and 1965; Langston taries of the lower Snohomish delta in the age of A.D. 1400±1640. We also interpret a and Blum, 1977; Baker and Langston, 1987; Puget Lowland of Washington reveal evi- sharp lithologic change, from olive-gray, Chleborad and Schuster, 1998). dence of at least three episodes of liquefac- rhizome-rich mud to grayer, rhizome-poor Paleoseismologic studies in southwest tion, at least one event of abrupt subsi- mud, ϳ1 m above the couplet, to indicate a Washington suggest that seven great earth- dence, and at least one tsunami since ca. second abrupt lowering of the marsh sur- quakes (M Ͼ 8) have occurred at the Cascadia A.D. 800. The 45 measured stratigraphic face during an earthquake ca. A.D. 1040± plate boundary since ca. 3500 yr B.P. (Atwater sections consist mostly of 2±4 m of olive- 1400, but no conclusive liquefaction struc- and Hemphill-Haley, 1997), including the gray, intertidal mud containing abundant tures have been identi®ed at this horizon. most recent (M ϳ 9) event in A.D. 1700 (e.g., marsh plant rhizomes. The most distinctive Two distinctive coarse-sand laminae, 30±80 Nelson et al., 1995; Satake et al., 1996). In stratigraphic unit is a couplet comprising a cm below the couplet, may record tsunamis the Puget Lowland, Bucknam et al. (1992) -0.5؊3-cm-thick, laminated, ®ning-upward, older than A.D. 800. showed that a large (M Ͼ 7) earthquake oc tsunami-laid sand bed overlain by 2؊10 cm Thus, study shows that in the past ϳ1200 curred on the Seattle fault ca. A.D. 900±930 of gray clay. We correlated the couplet, yr, this part of Washington's Puget Low- (Atwater, 1999). Recognition of these events which is generally ϳ2 m below the modern land has been subjected to stronger ground has boosted estimates of regional seismic haz- marsh surface, across an ϳ20 km2 area. shaking than in historic times, since ca. ard (Frankel et al., 1996) and demonstrates the Sand dikes and sand-®lled cracks to 1 m 1870. need for a more complete catalog of the num- wide, which terminate upward at the cou- ber, frequency, sources, magnitudes, and ef- plet, and sand volcanoes preserved at the Keywords: earthquakes, paleoseismicity, Pu- fects of large earthquakes affecting the Puget level of the sand bed record liquefaction at get Lowland, sedimentology, stratigraphy. Lowland. the same time as couplet deposition. Differ- Here we report geologic evidence for paleoearthquakes from the Snohomish River delta ences in the type and abundance of marsh INTRODUCTION plant rhizomes across the couplet horizon, near Everett in the northern Puget Lowland as well as the gray clay layer, suggest that (Figs. 1 and 3), where there is minimal his- The densely populated Puget Lowland of compaction during this liquefaction led to torical information on earthquake effects. Del- Washington State (Figs. 1 and 2) occupies a abrupt, local lowering of the marsh surface tas are particularly good sites for paleoseis- dynamic geologic setting in the forearc of the by as much as 50±75 cm. Radiocarbon ages mologic investigations because their young, North American plate above the Cascadia sub- show that the tsunami and liquefaction date loosely packed sediments are prone to coseis- duction zone. Complex plate interactions from ca. A.D. 800 to 980, similar to the age mic liquefaction, compaction, and subsidence, along this convergent continental margin are of a large earthquake on the Seattle fault, and because their low elevation and proximity the driving force for a signi®cant, yet poorly 50 km to the south. to a water body make them susceptible to tsu- understood earthquake hazard (Ludwin et al., We have found evidence for at least two, namis. Bank exposures along tidal distributar- 1991; Rogers et al., 1996). Sources for mod- and possibly as many as ®ve, other earth- ies of the lower Snohomish delta show evi- erate or larger earthquakes include slip on the quakes in the measured sections. At two or dence of at least three liquefaction events, Cascadia plate boundary (interplate), ruptures more stratigraphic levels above the couplet, inundation by at least one tsunami, and at least in the downgoing Juan de Fuca plate 60 km sand dikes locally feed sand volcanoes. Ra- ϳ one episode of abrupt subsidence, all since ca. below Puget Sound (intraplate), and move- diocarbon ages and stratigraphic position A.D. 800. We infer that the most distinctive ment on shallow crustal faults in the North stratigraphic horizon (ca. A.D. 800±1100) re- *E-mail: [email protected]. American plate such as the Seattle fault or cords strong ground motion and tsunami in- ²Corresponding author; e-mail: sjohnson@usgs. southern Whidbey Island fault (Figs. 1 and 2) undation associated with the A.D. 900±930 gov. (Gower et al., 1985; Johnson et al., 1994, Seattle fault earthquake.

GSA Bulletin; April 2001; v. 113; no. 4; p. 482±494; 9 ®gures; 3 tables; Data Repository item 2001034.

For permission to copy, contact Copyright Clearance Center at www.copyright.com 482 ᭧ 2001 Geological Society of America PALEOSEISMOLOGY ON THE SNOHOMISH DELTA

SNOHOMISH RIVER DELTA

The Snohomish River begins at the con¯u- ence of the Skykomish and Snoqualmie Rivers and empties into Possession Sound (Fig. 1). At its lower end, the Snohomish ¯ows through a wide (ϳ4 km) postglacial valley bounded by morainal deposits of the last (Fraser) glacia- tion. About 12 km upstream from Possession Sound, at an elevation of Ͻ1.5 m above sea level, the main channel divides into several distributaries, or sloughs (Fig. 3). Spring-tide range in the sloughs is as much as 4.5 m. Snohomish delta lowlands are now primarily undeveloped wetlands or are used for agriculture. Lumber mills and storage, marinas, sewage treatment plants, and a hazardous waste site are also located on the delta plain. The delta is crossed by an interstate highway (I-5), the Burlington-Northern railroad, and a busy local highway. Agriculture on the delta is dependent on a system of dikes and levees, construction of which began in 1876 (Dunnell and Fuller, 1975). The cities of Everett and Marysville, ¯anking the Snohomish delta, were settled in the late 1870s and early 1880s. The northwestern part of the delta is on the Tulalip Indian Reservation. Cutbanks along both the main river channel Figure 1. Generalized geologic map of the Puget Lowland region showing location of and the sloughs in the lower delta typically ex- Snohomish River delta (SRD, area shown in Fig. 3) and selected regional crustal faults or pose, at maximum low tide, 1±4 m of strata de- geophysical lineaments (heavy dashed lines). Abbreviations: CÐCultus Bay; DMFÐDevils posited in about the past 1500 yr (Figs. 4 and Mountain fault; EÐEverett; LWÐLake Washington; OÐOlympia; PÐPossession Sound; 5). These strata, in general, record the building PSÐPuget Sound; SÐSeattle; SFÐSeattle fault; SJÐSan Juan Islands; SRÐSnohomish up of the delta, from channel point-bar deposits River; SKRÐSkykomish River; SNRÐSnoqualmie River; SWFÐsouthern Whidbey Is- to intertidal mud ¯ats to supratidal marsh de- land fault; TÐTacoma; WÐWest Point; WIÐWhidbey Island. Faults based on Gower posits. Marsh deposits predominate in studied (1985) and Johnson et al. (1996, 1999a). outcrops, and in-growth-position plant material and detrital wood debris are abundant in these sediments. Although tidal laminae are visible on weathered surfaces, they have been largely dis- rupted by bioturbation associated with modern and relict marsh vegetation. Snohomish River delta channels and marshes appear not to have migrated much since ca. A.D. 800. Since a United States Coast Survey hydrographic map was made in 1884, there has been neither signi®cant lateral migration of channels nor progradation of the mouth. Sedimentary facies, including fossil plant material, in cutbanks are at approximately the same elevations as facies currently being deposited at nearby locations. Thus sediment supply appears to be in approximate balance with the combined effects of subsidence and slow sea-level rise. Vertical aggradation rates Figure 2. Schematic view of the Paci®c Northwest continental margin (no scale) showing are ϳ2 m/1 k.y., based on our radiocarbon distribution of three types of earthquakes that affect the region. dating (Table 1; Fig. 4).

Geological Society of America Bulletin, April 2001 483 BOURGEOIS and JOHNSON

SNOHOMISH SEDIMENTARY ENVIRONMENTS AND FACIES

The lower, modern Snohomish delta comprises ®ve basic subenvironments, in succession from deeper to shallower: (1) subtidal channels, (2) lower intertidal ¯ats and point bars, (3) upper intertidal ¯ats and point bars, (4) supratidal marsh, and (5) lower delta plain and levees. This same succession is present in late Holocene facies in outcrop, produced as distributary channels migrate laterally and as the delta aggrades. Facies are distinguished primarily by sediment texture and color, growth-position plant fossils (see Table 2), and sedimentary structures. Our observations of the modern delta environments and the outcrop facies, including modern and fossil vegetation, are summarized in the GSA Data Re- pository (Figs. DR1±DR31).

