Snowhomish Delta GSA 2001 - Paleoseismology

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

page 1 | 2 | 3 | 4

PALEOSEISMOLOGY

There is evidence at the lower Snohomish delta for several prehistoric, late Holocene earthquakes (labeled A through E from oldest to youngest) based on our examination of outcrops at about 45 localities (Figs. 3, DR 2, 3). Below we describe and discuss each of these events, in order of the strength of the evidence supporting a paleoseismic interpretation. We consider events B, C and E, all with liquefaction, more reliable than Events A1, A2 and D. We report the latter three events to help develop a regional catalogue of possible paleoseismic indicators.

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. Loc.11; Fig. 3), or contains roots of Sitka spruce and other trees and shrubs (e.g., Loc. 32, 29, 27; Fig. 3). The mud is cut by sand-filled dikes at many localities.

Surveying at five localities revealed that relief on the surface below the Event B couplet over lateral distances of 15 to 70 m is typically about 10 cm and ranges up to 40 cm. We also surveyed the differences in paleoelevation of the Event B couplet between some localities. For example, the couplet was deposited on a Scirpus-rich surface at Locality 8, 60 cm higher than the contemporaneous surface at Locality 6 where the couplet was deposited on a Triglochin-rich surface.

Event B sand bed. The thin sand bed typically forms a distinctive, laterally continuous sheet. At several localities (e.g., Loc. 6, 8, 21, 22; Fig. 3), the sand is laterally continuous for at least 30-50 m; also, a notch formed by erosion of this sand extends many tens of meters along slough cutbanks. The generally fine- to medium-grained sand bed ranges in thickness from a few mm to about 5 cm, and both thins and fines in the upstream direction (Fig. 7). It is commonly thicker in paleo-swales and thinner over what were topographic highs, such as logs. The sand is commonly graded and contains one or two olive-gray silty laminae toward its top. It was typically deposited on a vegetated surface of sedges, rushes, grasses and other herbaceous plants.

Event B gray clay. The Event B couplet gray clay is generally about 5 cm thick (range of 0 to ~20 cm), and is thickest in local paleo-swales. Unlike the underlying sand, the gray clay shows no systematic variation in thickness across the study area. The most distinctive aspects of this layer are its color and general lack of fossil plant material. It is commonly a medium light gray (N5.5) silty clay, markedly contrasting with underlying and overlying olive-gray mud. The base of the gray clay is typically sharp, and the lower portion of the bed commonly contains thin laminae of sand or coarse-silt. The gray clay grades upward to olive-gray mud.

Facies above the couplet. The olive-gray mud that typically overlies the couplet is generally indistinguishable in color and lithology from mud below the couplet. However, at many sites, there is a change in plant fossils from below to above the couplet (summarized in Table 3; e.g., Triglochin (and Carex) below to Carex only (no Triglochin) above, Sitka spruce below to no spruce above), generally indicating a lowering of the surface (discussed below under interpretation). At some localities which were delta-plain surfaces before Event B (e.g., 10, 11, 27, 32; Fig. 3), the sediment above the couplet is less peaty or more muddy than sediment below the couplet.

Benthic diatoms associated with the Event B couplet. Benthic diatom assemblages reflect the salinity and substrate preferences of dominant species. Benthic diatom species are present in the couplet sand and gray clay, in the underlying and overlying mud, and in the associated dike sand. Cursory study of diatom populations in samples from these sediments (Eileen Hemphill-Haley and Lisa Hodges, 1998, written communication; localities 2, 4, 8, 17; all on Steamboat Slough) indicates that both the mud and gray clay contain similar assemblages dominated by tidal marsh species, and the dike sand and couplet sand contain many subtidal species. No distinct change in diatoms was observed from below to above the couplet.

EVENT B INTERPRETATION
Evidence for liquefaction and ground failure contemporaneous with deposition of the couplet.

