University of Bologna

The 2012 Emilia Romagna (Italy) seismic sequence has been extensively studied given the occurrence of two mainshocks, both temporally and spatially close to each other. The recent literature accounts for several fault models, obtained with different inversion methods and different datasets. Several authors investigated the possibility that the second event was triggered by the first mainshock with elusive results. In this work, we consider all the available InSAR and GPS datasets and two planar fault geometries, which are based on both seismological and geological constraints. We account for a layered, elastic half-space hosting the dislocation and compare the slip distribution resulting from the inversion and the related changes in Coulomb Failure Function (CFF) obtained with both a homogeneous and layered half-space. Finally, we focus on the interaction between the two main events, discriminating the contributions of coseismic and early postseismic slip of the mainshock on the generation of the second event and discuss the spatio-temporal distribution of the seismic sequence. When accounting for both InSAR and GPS geodetic data we are able to reproduce a detailed coseismic slip distribution for the two mainshocks that is in accordance with the overall aftershock seismicity distribution. Furthermore, we see that an elastic medium with depth dependent rigidity better accounts for the lack of the shallow seismicity, amplifying, with respect to the homogeneous case, the mechanical interaction of the two mainshocks

The deformation and stress fields accompanying dyke injection in rift zones are described in terms of a vertical crack opening in response to internal overpressure, in plane-strain configuration. Previous numerical computations of the displacement field induced by shallow dykes are found to be affected, in some cases, by significant dis- tortion, mainly due to the limited extension of the computational domain, difficultiesin handling the singularities in the proximity of the free surface and an incorrect choice of the reference frame; in particular, it is found that no subsidence can be ascribed to the mere opening of a tensile crack in a homogeneous, elastic half-space. If analytical constant-dislocation models are employed, surface displacements compare reasonably well with crack-model solutions if the upper tip is not too shallow; however, constant dislocation solutions present unphysical singularities along the tips, which distort significantly the near-field stress pattern with respect to crack solutions. If the principal stress axes are computed from crack models, a broad region is found on both sides of the dyke where the induced stress has a nearly vertical intermediate axis while the com- pressive axis is normal to the dyke plane. Strike-slip earthquakes are expected to prevail in this region. Above the upper crack tip, a small region is present where the tensile stress is dominant and the intermediate stress is still vertical. In the proximity of the free surface, typically within a few tens of metres of it, induced tensile stresses are greater than the lithostatic pressure: open fissures might then develop in cohesionless soil or pre-faulted rock. The induced pressure in the host rock is found to be negative (suction) in the proximity of the ground surface and positive at greater depth: fluid flow within the aquifers can be significantly altered by this induced overpressure and by the anisotropic modification of the pre-existing permeability. According to the modified Coulomb failure criterion, in the short term the shallower region, characterized by suction, is strengthened, while the deeper, pressurized region is weakened. These results can explain in a straightforward way the abundance of strike-slip focal mechanisms in vol- canic areas, the switch between tensional and compressional axes inferred from focal mechanisms of earthquakes in connection with a dyke injection episode on Mount Etna, en echelon fracture systems observed at Piton de la Fournaise above the feeding dyke and precursory geochemical anomalies.

Campi Flegrei caldera (Italy) was affected by a new unrest phase during 2011-13. We exploit two COSMO-SkyMed datasets to map the deformation field, obtaining displacement rates reaching 9 cm/yr in 2012 in the caldera center. The resulting dataset is fitted in a geophysical inversion framework using finite element forward models to account for the 3D heterogeneous medium. The best-fit model is a North dipping mixed-mode dislocation source lying at ~5 km depth. The driving mechanism is ascribable to magma input into the source of the large 1982-84 unrest (since similar source characteristics were inferred) that generates initial inflation followed by additional shear slip accompanying the extension of crack tips. The history and the current state of the system indicate that Campi Flegrei is able to erupt again, and the advanced techniques adopted provide useful information for short-term forecasting.

We develop a mathematical model describing dyke propagation in proximity of an elastic discontinuity of the embedding medium. The dyke is modelled as a fluid-filled crack in plane strain configuration employing the boundary element method. The pressure gradient along the crack is assumed proportional to the difference between the densities of the host rock and the fluid. Mass conservation is imposed during propagation and fluid compressibility is taken into account. The path followed by the crack is found by maximizing the total energy release, given by the sum of the elastic and gravitational contributions. The mathematical simulations provide a sort of ‘refraction phenomenon’, that is a sudden change in the direction of propagation when the crack crosses the boundary separating different rigidities: if the dyke enters a softer medium, its path deviates towards the vertical, if the dyke enters a harder medium its path deviates away from the vertical and may even become arrested as a horizontal sill along the interface, if the rigidity contrast is large. Gravitational energy plays a major role during propagation; in particular, in proximity of layer boundaries, this role is enhanced by the shift of the centre of mass due to changes of dyke shape. Mathematical results were validated by laboratory experiments performed injecting tilted air-filled cracks through gelatin layers with different rigidities.

