Long-Term Records of Tsunamis
Contributors:
Ben Horton
Brady Rhodes
Martitia Tuttle
Harvey Kelsey
Lisa Doner
Alan Nelson
Marco Cisternas
I. Introduction
Damaging tsunamis are rare but costly. Consequently, studies of historical tsunamis—as recorded in written and video archives or instrumental sea-level data—are limited in what they can tell us about the scale and complexity of past tsunamis. Long-term (hundreds to thousands of years) records of tsunamis, as recorded by the sedimentary deposits they leave behind, help us understand tsunami processes by greatly expanding the range of tsunamis available for study. Understanding potentially catastrophic coastal processes, such as tsunamis, is key to assessing and mitigating the undesirable effects of both catastrophic and non-catastrophic processes on the ecology, geomorphology, and rapidly expanding human populations of coasts. Well preserved sequences of tsunami deposits can be used to estimate recurrence intervals of tsunami source events, including earthquakes, landslides, or volcanic eruptions, and to model source processes on meaningful timescales, for example, the long-term behavior of plate-boundary fault systems. Ancient tsunami deposits reveal the scale of catastrophic events such as meteor or asteroid impacts, for which there is no modern analog. By promoting cross-disciplinary collaboration in fields as diverse as biology, geomorphology, geophysics, marine geology, sedimentology, seismology, and human geography, research on long-term tsunami records may lead to unanticipated advances in multiple fields. Because study of long-term records improves our understanding of modern tsunami hazards, long-term tsunami research contributes to mitigation efforts around the world.
II. Previous Studies Using Long-Term Records
Tsunamis
The literature on tsunamis and tsunami deposits has rapidly expanded over the past decade (e.g. Tappin, 2004; Pelinovsky & Tinti, 2005) and the Indian Ocean tsunami of December 2004 will dramatically accentuate this trend. Inferring the character and scale of past tsunami source mechanisms (e.g., sea-floor faulting, mass movement) and the impact of tsunamis on former coasts is an iterative process involving modelling and field geology that begins with the identification of tsunami deposits. Such studies require integration of terrestrial and marine sedimentary records, observations of modern tsunamis, paleoenvironmental reconstructions, and computer simulations (Edwards, in press).
Modern Observations: Although tsunamis deposit beds of sediment that are anomalous in many settings, such as beds of marine sand within sequences of terrestrial peat or freshwater lake sediment (e.g. Dawson et al., 1991; Bondevik et al., 1997a, b,), rarely are the lithology and other characteristics of the beds unique. For example, storm surges may deposit marine sediment many meters above normal tidal levels (Dawson & Shi, 2000). Comparative studies examining storm and tsunami deposits are particularly valuable in this respect (e.g. Nanayama et al., 2000), as is documentation of the processes and impacts of sediment erosion and deposition during modern tsunamis (e.g. Maramai et al., 2005a, b). A recent example is the study by Goff et al. (2004) in New Zealand, which compares sediment deposited by a 15th century tsunami and a large storm that occurred in 2002. The two types of deposits show clear differences in their sedimentology??, bed continuity, and inland extent. Similarly, a study comparing tsunami deposits in southern Newfoundland resulting from the 1929 Grand Banks earthquake with deposits resulting from submarine slides and storms in New England during the 1991 Halloween storm found that the deposits differed in their landscape positions and sedimentary characteristics?? (Tuttle et al., 2004). On the basis of these and similar differences described in the literature for other tsunami and storm deposits, Tuttle et al. (2004) outlined preliminary criteria for distinguishing the two types of deposits. More detailed studies of tsunami deposits and their look-alikes will aid development of diagnostic criteria for the identification of prehistoric tsunami deposits (e.g. Goff et al., 2001; Smith et al., 2004; Bondevik et al., 2005; Williams et al., 2005). Replace vague terms above with specific sediment characteristics.
Models of processes capable of transporting coarse gravely sediment inland are much less developed for rocky coastlines than for sandy coasts (Felton & Crook, 2003). Consequently, interpretations of potential tsunami deposits on rocky coasts are more inferential (e.g. McMurtry et al., 2004; Scheffers, 2004). New data suggests that storm waves may transport larger clasts than previously thought (Edwards, in press). Recent studies of modern coarse storm deposits suggest previous modeling of the height of wave required to lift blocks of a given mass (e.g. Nott, 2003) overestimates wave height (e.g. Mastronuzzi & Sanso, 2004). For example, Williams & Hall (2004) graphically illustrate the power of storm waves in their description of boulders weighing several tons found as much as 50 m above sea level on cliff-tops along the Atlantic Irish coast. Such reassessments of the power of storm waves are prompting reinterpretation of some proposed tsunami deposits.
