2 Main objective To study the seismic response of a volcano to different magmatic processes To study the seismic response of a volcano to different magmatic processes Work axes To identify the portion of seismicity directly related to volcano processes To detect specific patterns of volcano tectonic seismicity with respect to tectonic seismicity and for different phases of volcanic activity, in particular during dyke propagation Mechanical implications for basaltic intrusion dynamics: Laboratory experiments Numerical modeling Monitoring of the stress history created by the dyke intrusion responsible for the nucleated seismicity
3 1. Volcano vs Tectonic seismicity Uncorrelated Seismicity Correlated Seismicity Modified Omori’s law [Utsu et al., 1995; Helmstetter and Sornette, 2002] Background activity, modeled by a homogeneous Poisson process Generated by an external loading Tectonic case: driven by plate- tectonic loading Volcanic case: driven by a volcanic process (pressure change, mass transfer...) Sequence of events following a “mainshock” (“Aftershock Cascade”) Triggered by earthquake interactions Average patterns reproducible by ETAS model Model of earthquake occurrence
4 Volcano vs tectonic seismicity Observations: Tectonic case and quiescent volcanoes: EQ occurrence => = and During magma intrusions: Magma rising => and Most portion of volcano seismicity during intrusion is related to volcano process (little noise)
5 Volcano vs Tectonic seismicity UncorrelatedCorrelated Vesuvius –dormant (Feb – Aug. 2006) t0 ≡ mainshock occurrence, defined as any event (independently of its magnitude) not preceded by another one within a time equal to the median of the Δt. t > t0 seismicity following the mainshock (aftershock cascade) Tectonic seismicity / ETAS model: Power law decrease of the seismicity rate following a “mainshock”, Go back to the background rate. Quiescent volcano: Volcano seismicity = Typical tectonic seismicity pattern
6 Volcano vs Tectonic Seismicity During dyke intrusions: Piton de la Fournaise (7 intrusions) – Etna 2002 intrusion – Miyakejima 2000 intrusion Piton de la Fournaise Etna intrusion MiyakejimaEtna non-intrusion Peculiar dyke seismicity patterns: weak, if any, Omori’s law pattern following mainshocks.
7 Volcano vs Tectonic seismicity Why seismicity during dyke injection looses its capacity of producing aftershocks? HP: Forcing rate generated by the intruding dyke is too high to allow for stress redistribution and full aftershock development within the solid matrix Analysis of the seismicity related to the 2000 Miyakejima dyke injection (the largest seismic swarm ever recorded)
Miyakejima dyke intrusion Coulomb stress loading: The overall change in normal stress generated by the dyke opening explains 80% of the total swarm seismicity (the near-dyke one) Seismic swarm accompanying the intrusion: June 26 – August 29
Miyakejima dyke intrusion Toda et al.,  explains the overall seismicity by the total change in shear stressing rate generated by the intrusion and a magma chamber inflation. ( stressing rate ↔ relaxation time)
Miyakejima dyke intrusion Time monitoring of the stressing history generating the 2000 seismicity Deterministic approach: RATE- and STATE- dependent friction law [Dieterich, 1994] Non-linear relationship between the stress state and the seismicity rate. We follow the evolution of the earthquake nucleation sources subjected to some stressing history. γ is a state variable evolving with time and stressing history. It determines the distribution of slip speeds preceding instabilities Magmatic intrusion
Miyakejima dyke intrusion Time monitoring of the characteristics of the seismic swarm accompanying the magmatic intrusion Stochastic approach: Epidemic Type Aftershock Sequence modeling (ETAS) Seismicity driven by magma movement --> Background activity and earthquake-interaction driven activity are non- stationary. Non-stationary ETAS formulation in λ 0 (t) and K 0
Miyakejima dyke intrusion Aftershock productivity HIGHER stressing rate higher component of events generated by the dyke intrusion lower component generated by earthquake interaction. Background fraction Stressing rate
Miyakejima dyke intrusion Power law relaxation of seismicity following the last eruption --> eruption behaves as a mainshock in generating the seismicity. Result confirmed by the seismicity pattern following the Etna 2002 eruption.
