Just a few months before the Indian Ocean tsunami of 26 December 2004, Phil Cummins forecast that such an event could result from a giant earthquake in central Sumatra. The Geoscience Australia seismologist's prediction, made in the organization's September newsletter, was accurate in all but one aspect: the magnitude-9.0 quake happened in northern Sumatra.

“I assumed it would take years to build a case and draw the attention and resources needed to study the area in detail,” says Cummins, who is leading Australia's effort to assess tsunami risks from the Indian Ocean. But the 2004 event boosted interest and since then, he says, “the human cost has become an overwhelming motivator for my work. It is a moral obligation.”

Unfortunately, he has little tangible evidence to build on. Geological surveys of the Indian Ocean are scarce, and those that are carried out are often done by petroleum geologists, so aren't focused on potential hazards. The limited monitoring of the region means that it is difficult to detect the surface-stress movements that precede a large quake. As a result, Cummins has spent several years combing historical accounts for indications of past tsunamis.

He tracked down as many accounts as he could find. Most had been made by British colonialists or merchants. One event stood out — an earthquake in 1762 at the northern tip of the Bay of Bengal. This area is difficult to study, because a 20-kilometre-thick layer of sediment streams down into it from the Himalayas. Most scientists presumed this fault to be inactive, because the direction of maximum stress is not aligned with the direction of deformation — or physical strain — as is typical. Most believe the region's greatest earthquake risk to be the fault running through central Myanmar.

But Cummins has combined historical and geological evidence to support the potential for a giant earthquake and tsunami hazard in the Bay of Bengal (see page 75). He cites observations of a British ship commander, made during a land survey in 1841, that detail evidence of a previous sudden uplift along the bay's southern coastal islands. Letters from Bangladesh's seaport city of Chittagong in the 1760s provide further evidence, describing a violent ground motion and fallen buildings. This suggests the area could be the northernmost portion of the fault.

These accounts convinced Cummins that a stretch of coast at least 700 kilometres long could have ruptured in 1762. “To produce a giant earthquake and tsunami you need to rupture a large, homogenous surface — which could be formed by the accumulated sediments,” he says.

Scientists have long used historical accounts to corroborate science, but have yet to exploit the information available for the Indian Ocean, says Cummins. There are probably other records, he adds, particularly in the non-English literature, that would prove valuable to review. He hopes that scientists will set their sights even further back in time. “The most important thing we can do now is systematically study all of the world's subduction zones for geological evidence of prehistoric earthquakes and tsunamis,” he says.

Cummins can surmise that a giant earthquake and tsunami is plausible in the northern Bay of Bengal, but he can't pin-point when such an event — which would endanger 1 million lives — might arise. He hopes his past success in predicting specific tsunami risks will help stimulate further study of the region.