Plunged into darkness, we were listening to the words of Havivra Da Ifrile, a young girl from Martinique. On 8 May, 1902, Mount Pelée, the volcano at the centre of the island, erupted. 30,000 people were killed, mostly by the pyroclastic lava flows that engulfed Saint-Pierre, the capital. Havivra was one of very people to survive and give an account of the eruption: she survived by hiding in a cave. One other survivor, Ludger Sylbaris, had the good fortune to be in jail, unconscious and drunk.

One study estimates that over 220,000 people have been killed by volcanic activity since 1783, at least 25% of them by lava flows. Currently, apart from volcanic tsunamis, pyroclastic density flows are the biggest killers in volcanic eruptions.

We’ve known about these currents for a long time – Pliny the Younger described their destruction of Pompeii in AD79 – yet we know little about how they behave. If you’re close enough to observe them in great detail, then you’re not likely to survive: they travel at speeds of up to 300mph and reach temperatures of 800˚ Celsius. They are surrounded by clouds of dust and ash, so we can only observe them from afar and by looking at the deposits they leave behind.

Dr Rebecca Williams, Lecturer and Geology Subject Group Head at the School of Environmental Sciences at the University of Hull, is a field volcanologist: she does study active volcanoes, but she prefers – for obvious reasons – to walk on very old, very dead ones. In her research, she tries to recreate the movement of pyroclastic flows. “We need to understand how these flows behave,” she says, “so that we can estimate danger more effectively.”

Her work takes her to Oregon, to Ecuador, and in particular to the island of Pantelleria, probably one of the least famous Italian volcanoes. The entire island is covered by lava flows from at least eight separate eruptions. The currents here were super-hot, making the resulting volcanic glass extremely hard and well preserved. “They’re among the best exposure of this kind of deposit anywhere,” said Rebecca.

Another useful feature of this volcano is that its magma reservoir was chemically zoned. The magma had different compositions at different layers: as an eruption began, it tapped magma at the top of the chamber; later magma comes from lower down. Rebecca can use those different chemical signatures to trace the current over time, as it spread around the island.

Most existing models of an eruption assume that lava flows more or less equally in all directions from the cone. Rebecca discovered that, in fact, the current behaved in much more complicated ways. “In its early stages, it was really feeble: it reached a 15-metre hill and was easily deflected. But then, through time, the current grew in size and force; it was able to swoop over hills several hundred metres high.”

Rebecca’s work has important implications for the production of hazard maps, which authorities draw up to help prepare for future eruptions. But her work is only one part of a highly complicated set of conditions. “Every volcano has its own heartbeat,” she says. And volcanology itself is, in her words, subject to silo-working. Geologists work in the field, or move – like Rebecca – into analogue modelling, using rice or beads; mathematicians make theoretical models; computer scientists build simulations; historians and social scientists have much useful information to offer. “We rarely look at eruptions together,” she told us: in the future, Rebecca wants to see much more interdisciplinary work, bringing all these sources of knowledge together to address specific risks.

One thing she has learnt is that putting scientists in direct contact with affected communities is often counter-productive; all the results of this research need to be mediated through civil authorities to ensure that people have the best chance of survival.

Read an interview with Rebecca Williams here.

Alan Barker is a writer, coach, training consultant and academic proofreader. Find out more about his work here.