We quote here two case studies which illustrate these problems very well. Both examples are extremely important in terms of insurance and reinsurance in respect of the Pacific coastline of the USA and Canada and also for coastlines bordering the North Atlantic and North Sea.

Along the coasts of the northern North Sea, Norwegian sea and north eastern Atlantic ocean a very prominent sand layer originally thought to have been deposited by a storm surge. More recently, it has been attributed to a large tsunami circa 7,100 14C years ago. This event was generated in the Norwegian sea as a result of the Second Storegga submarine slide, one of the world's largest underwater slides.

The detail with which the tsunami is known is impressive. Several studies have examined the sedimentology of the deposit, including its particle size, microfossil content and even the time of year it occurred. The diatom ecology of the layer has been examined and over 100 radiocarbon dates have been used to date the event.

Geological investigations of these tsunami deposits on the northern and eastern coastlines of Scotland as well as in uplifted lake basins along the west coast of Norway provide evidence of minimum tsunami runup. In eastern Scotland the minimum value of runup associated with this tsunami is in the order of 4-6 metres above contemporary high water mark. However this value should be treated with caution since tsunami flooding to higher elevations may have taken place yet did not leave a sedimentary record. A numerical model of the Second Storegga submarine slide has been developed and showed that the scale of tsunami runup along the Scottish and Norwegian coastlines very much depended upon the average landslide velocity. For example, it was noted that an average slide velocity of 20 m/s resulted in runup values onto adjacent coastlines of between 1-2 m. By contrast a modelled landslide velocity of 50 m/s resulted in runup values of between 5-14 m, values significantly in excess of the estimates for adjacent coastlines based on geological data. Harbitz (1992) concluded that a landslide velocity of 30 m/s provided the closest approximation to the estimated runup values based on geological data. However the weakness in this argument is that the geological data only provide minimum estimates of likely flood runup and therefore the related numerical models of the same tsunami will always underestimate the likely average value of the submarine slide velocity.

The occurrence of this tsunami is unusual since it appears to have been generated by one of the world's largest submarine sediment slides rather than by an earthquake. This serves to demonstrate that severe tsunamis can be generated by submarine slides in areas where major earthquakes do not normally occur.

The Storegga Slides example of an extremely large tsunami affecting most coastlines bordering the North Atlantic and North Sea regions (and possibly further afield also), poses a tremendous problem in terms of understanding and evaluating tsunami risk. To all intents and purposes no major tsunamis have struck the coastlines of these regions since this event took place approximately 7,100 years ago. A huge underwater landslide produced the tsunami in question. It is not known if an earthquake triggered the landslide or if methane gas release from sediments on the seabed produced the instability that triggered the underwater landslide. Whatever the reason there is no unequivocal evidence of any major catastrophic tsunami ever having taken place since. How then does one attempt to produce a statistical evaluation of the likely return period of a major tsunami?

The first problem is that the origin of this tsunami is not known. If earthquakes were involved in producing the underwater landslide then evaluation of offshore earthquake recurrence in this area may be important (see HSE Section). Above all, since the generating mechanisms are to all intents and purposes random, are we then to estimate the return period for a new catastrophic tsunami in the future on a random basis.

Additional geological data for this area partly contributes to understanding this problem. Thus, for example, we know that a much bigger underwater landslide took place in the very same area approximately 30,000 years ago, just prior to the last Ice Age in Northern Europe. We also know that a third underwater slide took place shortly after the second and is generally estimated to have taken place somewhere between 8,000 and 5,000 years ago. We therefore know that it is likely that three tsunamis were produced over a timescale of 30,000 years, but with the complication being that there is only a geological trace of one of these. It is also possible that if landslides are triggered on the seabed and they move slowly, there may not be a sufficiently rapid water displacement to generate a tsunami.

Detailed information from coastal Washington State indicate the former occurrence of a large tsunami that accompanied an episode of coseismic coastal submergence during a large earthquake which took place circa 300 years ago.

Sedimentary evidence for this tsunami is widespread throughout the Pacific west coast. Similar sheets of sand attributed to the 1964 great Alaska earthquake and tsunami have been described for coastal areas of British Columbia. Atwater and Moore also described stratigraphical evidence from the Puget Sound, Washington, for a palaeotsunami that flooded coastal areas circa 1000 years ago. At Cultus Bay a sand sheet between 5-15cm in thickness containing marine microfossils is enclosed within peat. It has been argued that the lack of other sand beds distinguishes this surge for most or all storms in the region during the past 2000 years.