Subtidal

The major subtidal channel environments of the modern delta (i.e., the axes of the main channel and its distributaries) are constantly submerged, so we did not directly observe and describe them. Lower point bars at low tide expose sand, minor gravel, and mud drapes. Thus we infer that the channel ¯oors are primarily sand and gravel. The subtidal facies is also rare in outcrop, presumably because it is below the present low tide level. By hand cor- ing, we encountered sand beneath exposed facies at a few localities. A sandy unit, inferred Figure 3. Map showing lower part of Snohomish River delta (Fig. 1 shows location). Dots to be a few meters below the surface, is the show localities where detailed stratigraphic information was collected. Dark lines show most likely source for sand-®lled dikes and freeway (Interstate 5) and major state highway (528). other liquefaction features described in this paper. Upper Intertidal in outcrops, and consists of bioturbated olive- Lower Intertidal gray mud with fossil vegetation in growth po- The modern upper intertidal zone is most sition. The most abundant plant fossils are Carex The modern lower intertidal zone has its prominent above lower intertidal ¯ats near the roots and stems, Triglochin rhizomes, and Scir- greatest areal extent near the delta front and delta front, and in upper point bar and lower pus rhizomes and stems. Detrital wood frag- along channel point bars. This subenviron- cutbank positions along distributary channels. ments are also common in this facies. ment is unvegetated and characterized by de- This subenvironment is characterized by depo- position of interbedded sand (®ne to coarse), sition of olive-gray mud and is commonly cov- Supratidal silt, and mud. The lower intertidal facies is ered with Carex lyngbyei. Other plants present, exposed at the base of many outcrops and in order of their typical ®rst appearance with in- The supratidal marsh, which makes up most along low banks adjacent to modern point creasing elevation in the tidal zone (Fig. DR1; of the modern lower delta area where not ar- bars. It comprises interlaminated silt, mud, see footnote 1), include Juncus balticus, Trig- ti®cially drained, is submerged only during and sand. lochin maritima, Lilaeopsis occidentalis, and extreme high tide and river ¯ooding, when Potentilla anserina. Deschampsia capitosa and mud may be deposited. The supratidal marsh 1GSA Data Repository item 2001034, stratigraph- Scirpus acutus, although most common on the surface is commonly littered with driftwood ic logs of upper Holocene strata and distribution of supratidal marsh, are locally present at intertidal and heavily vegetated, characterized by all of modern vegetation, Snohomish River delta, Wash- elevations. More salt-tolerant species such as the species in the upper intertidal zone plus ington, is available on the Web at http:// Distichlis spicata and Scirpus maritimus are pre- Deschampsia capitosa, thick stands of Scirpus www.geosociety.org/pubs/ft2001.htm. Requests may also be sent to Documents Secretary, GSA, sent as fossils in some intertidal-facies outcrops, acutus, and Typha latifolia, Rumex sp., and P.O. Box 9140, Boulder, CO 80301; e-mail: but were not observed on the modern marshes. unidenti®ed grasses. The upper ϳ30 cm of [email protected]. The upper intertidal facies is the dominant facies soil below the marsh surface is typically

484 Geological Society of America Bulletin, April 2001 PALEOSEISMOLOGY ON THE SNOHOMISH DELTA

overlies peaty mud in sharp contact and is clearly associated stratigraphically with sand dikes, sand volcanoes, and sand-®lled cracks. In the following we describe the sedimentary succession that includes this horizon and discuss its origin.

Event B Stratigraphy

Facies Below the Event B Sand-Clay Couplet The event B couplet was deposited on a vegetated surface at most localities. It is typically underlain by olive-gray (5Y4/1, 5Y3/2, 5Y3/1) mud, which contains plant fossils of Carex, Triglochin, and Scirpus acutus in growth position. At several localities, particularly those farther upstream, this mud is more peaty, more reddish (5YR3/2; e.g., locality 11, Fig. 3), or contains roots of Sitka spruce and other trees and shrubs (e.g., localities 32, 29, Figure 4. Composite drawing of sediments exposed in banks of Snohomish River delta 27, Fig. 3). The mud is cut by sand-®lled showing stratigraphic distribution of material used for dating and the calibrated radio- dikes at many localities. carbon dates (calendar years A.D., 2␴, 95% probability; see Table 1 and Fig. 9). Horizons Surveying at ®ve localities revealed that re- with features (e.g., sand dikes, sand volcanoes, prominent color changes) indicative of lief on the surface below the event B couplet prehistoric earthquakes are labeled A1, A2, B, C, D, and E. over lateral distances of 15Ϫ70 m is typically ϳ10 cm and ranges to 40 cm. We also sur- veyed the differences in paleoelevation of the weathered and oxidized. In bank exposures, PALEOSEISMOLOGY event B couplet between some localities. For the supratidal marsh facies is distinguished example, the couplet was deposited on a Scir- from the upper intertidal facies by its greater There is evidence at the lower Snohomish pus-rich surface at locality 8, 60 cm higher organic content, more pervasive fossil roots, delta for several prehistoric, late Holocene than the contemporaneous surface at locality and brownish to reddish mottling. Individual earthquakes (labeled AϪE from oldest to 6 where the couplet was deposited on a Trig- plant fossils are dif®cult to identify in this fa- youngest) based on our examination of out- lochin-rich surface. cies. Where supratidal marsh deposits have crops at ϳ45 localities (Figs. 3, DR2, and subsided below the water table into a more DR3 [see text footnote 1]). In the following Event B Sand Bed reducing environment, some of the mottling we describe and discuss each of these events The thin sand bed typically forms a distinctive, laterally continuous sheet. At several has been destroyed, and this facies is dif®cult in order of the strength of the evidence sup- places (e.g., localities 6, 8, 21, 22, Fig. 3), the to distinguish from upper intertidal facies. porting a paleoseismic interpretation. We consider events B, C, and E, all with liquefaction, sand is laterally continuous for at least 30±50 more reliable than events A1, A2, and D. We m; a notch formed by erosion of this sand ex- Delta Plain report the latter three events to help develop tends many tens of meters along slough cut- a regional catalog of possible paleoseismic banks. The generally ®ne- to medium-grained sand bed ranges in thickness from a few mil- The lower delta plain subenvironments, indicators. limeters to ϳ5 cm, and both thins and ®nes where not signi®cantly altered by modern ag- EVENT BÐEVIDENCE FOR A in the upstream direction (Fig. 7). It is com- riculture, is characterized by immature soils TSUNAMI, LIQUEFACTION, AND monly thicker in paleoswales and thinner over developed on a peaty mud substrate. Surfaces ABRUPT SUBSIDENCE what were topographic highs, such as logs. are vegetated by grasses, as well as shrubs and The sand is commonly graded and contains trees, including primrose, alder, crabapple, Nearly all high cutbanks along the four one or two olive-gray silty laminae toward its cottonwood, willow, and blackberries. The channels studied contain a distinctive, laterally top. It was typically deposited on a vegetated highest surfaces on the areas of the lower delta continuous stratigraphic horizon, which typi- surface of sedges, rushes, grasses, and other we studied are also occupied by a native co- cally weathers as a prominent to subtle, hori- herbaceous plants. nifer, Sitka spruce (Picea sitchensis), which zontal or subhorizontal notch in the outcrop, commonly appears unhealthy, with dead and ranging from 1.5 to 2.5 m below the modern Event B Gray Clay dying foliage. In outcrop, the delta plain facies marsh surface (Figs. 4±6). We used this ho- The event B couplet gray clay is generally is most easily distinguished by the presence rizon as our primary correlation unit. This ho- ϳ5 cm thick (range of 0 to ϳ20 cm), and is of growth-position root systems and trunks of rizon is typically a couplet composed of sand thickest in local paleoswales. Unlike the un- trees and woody shrubs. Detrital wood frag- (1±2 cm thick) overlain by gray clay (2±10 derlying sand, the gray clay shows no system- ments are also common. cm thick) (Fig. 6A). The sand-clay couplet atic variation in thickness across the study