Since 1993, geodetic data obtained by different techniques (GPS, EDM, SAR, levelling) have detected a consistent inflation of the Mt. Etna volcano. The inflation, culminating with the 1998– 2001 strong explosive activity from summit craters and recent 2001 and 2002 flank eruptions, is interpreted in terms of magma ascent and refilling of the volcanic plumbing system and reservoirs. We have modelled the 1993–1997 EDM and GPS data by 3-D pressurized sources to infer the position and dimension of the magma reservoir. We have performed analytical inversions of the observed deformation using both spheroidal and ellipsoidal sources embedded in a homogeneous elastic half-space and by applying different inversion methods. Solutions for these types of sources show evidence of a vertically elongated magma reservoir located 6 km beneath the summit craters. The maximum elevation of topography is comparable to such depth and strong heterogeneities are inferred from seismic tomography; in order to assess their importance, further 3-D numerical models, employing source parameters extracted from analytical models, have been developed using the finite-element technique. The deformation predicted by all the models considered shows a general agreement with the 1993–1997 data, suggesting the primary role of a pressure source, while the complexities of the medium play a minor role under elastic conditions. However, major discrepancies between data and models are located in the SE sector, suggesting that sliding along potential detachment surfaces may contribute to amplify deformation during the inflation. For the first time realistic features of Mt. Etna are studied by a 3-D numerical model characterized by the topography and lateral variations of elastic structure, providing a framework for a deeper insight into the relationships between internal sources and tectonic structures.

Deformation of the ground surface in volcanic areas is generally recognized as a reliable indicator of unrest, possibly resulting from the intrusion of fresh magma within the shallow rock layers. The intrusion process is usually represented by a deformation source such as an ellipsoidal pressurized cavity, embedded within a homogeneous and elastic half-space. Similar source models allow inferring the depth, the location and the (incremental) volume of the intrusion, which are very important parameters for volcanic risk implications. However, assuming a homogeneous and elastic rheology and, assigning a priori the shape and the mechanism of the source (within a very restricted “library” of available solutions) may bias considerably the inference of source parameters. In complete generality, any point source deformation, including overpressure sources, may be described in terms of a suitable moment tensor, while the assumption of an overpressure source strongly restricts the variety of allowable moment tensors. In particular, by assuming a pressurized cavity, we rule out the possibility that either shear failure may precede magma emplacement (seismically induced intrusion) or may accompany it (mixed tensile and shear mode fracture). Another possibility is that a pre-existent weakness plane may be chosen by the ascending magma (fracture toughness heterogeneity). We perform joint inversion of levelling and EDM data (part of latter are unpublished), collected during the 1982-84 unrest at Campi Flegrei caldera: a 43% misfit reduction is obtained for a general moment source if the elastic heterogeneities computed from seismic tomography are accouted for. The inferred source is at 5.2 km depth but cannot be interpreted as a simple pressurized cavity. Moreover, if mass conservation is accounted for, magma emplaced within a shallow source must come from a (generally deeper) reservoir, which is usually assumed to be deep enough to be simply neglected. At Campi Flegrei, seismic tomography indicates that the “deep” magma source is rather shallow (at 7-8 km depth), so that its presence should be included in any thorough attempt to source modeling. Taking into account a deflating source at 7.5 km depth (represented either as a horizontal sill or as an isotropic cavity) and an inflating moment source, the fit of both levelling and EDM data improves further (misfit reduction 80%), but still the best fitting moment source (at 5.5 km depth) falls outside the range of pressurized ellipsoidal cavities. The shallow moment source may be decomposed in a tensile and a shear dislocation. No clue is obtained that the shear and the tensile mechanisms should be located in different positions. Our favourite interpretation is in terms of a crack opening in mixed tensile and shear mode, as would be provided by fluid magma unwelding pre-stressed solid rock. Although this decomposition of the source is not unique, the proposed solution is physically motivated by the minimum overpressure requirement. An important implication of this new interpretation is that the magma emplaced in the shallow moment source during the 1982-84 unrest was not added to already resident magma at the same position.

