When Science Goes Wrong by Simon Levay Page B

attention had been paid to the volcano-tectonic earthquakes than to the long-period events as predictors of volcanic eruptions. It is certainly true that most volcanic eruptions are preceded by volcano-tectonic earthquakes. They may start months or years before an eruption, as magma begins to rise from deep magma chambers and collect nearer the surface. These earthquakes are commonly the long-term warning signs that tell volcanologists – and the local populations – that they need to pay attention to the volcano.
    Sometimes an intense swarm of volcano-tectonic earthquakes may immediately precede an actual eruption. This happened at the Rabaul caldera in Papua New Guinea in 1994, for example. The port city of Rabaul was evacuated within hours. Next day, the twin volcanoes that guard the harbour entrance both erupted, destroying much of the city. Thanks to the warning provided by the earthquake swarm, only five people died – one of them from a lightning strike.
    Yet, just as often, volcano-tectonic earthquakes don’t give timely warning of an eruption. An earthquake swarm may occur without a subsequent eruption, or an eruption may occur without a preceding swarm. The fallibility of volcano-tectonic earthquakes as short-term predictors of eruptions is the main reason why local populations are sometimes kept in a prolonged state of needless anxiety or, conversely, given false reassurance in the face of a looming catastrophe.
    During the 1980s, Bernard Chouet turned his attention to the long-period events as possible predictors of eruptions. Chouet had an insight about the nature of these events. He realised that their musicality – the pure tones with their added harmonics – must result from a resonance, that is to say, the reverberation of acoustic waves within a limited space. When Chouet analysed long-period events mathematically with this idea in mind, he was able to define the characteristics of the system that created them. Long-period events, he concluded, occur when a slug of magma jerks forward within a sheet-like crack in the rock. The motion of the magma is quickly arrested by the confining rock, triggering an acoustic wave. This is analogous to the ‘water-hammer’ that may be generated by household water pipes when a tap is turned off too quickly. ‘It’s as if you’re pinging it with a hammer blow,’ Chouet told me during a 2006 interview. ‘The acoustic pulse will travel through the fluid and hit a boundary and reflect, and keep going back and forth.’ Nevertheless, some of the acoustic energy escapes into the solid rock and radiates to the surface, where it can be detected by seismic instruments as a long-period event.
    According to this model, long-period events will occur when the path of the advancing magma is at least partially obstructed, rather than being open to the surface, just as water-hammer occurs in closed pipes. Obstructed magma is put under increasing pressure as more magma is forced up from the deep. Thus a series of long-period events may signal an increase in pressure that will end only when the pressure within the magma exceeds the weight of the overlying rock. At this point, an explosive eruption will occur.
    During the mid-1980s, Chouet looked through the seismographic records of past eruptions at a number of volcanoes. He found long-period events in the run-up to several eruptions, including the disastrous 1985 eruption of Nevado del Ruiz. Then, on January 2, 1991, he successfully predicted an eruption of the Redoubt volcano, which lies 100 miles southwest of Anchorage, Alaska, and 25 miles upriver from an oil-storage terminal. He made the prediction when there was a sudden increase in the frequency of long-period events – they began occurring every minute or so. In response to his prediction, the terminal was closed down and evacuated by 5pm that same day. Just two hours later the volcano erupted, and the resulting mudflow left parts of the terminal standing in three feet