Geological Society of America Bulletin, April 2001 485 BOURGEOIS and JOHNSON

Figure 5. Stratigraphic section at locality 21 (Fig. 3), Ebey Slough. Photograph shows middle part of 75 m exposure described at this locality. Lower arrow in photograph points to event B couplet. Upper arrow points to base of prominent color change associated with event D. Vertical scale is on line drawing. area. The most distinctive aspects of this layer Haley and Lisa Hodges, 1998, written commu- and was at least partially ®lled from below by are its color and general lack of fossil plant nication; localities 2, 4, 8, 17; all on Steamboat sand (Fig. 4). A vertical sand dike with an material. It is commonly a medium light gray Slough) indicates that both the mud and gray irregular margin ®lls one of these lateral (N5.5) silty clay, markedly contrasting with clay contain similar assemblages dominated by spreads (Fig. 8A). The top of this spread underlying and overlying olive-gray mud. The tidal marsh species, and the dike sand and cou- formed a paleoswale that was ®lled by a thick base of the gray clay is typically sharp, and plet sand contain many subtidal species. No dis- layer of couplet gray clay. At locality 2, a dike the lower portion of the bed commonly con- tinct change in diatoms was observed from be- below the couplet ends at the couplet horizon tains thin laminae of sand or coarse silt. The low to above the couplet. in a mounded lens of sand. Because the gray gray clay grades upward to olive-gray mud. clay above this sand lens thins over the sand Event B Interpretation lens, we interpret this feature to be a sand vol- Facies Above the Couplet cano and not a sill. This distinction is impor- The olive-gray mud that typically overlies the Evidence for Liquefaction and Ground tant because the sand member of the couplet, couplet is generally indistinguishable in color Failure Contemporaneous with Deposition being one of the mechanically weakest hori- and lithology from mud below the couplet. of the Couplet zons in the mud-dominated section, is intrud- However, at many sites, there is a change in Liquefaction features (cf. Obermeier, 1996; ed by sills at a few localities. At locality 17, plant fossils from below to above the couplet Obermeier and Pond, 1999), particularly sand for example, at least some of the abundant liq- (summarized in Table 3; e.g., Triglochin [and dikes (Figs. 4 and 8), are common in outcrop uefaction features are convincingly contem- Carex] below to Carex only [no Triglochin] along the lower Snohomish distributaries. poraneous with the couplet, but others, includ- above, Sitka spruce below to no spruce above), They range in width from a few millimeters ing sand sills intruded along the sandy part of generally indicating a lowering of the surface to Ͼ1 m, and contain ®ne to medium sand. the couplet horizon, are due to later (discussed in interpretation section). At some lo- Most dikes pinch out upward within the mud remobilization. calities that were delta-plain surfaces before section, making it possible to establish only event B (e.g., 10, 11, 27, 32, Fig. 3), the sedi- their lower age limit from the age of the high- Evidence for Abrupt Local Subsidence at ment above the couplet is less peaty or muddier est material intruded. Some dikes clearly cut the Stratigraphic Level of the Event B than sediment below the couplet. through and are thus younger than the couplet Couplet horizon. However, at four localities (2, 5, 6, Although there is rarely a lithologic change Benthic Diatoms Associated with the 17, Fig. 3) and probably at four others (21, from below to above the couplet, we interpret Event B Couplet 25, 32, 22, Fig. 3), dikes terminate in sand the changes in fossil vegetation (Table 3) to Benthic diatom assemblages re¯ect the salin- volcanoes at the couplet horizon, or occupy indicate localized subsidence generated by ity and substrate preferences of dominant spe- lateral spreads that terminate there; these fea- earthquake-induced compaction and liquefac- cies. Benthic diatom species are present in the tures were produced contemporaneously with tion. We attempted to estimate the magnitude couplet sand and gray clay, in the underlying the couplet. of this lowering from the thickness of sedi- and overlying mud, and in the associated dike At localities 5 and 6, ground failure of the ment between the disappearance of a fossil sand. Cursory study of diatom populations in paleosurface is indicated by lateral spreads, plant just below the couplet and its reappear- samples from these sediments (Eileen Hemphill- places where the marsh surface cracked open ance above the couplet.

486 Geological Society of America Bulletin, April 2001 PALEOSEISMOLOGY ON THE SNOHOMISH DELTA

TABLE 1. RADIOCARBON DATES FROM THE SNOHOMISH RIVER DELTA

Sample Radiocarbon Field Radiocarbon 13C/12C Calibrated Material Comment number laboratory locality age (½) age dated number (14C yr B.P.) SJ-98-1 122328 21 400 Ϯ 40 Ϫ25.9 A.D. 1430±1530, Bark-free twig Limiting maximum age, event E liquefaction 1550±1640 SJ-98-2 120284 21 430 Ϯ 70 Ϫ25.7 A.D. 1400±1640 Detrital wood Limiting maximum age, event E liquefaction and limiting minimum age of event D gray clay bed SJ-97-9 109811 21 790 Ϯ 60 Ϫ26.1 A.D. 1040±1090, Triglochin rhizomes Limiting maximum age of event D gray clay bed 1120±1140, 1150±1310 SJ-97-2 109810 16 920 Ϯ 50 Ϫ28.2 A.D. 1020±1220 Scirpus stems, rhizomes Limiting minimum age of event C liquefaction SJ-98-5 120285 21 1270 Ϯ 70 Ϫ25.5 A.D. 640±900, Triglochin rhizomes Limiting minimum age of event C liquefaction 920±950 SJ-97-1 109809 16 1120 Ϯ 50 ±30.4 A.D. 780±1020 Carex stems, rhizomes Limiting maximum age of event C liquefaction and limiting minimum age of event B couplet SJ-96±1 93255 4 1190 Ϯ 80 Ϫ25.0 A.D. 680±1000 Triglochin rhizomes Limiting minimum age of event B couplet SJ-96-20 104857 6 1240 Ϯ 70 Ϫ29.0 A.D. 660±900, Triglochin rhizomes Limiting maximum age of event B couplet 910±970 SJ-96-33 104858 9B 1120 Ϯ 50 Ϫ28.9 A.D. 780±1020 Triglochin rhizomes Limiting maximum age of event B couplet SJ-96±6 104856 1 1280 Ϯ 70 ±30.4 A.D. 640±900, Triglochin rhizomes Limiting maximum age of event B couplet 920±940 SJ-98-8 120286 32 1270 Ϯ 50 Ϫ28.3 A.D. 660±890 Picea sitchensis root Limiting maximum age of event B couplet M31A² QL-4920 27 1159 Ϯ 22 Not reported A.D. 800±990 Picea sitchensis stump Approximate age of event B couplet M31B² QL-4921 27 1211 Ϯ 22 Not reported A.D. 800±820 Picea sitchensis stump Approximate age of event B couplet and 850±980 JB-97-11 109814 8 1520 Ϯ 50 -27.3 A.D. 420±640 Triglochin rhizomes Approximate age of event A1 sand bed SJ-96±37 109808 11 1720 Ϯ 70 -29.3 A.D. 130±440, Carex rhizomes Limiting minimum age of event A2 sand bed 450±470, 500± 530 Notes: Unless indicated, laboratory data are from Beta Analytic Inc. Table 2 provides common name for plant material dated. Dates on Triglochin were obtained from the leaf-base and stem-base parts of the rhizome. Ages calibrated using OxCal v. 3.0 (Ramsey, 1995) and INTCAL98 radiocarbon age calibration (Stuiver et al., 1998). *Calendar year, 95.4% con®dence. ²Samples collected by Brian Atwater for high-precision dating at the University of Washington. M31A is rings 13±31 of small tree (ring 1 adjoins bark); M31B is rings 75±84 from larger stump. Calibrated high-precision ages include lab-recommended error multiplier of 1.6 and incorporate offset based on sampled tree rings.

TABLE 2. TYPICAL MARSH VEGETATION (GENERA AND SPECIES) ON THE SNOHOMISH RIVER DELTA

Latin name Common name Fossil characteristics Common habitat, observations, interpretation* Carex lyngbyei Lyngby sedge Dull gold ¯attened rhizomes and stems Pioneer on upper intertidal marsh, to supratidal, very common Deschampsia cespitosa Tufted hairgrass Not observed Supratidal (high marsh), brackish tolerant Distichlis spicata Seashore saltgrass Bright golden, woody, segmented rhizome Upper intertidal, saline indicator Juncus balticus Baltic rush Tough, thin rhizome, acute angle to stem Upper intertidal to supratidal, sometimes pioneer on low marsh Lilaeopsis occidentalis Western lilaeopsis Not observed Upper intertidal and supratidal, salt and brackish marshes Picea sitchensis Sitka spruce Tree root systems Pioneer conifer on delta plain, limited salt tolerance Potentilla anserina Paci®c silverweed Not observed Uppermost intertidal and supratidal marshes Rumex sp. Dock Not observed Supratidal, wet meadows, may tolerate brackish water Scirpus acutus Hardstem bulrush Large, dark red-brown rhizomes, round stems Abundant on supratidal marshes, present on emergent marshes, may tolerate brackish water Scirpus maritimus Seacoast bulrush Bulbous stem base, triangular stem Upper intertidal, high brackish or salt marshes Thuja plicata Western red cedar Not observed Delta plain, less tolerant of ¯ooding and salt than Sitka spruce Triglochin maritima Seaside arrowgrass Golden, v-shaped leaf bases Upper intertidal and supratidal, succeeds Carex on Snohomish Typha latifolia Common cattail Not observed Supratidal freshwater marshes, dense patches, associated with Scirpus acutus *Based on ®eld observations and Cooke (1997).

