TSUNAMI WARNING systems have two distinct phases, each with its own particular challenges.
First, readings from sensors and seismic monitoring stations have to be received and analysed, and a warning issued to the governments of the coasts at risk.
These governments then have to inform the threatened population in time for them to reach safe havens before the tsunami strikes.
The main problem is that not all undersea earthquakes cause tsunamis and not all tsunamis are caused by earthquakes.
Meteor strikes, land sliding into the sea and underwater sediment slides can all trigger long period surface waves capable of causing major damage to shorelines.
And even massive earthquakes similar to those triggered by California's San Andreas Fault can often produce little sea movement.
For example, when the tectonic plates are sliding sideways relative to each other there will be little vertical displacement.
A very different earthquake scenario took place off Sumatra on 26 December and experts have concluded that an earthquake of such a magnitude in this location was bound to be followed by a major tsunami event.
However it is not always possible to predict a tsunami accurately. For example, in Papua New Guinea in 1998, a much smaller earthquake, that would probably not have triggered an alert, is thought to have caused undersea mudslides.
These produced a very violent localised tsunami that killed at least 1,500 people.
By contrast, only 239 Japanese died when a 30m high tsunami smashed into the northern island of Hokkaido in 1993.
Most of the population at risk had been warned, had frequently rehearsed tsunami evacuations, and had access to stout shelters.
In this case the wave took only five minutes to arrive after the first warning.
Major tsunamis can cross the entire Pacific in 10 hours or so and still have enough energy to cause destruction.
This travel time gives scientists time to analyse seismic and pressure sensor data and give the scientists the opportunity to predict the likely scale of the waves when they arrive.