Science still learning how to limit tsunami misery
The modern world has never been hit in a single blow by a natural disaster as far reaching as the earthquake and tsunami of 26 December 2004.
It was not just the violence of the event - one estimate given to the BBC shortly after was that the energy released was the equivalent of more than 10,000 Hiroshima bombs.
Vulnerability was as important. Densely populated communities along the coastlines of northern Sumatra and around the Bay of Bengal had no protection of any kind against the waves that stormed ashore.
In the worst hit area, the port of Banda Aceh on the northern tip of Sumatra, and along the surrounding coastline, 170,000 were killed by the tsunami, about a quarter of the population.
Sumatra was already known to be vulnerable like no other region to natural disasters.
Nevertheless, the danger of a fault buckling the sea floor along a 1,600km seam from the northern end of Sumatra up to the Andaman Islands in the Bay of Bengal had been overlooked. No-one was prepared, and there was no warning system.
It is instructive to compare the Indian Ocean tsunami with the Japanese tsunami in 2011.
The mechanism was much the same, a massive undersea earthquake moving hundreds of kilometres of ocean floor and releasing strain built up over centuries of plate tectonic motion.
In Japan, too, experts were surprised particularly by the scale of the 2011 earthquake.
Nevertheless, with its history of recurrent earthquakes and deadly tsunamis, Japan was well prepared for the onslaught, with massive walls intended to keep out the waves, and tsunami alert systems, sirens and broadcasts warning people to evacuate to safe ground.
Without those preparations, there is no doubt tens of thousands more would have perished.
But the fact that nearly 20,000 still died showed that the protection was not enough.
The tsunami walls crumbled or were overtopped by the unexpectedly high waves.
Some people were lulled into a false sense of security. And evacuation plans were inadequate. Some buildings designated as safe zones for residents to retreat to in the event of a tsunami were not high enough, and the evacuees drowned in their refuges.
The Japanese warning system was based on the seismic signals radiated by the undersea grinding of the tectonic plates.
Those signals were then automatically compared to a library of scenario events computed long before, with wave heights calculated on the basis of the best available science.
But because the best available science had not included a magnitude nine earthquake, nor could it forecast 30m waves.
The responsible agency kept updating the forecasts as the event progressed, but with most of those in the waves' path already fled from their homes, and power cut off, it is not clear who heard the revisions.
In some ways, by 2011, Indonesia had a better approach than Japan, thanks to a collaboration with Germany called the GITEWS (the German-Indonesian Tsunami Early Warning System).
Ocean-bottom pressure sensors far out at sea could feel the weight of a tsunami wave passing overhead.
And GPS detectors on land which move during an earthquake can give an indirect measure of the movement on the sea floor.
The system got a real live test in 2012, when a pair of strong earthquakes stirred up a minor tsunami far offshore from Banda Aceh.
Parts worked, a reliable warning was issued within five minutes, although some of the deep-sea buoys had been damaged by fishermen using them to moor their boats.
Indonesia - land of disasters:
- 1797: Earthquake offshore Padang, at least magnitude 8.6. Tsunami swept 150 ton sailing ship 1km onshore.
- 1815: Eruption of Tambora killed more than 88,000 directly and later through starvation. The eruption was the equivalent of 800,000 Hiroshima bombs.
- 1833: Magnitude 8.8 earthquake, offshore Sumatra, one of the ten largest documented before 2004. Destroyed buildings up to 300km away. Powerful tsunami swept the western coast of Sumatra. Casualties "thousands".
- 1856: Eruption of Awu claimed at least 3,000 victims.
- 1861: Magnitude 8.4 earthquake off northern coast of Sumatra, generating 7m tsunami that affected 500 km of coastline.
- 1883: Eruption of Krakatoa in the Sunda Strait. The collapsing volcanic island produced tsunami waves rising to 30m, killing 36,000.
- 1919: Eruption of Kelud, East Java. Mudflows (Lahars) kill 5,000.
- 1963: Agung eruption, Bali. Pyroclastic flows kill 1,500.
- 1976: Earthquake in Papua, landslips and buildings collapses kill 5,000-9,000.
- 1992: Earthquake near Flores, tsunami. Kills 2,000.
- 2004: Banda Aceh earthquake, third most powerful on record. 170,000 fatalities on Sumatra.
- 2006: Magnitude 6.3 earthquake offshore, Yogyakarta, Java. 5,700 dead.
A more alarming test of Indonesian tsunami preparedness took place in 2009, when a moderately strong earthquake shook the major city of Padang.
Some 900km south of Banda Aceh, this port with a population of 800,000 or more was the city that most concerned seismologists.
Struck by a great earthquake in 1797, with a suspected tsunami of around 5m, a Dutch writer later recorded that "the waves of the sea ran with fury up the river".
Over the following 200 years, the tectonic forces had built up again, and a repeat was feared, with a lapse of only 25 minutes between the shaking and the first waves.
The earthquake of 30 September 2009 was not the feared repeat, but movement on a small sub-fault.
The shaking brought down buildings, causing 1,100 deaths and 2,000 injuries.
With memories of the 2004 tsunami still strong, the inhabitants tried to flee the city as best they could.
"Within 20 minutes, streets were clogged with vehicles, honking loudly, but hardly moving, producing a never seen chaos - dust everywhere," according to one account.
Half a million people were trying to escape through the crowded streets. Hours later the streets were still jammed.
Fortunately, there was no tsunami, and the automatic warning system - a prototype of GITEWS - successfully said so. But no-one heard, as communication systems were badly disrupted.
If there had been a tsunami, the toll might have rivalled 2004.
The threat of the big one has not diminished, either, the experts fear.
It is not clear how quickly tsunami preparations can be developed in the city.
Retreating to upper floors of solid buildings is often the best advice, but many of the designated refuges were damaged in the 2009 earthquake.
The problem with an earthquake, compared to other natural disasters, is the speed with which it develops.
The seeds of a hurricane will be seen days ahead out at sea by forecasters, who can track its development and approach to land.
When an earthquake occurs, the rupture progresses so fast - in three minutes in the case of Japan 2011, 10 minutes in the case of the Sumatran quake of 2004 - there is little time to react.
Tsunami waves take little longer, but there is time to respond only if the warning is acted on immediately.
Immediately after the 2004 disaster, Roger Bilham, of the University of Colorado, an expert on the earthquake hazards in the region, told the BBC: "I'm afraid the scientific community has let the world down."
There is no doubt the science is better now.
But tsunamis remain the deadliest of natural hazards, and knowing how to act before one happens is the best preparation.