A Tsunami About 1000 Years Ago in Puget Sound, Washington

Water surged from Puget Sound sometime between 1000 and 1100 years ago, overrunning tidal marshes and mantling them with centimeters of sand. One overrun site is 10 kilometers northwest of downtown Seattle; another is on Whidbey Island, some 30 kilometers farther north. Neither site has been widely mantled with sand at any other time in the past 2000 years. Deposition of the sand coincided—to the year or less—with abrupt, probably tectonic subsidence at the Seattle site and with landsliding into nearby Lake Washington. These findings show that a tsunami was generated in Puget Sound, and they tend to confirm that a large shallow earthquake occurred in the Seattle area about 1000 years ago.

A Tsunami About 1 000 Years Ago in Puget Sound, Washington Brian F. Atwater and Andrew L. Moore Water surged from Puget Sound sometime between 1000 and 1 1 00 years ago, overrunning tidal marshes and mantling them with centimeters of sand. One overrun site is 10 kilometers northwest of downtown Seattle; another is on Whidbey Island, some 30 kilometers farther north. Neither site has been widely mantled with sand at any other time in the past 2000 years. Deposition ofthe sand coincided-to the year or less-with abrupt, probably tectonic subsidence at the Seattle site and with landsliding into nearby Lake Washington. These findings show that a tsunami was generated in Puget Sound, and they tend to confirm that a large shallow earthquake occurred in the Seattle area about 1000 years ago.
The tsunami deposit at Cultus Bay forms a sheet of sand mostly 5 to 15 cm thick in an area at least 100 by 200 m (Figs. 1 and 2). There, wetland peat has built upward and bayward since a tidal marsh began to supplant a tidal flat about 2000 years ago. This peat contains the sand sheet, which we found in scores of auger borings and followed as a continuous bed along more than 100 m of a drainage ditch. Neither the auger borings nor the ditch revealed any other sand bed in the peat. The surface covered by the sand shows 2 m of relief: 1.5 m where the sand mantled a sloping marsh (12) and another 0.5 m where the sand covered colluvium of an adjacent hillside (Fig. 2). The median grain size, mostly about 0.1 mm, decreases landward and stratigraphically upward (13). The sand contains microscopic marine fossils (14).
Deposition of the sand sheet at Cultus Bay occurred sometime between 850 and 1250 years ago, and it happened while the site probably underwent little or no subsidence. We dated the sand sheet by obtaining radiocarbon ages on plant remains in growth position in the sand (Fig. 2, in ditch). The dated remains are rhizomes (below-ground stems) and attached leaf bases of arrowgrass (Triglochin manritimum), which at modem Cultus Bay thrives only in a 1--m range high in the intertidal zone.
Because additional arrowgrass rhizomes lie both below and above the sand, we suspect that the dated rhizomes grew upward through the sand sheet within years of its deposition. Such maintenance of arrowgrass would mean that deposition of the sand attended little or no subsidence of the Cultus Bay marsh (15).