Liquefaction features (cf. Obermeier, 1996; Obermeier and Pond, 1999), particularly sand dikes (Fig. 4,8), are common in outcrop along the lower Snohomish distributaries. They range in width from a few millimeters to more than 1 m, and contain fine to medium sand. Most dikes pinch out upward within the mud section, making it possible to establish only their lower age limit from the age of the highest material intruded. Some dikes clearly cut through and are thus younger than the couplet horizon. However, at four localities (2, 5, 6, 17; Fig. 3) and probably at four others (21, 25, 32, 22; Fig. 3), dikes terminate in sand volcanoes at the couplet horizon, or occupy lateral spreads that terminate there; these features were produced contemporaneously with the couplet.

At localities 5 and 6, ground failure of the paleo-surface is indicated by lateral spreads, places where the marsh surface cracked open and was at least partially filled from below by sand (Fig. 4). A vertical sand dike with an irregular margin fills one of these lateral spreads (Fig. 8A). The top of this spread formed a paleo-swale that was filled by a thick layer of couplet gray clay. At locality 2, a dike below the couplet ends at the couplet horizon in a mounded lens of sand. Because the gray clay above this sand lens thins over the sand lens, we interpret this feature to be a sand volcano and not a sill. This distinction is important because the sand member of the couplet, being one of the mechanically weakest horizons in the mud-dominated section, is intruded by sills at a few localities. At locality 17, for example, at least some of the abundant liquefaction features are convincingly contemporaneous with the couplet, but others, including sand sills intruded along the sandy part of the couplet horizon, are due to later remobilization.

Evidence for abrupt local subsidence at the stratigraphic level of the Event B couplet.

Although there is rarely a lithologic change from below to above the couplet, we interpret the changes in fossil vegetation (Table 3) to indicate localized subsidence generated by earthquake-induced compaction and liquefaction. We attempted to estimate the magnitude of this lowering from the thickness of sediment between the disappearance of a fossil plant just below the couplet and its reappearance above the couplet.

Locality 6 provides an example of an Event B fossil-vegetation change. The normal fossil plant succession in outcrop is Carex --> Triglochin --> Scirpus acutus (see also Table 2). At locality 6, the Carex --> Triglochin transition is just below the couplet, fossil Triglochin below the couplet abruptly disappears at the couplet horizon (replaced above by Carex), and Triglochin reappears in the section 50-75 cm above the couplet. We interpret that the paleo-surface at Locality 6 dropped perhaps 50-75 cm, below the level where Triglochin could thrive, and Carex reoccupied the surface until enough sediment accumulated to bring the land level back up to where Triglochin could reestablish.

At localities where Sitka spruce roots and trunks are present at and below the couplet (e.g., 27, 29, 32; Figs. 3, 6B), spruce disappears above the couplet and then reappears 30-100 cm higher in the section. This relationship suggests that trees were killed by subsidence, which increased submergence or salt exposure. New trees began to grow when sediment aggradation on the floodplain reached a level where standing water or salt exposure was sufficiently diminished.

Our cumulative evidence suggests that land lowering was probably variable across the delta, negligible at some sites, and half a meter or more at others. Among 28 localities where stratigraphic and plant-fossil successions bounding the couplet were examined, 15 showed plant-fossil evidence of an abrupt lowering of land level, and 13 provided no plant-fossil evidence of change (Table 3); no locality showed evidence of uplift. The absence of evidence for subsidence at many localities may either indicate no land-level change or a change that occurred within a particular plant's tolerance zone. The localized and variable subsidence at the time of Event B appears similar to that which occurred during the 1964 plate-boundary earthquake in south-central Alaska, where compaction- and liquefaction-induced subsidence of a few cm to more than 1 m occurred in many estuarine and alluvial environments (Plafker and Kachadoorian, 1966; McCulloch and Bonilla, 1970).

