Every time seismic disaster strikes, as it did in Christchurch New Zealand on 22 February 2011 to  kill at least 160 people and destroying a third of the city’s buildings, people long for some means to be forewarned of pending earthquakes early enough to escape collapsing buildings. Many approaches have been suggested over the years, such as changing water levels in wells, increased emission of radon and even the behaviour of animals in advance of major events. Ideally, seismic early warning tools should be generally applicable, easily implemented and possible to telemeter immediately to local and national authorities. Probably the best place to seek such a method is in the field of seismology itself, and one candidate recently emerged (Bouchon, M. et al. 2011. Extended nucleation of the 1999 M_w 7.6 Izmit earthquake. Science, v. 331, p. 877-880). This examines foreshocks of the tragic events at around midnight 16 August 2009 in NW Turkey that ripped along 150 km of the North Anatolian Fault to kill around 17 000 people. The seismological records of the Izmit earthquake are not good quality, but Michel Bouchon and his French and Turkish colleagues, experts on the event, were able mine the ‘blurred’ data using new techniques. What they found was a sequence of 18 small earthquakes up to 45 minutes before the main one, each of which showed remarkably similar seismogram traces. From them they were able to show that most of the foreshocks arose from the same place on the fault and involved the same kind of deformation; by slippage in a patch or nucleus only about a few hundred metres wide at 15 km depth on the main fault. At each successive foreshock the rate of slip can be shown to have speeded up, and in the final 2 minutes before the main earthquake the localised acceleration was at its fastest. Also the low-frequency ‘rumble’ associated with each shock steadily got more powerful. These features define a similar shape for each seismogram record in the foreshock sequence.

The Izmit data tally well with a theoretical scenario for the initiation of movement along a fault. As tectonic stress builds up it begins to be dissipated by slow creep that can focus on a small part of the fault. Since this weakens that patch, subsequent creep is likely to favour the same place which becomes a nucleus for later events. If the stress loading is large enough to presage an eventual rip along a greater section of the fault such a major event will probably propagate sideways from the nucleus weakened by creep. Given sensitive seismometers suitably placed along threatening faults zones linked by telemetry to a central unit, as might seem sensible anyway, automated analysis of foreshock records with the signature of spatially restricted creep that begin to show an accelerating sequence might give the 5 to 10 minutes of warning that are the minimum to reduce fatalities in major earthquakes. However, analysis of better data from some other earthquakes does not reveal the same features, but it is early days and similar patterns may emerge from yet others: fault systems behave in a range of ways depending on their tectonic settings. The other issue is the cost of installations and facilities and their maintenance over long periods – how could somewhere like Haiti find the resources. And sadly, some earthquakes, like that beneath Christchurch occur on faults that show no sign at the surface.

Related Articles

• New Zealand earthquake caused by new fault line (independent.co.uk)