Geologists know the Himalayas are in an active quake zone, but there were no predictions to save lives in Kashmir. Credit: © (AP Photo/B.K.Bangash)

When disaster strikes, as it did in Kashmir last week, it is natural to ask what could have been done to prevent it.

If scientists and engineers issue warnings that politicians neglect, as in August's flood of New Orleans, the question becomes more about government priorities than about gaps in understanding.

But it seems that no scientist forecast the devastation around Muzaffarabad, where the death toll from a magnitude-7.6 earthquake on 8 October is currently about 30,000 and rising. Where, one might ask, was the science when it was needed?

Scientists knew, of course, that the Himalayas are a tectonically active region that earthquakes are bound to hit. And they knew that these earthquakes could be shallow, which increases their destructive effect on buildings. But could anyone have known exactly what was going to happen, and when?

Even to ask that question is thought to concede to an outmoded idea of science as a description of a clockwork universe where everything is predictable. Few geologists believe that major earthquakes can be pinned down in advance to a particular location at a particular time. Yet without that ability, it is hard to see how mass evacuations can ever be contemplated - and evacuations seem in principle the best way of saving lives in poor regions where expensive earthquake-proofing is hard to come by.

Certainly uncertain

Earthquake prediction provides a happy hunting ground for amateurs, cranks, and outright publicity-seeking fakers." Charles Richter, inventor of the Richter scale

Two recent reports have emphasized this gloomy view. A team of geoscientists led by John Rundle at the University of California, Davis, has described a computer model of geological movements in the San Francisco bay area1. It was almost a hundred years ago that a rupture of the San Andreas fault destroyed much of San Francisco and killed more than 3,000 people, and everyone in northern California wants to know when the next big one will be.

The answer, after Virtual California was simulated for 40,000 years to collect the statistics of major earthquakes, is that a magnitude-7 event or worse will happen with 5% probability by 2009, and with a 55% chance by 2054. This is worth knowing, but, for all the sophistication of the model, pretty much what everyone suspected anyway.

The San Andreas fault is the most extensively mapped and monitored earthquake zone in the world. Vast arrays of strain gauges measure pressure build-up on the complex web of fractures in the Earth's crust, while satellite observations keep careful track of the deformations of the ground. So the haziness of these latest forecasts for this area is a stark reminder of what 'earthquake prediction' means these days.

Tardy tremor

The difficulties are similarly underlined by a study in Nature this week of the recent magnitude-6 earthquake at Parkfield in California by William Bakun, of the US Geological Survey in Menlo Park, and his co-workers2.

Parkfield has suffered from such an earthquake roughly every 30 years since 1857, so in the past years scientists have descended upon it with a huge range of instruments, waiting for the next one to hit some time (they predicted) in the 1990s.

The eventual arrival of the Parkfield earthquake over a decade 'late' has rather put paid to the idea that the tremors really were periodic, forcing Bakun and colleagues to admit that "reliable short-term earthquake prediction still is not achievable".

Hit and myth

Bakun and his team begin their report by saying that earthquake prediction is the "Holy Grail of seismology". They presumably mean by this that it is what everyone would like to achieve. But it could also imply, perhaps more aptly, that earthquake prediction may be just a romantic myth.

The problem is only partly that the complicated process by which rocks slip past each other is poorly understood. On top of this, the slip process may be inherently unpredictable.

Yet the topic of prediction, says seismologist Robert Geller of Tokyo University, is "fatally attractive to both scientists and the general public". Geller is not alone in his scepticism. Charles Richter, who devised the most famous scale for measuring the energy released by earthquakes, said in 1977 that "Journalists and the general public rush to any suggestion of earthquake prediction like hogs toward a trough." He added that prediction "provides a happy hunting ground for amateurs, cranks, and outright publicity-seeking fakers".

A little knowledge

This is not to say that earthquakes are a total mystery. Geologists warned four years ago that stresses in the Himalayas seemed to be building at a rate that threatens far larger quakes than that at Muzaffarabad: quakes that could lead to casualties of 200,000 from a single event. And advances in understanding have led to better assessments of where earthquakes will strike next, how and where they might cause aftershocks, how big they could be, and how best to prepare for them.

So what can we do? Hope for the highly populated regions of earthquake zones in developing countries has to come not from prediction, whatever that might mean, but from protection: better engineering practices to render buildings less prone to collapse, and better plans for emergency response.

At present, events that could cause minor damage in California or Tokyo can still be catastrophic in India and China. But changing that costs money. So even statistical forecasts of earthquake hazard have a role. They won't foresee individual events, but they can help to identify where scant resources should be most effectively concentrated.

Perhaps the Himalayas deserve a little more of that.