Tectonic subsidence is highly unlikely given the inferred amount of subsidence (locally more than 50 cm) and the location of the Snohomish delta. Elastic-dislocation models of great subduction zone earthquakes predict only about 10 cm of coseismic tectonic subsidence at this distance from the plate boundary (e.g., Wang et al., 1994; W.D. Stanley, written commun., 2000). The nearest known crustal fault, the southern Whidbey Island fault, lies 13 km to the southwest. Deformation models (e.g., Stein and Yeats, 1989) suggest that earthquake-generated slip of more than 4-5 m on this fault would be required to generate 50 cm of tectonic subsidence at the Snohomish delta. There is presently no paleoseismologic evidence for an event of this size and age (see below) on this nearby fault.

Origin of the couplet sand.

Several lines of evidence suggest the couplet sand bed is a tsunami deposit and not the deposit of a river flood or storm surge. The sand forms a thin, widespread layer over kilometers of outcrop, and it thins, fines, and ultimately disappears in the upstream direction (Fig. 7).

These features suggest a delta-wide submergence by a wave or waves directed upstream from Possession Sound. The protected geographic position of the Snohomish delta and the limited size and fetch of Possession Sound argues against a storm-surge origin for these waves. Further supporting a tsunami origin, the sand bed is graded and its top is indistinctly laminated, indicative of rapid deposition from suspension. Also, the sand was transported across a vegetated marsh surface where resuspension is nearly impossible, thus requiring initial rapid advection. Additionally, no other sand of comparable sheet-like geometry or grain size occurs in the post ~ A.D. 700 bank exposures on the Snohomish delta, indicating that the sand was deposited by a rare event and not by normal floods. The presence of subtidal diatoms in the couplet sand is also consistent with a tsunami interpretation. Perhaps most convincing, deposition of the sand was contemporaneous with formation of liquefaction structures, which indicate strong shaking. Although the sand layer may have a source locally in this erupted sand, the layer's widespread distribution, thinness, occurrence on vegetated surfaces, and presence in areas where no liquefaction structures could be found lead us to believe that the major sediment source was sand suspended from adjacent subtidal channels during a tsunami surge. Finally, bracketing radiocarbon ages overlap ages of a large Seattle fault earthquake (see below), which has been shown to have produced a northward-moving tsunami that deposited sand at Cultus Bay and West Point (Fig. 1; Atwater and Moore, 1992). At these localities, the deposit is typically a graded sand layer, and at West Point it is in places overlain by gray clay (B.F. Atwater, 1999, personal communication).

Origin of the couplet gray clay.

Subsidence that accompanied Event B (see Table 3) was probably the key control on deposition of the gray clay. The absence of plant fossils is the key distinction between this unit and bounding mud units. In our interpretation, subsidence created uneven "accommodation space" across the delta, which was rapidly filled by deposition of clay transported by normal tidal currents and floods. This rapid deposition preceded and precluded re-colonization of normal marsh vegetation. The variable thickness of the gray clay reflects both its depositional thickness and the amount of post-depositional alteration (to olive-gray mud) by plant colonization and pedogenesis. Thus, the thickness of gray clay does not provide an accurate indication of the amount of subsidence at specific sites.

The following are arguments against an alternative explanation that we considered: that the gray clay is part of the tsunami deposit. First, unlike the couplet sand, there is no systematic geographic variation in thickness or grain size of the gray clay. Second, the base of the clay is sharp; the lower part of the gray clay commonly has one or two fine-sand or silt laminae, but they are easily distinguished from laminae in the underlying sand unit, which are olive gray. Finally, diatoms in the gray clay are normal tidal marsh diatoms, resembling assemblages in muds below and above.

Age of Event B.

Two high-precision radiocarbon dates on stumps of spruce trees inferred to have been killed by earthquake-induced subsidence individually suggest Event B occurred between A.D. 800 and 980 (Table 1; Figure 9). When these two dates are combined using OxCal (Ramsey, 1995), the age range narrows to A.D. 850 to 980 (95% probability). This age range is consistent with four other limiting maximum ages, one from the distal end of the root of a spruce tree that died at the time of the event, and three from Triglochin rhizomes sampled within 10 cm below the base of the couplet, and with two additional limiting minimum ages from Triglochin and Carex plant material slightly above the couplet.

