Since the most important causes of tsunamis (earthquakes in particular) are inherently unpredictable, in the sense that it is not possible to give an accurate time and location for a future source event, specific predictions or forecasts of tsunamis are also not going to be possible in the foreseeable future. The system of watches and warnings put in place after a potential source event such as an earthquake (see Tsunami warning systems) will remain the standard approach for some time to come. However, in the more general context of probabilistic prediction for purposes of hazard zonation, disaster mitigation planning, and the setting of insurance premiums, two developments in particular offer considerable promise.

First, high resolution digital bathymetric maps of continental shelf areas are beginning to appear, with horizontal resolutions of the order of hundreds of metres and vertical resolutions of as little as a few metres. These are produced by means of multibeam bathymetric sonar surveys. Present large-scale compilations only cover a few areas, such as the East Coast of the USA (although corresponding maps for the Pacific Coast, the Gulf of Mexico and Caribbean territories of the USA are also due to appear in the next five years). These maps, together with similar but lower-resolution maps of the deep ocean, will allow accurate calculation of tsunami propagation paths and also identification of areas of refraction and other phenomena which lead to focussing of wave energy upon particular sections of the coastline.

Secondly, the future availability of such maps, together with that of cheap large-scale computing power, will allow larger numbers of numerical simulations of tsunamis to be produced. Whilst at present these are largely restricted to post-event modelling of tsunamis (where the availability of actual runup data from post-event surveys and eyewitness accounts allows detailed constraints to be put on the models: see for example the simulations at http://corona.pmel.noaa.gov/~tsunami/PNG/png_animation.html, a future application would be the detailed simulation of large numbers of model tsunamis upon coastlines of interest to allow the probabilistic prediction of tsunami hazards at, for example, concentrations of value-at-risk such as coastal towns and tourist complexes or at hazardous installations. The central problem in such simulations will, however, be an understanding of the physics of tsunami generation at the source.