Lab report

Could the Kashmir earthquake have been predicted? And why are American scientists rebuilding the 1918 virus—one of the most virulent ever?
November 20, 2005
Predicting earthquakes

Could the Muzaffarabad earthquake in Kashmir, for which the death toll stands at nearly 40,000, have been foreseen? And if it could have been, would it have made much difference?

Earthquake fatalities depend not only on the quake's magnitude—a measure of how much energy it releases—but on geological and demographic factors. In densely populated Kobe in 1995, a magnitude-6.9 quake claimed over 6,000 lives, despite Japan's relatively advanced building technologies, because the region's soft rocks liquefied. In contrast, the 1989 Loma Prieta earthquake in California was magnitude 7.1 and the epicentre was just ten miles northeast of Santa Cruz, yet only 68 people died.

Earthquakes in less developed countries are usually no bigger than those in rich countries, but are often far more lethal because buildings are not made to withstand them. Without the resources for quake-proofing, the only option is prediction and evacuation. Some argue that this is possible. In 1975, Chinese seismologists claimed that they successfully predicted a 7.3 quake in Haicheng, northeast China. Although more than 1m people lived near the epicentre, they were evacuated and there were "very few" fatalities (about 1,300, which may indeed count as very few in China in 1975). Whether this was a genuine prediction or party propaganda is still disputed. The foresight seemed to fail for the 1976 Tangshan quake, which killed at least 240,000.

Charles Richter, who devised the earthquake magnitude scale, was among the sceptics. In 1977 he grumbled that "journalists and the general public rush to any suggestion of earthquake prediction like hogs toward a full trough." Earthquake prediction, he said, "provides a happy hunting ground for amateurs, cranks and outright publicity-seeking fakers."

Many scientists still feel that way. Despite claims that major earthquakes on a given fault recur periodically, the latest thinking tends to prefer the idea that these ruptures are similar to stock-market fluctuations (another favourite hunting ground for phantom patterns) and avalanches: they are inherently unpredictable, so that you can never tell when they will come or how big they'll be. The technical term for this is "self-organised criticality."

But that's not to say the earth is merely capricious, or that prediction is hopeless. It is possible to monitor the build-up of strain on a fault and so to estimate how close it is to slipping. But that requires a lot of information to be gathered over space and time, and only the San Andreas fault is studied with this degree of attention.

A new approach, reported in October, is to treat earthquake prediction like weather forecasting and climate prediction: to run computer simulations of the system and look for the statistically most likely outcomes. A team led by Californian geologists has created Virtual California, a detailed computer model of the San Andreas fault system, and used it to predict seismic activity in the San Francisco bay. They say that a magnitude-7 quake, like the one that devastated San Francisco nearly 100 years ago, will occur with 5 per cent probability before 2009, and with 55 per cent probability before 2054.

If that's the best we can do for this most studied of all fault lines, there seems scant hope that poor countries like Pakistan will receive clear advance warning before the earth moves.

Why rebuild a deadly virus?

Without wishing to feed the panic over bird flu, it remains a chilling fact that more people died in the flu pandemic of 1918-19 than in the entire first world war. It was the worst epidemic in world history, more devastating than the black death. This prompts the question: why have US scientists rebuilt the virus responsible for it from scratch?

That's what a team in Atlanta has just announced in Science. To make the virus from its DNA, they used the genetic sequence deduced by a group from the US Armed Forces Institute of Pathology. The viral DNA was reconstructed for analysis from extracts of the lung tissue of a human victim from 1918 whose body was frozen in Alaskan permafrost. The researchers say that the information gleaned from watching how the "1918 virus" gets transmitted among infected mice will help them to understand and combat the virulence of dangerous new strains like avian flu. (The 1918 virus is "bird-like": it is another strain that apparently evolved to jump from birds to mammals.)

But other biosafety experts have questioned the wisdom both of resurrecting the deadly virus and of making its genome data freely available. One American bacteriologist called it "perhaps the most effective bioweapons agent now known." Another said "this would be extremely dangerous should it escape, and there is a long history of things escaping." Such an accident is not the only way for the virus to get out there. The genetic sequence is in the GenBank database, accessible to anyone, and one can order DNA with just about any sequence from mail order companies. That's how a team of virologists made the polio virus in 2002. Some DNA-making companies screen their orders to check if they are dishing out pathogen genes; others do not, and there is no legal requirement for them to do so. It's an unresolved dilemma of an age when life can be synthesised: do the benefits justify the risks?