Locality 6 provides an example of an event land level back up to where Triglochin could lowering was probably variable across the del- B fossil-vegetation change. The normal fossil reestablish. ta, negligible at some sites, and 0.5 m or more plant succession in outcrop is Carex → Trig- At localities where Sitka spruce roots and at others. Among 28 localities where strati- lochin → Scirpus acutus (see also Table 2). trunks are present at and below the couplet graphic and plant-fossil successions bounding At locality 6, the Carex → Triglochin transi- (e.g., 27, 29, 32; Figs. 3 and 6B), spruce dis- the couplet were examined, 15 showed plant- tion is just below the couplet, fossil Triglochin appears above the couplet and then reappears fossil evidence of an abrupt lowering of land below the couplet abruptly disappears at the 30±100 cm higher in the section. This rela- level, and 13 provided no plant-fossil evi- couplet horizon (replaced above by Carex), tionship suggests that trees were killed by sub- dence of change (Table 3); no locality showed and Triglochin reappears in the section 50±75 sidence, which increased submergence or salt evidence of uplift. The absence of evidence cm above the couplet. We interpret that the exposure. New trees began to grow when sed- for subsidence at many localities may indicate paleosurface at locality 6 dropped perhaps 50± iment aggradation on the ¯ood plain reached no land-level change or a change that occurred 75 cm, below the level where Triglochin could a level where standing water or salt exposure within a particular plant's tolerance zone. The thrive, and Carex reoccupied the surface until was suf®ciently diminished. localized and variable subsidence at the time enough sediment accumulated to bring the Our cumulative evidence suggests that land of event B appears to be similar to that of the

Geological Society of America Bulletin, April 2001 487 BOURGEOIS and JOHNSON

Figure 6. (A) Event B couplet at locality 4 on Steamboat Slough (Fig. 3). Arrow points to sand bed at base of couplet. T shows Triglochin rhizome. Note color change from darker mud below sand bed to lighter clay (the couplet gray clay) above the sand bed. Knife handle is 12 cm long. (B) Event B sand bed at locality 32 on the Snohomish River. Horizontal arrow points to event B couplet. Vertical arrow points to location of tree root sampled for radiocarbon dating (SJ-98±8; Table 1). Shovel handle is 50 cm long.

lometers of outcrop, and it thins, ®nes, and ultimately disappears in the upstream direction (Fig. 7). These features suggest a delta- wide submergence by a wave or waves di- rected upstream from Possession Sound. The protected geographic position of the Snohom- ish delta and the limited size and fetch of Pos- session Sound argue against a storm-surge origin for these waves. That the sand bed is graded and its top is indistinctly laminated, indicative of rapid deposition from suspen- sion, further supports a tsunami origin. The sand was transported across a vegetated marsh surface where resuspension is nearly impos- sible, thus requiring initial rapid advection. In addition, no other sand of comparable sheet- like geometry or grain size occurs in the post ca. A.D. 700 bank exposures on the Snohom- ish delta, indicating that the sand was deposited by a rare event and not by normal ¯oods. Figure 7. Graphs showing upstream change in thickness (vertical bars) and grain size The presence of subtidal diatoms in the cou- (solid dots) of the event B sand bed. plet sand is also consistent with a tsunami interpretation. Perhaps most convincing, depo- 1964 plate-boundary earthquake in south-cen- the southern Whidbey Island fault, is 13 km sition of the sand was contemporaneous with tral Alaska, where compaction- and liquefac- to the southwest. Deformation models (e.g., formation of liquefaction structures, which in- tion-induced subsidence of a few centimeters Stein and Yeats, 1989) suggest that earth- dicate strong shaking. Although the sand layer to more than 1 m occurred in many estuarine quake-generated slip of more than 4±5 m on may have a source locally in this erupted sand, and alluvial environments (Plafker and Kacha- this fault would be required to generate 50 cm the layer's widespread distribution, thinness, doorian, 1966; McCulloch and Bonilla, 1970). of tectonic subsidence at the Snohomish delta. occurrence on vegetated surfaces, and pres- Tectonic subsidence is highly unlikely giv- There is no paleoseismologic evidence for an ence in areas where no liquefaction structures en the inferred amount of subsidence (locally event of this size and age (see following) on could be found lead us to believe that the ma- more than 50 cm) and the location of the Sno- this nearby fault. jor sediment source was sand suspended from homish delta. Elastic-dislocation models of adjacent subtidal channels during a tsunami great subduction zone earthquakes predict Origin of the Couplet Sand surge. Finally, bracketing radiocarbon ages only ϳ10 cm of coseismic tectonic subsidence Several lines of evidence suggest that the overlap ages of a large Seattle fault earthquake at this distance from the plate boundary (e.g., couplet sand bed is a tsunami deposit and not (see following), which was shown to have Wang et al., 1994; W.D. Stanley, 2000, written the deposit of a river ¯ood or storm surge. The produced a northward-moving tsunami that commun.). The nearest known crustal fault, sand forms a thin, widespread layer over ki- deposited sand at Cultus Bay and West Point

