Volcano Observatory Best Practice Workshop Near Term Eruption Forecasting Near Term Eruption Forecasting Erice, Sicily (IT), September 2011 CALDERAS Problems and challenges for near-term eruption forecasting Paolo Papale INGV, Italy

Caldera-forming eruptions are the largest eruptions on Earth. For example, the Fish Canyon eruption in southwestern Colorado (United States) about 28 million years ago erupted more than 5,000 km 3 of magma from La Garita caldera. That's enough magma to bury the entire state of California to a depth of nearly 12 m! Caldera-forming eruptions are the largest eruptions on Earth. For example, the Fish Canyon eruption in southwestern Colorado (United States) about 28 million years ago erupted more than 5,000 km 3 of magma from La Garita caldera. That's enough magma to bury the entire state of California to a depth of nearly 12 m! At least 1,299 episodes of unrest have occurred at 138 calderas greater than 5 km in diameter during historical time. At least 1,299 episodes of unrest have occurred at 138 calderas greater than 5 km in diameter during historical time. In a typical year some form of unrest (earthquakes, ground deformation, change in fumarole activity, or eruptions) occurs at about 18 large calderas worldwide, and eruptions occur within or near at least five of them. In a typical year some form of unrest (earthquakes, ground deformation, change in fumarole activity, or eruptions) occurs at about 18 large calderas worldwide, and eruptions occur within or near at least five of them. source: CALDERAS: some general facts

CALDERAS: why are they different? The structure of calderas is profoundly different from that of stratovolcanoes The structure of calderas is profoundly different from that of stratovolcanoes “negative” as opposed to “positive” edifice “negative” as opposed to “positive” edifice boarder faults boarder faults chaotic rock assemblage chaotic rock assemblage development of large geothermal circulation development of large geothermal circulation resurgency resurgency compressional/extensional portions compressional/extensional portions several distinct post-collapse vents several distinct post-collapse vents … …

CAMPI FLEGREI, Southern Italy main caldera border internal caldera post-collapse eruptive vents resurgent block extensional setting compressional setting maximum uplift area extensive degassing

CALDERAS: why are they different? They often display unrest dynamics that if observed at central volcanoes, they would almost certainly culminate into an eruption Observations that are often reported as “critical” for near-term eruption forecast: acceleration in seismicity acceleration in seismicity acceleration in deformation acceleration in deformation increase of gas fluxes, especially CO 2 flux (and concentration) increase of gas fluxes, especially CO 2 flux (and concentration) (re: first two days of VOBP workshop) Are they equally diagnostic / critical at calderas?

CAMPI FLEGREI Vertical displacement during last 2 centuries In 1983 about 40,000 people were evacuated from the town of Pozzuoli, officially, due to the risk of structural collapses as a consequence of rapid ground displacement and seismic swarms > 8,000 earthquakes recorded  3.5 m of ground uplift

From: Chiodini et al., Campi Flegrei, Italy,

RABAUL eruption, 1994 “The eruption began on September 18 after less than a day of intense seismicity…” “The people who lived there were reminded of the inevitability of an eruption by intense earthquake activity and uplift of the ground within the caldera in the mid-1980's.” “However, despite warnings and a declared stage-2 emergency in 1983 and 1984, Rabaul did not erupt and, in fact, activity waned and remained at low levels until hours before the latest eruption broke out…” Source:

Vent location is definitely more uncertain than for central volcanoes CALDERAS: why are they different? map of the probability of venting for a next eruption at Campi Flegrei from Selva et al., km

RABAUL, 1994: several vents up to km apart were active “At times on September 19, there may have been as many as five active vents along the caldera rim, including several that began below the sea...” Source:

CALDERAS: why are they different? Many caldera depressions are partially or totally filled with water Phreatic explosions and phreatomagmatic eruptions can be frequent at calderas Geothermal circulation is usually well developed below caldera floor

CALDERAS: some “hot” questions what’s the origin of unrest at calderas, and why so often large unrest dynamics do not culminate into an eruption? what’s the origin of unrest at calderas, and why so often large unrest dynamics do not culminate into an eruption? During last 3 decades a number of interpretations have been proposed for the crisis at Campi Flegrei. Based on signal inversion and forward modeling, ground displacement has been alternately interpreted as mainly due to: increased heat/fluid flow in the geothermal system increased heat/fluid flow in the geothermal system emplacement of a shallow magma body emplacement of a shallow magma body