Paleoenvironmental reconstructions: Investigations of past tsunamis can be used to develop and test models of tsunami generation and runup, and identify regions at greatest risk from future tsunamis (Edwards, in press). The number, timing, and height of tsunami waves reflect the source mechanism. For example,for fault-related tsunamis runup height rarely exceeds twice the fault slip (Okal & Synolakis, 2004). Consequently, maximum water levels of 25-30 m recorded in Sumatra (Stein & Okal, 2005) suggest the recent Indian Ocean tsunami was generated by fault slip of 12-15 m. Mass movements have the potential to produce even higher waves, but tsunamis produced by submarine slides and slumps are particularly difficult to model (Pelinovsky & Tinti, 2005). Modeling is further complicated when large earthquakes with sea-floor displacement on faults also trigger mass movements.
Smith et al. (2004) comprehensively review evidence for the Storegga Slide tsunami from 32 coastal sites in northern Britain. At most sites, evidence consists of widespread but anomalous sand beds containing marine microfossils that date from ca. 8000 cal. yr BP. Sedimentological evidence?? from these sites confirms that the beds were deposited by more than one wave, supporting Bondevik et al.’s (2003) interpretation of multiple waves from the characteristics of tsunami deposits in lakes? on the Shetland Isles. Tooley & Smith (2005) show that tsunami deposits can be identified even within a sequence of high-energy deposits; they describe two fining-up sequences in coarse sand and gravel from eastern? Scotland.
Reconstructing runup height from a tsunami deposit requires information on the maximum elevation attained by tsunami waves, and the elevation of sea level at the time of the tsunami. Runup heights based on the inland extent of anomalous sand beds are minimum estimates because water levels exceed the elevation of deposited sediment (Dawson, 1999; Dawson & Shi, 2000; Tuttle et al, 2004). In many studies, the precision of inundation heights of past tsunami is further limited by estimates of former sea-level at the time of deposition based on geophysical models of glacial-isostatic adjustment of the crust (e.g. Bondevik et al., 2003). In an interesting development, Smith et al. (2004) use the tsunami deposits as a time horizon. By locating the inland limit of intertidal sediment capped by tsunami deposits, they reconstruct the shoreline position at the time of the tsunami, and from the position identify subsequent patterns of coastal retreat. Smith et al. (2004) attribute the variable pattern of runup height to differences in wave erosion, and to differences in tide level at the time of the tsunami.
Kelsey et al.’s (2005) study of Bradley Lake, on the southern Oregon coast, revealed a 7000-yr-long record of local plate-boundary earthquakes and accompanying tsunamis on the Cascadia subduction zone. At least 12, and probably 13, tsunamis deposited landward-thinning sheets of sand , derived from nearshore, beach, and dune environments to the west, in the lake. Kelsey et al. (2005) calculate that the tsunamis rose at least 5–8 m above sea level, and that the cumulative duration of each tsunami was at least 10 min. Between 4600 and 2800 cal yr B.P., tsunamis occurred at the average frequency of ~3–4 every 1000 yr. Then, starting ~2800 cal yr B.P., there was a 930–1260 yr interval with no tsunamis. That gap was followed by a ~1000 yr period with 4 tsunamis. In the last millennium, a 670–750 yr gap preceded the A.D. 1700 earthquake and tsunami. Kelsey et al. (2005) suggest that the A.D. 1700 earthquake may be the first of a new cluster of plate-boundary earthquakes and accompanying tsunamis.
Determining the frequency of tsunamis through dating tsunami deposits is complicated by the difficulties in developing tsunami histories from sedimentary sequences (Edwards, 2003, in press; e.g.,). With a few notable exceptions (e.g., Satake et al., 2003; Ollerhead et al., 2001; Clague et al.,), dating of tsunami deposits relies on 14C ages, which typically have errors of at least many decades and commonly hundreds of years (Nelson et al., 1996; Atwater and Hemphill-Haley, 1997; Witter et al., 2003; Kelsey et al., 2005). For example, the uncertainty in the time of the Storegga Slide tsunami in the North Sea is still hundreds of years despite many tens of 14C ages due to dating errors and other problems such as erosion of slide deposits. Fish skeletons and plant macrofossils preserved within Storegga slide deposits suggest that the tsunami occurred in late autumn (Bondevik et al., 1997; Dawson & Smith, 2000). A related problem is that many radiometric 14C ages on bulk, organic-rich sediment are commonly younger than the time of sediment deposition (Nelson, 1992; Bondevik et al., 2003, in press). Nevertheless, even though the times of individual tsunamis may be uncertain, errors in radiocarbon dating are rarely a significant source of uncertainty in determining the average recurrence of tsunamis from long tsunami records.