Correlation with a large earthquake on the Seattle fault.

The A.D. 850 to 980 range includes the time of a large earthquake on the Seattle fault (Bucknam et al., 1992), the age of which has recently been narrowed to A.D. 900-930 (Atwater, 1999). This earthquake produced a large, northward-moving tsunami in Puget Sound (Atwater and Moore, 1992) which we infer deposited the Event B couplet sand at the Snohomish delta. Although Bucknam et al. (1992) and Sherrod (1998) have found evidence for other crustal earthquakes at about this time in the southern Puget Lowland, these earthquakes apparently did not produce tsunamis in Puget Sound. The inferred magnitude of the Seattle fault earthquake (M >7), based on other considerations (Bucknam et al., 1992), would have been sufficient to generate liquefaction and ground failure on the Snohomish delta 50 km away (Ambraseys, 1988; Obermeier and Pond, 1999).

Other possible sources of earthquakes that could have produced Event B features include 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 al., 1999b) about 35-40 km to the north (Fig. 1). However, no late Holocene earthquakes have yet been attributed to these structures. Tsunami deposits from large offshore plate-boundary earthquakes (e.g., Atwater and Hemphill-Haley, 1997) have also not been recognized in Puget Sound, and tsunami models (Murty and Hebenstreit, 1989) suggest they would not significantly inundate this region. Deep earthquakes in the downgoing plate (Fig. 2) would not offset the sea floor and are thus not primary tsunami sources.

It is possible that the A.D. 900-930 earthquake on the Seattle fault or another large earthquake around that time triggered large landslides along Possession Sound (Fig. 1), which could displace sufficient water to produce a tsunami large enough to submerge the Snohomish delta. Chleborad (1994) cites accounts of a 2.5-m-high "tidal wave" caused by a large landslide near Tacoma (Fig. 1) three days after the 1949 (M 7.1) deep earthquake between Tacoma and Olympia. There is evidence for large paleo-landslides in Possession Sound (based on high-resolution seismic-reflection data; Karlin et al., 1996), but they have not been dated.

EVENTS C AND E --EVIDENCE FOR LIQUEFACTION
Sand Dikes, Volcanoes, and Sills

Out of 33 sites where we described a section that included the Event B couplet, at least six include sand dikes that cut through, or are present above, the couplet. These dikes are generally 0.5 to about 3 cm wide and penetrate a few tens of centimeters of section. In a few cases, sand from Event B appears to have been remobilized; also, some sills have been intruded along the Event B sand horizon. These sills are distinguished from the inferred tsunami sand by a lack of lamination, more variability in thickness, and their association with sand feeder dikes.

As previously noted, the age of liquefaction features can be difficult to pin down because they pinch out upward or emerge from plane of the outcrop below the level of the land surface at the time of intrusion. However, at locality 21 at Ebey Slough (Fig. 3), we found two different horizons above the Event B level where sand dikes feed horizontal to mounded sand lenses (Fig. 4). The sand lenses at the lower of these two horizons (Event C) extend laterally for as much as 1 m, and are as much as 1 cm thick. The sand lens in the upper horizon (Event E) extends laterally for 2 m and is as thick as 2.5 cm. We consider these lenses as sand volcanoes, rather than sills, because the layers are thin and irregular, appear to drape growth-position plants, and occur within horizons where there is no mechanically weak zone such as a preexisting sand layer that could control their stratigraphic position.

Locality 16 also contains liquefaction structures younger than Event B, but we did not recognize any sand volcanoes. We did find a 5-cm-wide sand dike connected to a 2-m-long sill (3-5 cm thick), in strata younger than the couplet. We constrained the age of this sand sill by dating fossil plant material directly below the sill (limiting maximum age), and fossil marsh-plant stems that grew through the sand sill (limiting minimum age) (Table 1; Figs. 4, 9). Based on the similarity of the two ages (see below), we tentatively correlate the sill-forming liquefaction event at locality 16 with Event C at locality 21.