488 Geological Society of America Bulletin, April 2001 PALEOSEISMOLOGY ON THE SNOHOMISH DELTA

TABLE 3. VEGETATION CHANGES AT EVENT B HORIZON root of a spruce tree that died at the time of Locality Vegetation below event B Vegetation above event B Interpretation of vegetation the event, and three from Triglochin rhizomes number (evidence of drop, no drop) sampled within 10 cm below the base of the Steamboat Slough couplet, and with two additional limiting min- 6 Triglochin Carex; T at ϩ75 cm Drop imum ages from Triglochin and Carex plant 8 Scirpus acutus Carex Drop 17 Carex No Carex, Carex at ϩ75 cm Drop material slightly above the couplet. 9 Triglochin Triglochin, Scirpus acutus* No drop 1 Triglochin Carex and Scirpus acutus* Drop 2 S, T, and C Scirpus acutus No drop, normal succession? Correlation With a Large Earthquake on 3 Carex Carex; Triglochin at ϩ25 cm No drop, normal succession the Seattle Fault 4 Carex, Scirpus acutus? Triglochin, Scirpus acutus?* No drop, normal succession? The A.D. 850Ϫ980 range includes the time 5 Carex Carex No drop 11 Peaty, reddish mud Olive-gray mud Drop, but no obvious roots or rhizomes of a large earthquake on the Seattle fault 10 Reddish woody peat Olive-gray mud Drop, but no obvious roots or rhizomes (Bucknam et al., 1992), the age of which has 28 Scirpus acutus at Ϫ50 Spruce at ϩ40 cm Normal succession? 27 Spruce Peaty mud, spruce at ϩ100 cm Drop recently been narrowed to A.D. 900±930 (At- water, 1999). This earthquake produced a Union Slough 13 Yellowish-brown peaty Scirpus acutus*? Drop large, northward-moving tsunami in Puget 15 Carex ϩ Triglochin Carex, Scirpus acutus* Drop Sound (Atwater and Moore, 1992), which we 36 Scirpus acutus Scirpus acutus No drop infer deposited the event B couplet sand at the 35 Scirpus acutus Scirpus acutus No drop 29 Spruce No spruce; spruce at ϩ30 cm Drop Snohomish delta. Although Bucknam et al. Ebey Slough (1992) and Sherrod (1998) found evidence for 21 Carex, Triglochin Carex, Triglochin No drop other crustal earthquakes at about this time in 22 Scirpus acutus Carex; Scirpus at ϩ100 cm Drop the southern Puget Lowland, these earth- 23 Carex Carex, Scirpus acutus* No drop? 24 Carex, Scirpus acutus Carex, Scirpus acutus No drop quakes apparently did not produce tsunamis in 25 Scirpus acutus Scirpus acutus No drop Puget Sound. The inferred magnitude of the 26 Carex Carex No drop Seattle fault earthquake (M Ͼ7), based on oth- Snohomish River er considerations (Bucknam et al., 1992), 33 Triglochin Triglochin No drop would have been suf®cient to generate lique- 32 Spruce No spruce, Triglochin at ϩ60 cm Drop 31 Scirpus, Triglochin, Juncus Triglochin, Distichlis Drop faction and ground failure on the Snohomish Notes: CÐCarex lyngbei; TÐTriglochin maritimum; SÐScirpus acutus (see Table 2). ``ϩ 75 cm'' indicates delta 50 km away (Ambraseys, 1988; Ober- plant fossils appear 75 cm above base of Event B couplet. Carex is virtually always present; it is generally noted meier and Pond, 1999). here only when by itself or most abundant. Scirpus* ϭ interpreted Scirpus pioneer, out-of-place in normal suc- Other possible sources of earthquakes that cession, which is C→ T→ S. Interpretation of ``drop'' or ``no drop'' in this table is based mostly on fossil vegetation. Most ``no drop'' localities contain a thick gray clay bed at the top of the Event B couplet, which we also consider could have produced event B features include to be evidence of abrupt subsidence. the southern Whidbey Island fault (Johnson et al., 1996) 13 km to the southwest and the Devils Mountain fault (Gower et al., 1985; Johnson et (Fig. 1; Atwater and Moore, 1992). At these The following are arguments against an alter- al., 1999b) ϳ35±40 km to the north (Fig. 1). localities, the deposit is typically a graded native explanation that we considered, i.e., that However, no late Holocene earthquakes have yet sand layer, and at West Point it is in places the gray clay is part of the tsunami deposit. First, been attributed to these structures. Tsunami de- overlain by gray clay (B.F. Atwater, 1999, per- unlike the couplet sand, there is no systematic posits from large offshore plate-boundary earth- sonal commun.). geographic variation in thickness or grain size of quakes (e.g., Atwater and Hemphill-Haley, the gray clay. Second, the base of the clay is 1997) have also not been recognized in Puget Origin of the Couplet Gray Clay sharp; the lower part of the gray clay commonly Sound, and tsunami models (Murty and Heben- Subsidence that accompanied event B (see has one or two ®ne-sand or silt laminae, but they streit, 1989) suggest that they would not signif- Table 3) was probably the key control on de- are easily distinguished from laminae in the un- icantly inundate this region. Deep earthquakes position of the gray clay. The absence of plant derlying sand unit, which are olive-gray. Finally, in the downgoing plate (Fig. 2) would not offset fossils is the key distinction between this unit diatoms in the gray clay are normal tidal marsh the sea¯oor and are thus not primary tsunami and bounding mud units. In our interpretation, diatoms, resembling assemblages in muds below sources. subsidence created uneven accommodation and above. It is possible that the A.D. 900±930 earth- space across the delta, which was rapidly quake on the Seattle fault or another large ®lled by deposition of clay transported by nor- Age of Event B earthquake around that time triggered large mal tidal currents and ¯oods. This rapid de- Two high-precision radiocarbon dates on landslides along Possession Sound (Fig. 1), position preceded and precluded recoloniza- stumps of spruce trees inferred to have been which could displace suf®cient water to pro- tion of normal marsh vegetation. The variable killed by earthquake-induced subsidence in- duce a tsunami large enough to submerge the thickness of the gray clay re¯ects both its de- dividually suggest that event B occurred be- Snohomish delta. Chleborad (1994) cited ac- positional thickness and the amount of post- tween A.D. 800 and 980 (Table 1; Fig. 9). counts of a 2.5-m-high so-called tidal wave depositional alteration (to olive-gray mud) by When these two dates are combined using caused by a large landslide near Tacoma (Fig. plant colonization and pedogenesis. Thus, the OxCal (Ramsey, 1995), the age range narrows 1) three days after the 1949 (M 7.1) deep thickness of gray clay does not provide an ac- to A.D. 850Ϫ980 (95% probability). This age earthquake between Tacoma and Olympia. curate indication of the amount of subsidence range is consistent with four other limiting There is evidence for large paleolandslides in at speci®c sites. maximum ages, one from the distal end of the Possession Sound (based on high-resolution

Geological Society of America Bulletin, April 2001 489 BOURGEOIS and JOHNSON

limiting ages for the Seattle fault event are correct, statistical analysis using OxCal (Ram- sey, 1995) of the sequence of age distributions bracketing events B and C indicates (95% probability) that event C occurred between A.D. 910 and 990. Event C liquefaction might have been caused by an earthquake generated on a crustal fault, on the plate-boundary thrust fault, or within the downgoing plate (Fig. 2). As noted here, there is currently no evidence for late Holocene rupture on the potentially active crustal faults (southern Whidbey Island fault, Devils Moun- tain fault) that are closest to the Snohomish delta (Fig. 1). In the southern Puget Lowland, however, Bucknam et al. (1992) and Sherrod (1998) described regions of abrupt uplift and subsidence that also occurred ca. A.D. 900, and argued that the pattern of crustal movement supports at least one large upper crustal earthquake other than the ca. A.D. 900 Seattle fault event at about this time in the region. The inferred sources for these additional possible crustal earthquakes are Figure 8. (A) Irregular wall of 1-m-wide sand dike at locality 5, Steamboat Slough. Note ϳ75Ϫ120 km away from the Snohomish delta; mud clasts near dike margins. (B) Thin sand dikes (shown by arrows) cutting mud exposed liquefaction at these distances would require on horizontal surface at locality 16 on Steamboat Slough. earthquakes of minimum magnitude ϳ6.5Ϫ7 (Ambraseys, 1988; Obermeier and Pond, 1999). Event C liquefaction could also have been seismic re¯ection data; Karlin et al., 1996), as 1 cm thick. The sand lens in the upper ho- caused by a great, plate-boundary earthquake. but they have not been dated. rizon (event E) extends laterally for 2 m and Based on work along the Paci®c Coast of is as thick as 2.5 cm. We consider these lenses southwest Washington, Atwater and Hem- EVENTS C AND EÐEVIDENCE FOR as sand volcanoes, rather than sills, because phill-Haley (1997) documented a subduction- LIQUEFACTION the layers are thin and irregular, appear to zone event that occurred between ca. A.D. 700 drape growth-position plants, and occur within and A.D. 1100. There is no consensus on the Sand Dikes, Volcanoes, and Sills horizons where there is no mechanically weak strength of ground motions such an event zone such as a preexisting sand layer that could produce as far inland (ϳ170 km) as the Of 33 sites where we described a section could control their stratigraphic position. Snohomish delta (e.g., Obermeier, 1995; Silva that included the event B couplet, at least six Locality 16 also contains liquefaction struc- et al., 1998; Cohee and Somerville, 1998). To include sand dikes that cut through, or are pre- tures younger than event B, but we did not date, no one has conclusively linked paleoli- sent above, the couplet. These dikes are gen- recognize any sand volcanoes. We found a 5- quefaction structures this far inland to a plate- erally 0.5 to ϳ3 cm wide and penetrate a few cm-wide sand dike connected to a 2-m-long boundary earthquake. tens of centimeters of section. In a few cases, sill (3±5 cm thick), in strata younger than the It is also conceivable that event C liquefaction sand from event B appears to have been re- couplet. We determined the age of this sand was caused by a strong, deep, intraplate earth- mobilized; in addition, some sills have been sill by dating fossil plant material directly be- quake (Fig. 2) similar to the 1949 M 7.1 Olym- intruded along the event B sand horizon. low the sill (limiting maximum age), and fos- pia and 1965 M 6.5 Seattle-Tacoma events These sills are distinguished from the inferred sil marsh-plant stems that grew through the (Langston and Blum, 1977; Baker and Langston, tsunami sand by a lack of lamination, more sand sill (limiting minimum age) (Table 1; 1987). Each of these earthquakes produced local variability in thickness, and their association Figs. 4 and 9). On the basis of the similarity sand boils and other evidence of liquefaction and with sand feeder dikes. of the two ages (see following), we tentatively ground failure in southern and central Puget As previously noted, the age of liquefaction correlate the sill-forming liquefaction event at Sound above their hypocenters (Chleborad and features can be dif®cult to determine because locality 16 with event C at locality 21. Schuster, 1998). Neither earthquake produced they pinch out upward or emerge from plane liquefaction at the Snohomish delta. However, a of the outcrop below the level of the land sur- Age of Event C and Possible Earthquake deep, liquefaction-producing, intraplate earth- face at the time of intrusion. However, at lo- Sources quake could have occurred below the Snohom- cality 21 at Ebey Slough (Fig. 3), we found ish delta ca. A.D. 910Ϫ990. Because this type two different horizons above the event B level Stratigraphic relationships (Fig. 4) indicate of earthquake will not result in tectonic changes where sand dikes feed horizontal to mounded that event C liquefaction is younger than event in land level or a tsunami, liquefaction may be sand lenses (Fig. 4). The sand lenses at the B and thus postdates A.D. 900, the inferred its only record. lower of these two horizons (event C) extend limiting maximum age for the large Seattle Given regional evidence for one or more, laterally for as much as 1 m, and are as much fault event (Atwater, 1999). Assuming that the ca. A.D. 900, strong upper crustal earthquakes