Dating errors are of much greater concern when the ages of tsunami deposits are used as a diagnostic characteristic when correlating deposits to show their tsunami origin. For example, Williams et al. (2005) attempt to fingerprint the sources of nine muddy sand beds preserved within a tidal marsh in Washington State, USA by comparing bed ages with ages for tsunami deposits at other sites in the region. The authors conclude that four to six of the beds are probably tsunami deposits, but that the tsunamis probably had various sources, such as mass movements and distant as well as and nearby earthquakes. Correlation of deposits laid down by tsunamis accompanying plate-boundary earthquakes is complicated by the fact that the frequency of high tsunamis on many plate-boundary coasts is of similar magnitude to the errors on typical tsunami deposit ages (Nelson, 1992; Williams et al., 2005).
Model simulation and marine data: Computer models are used to simulate patterns of wave propagation, height, and other characteristics of tsunamis from different source mechanisms (McMurty et al., 2004; Fryer et al., 2004; Løvholt et al., 2005). For example, Bondevik et al. (2005) compare the extensive set of geologic runup data from Norway, Scotland and the Shetland Islands, with numerical simulations of the Storegga slide. Their best-fit model suggests that sea levels along the Norwegian coast fell by 20 m during the first 30 minutes following the slide. The simulation also predicts the generation of multiple waves, matching inferences about tsunami characteristics based on sedimentological data. In an alternative application, Okal (2005) uses a model to choose between two sources for the 1906 Pacific-wide tsunami. Although the two tsunamis occurred within 30 minutes of each other, the modeled tsunami matches only the far-field characteristics of the tsunami generated in Chile.
Additional information on tsunami sources is provided by marine seismic data. The location, extent and architecture of the Storegga slide is now well-constrained by geophysical surveys (Haflidason et al., 2004). The rest of this paragraph needs a better home. It is not about marine seismic data. The triggering mechanism is still under investigation, and Bryn et al. (2005) suggest a strong earthquake is a likely cause. This may have been facilitated by excess porewater pressure in the sediments brought about by high rates of deposition. Solheim et al. (2005) suggest that immediately following deglaciation, rapid sedimentation coupled with glacio-isostatic seismicity could produce conditions favourable for slope failure. They report seven large pre-Holocene slides in the area that appear to form a complex of failures related to the glacial-interglacial cycle. The recent discovery on the Shetland Islands of two tsunami-like deposits post-dating the Storegga Slide suggests that this instability persists throughout the Holocene (Bondevik et al., in press). Hutton & Syvitskli (2004) model sediment failures under changing sea levels, and note that whilst most occur during sea level falls or lowstands, the largest volume failures are associated with rising sea levels and highstands.
In fact, the nature of coastal sedimentation is increasingly being tied with tsunami risk. Using a catalogue of historic tsunamis in the Pacific Ocean, Gusiakov (2005) shows that sedimentation has a strong control on the likelihood that an earthquake will generate a tsunami. In this paper, written almost a year before the Indian Ocean tsunami of 2004, the extremely high efficiency of earthquakes in the western part of Indonesia is highlighted. In a recent paper, Syvitski et al. (2005) use a model to examine how the flux of sediment into the global coastal ocean has changed due to human activities. One of their results is that, in contrast to many other parts of the globe, Indonesian rivers now deliver much more sediment to their coastal waters than before. This research highlights the need for a greater understanding of land-ocean sediment fluxes and processes operating on the shelf (Long, 2003). I had a hard time understanding this paragraph and why it is important for this report. It needs to be deleted or completely rewritten.
Tsunami deposits and earthquake-related land-level and sea-level changes
A major factor in the long-term preservation of tsunami deposits is the direction and rate of vertical movements of shoreline areas that reflect crustal deformation during cycles of tectonic strain accumulation and their release during earthquakes, particularly near plate-boundary faults (Clague, 1997). Tsunami deposits are best preserved where coasts subside (and relative sea level rises) during large or great earthquakes because the deposits are rapidly buried by intertidal sediment (e.g. Atwater, 1987; Atwater & Hemphill-Haley, 1997). Where either coseismic or postseismic uplift exceed net subsidence or where long-term regional sea level falls (e.g., Atwater et al., 1992; Cisternas et al., in press), shorelines will rise relative to the sea and tsunami deposits may only be preserved in special environments such as lakes (Bondevik et al., 1997; Kelsey et al., 2005).