Assessment of risk and providing scientific advice at

Assessment of risk and providing scientific advice at the eruption of the Soufrière Hills Volcano. Montserrat Integrated Uncertainty Modelling for Decision Making Workshop, 2 April 2009 Steve Sparks

Location

A regional history of volcanic disasters in the Eastern Caribbean Mt Pelée, Martinique 1902 29, 000 people die when political priorities take precedence over public concerns

Then, Guadeloupe, 1976…… …. a volcanic crisis leads to a major evacuation, but the eruption is stillborn; scientists are embroiled in public controversy, severe criticism and recriminations

Soufrière Hills Volcano Eruption, July 1995 to 2009

Collapse and Fill

1997 Map of pyroclastic flow deposits

25 June 1997 20 people killed

Destruction of Plymouth

Galways Wall Crisis

21 st December 1997

Volcanic Blast: Destruction

Living with an erupting volcano: hazard zones for crisis micro-management “…this island is exactly the wrong size for an eruption…”

Map simplified in July 1997 and has remained much the same ever since

Prompted by the Guadeloupe 1976 experience…. …. in Montserrat, we put in place a formalised procedure for providing scientific advice to the authorities using a procedure developed originallyfor the European Space Agency Roger Cooke (Delft) Expert Elicitation Method

Risk Assessment Panel formed December 1997 • Assess the hazards and risks at the volcano every 6 months • Use of observations, models and expert elicitation • Output as a report, event trees, and risk charts expressed as probabilities • Risk reported in terms of Chief Medical Officers scale and analogies familiar to public Scientific Advisory Committee (SAC) formed July 2003 • Under UK guidelines for SAC (Sir Robert May CSA) Advice given to Governments of UK and Montserrat

MVO structured elicitation procedure for scientific advice Pooled estimates and opinions: • What is the chance a major explosion in next 6 months? • What is the chance that a village 6 km NW of the volcano will be inundated by a pyroclastic flow? Central estimate and bounds

An optimal decision on any question of interest can be obtained from the weighted sum of the opinions of a group of experts: The art of being a good expert is to get your net just the right size for catching the ‘scientific fish’……

“The method is considered superior to other schemes for pooling judgements, using asymptotically strictly proper scoring rules that reward good calibration and penalise low informativeness” Note: Roger Cooke to be distinguished BRISK visitor in September 2009

UK Individual Risk Scale HIGH >1 in 100 MODERATE 1 in 100 to 1 in 1000 LOW 1 in 1000 to 1 in 10000 VERY LOW 1 in 10000 to 1 in 100000 MINIMAL 1 in 100000 to 1 in a million

Quantitative risk assessment Pyroclastic flow hazards on Montserrat

Lower Belham valley, Montserrat Evacuated 8 October 2002

Quantitative risk assessment Pyroclastic flow hazards on Montserrat Dome grows in directed lobes and surges of growth cause switches in direction every few weeks or months Most collapses occur in direction of growth; flows go down adjacent valleys

Quantitative risk assessment Pyroclastic flow hazards on Montserrat Runout distance of flows proportional to (volume)1/3

Ingredients of model I • Probability of collapse to the northwest • Probability of pyroclastic flow reaching area (>3 million cubic metres) • Surge cloud behaviour • Number of people in area (casualties) • 6 month time period 20 m/s

Ingredients of model II Evaluation of probabilities and their uncertainties by: • Models • Empirical evidence • Expert elicitation methods • Monte Carlo sampling of uncertainties 15% chance of pfs getting to Belham Valley 70% chance of collapse to the east Willy Aspinall: facilitator

12 July 2003 July 12 2003 What happened?

Largest historic dome collapse (210 million cubic metres) on 12 th July 2003 13 th July 2003 Risk reduced!

Pros & cons of structured opinion elicitation for decision-making - for volcanic crisis management Advantages: Disadvantages: Formal structured and systematic approach Concept and principles of subjective probability are not familiar to many scientists Inclusive: whole team involved Un-biased: individual polling procedure encourages expression of true opinion Exhaustive: all sources of uncertainty are treated fully and explicitly Neutral: de-personalises provision of scientific advice Transparent: approach accords with new British government guidelines for scientific advice, and requirements to pool wide range of expertise Individual “calibrations” are more difficult to justify in context of volcanological hazard assessment than in some other disciplines Requires specialist “facilitator” to ensure correct implementation

Health effects

Modelling Ash Health Risk (DOH) • Stochastic Volcano model • Atmospheric ash deposition • Rainfall model • Erosion model • Human exposure model • Epidemiological model Monte Carlo ensemble models Thea Hincks Ph. D

% trials exceeding exposure Results: 20 year cumulative cristobalite exposure mg. m-3. year w s c f Montserrat assessed as “safe” by DOH cumulative cristobalite exposure mg. m-3. year

Thank you for your attention