In addition to the refinement of tsunami propogation models using computer simulations and the collection of data at ocean sensor sites, referred to in Tsunami Warning Systems and Tsunami Prediction, a number of important developments are likely in the next decade or so. An impression of some of the current research issues and problems of interest can be gained from the abstracts of the journal "Science of Tsunami Hazards", the journal of the Tsunami Society, at http://www.ccalmr.ogi.edu/STH/.

Of these, perhaps the following are most important (in addition to the propagation and impact modelling efforts) for the development of frequency - magnitude distributions for tsunamis produced by different (especially non-earthquake) source mechanisms, prediction of tsunami hazard zones and return times for the larger tsunamis not represented in the historical record, and other aspects of interest to the insurance industry.

- Further investigations of tsunami deposits and other evidence for palaeotsunamis: interpretation of these deposits in terms of the magnitude of the tsunamis, and velocities of the surges and bores from which they were deposited. Application of new dating techniques to a greater variety of materials in these deposits will allow better assessment of the frequency of tsunamis in regions such as the Atlantic where tsunamis are rare.

- Accurate dating of the large terrestrial events (continental slope sediment landslides, oceanic island lateral collapses in particular) that are suspected to produce giant tsunamis, and correlation where possible of such events with transoceanic tsunami deposits. A particular issue in relation to these events is that of whether climatic or environmental factors may affect their frequency. Suggestions of particular interest include the proposal that changes in sea level and sedimentation rates may trigger continental slope failures through gas hydrate decomposition (Kayen & Lee, 1993) or overloading with rapidly deposited sediment; and that environmental factors such as changes in rainfall patterns may affect the water tables within volcanoes and thus make their collapse and the consequent tsunamis more frequent (Elsworth & Day, 1999; research in progress). The latter model is potentially particularly important because it may imply short-term increases in the collapse-generated tsunami hazard over the next centuries.

- More accurate evaluation, most probably from space surveys of asteroid sizes, orbits and abundances as much as from investigations of past events, of the frequencies of impact of asteroids of different sizes in the size range (c. 200 m to c. 1 km diameter) at which impact-generated tsunamis are the most important damage mechanism (Morrison et al., 1994).

- More accurate physical models for the generation of tsunamis by non-earthquake mechanisms, in particular by fast-moving landslides entering the sea; by submarine landslides involving the motion of rigid blocks or groups of blocks (such as the megablocks in the Hawaiian volcano landslides); and by hypervelocity asteroid impacts. At present there are order-of-magnitude uncertainties in the efficiency of these processes, with consequent effects upon estimates of the frequency magnitude distributions of tsunamis produced by them (Figure). These are being reduced by numerical models of the landslide motion, coupled to wave generation models (for an example, see Tinti et al., 1999).