Can we predict when a volcano will erupt?

Sometimes. Volcano forecasting is an interdisciplinary monitoring and research effort to predict the time and severity of an eruption, and risk and uncertainty are central to it. The warnings come from the volcano itself. Seismic activity always occurs as volcanoes awaken, and in particular long-period earthquakes, believed to indicate increased gas pressure in the plumbing, are a signal when they become dominant. Sulfur dioxide gas is another sign: more of it heralds more magma near the surface. And a volcano can swell: before St. Helens erupted in 1980, its north side was bulging upwards.

The success story is Pinatubo in 1991. Magma rising from more than 32 km down triggered small earthquakes and steam explosions in March and April, and on 28 May sulfur dioxide emissions had risen to 5,000 tonnes, ten times the earlier amount. The volcano first erupted on 12 June. The early predictions led to the evacuation of tens of thousands of people, saving many lives. In December 2000 scientists in Mexico City predicted an eruption of Popocatépetl within two days; the government evacuated tens of thousands of people, the volcano erupted 48 hours later in its largest eruption for a thousand years, and no one was hurt.

The failure is Nevado del Ruiz in 1985. A small eruption produced an enormous lahar that destroyed the town of Armero, causing an estimated 25,000 deaths. Scientists did not know precisely when the eruption would occur, and the authorities would not take costly preventive measures without clear warnings. Hazard maps had been distributed more than a month before, but the Colombian Congress criticised the agencies for scaremongering. Afterwards, scientists noticed long-period earthquakes in the final hours. Bernard Chouet said the volcano was "screaming", but nobody could read the signal.