The dating of the Cascadia episode of coseismic subsidence and the related tsunami led Satake et al. (1996) to conclude on the basis of historical records that the circa 300 years BP palaeoearthquake may have generated a trans-Pacific tsunami that struck the coastline of Japan on 26th January 1700. Satake et al. (1996) concluded that the reports of waves striking the Japan coast at this time were more characteristic of a far-field tsunami rather than a storm surge and noted that most storm surges in Japan are generated by typhoons that normally take place between August and October. This observation together with the fact that the historical reports appear to indicate an apparently uniform distribution of water heights along the Japan coastline at this time supported the notion of a palaeotsunami having taken place rather than a storm surge. It is possible that this tsunami may have been a trans-Pacific tsunami and affected coastlines as far apart as Alaska, New Zealand and southern Chile. Goff (pers. comm.) describes some evidence from Polynesia and New Zealand of a major tsunami at this approximate time. Certainly an event sufficiently large that was capable of causing serious destruction simultaneously along the coast of Japan and also along the Pacific coastline of Canada and the United States is likely to have been associated with wave propagation throughout the Pacific Basin.

As previously mentioned the chronology of historic tsunamis for countries bordering the Pacific is restricted to a series of historic documents that extend no further back in time than the 18th century. There are myths and legends from ancient civilisations in western South America of destructive tsunami. Similarly, there are reports of great sea floods by Maori communities in New Zealand. By contrast, the Japanese historical record is considerably longer, but here the proximity of a deep ocean trench and the high frequency of offshore earthquakes close to Japan imply a tsunami record for Japan which has a major component related to local events. It would be fair to say that geological investigations of prehistoric earthquakes for the coastlines bordering the Pacific are not sufficiently advanced at the present stage to enable scientists to discriminate between local tsunamis affecting individual sections of coastlines or even countries and giant trans-Pacific tsunamis that caused destruction along most coastlines bordering the Pacific Basin.

How therefore does one evaluate tsunami risk for coastlines and properties bordering the Pacific West Coast of the United States and Canada? We have a situation where there are historical records extending to circa 1740-1750 AD together with detailed information on a prehistoric event that is known to have taken place on 26th January 1700 AD. American data for this event is based on geological information that is imprecise and certainly is no more accurate than +/- 5 years. However, the Japanese data is very precise since it is based on historical information of a major trans-Pacific tsunami having taken place. A set of scientific arguments has been used to correlate between the two reports of tsunami flooding.

Records of prehistoric tsunamis for the Pacific West Coast of Canada and USA prior to 1700 AD are extremely fragmentary. There are a number of reports that describe stratigraphical evidence from Washington State for a tsunami that took place approximately 1000 years ago. For the majority of time for, say, the last 10,000 years there is almost no information. Therefore, is it a valid exercise to interpret that since that an exceptionally destruction tsunami took place approximately 300 years ago in this area, that the return period for a future high magnitude tsunami is in the order of 300 years? Or is it the case that the destructive tsunami of 1700 AD was the most destructive in the previous 10,000 years? The fact that there are no satisfactory answers to these questions makes its exceptionally difficult to evaluate the return period of tsunamis for this area of the world. Suffice it is to say, however, that these coastlines have been periodically affected by exceptionally destructive tsunamis and there is a possibility that this may happen again in the future. The event of 1700 AD is also important since the generating mechanism is thought to have been an offshore earthquake in the general area of the Cascadia Subduction Zone, a deep ocean trench located immediately to the W of western Canada and the United States.

The same problem is also applicable to the coastlines of Portugal and southern Spain. Here, an exceptionally destructive tsunami took place on November 1st 1755 AD. The earthquake and the tsunami together led to loss of life of approximately 50,000-60,000 people. Geological studies of sediments along the coastline of southern Portugal appear not to indicate any geological traces of any older destructive tsunamis, certainly not for over the last circa 4,000 years. We are also faced with a similar problem as in Cascadia that an exceptionally destructive event took place only 250 years ago. How therefore does one attempt to predict into the future and estimate future tsunami risk since the return period may not be once in 250 years but may be considerably longer?