Highly fractionated, silica-rich magma poses the greatest danger of explosive volcanic eruption, characterised by glowing pyroclastic flows that produce the strange rock ignimbrite. For example, in the Andes, ignimbrites extend for large distances from the calderas that emitted them. Fortunately rhyolite eruptions are rare, but that poses a scientific problem – they have not been as well studied as more common magmatic phenomena. Until May 2008 the latest rhyolite eruption had been in Alaska during 1912. In 2008 the Chilean volcano Chaitén erupted for the first time in 9 thousand years. There was no warning. Andesitic and dacitic volcanoes are restless for months before an eruption, though that is not much comfort as exactly when they ‘go off’ is still unpredictable. But any warning helps prepare local populations for the worst. A volcanoes precursory rumblings and shakings reflect the slow upward movement of magma. In the case of Chaitén, magma rose at about 1 m s^-1 that flabbergasted the volcanologists who rushed to study such a rare event (Castro, J.M. & Dingwell, D.B. 2009. Rapid ascent of rhyolitic magma at Chaitén volcano, Chile. Nature, v. 461, p. 780-783). The magma rose 5 km from its source in less than 4 hours. It is generally thought that the more silicic magma is, the more viscous and sluggish, which is certainly the case for rhyolite when it emerges: the melting of impurities in a coal fire produces a very silica-rich melt but such slag certainly does not dribble out of the fire box to pool on the hearth. High viscosity allows an erupting magma to retain gas escaping from solution as pressure drops, which is the source of the catastrophic blasts of massive ignimbrite events. Below the surface the Chaitén magma behaved in an extremely fluid manner, perhaps because it contained so much dissolved gas that it became a fluid froth at quite shallow depth. This unique observation is deeply disturbing for populations living in areas blanketed by ancient ignimbrites, as in the Andes. The very worst terrestrial events imaginable are ignimbrite eruptions that can blast out at such high velocities as to groove the ground and carry over thousands of km² in matter of minutes. Without warning, there is no escape.

Wenchuan earthquake (May 2008) analysed

On 12 May 2008 a magnitude 7.90 earthquake killed more than 80 thousand people and left many more injured and homeless in the Wenchuan area of Sichuan province China. In the worst affected areas up to 60% of the population were killed. The catastrophe occurred at the densely populated western boundary of the Sichuan basin with the Tibetan Plateau, and involved surface displacement that propagated rapidly north-eastwards along a 235 km long zone. There was virtually no warning sign and although crossed by major faults, high-magnitude seismicity was a rarity in the area. Several satellites now repeatedly deploy synthetic aperture radar sensing along their ground swath, so that interferometric methods (InSAR) are able to assess ground motions between separate times of overpass, with sub-centimetre precision. Together with direct measurement of motions at GPS ground stations, InSAR allows an unprecedented ‘post-mortem’ of this dreadful event (Shen, Z-K et al. 2009. Slip maxima at fault junctions and rupturing of barriers during the 2008 Wenchauan earthquake. Nature Geoscience, v. 2, p. 718-724). The structural architecture of the surrounding area is of five fault-bounded blocks that jostled during the event, resulting in profound shifts in the geometry of motion along two parallel faults that ruptured. The event was so sudden and large because what would otherwise have been barriers to propagation of strain failed at the same time. All the strain cascaded through several fault segments. This is not a scenario that could have been easily predicted, the authors judging it to have been a once-in-4000 years concatenation of crustal failure.

Seismic unpredictability is something that seismologists now recognise (Chui, G. 2009. Shaking up earthquake theory. Nature, v. 461, p. 870-872). Active faults turn out not to be ‘creatures of habit’, and nor can we assume that long-quiet segments are the most likely to fail in future. Ominously, there is a growing body of evidence that great earthquakes are able somehow to trigger others, often far distant. An example is the giant Sumatra-Andaman event of 26 December 2004, tsunamis from which caused a toll of hundreds of thousand lives around the Indian Ocean. It was followed quickly by swarms of small tremors on the San Andreas Fault 8000 km away. Rapid successions of great earthquakes around the world, such as the October 2005 Pakistan earthquake 9 months after that in the Indonesian area, can no longer be regarded as ‘bad luck’. Seismic waves are able to weaken far-off segments of active faults.