Abstract. Dike intrusions often cause complex ground displacements that are not sufficiently explained by simple analytical models. We develop a method to find complex and realistic dike geometries and overpressures from interferometric synthetic aperture radar (InSAR) data. This method is based on a combination of a boundary element method with realistic topography and a neighbourhood algorithm inversion. Dike model geometry is roughly a quadrangle with its top reaching the ground. The inversion has two stages: search and appraisal. The appraisal stage involves calculations of model marginal probability density functions using misfit values calculated during the search stage. The misfit function takes into account the variance and correlation of data noise. Synthetic tests show that a model is successfully retrieved within predicted narrow confidence intervals. We apply the method on InSAR data of the February 2000 flank eruption at Piton de la Fournaise and get a trapezoid dike dipping seaward (61.0 • -67.3 • ) with its bottom passing 800-1000 m beneath the summit. A model with a basal slip plane does not better explain observed asymmetric displacements and thus this asymmetry is solely attributed to the dipping dike. The dike lies above a narrow band of pre-eruption seismicity, suggesting that lateral magma propagation occurred. Neglecting topography results in poor modeling at depth and in overestimations of overpressure (or opening), height (both about 30%), and volume (about 20%).Citation: Fukushima, Y., V. Cayol, and P. Durand (2005), Finding realistic dike models from interferometric synthetic aperture radar data:

No abstract

[1] At basaltic volcanoes, magma is transported to the surface through dikes (magmafilled fractures), but the evolution of these dikes as eruptions proceed is rarely documented. In March 1998, after five and a half years of quiescence, Piton de la Fournaise volcano (Réunion Island) entered into a new eruptive phase characterized by intense eruptive activity. Coeruptive displacements recorded by interferometric synthetic aperture radar (InSAR) for the first five eruptions of the cycle are analyzed using 3-D boundary element models combined with a Monte Carlo inversion method. We show that the eruptions are associated with the emplacement of lateral dikes rooted at depths of less than about 1000 m, except for the first March 1998 event where an additional deeper source is required. The dikes are located above preeruptive seismic swarms. This is consistent with nearly isotropic stress caused by repeated dike intrusions and low confining pressure enhanced by the presence of pores in the shallowest 1000 m of the edifice. The volumes of the modeled dikes represent 17% of the volume of emitted lava, showing that exogenous growth plays a major role in building the volcano. By taking into account the preeruptive seismicity and tilt data together with the results of InSAR data modeling, we find that dikes first propagate vertically from a source region below sea level before being injected laterally at shallow depth. This behavior is consistent with the presence of levels of neutral buoyancy at shallow depth in the edifice.

No abstract