High velocity – High Density Inversion of P-wave velocity and gravity at Campi Flegrei 3D integrated v P model of Campi Flegrei From A. Zollo and co-workers Seismic attentuation tomography From De Siena et al., 2010

m  0.1 km 3

From Di Vito and Orsi, 2009 Magnitude of last 5 ka activity at Campi Flegrei 0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0, eruzione volume (DRE) eruzione volume (km 3 DRE) eruzione 0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0, eruzione volume (DRE) 0.1 km km 3 AD 1538 Monte Nuovo

top of carbonatic basement composition of the gas phase (wt% CO 2 ) pressure (MPa) depth (km) Top of carbonatic basement Seismic discontinuity Vesuvius Top of carbonatic basement pressure (MPa) depth (km) Seismic discontinuity Campi Flegrei composition of the gas phase (wt% CO 2 )

CO 2 ~ wt% in the gas phase From Rutherford, this project and previous GNV Campi Flegrei Project Plinian phase D1 of the 4100 BP Agnano Monte Spina eruption 1.5 – 3 km depth

Agnano Monte Spina eruption A few tens of hours before discharge shallow phonolite deep trachyte From Rutherford, INGV-DPC Projects /17 and /V3_2

AVERNO Eruption IC Eruption AMS Eruption Chemical and isotopic evidence of mixing-mingling preceeding many CF eruptions AMS Eruption From: Civetta, 2009

Schematic view of Campi Flegrei system in the past 5 ka. heterogeneous, mostly shoshonitic, CO 2 -rich, large (>100 km 3 ) magma reservoir large geothermal circulation

CALDERAS: some “hot” questions what’s the origin of unrest at calderas, and why so often large unrest dynamics do not culminate into an eruption? what’s the origin of unrest at calderas, and why so often large unrest dynamics do not culminate into an eruption? how to discriminate between unrest leading or not leading to an eruption? (e.g., dominantly due to the action of magma vs. dominantly due to the action of geothermal fluids) how to discriminate between unrest leading or not leading to an eruption? (e.g., dominantly due to the action of magma vs. dominantly due to the action of geothermal fluids) how to relate observations to expected vent location? how to relate observations to expected vent location? how to deal with often decades-long unrest dynamics? how to deal with often decades-long unrest dynamics? how long in advance will the signals allow robust forecast? how long in advance will the signals allow robust forecast? how to evaluate the hazard (and risk) related to often intense unrest dynamics? how to evaluate the hazard (and risk) related to often intense unrest dynamics?

Schematic view of Campi Flegrei system in the past 5 ka. heterogeneous, mostly shoshonitic, CO 2 -rich, large (>100 km 3 ) magma reservoir large geothermal circulation what controls the size of an eruption at calderas? or, do we need large magma bodies at shallow depth for a new caldera-forming eruption?

More general interpretation issue: Whether unrest at calderas (e.g., Long Valley, Yellowstone, Campi Flegrei, …) simply punctuates long periods of quiet or is the early warning sign of future eruptions is an important but still unanswered question

Short-term volcanic hazard forecast at calderas is generally characterized by uncertainties larger than for central volcanoes!

Boolean parameters are represented by “YES” “Gray areas” correspond to variable probability of being in the adjacent states, depending on the measured values ELICITATION V BACKGROUND Gray area UNREST Gray area MAGM. UNREST Gray area ERUPTION VT (M > 0.8) [ev/day] 5 15 LP/VLP/ULP [ev/month ] 2 10 Rate uplift [cm/month ] Uplift [cm] T Pisciarelli VLP/ULP 1 5 Deep VT (M > 0.8) [ev/day] 2 20 Deep LP (> 2 Km) [ev/day] 3 20 Disp. Hypocenters [km] 1 3 Tremor YES Deep Tremor (>3.5 Km) YES Acc. seismic events YES Acc. RSAM YES New fractures YES Macr. (dm) variation in def. YES Migr. max uplift YES Ext degassing YES Magm. comp. gases YES HF - HCl - SO2 YES Phreatic activity YES Red parameters: Seismicity Green parameters: Deformation Blue parameters: Geochemistry Campi Flegrei – Pre-eruptive Event Tree after Selva et al., 2011 DELPHI METHOD

unrest magmatic eruption Application to Campi Flegrei crisis after Selva et al., 2010 Probability estimates: beyond the color codes

Colour code Continuous probabilities (with uncertainties) Green:normal Yellow: watch Orange: attention Red: crisis Artificial discretization forces actions to be strictly tied to evaluation from scientists Scientists become de-facto decision-makers Correctly communicate the uncertain nature of predictions Allow a clear distinction of roles and responsibilities between scientists and decision-makers

GLOBAL VOLCANO MODEL(S) after Longo et al., 2010 vertical displacement (m) 0246 time (hours) after Voight et al., 2006 ULP ground oscillations