An alternative approach to determining the long-term frequency of the larger tsunamis in particular is that of investigating geological evidence for both tsunamis themselves and for the causative events. The former approach is often most relevant in the case of earthquake-generated tsunamis, whereas large submarine landslides and volcano lateral collapses produce highly distinctive deposits which can be readily identified in the geological record.

Tsunami deposits

Known occurrences of tsunami deposits fall into two main categories: sand sheets and boulder beds.

Tsunami sand sheets in coastal marshes: the example of Cascadia.

The northern Pacific coastline of the United States, in Oregon and Washington, contains a large number of coastal estuaries, lagoons and marshes which under normal circumstances are isolated from the ocean by sand dunes on coastal spits. For much of the time these coastal areas are marked by the gradual accumulation of mud and the growth of salt-tolerant woodland on top of the mud. However, excavations have shown that interbedded with the muds are laterally extensive - extending inland by distances of several hundreds of metres to kilometres - sheets of sand. Furthermore, the deposition of each sand sheet is followed immediately by a return to deeper-water estuarine conditions, indicating sudden subsidence of the coast by up to 2 to 3 metres.

This very characteristic sequence is repeated up to a dozen times along parts of this coastline. Investigations by Atwater and others (summarized in Atwater et al., 1997) have shown that each repetition marks the occurrence of a large subduction zone earthquake offshore from Oregon and Washington, leading to coseismic coastal subsidence (see diagram generation of a tsunami by a subduction zone earthquake to see why) and the generation of a tsunami which impacts the coastline, strips off sediment from the coastal beaches and dunes, and deposits this in the lagoons and estuaries as the sand sheets.

C-14 dating of trees killed by the earthquakes and coseismic subsidence, and of organic material in the intervening muds has shown that these dozen earthquake tsunami events have occurred over the past 7000 years or so, with recurrence intervals of about 300 to 700 years and the last event about 300 years ago, consistent with the inference that this event produced the 26th January, 1700 A.D. tsunami in Japan. Tsunami deposits and archaeological evidence subsidence from this event and from some of the others are found all along the Oregon - Washington coastline and as far north as central Vancouver Island, implying that the source events were most probably giant subduction zone earthquakes with rupture lengths of several hundred kilometers and Mw ? 9. The recurrence interval and the size of the 26 January 1700 tsunami in Japan are also consistent with this, as is the distribution of turbidite sands offshore. Interestingly, the coastal tsunami sedimentation record in Cascadia is accompanied by Native American legends and episodes of village abandonment at coastal settlements excavated by archaeologists which are also consistent with the occurrence of major tsunamis along this coastline.

The record from Oregon and Washington is perhaps the most complete and impressive record of prehistoric tsunami deposits from anywhere in the world. Comparable deposits have however been found in estuaries and on coastal plains in a number of places around the world (for example, the Storegga landslide tsunami deposits in Northern Scotland), and also compared with more recent deposits such as those from the 1755 Lisbon tsunami in Portugal. These studies indicate that:

1. The deposits are highly distinctive, as a result of rapid deposition from fast-flowing, sediment laden water in a single flooding (or at most a few floodings), and are easily distinguished by both their sedimentology and large lateral extent from storm deposits.
2. The preservation of the deposits is patchy, most especially toward the landward limit of the inundation zone, and highly dependent upon the occurrence of favourable conditions for preservation such as in lagoons or estuaries.
3. There is an abundance of material in the deposits suitable for C-14 radiometric dating and dating by other means.

These results mean that although under favourable conditions such deposits provide good indications of the occurrence of tsunamis and even of typical recurrence intervals where a number of successive tsunami deposits have been found (as in Cascadia), their absence (especially on rapidly eroding coastlines) does not imply an absence of tsunamis in the past. Furthermore, they do not usually provide a good indication of inundation distances or tsunami magnitudes, because of the loss of deposits by erosion from the original landward edge of the deposit, or because these were never areas of tsunami sand deposition in the first place. The extent of the remaining deposits only provides a minimum value for the inundation distance of the source tsunami waves. Since these sand sheet deposits are only found on flat coastlines, they never provide runup values comparable to those usually recorded for historical tsunamis, making comparison of the two sets of records difficult.

Boulder Beds

Individual large boulders and localized boulder beds, along with evidence of catastrophic erosion events, have been found along a number of rocky coastlines. Examples from Australia and the Bahamas are described by Young & Bryant (1992), Nott (1997) and Hearty (1997). These have been interpreted as the product of the impact of giant waves (most plausibly tsunamis).

Fast-moving tsunamis (breaking waves, surges and bores) up to several tens of metres in height

Perhaps the most famous examples, however, are the coral gravel and boulder beds from the Hawaiian islands, described by Moore & Moore (1984) and interpreted as the products of giant tsunamis with wave runups of up to 375 metres, produced by oceanic island lateral collapses on the island of Hawaii itself, or asteroid impacts. These deposits have proved controversial, not least because of the alternative suggestion that they are the remnants of beaches displaced to their present position by uplift of the entire islands since their deposition. However, more recently comparable boulder deposits have been found in the Canary Islands, at 100 metres or more above sea level; and it is known that the Canary Islands have not undergone uplift or subsidence in the relevant period.

A more fundamental problem in the use of these deposits for tsunami hazard analysis is that they are invariably remnants of originally more extensive deposits, or isolated patches of sediment left by largely erosional waves. They therefore provide only limited information (absolute minima) on runup heights or inundation distances, hence upon the magnitudes of the tsunamis involved, and it is likely that comparable tsunamis have occurred without producing any deposits that remain at the present day. They are nonetheless extremely valuable in indicating the past occurrence of giant tsunamis far larger than any in the historical record, with runup heights measured in hundreds of metres.