490 Geological Society of America Bulletin, April 2001 PALEOSEISMOLOGY ON THE SNOHOMISH DELTA

Figure 9. Probability distribution histograms (using OxCal v. 3.0, Ramsey, 1995; based on Stuiver and Reimer, 1993) of the calibrated radiocarbon ages of the Snohomish River samples (Table 1) compared to known or inferred paleoseismic events in Puget Sound and the Paci®c coast of southwest Washington. Geologic data from the southern Puget Lowland area (SP) suggest that more than one crustal earthquake shook the region about the time of the Seattle fault (SF) earthquake (Bucknam et al., 1992; Sherrod, 1998; Atwater, 1999). LW events indicate approximate time (uncertainty is unknown) of turbidite deposition in Lake Washington, which Karlin and Abella (1996) inferred is related to earthquake-induced slumping. F indicates time of signi®cant earthquake-induced liquefaction on the Fraser delta (Clague et al., 1997). Y, W, U, and S are buried soils in southwest Washington inferred to record rapid subsidence during subduction-zone earthquakes (Atwater and Hemphill-Haley, 1997). in the southern Puget Lowland (Bucknam et boundary thrust fault, or within the downgoing D. Generally, event D is recorded by a well- al., 1992; Sherrod, 1998), it seems most likely plate (Fig. 2). de®ned, anomalous abrupt stratigraphic to us that event C liquefaction was caused by There is no evidence of rupture on any of change from more olive-colored, massive, such an earthquake. If so, our evidence indi- the active or potentially active crustal faults of plant-rich sediment below to grayer, more cates that at least one of these earthquakes the Puget Lowland (Fig. 1) in the past 450 yr. laminated, less plant-rich sediment above closely postdates the Seattle fault event. Given the nascent status of regional paleo- (Fig. 4). Fossil vegetation changes at this seismologic investigations, however, a large, sharp lithologic contact are similar to those Age of Event E and Possible Earthquake crustal-fault earthquake in the Puget Lowland observed at the event B horizon (Table 3). For Sources at this time cannot be ruled out. We consider example, at locality 21, Triglochin and Scir- it unlikely that event E liquefaction is a distal pus disappear at the contact and Scirpus re- Two radiocarbon ages on wood provide lim- effect of the great A.D. 1700 Cascadia earth- appears 60 cm above the contact (Fig. 5). At iting maximum ages on event E liquefaction (Ta- quake. Radiocarbon ages for this plate-bound- locality 2, Scirpus disappears at the contact ble 1; Figs. 4 and 9). One (calibrated age of A.D. ary earthquake (Nelson et al., 1995) are gen- and reappears 50 cm above the contact. At lo- 1400Ϫ1640) was from a 2-cm-diameter detrital erally younger than the age we infer for event cality 4, Carex, Triglochin, and Scirpus occur wood fragment 11 cm below the sand volcano, E, and there is no evidence that the A.D. 1700 below the contact but only Carex occurs the other (calibrated age of A.D. 1430Ϫ1640) Cascadia earthquake produced liquefaction in above. At locality 29, spruce tree roots occur was from a 2-mm-diameter piece of bark-free the Puget Lowland. below the contact but disappear above it. Our twig within the sand volcano. Because of the correlation of the event D horizon across the fresh appearance and delicate nature of this twig, EVENT DÐPOSSIBLE ABRUPT delta is tentative, based on lithologic similar- we think that this younger date may closely ap- SUBSIDENCE ity and on stratigraphic distance above event B, typically 50±100 cm. proximate the time of liquefaction; therefore we tentatively infer an age of ca. A.D. 1430±1640 Sharp and Anomalous Stratigraphic Change Inferred Rapid Subsidence for event E. As with event C, the liquefaction may have been forced by a prehistoric earth- Outcrops at 10 or more localities display We interpret the lithologic change at event quake generated on crustal faults, on the plate- the stratigraphic horizon that we label event D as possible evidence of rapid subsidence as-

Geological Society of America Bulletin, April 2001 491 BOURGEOIS and JOHNSON sociated with a prehistoric earthquake. Evi- which they suggest may have been caused by lowing criteria. First, sand layers of this grain dence includes the noted color change from earthquake-induced slumping at the lake mar- size and thickness are anomalous in upper in- olive-gray to more gray and changes in fossil gin. At present, this is the only possible earth- tertidal ¯at and supratidal marsh facies of the vegetation. In a normal aggrading succession, quake in the Puget Lowland that may correlate Snohomish delta. Second, the sand appears to grayer facies are deeper in the section and rep- with event D. ®ne and then disappear in the landward direc- resent a topographically lower environment. tion, consistent with upstream sediment trans- Interpretation that this subsidence was abrupt EVENTS A1 AND A2ÐPOSSIBLE port. Third, the sand, although only present in is based on the unusually sharp boundary at TSUNAMI DEPOSITS patches, appears to have a thin, sheet-like the base of this horizon, and the occurrence of character. Finally, the sand is present close to Below the event B couplet, we found two fossil-poor gray clay, which we interpret to the mouth of Steamboat Slough, which opens unusual sand layers. These sand layers are indicate rapid ®lling of newly created accom- directly to Possession Sound. only exposed at the base of a few outcrops at modation space (as for event B). We also speculate that the A1 sand is a tsu- lowest tides, and they have only been docu- Alternatively, this layer could have been nami deposit on the basis of its unusual grain mented at a few of our localities. Neverthe- produced by normal subsidence coupled with size, deposition within marsh facies, and its less, it is important to describe them and spec- a change in ¯ood frequency or sediment load. unique occurrence. Although A1 and A2 do However, the normal stratigraphic succession ulate on their origin, because they may ultimately have more importance when com- not both occur in the same outcrop, we ten- we observed throughout the delta indicates tatively infer A1 to be older than A2 because that, except in anomalous circumstances such bined with data from other sites as the paleoseismologic catalogue for the Puget Lowland A1 is about 0.5 m farther below the event B as event B, plants can live through such couplet than A2. Although this interpretation changes. In addition, in possible support of grows. The sand layers are labeled A1 and A2 because we have not found them in the same is also suggested by radiocarbon ages (see fol- our interpretation, there are intrusive liquefac- lowing; Fig. 9), it is possible that the A1 and tion structures that penetrate to about the locality and they could represent the same event. A2 sand layers were deposited in the same event D stratigraphic horizon at a few locali- event. ties. None of these structures, however, has Events A2 and A1ÐDescription been de®nitely correlated with event D. Age and Possible Correlation of the A1 Horizon A2 is present ϳ30 cm below the and A2 Sand Beds Age of Event D and Possible Earthquake event B couplet at locality 8 on Steamboat Sources Slough (Fig. 3). It consists of a thin (Ͻ0.5 cm), coarse-grained to granule-rich sand. A Triglochin rhizomes at the level of the A2 Triglochin rhizomes from within a few cen- similar sand layer is present ϳ45 cm below sand bed at locality 8 yield a calibrated radio- timeters below the distinctive event D gray the event B couplet at locality 15 at Union carbon age range of A.D. 420Ϫ640 (Table 1; clay horizon at locality 21 (Figs. 3±5) yielded Slough (Fig. 3). At locality 8, the sand is pre- Figs. 4 and 9), providing a closely limiting a calibrated age of A.D. 1040Ϫ1310 (Table 1; sent at sites at least 40 m apart; at locality 15, age of A2 sand deposition. Carex stems and Figs. 4 and 9), providing a limiting maximum it was traced laterally in cores for 50 m. At rhizomes that penetrate the A1 sand bed at age on the timing of event D. Detrital wood three other Steamboat Slough localities within locality 11 yielded a calibrated age of A.D. fragments from 30 to 35 cm above the top of 2 km of the current river mouth, there is a 130Ϫ530, providing a limiting minimum age the horizon at this locality yielded a calibrated distinctive, 0.5Ϫ1.3-cm-thick, very ®ne to for A1 sand deposition. The calibrated ages age of A.D. 1400 1640 (SJ-98±2; Table 1; Ϫ ®ne-grained sand bed 25±50 cm below the for these samples overlap slightly (Table 1; Fig. 9), providing a probable limiting mini- event B couplet. These sand layers are bound- Fig. 9), but stratigraphic relationships suggest mum age on the time of the event. Because ed by homogeneous, fossiliferous, olive-gray that A1 is older. we infer that the gray clay unit was rapidly mud of the upper intertidal marsh facies. No Our speculation that the A1 and A2 sand deposited on the Triglochin-vegetated sub- sand layer in this stratigraphic position has beds are tsunami deposits requires a big im- strate, we think that the lower bracketing age been identi®ed in outcrops along Ebey Slough pulse in Puget Sound, either by tectonic de- approximates the age of event D gray-clay de- or the Snohomish River (Fig. 3), and it was formation or by a massive landslide. There is position and therefore infer an age of ca. A.D. not present at upstream localities along any 1040Ϫ1400 (Table 1) for event D. no independent evidence for a correlative slough. We tentatively correlate these anoma- earthquake on a nearby crustal fault (e.g., Se- Event D abrupt subsidence is probably re- lous sand layers on the basis of their similar lated to earthquake-induced ground shaking attle fault, southern Whidbey Island fault) that stratigraphic position. might have produced the sea¯oor deformation. and compaction. Although tectonic subsidence Horizon A1 occurs at locality 11, ϳ3km cannot be ruled out, the study area is outside There has been large-scale prehistoric land- upstream from the mouth of Steamboat sliding just 15 km to the south in Possession the expected deformation ®eld for inferred late Slough. It forms a thin patchy lamina of me- Sound off southeast Whidbey Island (Fig. 1; Holocene crustal faults in the Puget Lowland. dium-grained sand ϳ95 cm below the event Karlin et al., 1996), but it has not yet been Furthermore, there is no known evidence of a B couplet within muddy intertidal marsh fa- dated. Clague et al. (1997) reported ca. A.D. large crustal earthquake in the Puget Lowland cies. It is the only sand we found in the 175 200Ϫ410 earthquake-induced liquefaction at this time. The inferred age ranges of event cm of outcrop below event B at this site. D and the event W plate-boundary earthquake from the Fraser River delta in southwestern (Atwater and Hemphill-Haley, 1997) overlap Event A2 and A1ÐInterpretation British Columbia, ϳ140 km north of the Sno- only slightly (Fig. 9). Karlin and Abella homish delta, and plate-boundary earthquake (1996) reported a ca. A.D. 1200 turbidite silt We speculate that the A2 sand beds repre- event S of Atwater and Hemphill-Haley layer in Lake Washington (Figs. 1 and 9), sent a tsunami deposit on the basis of the fol- (1997) has a similar age (Fig. 9).