Age of Event C and Possible Earthquake Sources

Stratigraphic relationships (Fig. 4) indicate that Event C liquefaction is younger than Event B and thus postdates A.D. 900, the inferred limiting maximum age for the large Seattle fault event (Atwater, 1999). Assuming the limiting ages for the Seattle fault event are correct, statistical analysis using OxCal (Ramsey, 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 above, there is currently no evidence for late Holocene rupture on the potentially active crustal faults (southern Whidbey Island fault, Devils Mountain fault) that are closest to the Snohomish delta (Fig. 1). In the southern Puget Lowland, however, Bucknam et al. (1992) and Sherrod (1998) have described regions of abrupt uplift and subsidence that also occurred about A.D. 900, and argue that the pattern of crustal movement supports at least one large upper crustal earthquake other than the ~AD 900 Seattle fault event at about this time in the region. The inferred sources for these additional possible crustal earthquakes are about 75 to 120 km away from the Snohomish delta; liquefaction at these distances would require earthquakes of minimum magnitude about 6.5 to 7 (Ambraseys, 1988; Obermeier and Pond, 1999).

Event C liquefaction could also have been caused by a great, plate-boundary earthquake. Based on work along the Pacific Coast of southwest Washington, Atwater and Hemphill-Haley (1997) documented a subduction-zone event that occurred between about A.D. 700 and A.D. 1100. There is presently no consensus on the strength of ground motions such an event could produce as far inland (~170 km) as the Snohomish delta (e.g., Obermeier, 1995; Silva et al., 1998; Cohee and Somerville, 1998). To date, no one has conclusively linked paleo-liquefaction structures this far inland to a plate-boundary earthquake.

It is also conceivable that Event C liquefaction was caused by a strong, deep, intraplate earthquake (Fig. 2) similar to the 1949 M 7.1 Olympia and 1965 M 6.5 Seattle-Tacoma events (Langston and Blum, 1977; Baker and Langston, 1987). Each of these earthquakes produced local sand boils and other evidence of liquefaction and ground failure in southern and central Puget Sound above their hypocenters (Chleborad and Schuster, 1998). Neither earthquake produced liquefaction at the Snohomish delta. However, a deep, liquefaction-producing, intraplate earthquake could have occurred below the Snohomish delta about A.D. 910 to 990. Because this type of earthquake will not result in tectonic changes in land level or a tsunami, liquefaction may be its only record.

Given regional evidence for one or more, ~A.D. 900, strong upper crustal earthquakes in the southern Puget Lowland (Bucknam et al., 1992; Sherrod, 1998), it seems most likely to us that Event C liquefaction was caused by such an earthquake. If so, our evidence indicates that at least one of these earthquakes closely post-dates the Seattle fault event.

Age of Event E and Possible Earthquake Sources

Two radiocarbon ages on wood provide limiting maximum ages on Event E liquefaction (Table 1; Figs. 4, 9). One (calibrated age of A.D. 1400 to 1640) was from a 2-cm-diameter detrital wood fragment 11 cm below the sand volcano, the other (calibrated age of A.D. 1430 to 1640) was from a 2-mm-diameter piece of bark-free twig within the sand volcano. Because of the fresh appearance and delicate nature of this twig, we think this younger date may closely approximate the time of liquefaction; therefore we tentatively infer an age of ~A.D. 1430-1640 for Event E. As with Event C, the liquefaction may have been forced by a prehistoric earthquake generated on crustal faults, on the plate-boundary thrust fault, or within the downgoing plate (Fig. 2).

There is presently no evidence of rupture on any of the active or potentially active crustal faults of the Puget Lowland (Fig. 1) in the last 450 years. Given the nascent status of regional paleoseismologic investigations, however, a large, crustal-fault earthquake in the Puget Lowland at this time cannot be ruled out. We consider it unlikely that Event E liquefaction is a distal effect of the great A.D. 1700 Cascadia earthquake. Radiocarbon ages for this plate-boundary earthquake (Nelson et al., 1995) are generally younger than the age we infer for Event E, and there is presently no evidence that the A.D. 1700 Cascadia earthquake produced liquefaction in the Puget Lowland.