The sand sheet at Cultus Bay is better explained by a tsunami than by a flood or storm. The landward fining and salt water fossils of the sand implicate a surge from A large earthquake probably happened between 500 and 1700 years ago on the Seattle fault (1), which has been inferred to extend westward across Puget Sound from downtown Seattle (2). The main evidence for the earthquake consists of terraces that record meters of abrupt uplift at Puget Sound (1). If abrupt enough to have accompanied an earthquake, such uplift should have generated a tsunami in Puget Sound. In this report, we show that a tsunami originated in Puget Sound between 1000 and 1100 years ago (3) and that it probably was generated by an earth-  (4,5), Japan (6), and British Columbia (7), and ancient examples have been inferred for Chile (5), Japan (6), Scotland (8), Alaska (9), and the Pacific coast of Washington and Oregon (10,11). In (1), a landslide at Lake Washington between 1000 and 1100 years ago (16), rock avalanches in the Olympic Mountains between 1000 and 1300 years ago (17), a ground-water eruption along the Pacific coast of Washington between 900 and 1300 years ago (11), and abrupt subsidence at West Point between 1000 and 1100 years ago [see figure 1 of (1)]. The tsunami deposit at West Point punctuates a sequence of mostly intertidal deposits exposed in a sewer excavation 150 m long (Fig. 3). Sand and gravel low in the excavation represent a beach (18) on which people built fires and discarded shells before 2000 years ago (unit A, Fig. 4). The beach was eventually buried by silty debris flows from an adjacent hillside and by intertidal mud, peat, and sand. The first phase of intertidal burial, marked by unit B, lasted about 1000 years and concluded with a West #I marsh dominated by saltgrass (Distichfis spicata) and bulrush (Scirpus maritimus). Next came the tsunami, which deposited the only widespread, tabular body of sand in the excavation (19). At that point the marsh and the toe of a debris flow became a short-lived tidal flat. This tidal flat, recorded by unit C, aggraded rapidly until it became a saltgrass marsh, recorded by unit D (20 Properties of the sand sheet at West Point vary with the kind of land that the sand covered. Where deposited on a marsh, the sheet ranges from 4 to 6 cm thick, shows little or no evidence of basal scour, grades stratigraphically upward from 0.5mm sand to 0.1-mm sand, locally contains a basal lamina of 0.1-to 0.2-mm sand, includes sparse transported bivalves and barnacles, and surrounds culms (aboveground stems) of saltgrass and bulrush that are rooted just below the sand and extend vertically into tidal-flat mud as much as 10 cm above the top of the sand. On debris flows, the sand thickens to 40 cm in swales, disappears on rises, and contains angular clasts of debris-flow silt and rounded pebbles. The highest deposit that we assign to the sand sheet (below H, Fig. 4) contains microscopic marine fossils (21). Presentday relief on the sand-mantled marsh and debris flows totals 1.5 m, but this value exceeds initial relief if unit B has been compacted by its overburden.
Land at West Point underwent at least 1 m of abrupt, largely tectonic subsidence that coincided, within months, with deposition of the sand sheet. The subsidence sufficed to make room for the 1 to 1.5 m of tidal-flat deposits that widely accumulated on the sand-mantled marsh (22). Both the subsidence and some of the consequent tidal-flat deposition happened too abruptly for the saltgrass and bulrush culms to decompose before being buried by tidal-flat mud. Having initiated rapid tidal-flat deposition, the subsidence cannot have preceded deposition of the sand sheet by many months, for the sand accumulated on a marsh, not on a tidal flat. Nor did the subsidence follow sandsheet deposition by more than one growing season: None of the marsh plants survived long enough after subsidence to grow rhizomes or tubers into the sand sheet, toward the aggrading tidal flat. We doubt that much of the subsidence resulted from landsliding or compaction because none of the subsided land appears to have been rotated toward the hillside and because the subsided hillside consists of scarcely compressible diamict and silt (Fig. 4).
The probable age of the sand sheet at West Point is between 1000 and 1100 years ago. This century contains the 95% confidence interval for the time of deposition, as shown by a high-precision radiocarbon age on standing, rooted bulrush culms (S, Fig.  4) (23) and as further shown by radiocarbon ages and matched ring-width patterns of Douglas fir. A Douglas fir log (L, Fig. 4) was deposited with the sand sheet: it rests on patches ofthe sand and on toppled, flattened bulrush culms. Bark on the trunk and on flexible limbs suggests that this fir died close to its time of deposition. A conventional radiocarbon age (Fig. 4) brackets the time of 1616 death between 850 and 1350 years ago. Matching of ring-width patterns shows that death of the West Point fir coincided, to the half year or less, with a landslide into Lake Washington (16). High-precision radiocarbon ages show that this landslide occurred between 1000 and 1100 years ago (16). A tsunami explains the sand sheet at West Point because the sand contains marine fossils, mantles a former tidal marsh, ascends and incorporates hillside deposits, and dates to within months of subsidence and landsliding in the Seattle area. The tsunami probably originated in Puget Sound: Not only does the tsunami correlate closely with local subsidence and landsliding, it also left abundant deposits at sites that show no obvious sign of the largest tsunamis that probably struck the Pacific coast of Washington in the past 2000 years (24). We equate the tsunami at West Point with the one at Cultus Bay because the sand sheets at Cultus Bay and West Point resemble one another in graded bedding and radiocarbon age and because we recognized no evidence for any other tsunami in the past 2000 years at either site.
