Pinatubo blew itself apart

In June 1991, a volcano in the Philippines destroyed its own summit, leaving a chasm 2.5 kilometres wide where the peak had been.

A very close call

More than 800 people died, mostly under roofs that collapsed from rain-soaked ash. Around 250,000 lived in the volcano's shadow.

An educated guess

A rapid geological assessment told scientists this volcano only erupts big. That became the forecast that triggered the evacuation.

“Oh, crap, when this thing erupts, it only erupts big.”

Nothing like a weather report

The team could not say an explosive eruption would happen on June 12, or how it would unfold once it started. That imprecision still holds at nearly every monitored volcano.

The field has moved on

Instruments got better, machine learning made the data tractable, and scientists now understand far more about the magmatic plumbing driving eruptions.

So could we forecast one?

Could anyone say a volcano has an 80 percent chance of erupting next week, with a given force, down a given flank? Some volcanologists think yes.

“The short answer, otherwise I wouldn't be doing this, is yes.”

Weather is always happening

Meteorologists watch an atmosphere that never switches off. Magma sits kilometres down, and most active volcanoes erupt once every few decades.

Every volcano is its own animal

Plumbing, magma chemistry, eruption cadence and style all differ by site. And no single trigger sets one off.

Order inside the chaos

Pressure, rock strength, gas and crystal content, magma depth, plate motion. Each nudges a volcano toward a paroxysm or a fizzle.

“Geology is chaotic.”

We can hear the orchestra

Seismometers catch rock cracking as magma climbs, satellites track the crust shifting, gas detectors sniff magma nearing the surface.

Unrest is not a promise

Alert systems flag escalating activity, not imminent eruption. Only half of the unrest that looks like a coming eruption actually becomes one.

“Only 50 percent of volcanic unrest that looks like it's going to be an eruption ends up in an eruption.”

Some volcanoes ambush you

A pocket of trapped water heated by nearby magma can rupture into a steam blast with no warning signs, then release the magma behind it.

The ones we can call

At Stromboli, Etna, Kilauea and Iceland's Reykjanes, repeat performances let scientists pin down where lava will surface, sometimes within the hour.

“We have systems that can tell us that in a few hours, the volcano will erupt.”

An hour is not enough

Those precise calls are unusual, and they come from frequently active volcanoes unlikely to explode. An hour's notice rarely gets a town to safety.

The missing physics

Nobody can yet say what flips a magma reservoir from stable to catastrophic failure. Find that, and it should apply everywhere.

“We don't even fully understand the underlying physics.”

We only know the aftermath

Navier-Stokes and the heat equation already predict how lava and pyroclastic flows travel and cool. That only helps once the eruption has begun.

“Once the rain starts to fall, we can forecast what watersheds might flood.”

Patterns without causes

Warnings today rest on correlations like rising seismicity. When the patterns shift, and they often do, the correlation stops being useful.

“If you understand what those patterns mean, when those patterns change, we're not that stuck.”

Expecting the unexpected

The Ex-X project is wiring Eastern Caribbean volcanoes with hundreds of seismometers and fibre-optic cable, with machine learning reading the faint signals.

Drilling into the magma

Lab experiments now recreate planet-birth conditions, and Iceland's Krafla Magma Testbed aims to drill down and watch a magma body directly.

“You can't just make a magma chamber at the surface of the Earth.”

A geologic Manhattan Project

Instrument many diverse volcanoes for decades, derive the shared laws, then build one generic model any volcano can be poured into. Only a handful have permanent networks today.

“It's time for a big push.”