492 Geological Society of America Bulletin, April 2001 PALEOSEISMOLOGY ON THE SNOHOMISH DELTA

DISCUSSION bled with a diverse approach including anal- rah Brown, Edward Cranswick, Taber Hersum, Rog- ysis of ancient land-level changes (e.g., er Lewis, Andy Moore, Juan Carlos Moya, Leslie Phillips, Linda Smith, Faith Taylor, and Jen Zwei- This paleoseismologic investigation of the Bucknam et al., 1992; Sherrod, 1998), tsuna- bel. Eileen Hemphill-Haley provided preliminary Snohomish River delta adds to a slowly mi deposits (e.g., Atwater and Moore, 1992), diatom analyses and advice. Lisa Hodges also ex- emerging prehistoric record of strong earth- landsliding (e.g., Jacoby et al., 1992; Schuster amined several Snohomish diatom samples for her quakes and ground shaking in Washington's et al., 1992), liquefaction (e.g., Clague et al., University of Washington senior thesis, under the direction of Mark Holmes and Brian Sherrod. Puget Lowland. This nascent record strongly 1997), and lacustrine turbidite deposition suggests that in the past ϳ1200 yr, the region (Karlin and Abella, 1992, 1996; Karlin et al., has been subjected to stronger earthquakes 1996). Continuation of this diverse approach REFERENCES CITED and ground shaking than in historic times is essential to completion of this record and to Ambraseys, N.N., 1988, Engineering seismology: Earth- (since ca. 1870). Completing this record reliable earthquake hazard assessment in the quake Engineering and Structural Dynamics, v. 17, p. should constrain the number, sources, frequen- Puget Lowland. 1±105. cy, and magnitude of these large seismic Atwater, B.F., 1999, Radiocarbon dating of a Seattle earthquake to A.D. 900±930 [abs.]: Seismological Re- events and should have signi®cant impact on CONCLUSIONS search Letters, v. 70, p. 232. regional earthquake hazard assessments. Atwater, B.F., and Hemphill-Haley, E., 1997, Recurrence Important questions that a more complete A paleoseismologic investigation of the intervals for great earthquakes of the past 3,500 years at northeastern Willapa Bay, Washington: U.S. Geo- paleoseismologic catalog can address include Snohomish delta has revealed evidence of at logical Survey Professional Paper 1576, 108 p. the following. least three episodes of liquefaction, at least Atwater, B.F., and Moore, A.L., 1992, A tsunami about one abrupt subsidence event, and at least one 1,000 years ago in Puget Sound, Washington: Science, 1. During the Holocene, how many other v. 258, p. 1614±1617. crustal faults in the Puget Lowland in addition tsunami since ca. A.D. 800. The most distinc- Baker, G.E., and Langston, C.A., 1987, Source parameters to the Seattle fault have ruptured? Evidence tive stratigraphic unit, produced during event of the 1949 magnitude 7.1 south Puget Sound Wash- ington earthquake as determined from long-period described for the Snohomish delta and in B, can be correlated over the entire ®eld area body waves and strong ground motions: Seismological Bucknam et al. (1992) and Sherrod (1998) ar- and provides evidence for three earthquake- Society of America Bulletin, v. 77, p. 1530±1557. gues for two or more crustal earthquakes ca. related phenomena: strong shaking leading to Bucknam, R.C., Hemphill-Haley, E., and Leopold, E.B., 1992, Abrupt uplift within the past 1,700 years at A.D. 900±950, and Snohomish data possibly liquefaction, abrupt subsidence, and a tsuna- southern Puget Sound, Washington: Science, v. 258, indicate one or two younger crustal events. mi. Radiocarbon ages indicate that event B p. 1611±1614. has an age of A.D. 800 980, similar to the Chleborad, A.F., 1994, Modeling and analysis of the 1949 Given the nascent status of investigations, we Ϫ Narrows Landslide, Tacoma, Washington: Association can expect more such earthquakes to be added age of a large prehistoric earthquake on the of Engineering Geologists Bulletin, v. 31, p. 305±327. to this catalogue. The clustering of the two ca. Seattle fault ϳ50 km to the south. The age Chleborad, A.F., and Schuster, R.L., 1998, Ground failure associated with the Puget Sound region earthquakes A.D. 900±950 events also raises the possibil- coincidence and the presence of a tsunami of April 13, 1949, and April 29, 1965, in Rogers, ity that movement on one fault can increase sand on the Snohomish delta that has also A.M., et al., eds., Assessing earthquake hazards and tectonic load and trigger movement on other been described at other localities lead us to reducing risk in the Paci®c Northwest: U.S. Geologi- cal Survey Professional Paper 1560, p. 373±439. Puget Lowland crustal faults. conclude that the event B features were caused Clague, J.J., Naesgaard, E., and Nelson, A.R., 1997, Age 2. With regard to local earthquakes produc- by the earthquake on the Seattle fault in A.D. and signi®cance of earthquake-induced liquefaction 900±930. near Vancouver, British Columbia, Canada: Canadian ing liquefaction, the Snohomish data indicate Geotechnical Journal, v. 34, p. 53±62. that the ca. A.D. 900 Seattle fault earthquake Two younger sets of sand dikes that locally Cohee, B., and Somerville, P., 1998, Simulated strong produced signi®cant ground shaking almost feed sand volcanoes are present stratigraphi- ground motions for Magnitude 8 earthquakes on the Cascadia subduction zone, in Rogers, A.M., et al., 50 km away, to the north. How widespread cally above the event B couplet at horizons eds., Assessing earthquake hazards and reducing risk are the effects of this strong earthquake, and representing event C and event E. Radiocar- in the Paci®c Northwest: U.S. Geological Survey Pro- what could be expected from a similar earth- bon dates and other stratigraphic information fessional Paper 1560, p. 325±344. Cooke, S.S., ed., 1997, A ®eld guide to the common wet- quake in the future? Moreover, both the 1949 suggest that event C has an age of ca. A.D. land plants of western Washington and northwestern (M 7.1) and 1965 (M 6.5) intraplate earth- 910±990; we tentatively attribute this event to Oregon: Seattle, Washington, Seattle Audubon Soci- a shallow crustal earthquake in the southern ety, 417 p. quakes caused local liquefaction in the Puget Dunnell, R.C., and Fuller, J.W., 1975, An archaeological Lowland, but not at the Snohomish delta. How Puget Lowland. Radiocarbon ages indicate survey of Everett Harbor and the lower Snohomish frequently does this kind of earthquake occur, that event E is younger than A.D. 1450. The estuary delta: U.S. National Park Service Document CX-9000-4-0101, 110 p. and can its effects in the geologic record be section above the event B couplet also com- Frankel, A., Mueller, C., Barnhard, T., Perkins, D., Leyen- distinguished from effects of shallow crustal monly includes a second sharp lithologic decker, E.F., Dickman, N., Hanson, S., and Hopper, earthquakes or plate-boundary events? change from olive-gray, rhizome-rich mud to M., 1996, National seismic hazards maps, June 1996 documentation: U.S. Geological Survey Open-File 3. How much ground motion or deforma- grayer, rhizome-poor mud, which may indi- Report 96±532, scale, 1:7 000 000. tion in the Puget Lowland is induced by large cate a second subsidence event associated Gower, H.D., Yount, J.C., and Crosson, R.S., 1985, Seismotec- with an earthquake (event D). tonic map of the Puget Sound region, Washington: U.S. plate-boundary earthquakes? There is no con- Geological Survey Map I-1613, scale 1:250 000. clusive paleoseismologic evidence in the Pu- Jacoby, G.C., Williams, P.L., and Bucknam, R.C., 1992, get Lowland for the A.D. 1700 plate-boundary ACKNOWLEDGMENTS Tree-ring correlation between prehistoric landslides and abrupt tectonic events in Seattle, Washington: Sci- event. Our investigation of the Snohomish We thank Brian Atwater, Bob Bucknam, Sue ence, v. 258, p. 1621±1623. delta provides no compelling evidence for Cashman, John Clague, Alan Nelson, Brian Sher- Johnson, S.Y., Potter, C.J., and Armentrout, J.M., 1994, Or- earthquake-induced effects in A.D. 1700. Cor- rod, and Robert Witter for constructive reviews and igin and evolution of the Seattle basin and Seattle stimulating discussions. Field assistance was pro- fault: Geology, v. 22, p. 71±74. relation with older plate-boundary earth- Johnson, S.Y., Potter, C.J., Armentrout, J.M., Miller, J.J., vided especially by Brian Atwater, Tatiana Pinegina, quakes is possible but inconclusive (Fig. 9). Finn, C., and Weaver, C.S., 1996, The southern Whid- and Brian Sherrod. Others who accompanied and bey Island fault, an active structure in the Puget Low- The regional paleoseismologic record need- aided us in the ®eld include Hans Abrahamson, land, Washington: Geological Society of America Bul- ed to answer these questions is being assem- Boyd Benson, Stein Bondevik, Judy Boughner, Sa- letin, v. 108, p. 334±354.