A large earthquake on the Seattle fault probably generated the tsunami by causing abrupt uplift south of the fault and complementary subsidence to the north (1). Such movement would have caused water in Puget Sound to surge northward across the fault. As it approached the West Point and Cultus Bay marshes, the tsunami probably encountered sandy shallows ancestral to modem tidal flats (Figs. 1 and 3). Sand thus suspended could have settled onto the marshes as the tsunami slowed across them. If tsunami deposits at West Point and Cultus Bay record every large earthquake on the Seattle fault in the past few thousand years, only one large earthquake has occurred on that fault since 2000 years ago. technical Data Report, West Point Treatment Plant Secondary Treatment Facilities, Liquids Stream, v. 19" (Metro, Seattle, WA, 1991). 19. Unlike the peat at Cultus Bay, the tidal deposits above unit A at West Point contain many sand bodies in addition to the sand sheet. These sand bodies are low in unit B and high in unit C and in unit D. None of them, however, show the conformable base, tabular shape, graded bedding, or wide extent of the sand sheet. Rather, the additional sand bodies above unit A are disconformable lenses of interbedded fine to very coarse sand and are restricted to the western half of the excavation. 20. We infer that unit C was deposited rapidly because it shows no upward decrease in radiocarbon ages measured on individual sticks (Fig. 4) and because, in contrast to the probably bioturbated intertidal mud in unit B, unit C is distinctly bedded. 21. Trochammina sp. (14). 22. Our assumptions are that (i) the marsh deposits below and above unit C represent similar positions high in the intertidal zone and (ii) relative sea level changed little during deposition of unit C. We interpret unit C as having formed low in the intertidal zone because it contains many transported shells of mussels (Mytilus edulis), clams (Macoma sp.), and barnacles in the western half of the excavation and because it abounds in well-preserved leaves and sticks yet lacks growth-position rhizomes. This fossil assemblage signifies salt water too deep for the growth of tidal-marsh plants. 23. Scirpus maritimus culms surrounded by the sand Holocene sediments in Lake Washington contain a series of turbidites that were episodically deposited throughout the lake. The magnetic signatures of these terrigenous layers are temporally and areally correlatable. Large earthquakes appear to have triggered slumping on the steep basin walls and landslides in the drainage area, resulting in turbidite deposition. One prominent turbidite appears to have been deposited about 1 100 years ago as the result of a large earthquake. Downcore susceptibility patterns suggest that nearsimultaneous slumping occurred in at least three separate locations, two of which now contain submerged forests. Several other large earthquakes may have occurred in the last 3000 years.
Recently, there has been increasing concern that the Pacific Northwest may be subject to infrequent great earthquakes caused by subduction of the Juan de Fuca plate under North America (1-3). The Puget Sound region itself is also seismically active, and occasional large earthquakes have occurred in recent times, such as the magnitude 7.1 Olympia earthquake in 1949 and the magnitude 6.5 Seattle earthquake in 1965 (4,5).
