Volcanic Unrest: Choosing an Alert Level Under Uncertainty
A 2,500 m stratovolcano last erupted 120 years ago and now shows renewed unrest. Background seismicity of about 15 shallow volcanic earthquakes per day rose to 320 per day over six weeks, with hypocentres migrating from 8 km to 3 km depth. A magnitude 4.2 event was felt in nearby towns. GNSS stations record 55 mm of uplift since January, centred 4 km west of the summit. SO₂ flux rose from 180 to 1,400 tonnes per day, and the summit crater lake warmed by 6°C. No eruption has occurred. About 4,200 residents live within 15 km, and 900 tourists per day use a single access road that crosses a lahar path. Raising the alert level to 3 would close the road and cost about $2.8 million per week in lost tourism and evacuation preparation. The volcano observatory must decide within 48 hours.
1. Interpret the monitoring dataset. What process best explains the changes, and which signals are most diagnostic?
Read the worked answer
The pattern is best explained by magma intrusion into the shallow edifice, with associated hydrothermal and gas changes.
- Seismicity: the rise from 15 to 320 events/day and the migration from 8 km to 3 km depth indicate brittle failure as material moves upward. Migration is more diagnostic than the raw count, because it shows a source moving toward the surface.
- Deformation: 55 mm of uplift over about six months, centred 4 km west of the summit, is too large for weather or groundwater alone and indicates volume addition at depth. The centre locates the intrusion.
- SO₂: the rise from 180 to 1,400 t/d is the strongest evidence of fresh magma, because SO₂ is released from magma and is not produced in quantity by shallow hydrothermal boiling (which is dominated by H₂S and steam).
- Crater lake warming: consistent with increased heat and gas flux.
The M4.2 event shows the system can produce damaging earthquakes even without eruption. The key judgement is that the signals are multi-parameter and co-located; a single anomaly would be weak, but gas + deformation + migrating seismicity together strongly indicate magma ascent.
2. Is an eruption certain? Explain the uncertainty and how it should affect the alert-level decision.
Read the worked answer
No - an eruption is not certain. Many intrusions stall and never erupt; magma can degas and solidify at depth. The data show that magma has moved and that the system is pressurized, not that an eruption will occur. There is also a real alternative: a hydrothermal system heated by the intrusion could generate phreatic explosions with little warning, which are dangerous precisely because they are small and unpredictable.
What the data do support is an increased probability of eruption or explosion, and an increased probability of hazards (earthquakes, lahars from destabilized material, gas) regardless of eruption. The decision should therefore be framed as risk management, not prediction.
Because the road crosses a lahar path and evacuation takes time, waiting for certainty removes the option of evacuating safely. The expected loss from a false alarm (about $2.8M per week, reversible) is much smaller than the expected loss from a missed event with 4,200 residents and 900 tourists on one road. Raising the alert level and restricting access is justified even at, say, a 20–30% eruption probability, provided the communication is honest about that probability.
3. Outline how the observatory should communicate the decision to residents and tourists, and identify one way the message could fail.
Read the worked answer
The message should state three things plainly: what is known (magma has moved to shallow depth; SO₂ and uplift are high), what is uncertain (an eruption may or may not occur; phreatic explosions are possible with little warning), and what to do (road closure, evacuation preparation, where to get updates). It should give a review time - for example, a reassessment in 72 hours - so the alert does not feel open-ended.
It must avoid both extremes: "no need to worry" invites complacency, while "eruption is imminent" destroys trust if nothing happens. Using probability language and showing the monitoring data publicly helps people see the basis for the decision.
How it could fail: if the alert is raised and nothing happens, residents may discount the next warning (the "cry wolf" effect), especially if authorities overstate certainty. Trust is also damaged if tourists receive a different message from residents, or if the single access road is closed without a clear alternative. A second failure mode is jargon: "alert level 3" means nothing without explanation of what people should actually do.