Geological Society of America Bulletin, April 2001 493 BOURGEOIS and JOHNSON

Johnson, S.Y., Dadisman, S.V., Childs, J.R., and Stanley, Murty, T.S., and Hebenstreit, G.T., 1989, Tsunami ampli- Rogers, A.M., Walsh, T.M., Kockelman, W.J., and Priest, W.D., 1999a, Active tectonics of the Seattle fault and tudes from local earthquakes in the Paci®c Northwest G.R., 1996, Earthquake hazards in the Paci®c North- central Puget Sound, WashingtonÐImplications for region of North America, Part 2ÐStrait of Georgia, westÐAn overview: U.S. Geological Survey Profes- earthquake hazards: Geological Society of America Juan de Fuca Strait, and Puget Sound: Marine Geol- sional Paper 1560, p. 1±54. Bulletin, v. 111, p. 1042±1053. ogy, v. 13, p. 189±209. Satake, K., Shimazaki, K., Tsuji, Y., and Ueda, K., 1996, Johnson, S.Y., Dadisman, S.V., Mosher, D.C., Blakely, R.J., Nelson, A.R., Atwater, B.F., Bobrowsky, P.T., Bradley, Time and size of a giant earthquake in Cascadia in- and Childs, J.R., 1999b, Neotectonics of the Devils L.A., Clague, J.J., Carver, G.A., Darienzo, M.E., ferred from Japanese tsunami record of January 1700: Mountain fault and northern Whidbey Island fault, Grant, W.C., Krueger, H.W., Sparks, R., Stafford, Nature, v. 379, p. 246±249. eastern Strait of Juan de Fuca and northern Puget T.W., and Stuiver, M., 1995, Radiocarbon evidence for Schuster, R.L., Logan, R.L., and Pringle, P.T., 1992, Pre- historic rock avalanches in the Olympic Mountains, Lowland, Washington [abs.]: Seismological Research extensive plate-boundary rupture about 300 years ago Washington: Science, v. 258, p. 1620±1621. Letters, v. 70, p. 220. at the Cascadia subduction zone: Nature, v. 378, p. Sherrod, B.L., 1998, Late Holocene environments and Karlin, R.E., and Abella, S.E.B., 1992, Paleoearthquakes in 371±374. earthquakes in southern Puget Sound [Ph.D. thesis]: the Puget Sound region recorded in sediments from Obermeier, S.F., 1995, Preliminary estimates of the strength Seattle, University of Washington, 155 p. Lake Washington, U.S.A.: Science, v. 258, p. of prehistoric shaking in the Columbia River Valley Silva, W.J., Wong, I.G., and Darragh, R.B., 1998, Engi- 1617±1620. and the southern half of coastal Washington, with em- neering characterization of earthquake strong ground Karlin, R.E., and Abella, S.E.B., 1996, A history of Paci®c phasis for a Cascadia subduction zone earthquake motions in the Paci®c Northwest, in Rogers, A.M., et Northwest earthquakes recorded in Holocene sedi- about 300 years ago: U.S. Geological Survey Open- al., eds., Assessing earthquake hazards and reducing ments from Lake Washington: Journal of Geophysical File Report 94±589, 46 p. risk in the Paci®c Northwest: U.S. Geological Survey Research, v. 101, p. 6137±6150. Obermeier, S.F., 1996, Using liquefaction-induced features Professional Paper 1560, p. 313±324. Karlin, R.E., Watters, R.J., Grass, M.J., and Prunier, C.F., for paleoseismic analysis, in McCalpin, J.P., ed., Pa- Stein, R.S., and Yeats, R.S., 1989, Hidden earthquakes: Sci- 1996, Seismic landslides in Puget Sound (SLIPS) II: leoseismology: San Diego, California, Academic enti®c American, v. 250, no. 6, p. 48±57. Quaternary faulting and submarine mass-wasting Press, p. 331±396. Stuiver, M., and Reimer, P.J., 1993, Extended 14C data base [abs]: Eos (Transactions, American Geophysical Obermeier, S.F., and Pond, E.C., 1999, Issues in using liq- and revised CALIB 3.0 14C age calibration: Radio- Union), v. 77, p. F500. uefaction features for paleoseismic analysis: Seismo- carbon, v. 35, p. 215±230. Stuiver, M., Reimer, P.J., Bard, E., Beck, J.W., Burr, G.S., Langston, C.A., and Blum, D.E., 1977, The April 29, 1965, logical Research Letters, v. 70, p. 34±58. Hughen, K.A., Kromer, B., McCormac, F.G., van der Puget Sound earthquake and the crustal and upper Plafker, G., and Kachadoorian, R., 1966, Geologic effects Plicht, J., and Spurk, M., 1998, INTCAL98 radiocar- mantle structure of western Washington: Seismologi- of the March 1964 earthquake and associated seismic bon age calibration, 24,000±0 cal BP: Radiocarbon, cal Society of America Bulletin, v. 67, p. 693±711. sea waves on Kodiak and nearby islands, Alaska: U.S. v. 40, p. 1041±1084. Ludwin, R.S., Weaver, C.S., and Crosson, R.S., 1991, Seis- Geological Survey Professional Paper 543-D, 46 p. Wang, K., Dragert, H., and Melosh, H.J., 1994, Finite el- micity of Washington and Oregon, in Slemmons, Pratt, T.L., Johnson, S.Y., Potter, C.J., Stephenson, W.J., and ement study of uplift and strain across Vancouver Is- D.B., et al., eds., Neotectonics of North America: Finn, C., 1997, Seismic-re¯ection images beneath Pu- land: Canadian Journal of Earth Sciences, v. 31, p. Geological Society of America Decade Map Volume get Sound, western Washington StateÐThe Puget 1510±1522. 1, p. 77±98. Lowland thrust system hypothesis: Journal of Geo- McCulloch, D.S., and Bonilla, M.G., 1970, Effects of the physical Research, v. 102, p. 27469±27490. MANUSCRIPT RECEIVED BY THE SOCIETY JULY 12, 1999 earthquake of March 27, 1964, on the Alaska Rail- Ramsey, C.B., 1995, Radiocarbon calibration and analysis REVISED MANUSCRIPT RECEIVED FEBRUARY 22, 2000 MANUSCRIPT ACCEPTED APRIL 14, 2000 road: U.S. Geological Survey Professional Paper 545- of stratigraphyÐThe OxCal Program: Radiocarbon, v. D, 161 p. 37, p. 425±430. Printed in the USA

494 Geological Society of America Bulletin, April 2001