Skip to content

Forecasting the position and timing of eruptive vents using a particle filter method.

Léa Zuccali 1,2, Virginie Pinel1, Yajing Yan2

  • Affiliations: 1Univ, Grenoble Alpes, Univ. Savoie Mont Blanc, CNRS, IRD, Univ. Gustave Eiffel, ISTerre, Grenoble, France 2LISTIC, Université Savoie Mont Blanc, 74940 Annecy, France

  • Presentation type: Poster

  • Presentation time: Monday 16:30 - 18:30, Room Poster Hall

  • Poster Board Number: 227

  • Programme No: 2.4.17

  • Theme 2 > Session 4

Abstract

To assess volcanic risks, we need to determine as early as possible if, where and when the magma propagating from the storage zone will reach the surface. Data assimilation is an efficient way of combining a dynamic magma propagation model with recorded geophysical observations. Here we develop a particle filter method in the two-dimensional case of magma propagation beneath a caldera in an extensional stress field. The dynamical model provides the position, size and geometry of the magma intrusion from a set vector of 7 parameters : the initial magma position at depth, its viscosity and driving pressure, the volume of magma injected, the crustal rigidity and the local stress field. Surface displacements induced are estimated using the Okada dislocation model. The Particle Filter Method, based on the Monte Carlo principle, aims to provide an increasingly accurate estimate of the model state throughout the assimilation cycles (forecast--analysis). At the forecast step, the dynamical model is run forward for a large number of model state vectors (namely particles). At the analysis step the likelihood of the particles is calculated by comparing the theoretical displacements to the observed displacements. Then, particles with higher likelihood are retained, while the others are resampled by the Stochastic Universal Sampling method. The assimilation strategy has first been tested with synthetic data. 100 particles are generally enough to accurately forecast the location and timing of arrival. The time between assimilation steps must remain less than 1% of the total magma propagation time to ensure correct prediction.