EVENT D -- POSSIBLE ABRUPT SUBSIDENCE
Sharp and Anomalous Stratigraphic Change

Outcrops at 10 or more localities display the stratigraphic horizon that we label Event D. Generally, Event D is recorded by a well-defined, anomalous abrupt stratigraphic change from more olive-colored, massive, plant-rich sediment below to grayer, more laminated, less plant-rich sediment above (Fig. 4). Fossil vegetation changes at this sharp lithologic contact are similar to those observed at the Event B horizon (Table 3). For example, at locality 21, Triglochin and Scirpus disappear at the contact and Scirpus reappears 60 cm above the contact (Fig. 5).

At locality 2, Scirpus disappears at the contact and reappears 50 cm above the contact. At locality 4, Carex, Triglochin, and Scirpus occur below the contact but only Carex above. At locality 29, spruce tree roots occur below the contact but disappear above it. Our correlation of the Event D horizon across the delta is tentative, based on lithologic similarity and on stratigraphic distance above Event B, typically 50-100 cm.

Inferred Rapid Subsidence

We interpret the lithologic change at Event D as possible evidence of rapid subsidence associated with a prehistoric earthquake. Evidence includes the noted color change from olive-gray to more gray and changes in fossil vegetation. In a normal aggrading succession, grayer facies are deeper in the section and represent a topographically lower environment. Interpretation that this subsidence was abrupt is based on the unusually sharp boundary at the base of this horizon, and the occurrence of fossil-poor gray clay, which we interpret to indicate rapid filling of newly created accommodation space (as for Event B).

Alternatively, this layer could have been produced by normal subsidence coupled with a change in flood frequency or sediment load. However, the normal stratigraphic succession we observed throughout the delta indicates that, except in anomalous circumstances such as Event B, plants can live through such changes. Also, in possible support of our interpretation, there are intrusive liquefaction structures that penetrate to about the Event D stratigraphic horizon at a few localities. None of these structures, however, has been definitely correlated with Event D.

Age of Event D and Possible Earthquake Sources

Triglochin rhizomes from within a few cm below the distinctive Event D gray clay horizon at locality 21 (Figs. 3, 4, 5) yielded a calibrated age of A.D. 1040 to 1310 (Table 1; Figs. 4, 9), providing a limiting maximum age on the timing of Event D. Detrital wood fragments from 30 to 35 cm above the top of the horizon at this locality yielded a calibrated age of A.D. 1400 to 1640 (SJ-98-2; Table 1; Fig. 9), providing a probable limiting minimum age on the time of the event. Because we infer that the gray clay unit was rapidly deposited on the Triglochin-vegetated substrate, we think that the lower bracketing age approximates the age of Event D gray-clay deposition and therefore infer an age of ~A.D. 1040 to 1400 (Table 1) for Event D.

Event D abrupt subsidence is probably related to earthquake-induced ground shaking and compaction. Although tectonic subsidence cannot be ruled out, the study area lies outside the expected deformation field for inferred late Holocene crustal faults in the Puget Lowland. Furthermore, there is no known evidence of a large crustal earthquake in the Puget Lowland at this time. The inferred age ranges of Event D, and the Event "W" plate-boundary earthquake (Atwater and Hemphill-Haley, 1997) overlap only slightly (Fig. 9). Karlin and Abella (1996) report a ~1200 A.D. turbidite silt layer in Lake Washington (Figs. 1, 9), which they suggest may have been caused by earthquake-induced slumping at the lake margin. At present, this is the only possible earthquake in the Puget Lowland that may correlate with Event D.

EVENTS A1 and A2 --POSSIBLE TSUNAMI DEPOSITS DISCUSSION CONCLUSIONS ACKNOWLEDGMENTS REFERENCES