Some of these problems, particularly those related to false alarms and the unreliability of tide gauges, are presently being addressed:

1) Deepwater buoy and pressure sensor systems are being installed, most notably off the west coast of the USA, Hawaii and Japan. The American system is described at http://www.pmel.noaa.gov/tsunami-hazard/warning.html. These are designed to detect the very small pressure changes on timescales of tens of minutes that are produced on the ocean floor by the passage of tsunami waves. Since they are not affected by the near shore phenomena that affect tide gauges, and because they can be positioned at will on the routes that tsunamis from major sources follow to impact major population sensors, these offer the potential for much more accurate upgrading of tsunami watches to tsunami warnings. Furthermore, the data collected from them on tsunami amplitudes and wave forms in the deep ocean will greatly aid the process of tsunami modeling.

2) Rapid modeling of the tsunami generation process by interpolation. Computer simulations of tsunami generation involve very large amounts of computational power and considerable time, even on the most powerful machines. However, efforts by the Japan Meteorological Agency (Tatehata, 1997) and the Pacific Marine Engineering Laboratory, amongst others, are now producing large sets of model tsunamis, produced by earthquakes of different magnitudes, focal plane mechanisms, focal depths and epicentral locations on a regular grid. The PMEL programme, the Method Of Splitting Tsunamis, is described at http://www.pmel.noaa.gov/pubs/PDF/tito1927/tito1927.pdf. It is a relatively simple matter to interpolate between these precomputed model tsunamis to produce a rapid estimate of the likely magnitude of a tsunami event, as well as wave arrival times. Future work along these lines may allow tsunami alerts to contain estimates of wave runups at key locations as well as times of arrival for the first wave. At present, however, the simulations commonly under- or over- estimate runup values.