14 Intrusion Data Piton de la Fournaise ( ), 7 seismic crises (durations: h), 6 of them ending up into an eruption. Temporal histories extracted from analog signals (time, Md, no location). Etna (2002), crisis preceding the eruption, duration: 6.3 h Miyakejima (2000), seismic swarm accompanying the June-July 2000 intrusion, duration: 281.2h
Seismic activity during intrusion No clear accelerating/decelerating patterns of seismicity Fluctuations dues to the undersampling of the Poisson process Constant occurrence rate is repruducible by a homogeneous Poisson process (constant mean) Random draws with the same dimension as seismic series N (m≥mc) N(m≥mc)
16 Energy rate during intrusion Fluctuating around a mean value, without any accelerating/decelerating pattern. Fluctuations compatibles with those of a Gutenberg-Richter law with constant b value. Start injection End injection
17 Seismicity during dyke intrusion Stationary seismicity rate Stationary energy rate Independently of: Intrusion duration (0.5 h to 11 days) Maximum magnitude (1.6 to 5.6) Dyke size (~1km to 15 km) Emitted lava volume DYKE INTRUSION: SCALE INDEPENDENT STATIONARY PROCESS SMALL SCALE HETEROGENEITIES, STEP-WISE PROPAGATION (lab simulations), AND TWO-PHASES INTRUSION NOT SOLVED Constant damage rate ~ Constant injection flux Seismicity during dyke intrusion is a response of the edifice on its all
18 Generic model for the intrusion Dyke propagation: STATIONARY At lab scales: stationarity reproducible by strain driven experiments variable loading: Tesile mode I on paper Strain driven peeling Strain driven tensile (mode I) Secondary creep Secondary creep on rocks Sec. Creep DYKE PROPAGATION ~ STRAIN-DRIVEN PROCESS (VARIABLE LOADING) → SECONDARY CREEP
19 Dynamics of magma injection Observed stationary seismicity prevent from distinguishing the recurrently observed two-phase intrusions → HP: constant magma flux at the dyke entry Verify that the constant flux model is compatible with the observations. vertical injection → effect of a lithological discontinuity (?) → change to a horizontal magma injection near the surface - Same flux injection as the vertical “conduit” - ¿ Can we recover the observed velocity scaling between vertical and horizontal dyke propagation? - ¿ How to model the change in direction? z Magma reservoir Dyke Volcanic edifice h
20 Numerical model of stationary volumetric magma flux into a dyke rising vertically from a superficial reservoir (test on Piton de la Fournaise, input data from previous studies): - Rising time - Volume of magma injected into the dyke - Average propagation velocity RESULTS COMPATIBLE WITH OBSERVATIONS Vertical magma rising from a superficial reservoir. Processes: - Buoyancy - Varying pressure at the dyke inlet - Lithostatic loading Horizontal migration. Processes - Effect of a density discontinuity - Topography - Flux at the entry Dynamics of magma injection z Magma reservoir Dyke Volcanic edifice h
Miyakejima dyke intrusion ¿Any correlation between the dyke propagation velocity and the observed seismicity rate? Dyke propagation period: June 26 to July 8 [Ueda et al., 2005] Migration of hypocenters Time since June 26
Miyakejima dyke intrusion Migration of earthquake center of mass during dyke propagation ~ propagation velocity of the dyke. 200 event (m ≥ m c ) non-overlapping windows.
Miyakejima dyke intrusion Correlation between the dyke propagation velocity and the observed seismicity ~ constant propagation velocity correlation between the observed seismicity rate. Cumulative seismicity Dyke propagation speed Inversion of sense
24 Conclusions On volcanoes, ↑ in seismicity are primarily due to background seismicity rate changes (strongest during dyke intrusions). Dyke intrusions: external forcing rate > tectonic loading rate. This prevents earthquakes interactions (Omori’s law) to fully develop. Magma intrusions can be modeled as “silent” or “slow” earthquakes in terms of stress history. Constant seismicity rate → Dyke intrusion is a stationary process → Constant volumetric flux at the dyke inlet Stationarity → Dyke intrusion ~ strain driven process with variable loading (i.e. secondary creep) → ¿ pressure decrease at the base of the dyke? → FINITE SIZE OF THE RESERVOIR