The history of earthquake activity in Puget Sound has been difficult to decipher because the area lacks traditional morphologic indicators such as well-defined fault scarps from which the timing and areal extent of past earthquakes can be interpreted. Perhaps one of the most promising ways of assessing paleoseismicity is to study continuously deposited sedimentary sequences in lakes and fiords, where basin topography might be conducive to slumping and associated turbidite activity during a major R. E. Karlin earthquake. In addition to slumping from the sides, such bodies of water might also contain evidence of large changes in sediment input caused by earthquake-induced landslides in the drainage basin. Of course, seismically induced changes must be differentiated from changes due to climatic influences (floods, lake level variations) and nonseismic geotechnical effects (delta overloading and slope failure). Lake Washington, bounding the eastern side of Seattle, lies in a steep-sided glacially sculpted valley. The oligotrophic lake averages about 34 m deep and has a subdued W-shaped cross section with marginal elongate troughs 3 to 4 m deeper than in the center of the lake (6). The lake sediments consist of a thick sequence of blue glacial clay of indeterminate thickness that is overlain by 7 to 17 m of Holocene limnic peat or gyttja with a basal radiocarbon age of 13,400 years before the present (yr B.P.) (6,7). The limnic section contains the Mazama ash with a radiocarbon age of -6850 yr B.P. and distinctive post-1916 A.D. laminations that serve as key marker beds. The sediments are anoxic (8), so SCIENCE * VOL. 258 * 4 DECEMBER 1992 sediment disturbance due to bioturbation is minimal. The lake contains three sunken forests (Fig. 1) that were emplaced by massive block slides with trees still in growth position. The submergence of the forests, lying at the north and south ends of the lake, was originally dated at about 1160 14C yr B.P. (9), and more recently, by highresolution dates on rings of standing drowned trees (10).
Sediments from a series of gravity and piston cores taken throughout the lake contain a record of quasi-periodic sedimentary disturbances that may represent turbidity flows or rapid changes in mass flux from the drainage area. Here we report sedimentologic and paleomagnetic analyses of a suite of ten 3-m-long gravity cores that span the last 3000 years and discuss spatial and temporal patterns of sedimentation that constrain the timing, sources, and causes of these disturbances.
Because the magnetic properties of sediments are sensitive to small changes in the concentration and grain size of magnetic minerals, measurements of magnetic susceptibility (X) are an extremely useful remote sensing technique for correlating cores and rapidly identifying lithologic and textural changes. As shown in Fig. 1, susceptibility profiles (11) of the cores show a high degree of intercore correlation. The shape, position, and magnitudes of the X peaks are in close agreement for all cases, and several features can be traced across the lake. The magnetic spikes appear to define terrigenous clay and silt layers, which signify short, intense periods of rapid mud accumulation. One interval at 10 to 30 cm is probably the clay and silt layer deposited as a result of the 3 m lowering of the lake level and opening of the Lake Washington Ship Canal in 1916 A.D. (6,12).
Another dominant susceptibility peak at 80 to 110 cm is present in all cores. The peak is sharp at the base and gradational toward the top, a pattern suggestive of a turbidite because hydraulic sorting causes upward fining and concentration of the heavy magnetic minerals in the coarse basal layer. X-radiographs show a distinctive opaque layer 8 to 10 cm thick at this depth. Visual examination and detailed grain-size analyses on TT195 cores 8, 14, and 15 confirm that the layer shows graded bedding and thus has the characteristics of a distal turbidite.
The X intensities for the horizon at 80 to 110 cm shows a distinctive dependence on location. Magnitudes are highest in the northern and southern cores and lowest in the central cores. A lone exception to this trend is in the core (TT195-5 gc) taken on the western edge of the central basin by Madison Park. This overall pattern suggests that there were multiple detrital sources for 1617 yielded a high-precision age of 1108 ± 16 14C yr before A.D. 1950 (QL-4623), which corresponds to a 95% confidence interval between A.D. 885 and 990 (3). This interval probably includes the time when the dated culms lived. Culms of modern S. maritimus at Puget Sound live less than 1 year and rarely stand dead for more than 2 years. By analogy, the dated culms lived within 3 years of the abrupt subsidence that killed the S. maritimus and that coincided, within months, with deposition of the sand sheet. 24. Oceanic tsunamis produced sand sheets along the southern Washington coast 300 and 1400 to 1900 years ago (10, 11)-times when little or no sand accumulated at our Cultus Bay and West Point sites. 25. Laboratory numbers, from greatest to least age: Beta-51806, -48232, and -48231; USGS-3090; Beta-51805. Ages calculated with an assumed 813C value of -25 per mil except for USGS-3090 (1040 ± 35 14C yr B.P.), which was calculated with a measured value of -26.8 per mil. Use of