Dikes and sills are the moving building blocks of the plumbing system of volcanoes and play a fundamental role in the accretionary processes of the crust. They nucleate, propagate, halt, resume propagation, and sometimes change trajectory with drastic implications for the outcome of eruptions (Sigmundsson et al., 2010). Their dynamics is still poorly understood, in particular when different external influencing factors are interacting. Here we apply a boundary element model to study dike and sill formation, propagation and arrest in different scenarios. We model dikes as finite batches of compressible fluid magma, propagating quasi-statically in an elastic medium, and calculate their trajectories by maximising the energy release of the magma-rock system. We consider dike propagation in presence of density layering, of density plus rigidity layering, of a weakly welded interface between layers, under the action of an external stress field (of tectonic or topographic origin). Our simulations predict sill formation in several situations: i) when a horizontal weak interface is met by a propagating dike; ii) when a sufficiently high compressive tectonic environment is experienced by the ascending dike and iii) in case a dike, starting below a volcanic edifice, propagates away from the topographic load with a low dip angle. We find that dikes halt and stack when they become negatively buoyant and when they propagate with low overpressure at their upper tip toward a topographic load. Neutral buoyancy by itself cannot induce dikes to turn into sills, as previously suggested.

Monitoring of quiescent volcanoes, such as Campi Flegrei (Italy), involves the measurement of geochemical and geophysical parameters that are expected to change as eruptive conditions approach. Some of these changes are associated with the hydrothermal activity that is driven by the release of heat and magmatic fluids. This work focuses on the properties of the porous medium and on their effects on the signals generated by the circulating fluids. The TOUGH2 porous media flow model is applied to simulate a shallow hydrothermal system fed by a source of magmatic fluids. The simulated activity of the source, with periods of increased fluid discharge, generates changes in gas composition, gravity, and ground deformation. The same boundary conditions and source activity were applied to simulate the evolution of homogeneous and heterogeneous systems, characterized by different rock properties. Phase distribution, fluid composition, and the related signals depend on the nature and properties of the rock sequence through which the fluids propagate. Results show that the distribution of porosity and permeability affects all the observable parameters, controlling the timing and the amplitude of their changes through space and time. Preferential pathways for fluid ascent favor a faster evolution, with larger changes near permeable channels. Slower changes over wider areas characterize less permeable systems. These results imply that monitoring signals do not simply reflect the evolution of the magmatic system: intervening rocks leave a marked signature that should be taken into account when monitoring data are used to infer system conditions at depth.

Pressurized cavities are commonly used to compute ground deformation in volcanic areas: the set of available solutions is limited and in some cases the moment tensors inferred from inversion of geodetic data cannot be associated with any of the available models. Two different source models (pure tensile source, TS and mixed tensile/shear source, MS) are studied using a boundary element approach for rectangular dislocations buried in a homogeneous elastic medium employing a new C/C++ code which provides a new implementation of the dc3d Okada fortran code. Pressurized triaxial cavities are obtained assigning the overpressure in the middle of each boundary element distributed over the cavity surface. The MS model shows a moment domain very similar to triaxial ellipsoidal cavities. The TS and MS models are also compared in terms of the total volume increment limiting the analysis to cubic sources: the observed discrepancy (~10%) is interpreted in terms of the different deformation of the source interior which provides significantly different internal contributions (~30%). Comparing the MS model with a Mogi source with the some volume, the overpressure of the latter must be ~37% greater than the former, in order to obtain the same surface deformation; however the outward expansion and the inner contraction separately differ by ~±10% and the total volume increments differ only by ~2%. Thus, the density estimations for the intrusion extracted from the MS model and the Mogi model are nearly identical.

The ground deformation produced by a spherical overpressure source in a heterogeneous elastic and/or viscoelastic medium is investigated by numerical models based on the finite element method. Sources are assumed to be located at different depths beneath Mount Etna, Sicily, Italy, the structure of which is approximated as axially symmetric. Finite element modelling allows to incorporate in the analysis realistic features such as topographic relief and the laterally heterogeneous multi-layered structure inferred from seismic tomography. In order to avoid introducing artifacts in the solution, great care was taken to calibrate the computational domain necessary to reproduce analytical results accurately. An elastic analysis, performed initially, shows significant changes of the deformation field with respect to homogeneous half-space solutions: topography induces slight but detectable changes in the deformation field; in particular the maximum value of the vertical component is shifted away from the symmetry axis. When introducing the elastic heterogeneities, the ground deformation is found to be more confined to the proximity of the axis and its amplitude is mostly sensitive to the presence of low rigidity layers above the source. The ratio of maximum radial to vertical deformation is significantly larger for deeper sources. A further development of the model includes the study of inelastic properties assuming a Maxwell viscoelastic rheology for different layers. If the viscoelastic rheology is applied only to layers deeper than the source, the solutions are affected in different ways according to the distance of the source from the viscoelastic layer. If a viscoelastic layer is present above the source, a very large amplification (by more than 100%) of the surface deformation is predicted by the model; moreover, uplift transients are found to be followed by subsidence, without invoking any decrease in source overpressure. The most striking effects are observed when the source is embedded within a viscoelastic layer: in this case a static equilibrium configuration is not attained and, in the long term, both components of deformation reverse their signs in proximity to the axis. Furthermore, the surface deformation becomes nearly independent of source depth, in the long term. Simple physical explanations are proposed for the different cases.