Complex source geometries in volcano seismology

The seismic source for small to moderate earthquakes is usually described by a point source on a planar geometry, where the amplitudes
observed are linearly dependent on the seismic moment and the waveforms are predicted by the seismic theory. Particularly in volcanic settings, there is increasing evidence of non-planar ruptures which follow complex geometries instead, such as ring faults (conduits and calderas) and dyke faults. I propose and describe the action of complex sources as a superposition of point sources aligned along with ring structures and dykes. Synthetic seismograms are calculated and their magnitudes and waveforms analysed, finding that moment tensor inversions systematically underestimate the seismic moment or magnitude, the displacement at the fault is misinterpreted and the source dynamics follow mainly isotropic behaviour. For long wavelengths, I can treat the waves as coherent and a moment tensor inversion under a point source approach is applicable. However, this source parameters need to be carefully analysed and eventually corrected for a complex source. The correction factor for each different source studied can be calculated, thus, a corrected value for the seismic moment is available under these conditions. To test the results obtained, low-frequency events at Soufrière Hills are considered, in which the rupture is produced by brittle behaviour of magma within a conduit, the seismic moment correction is applied to the slip maintaining the area as constant, enhancing those values to match geological observations in rhyolitic volcanoes. Furthermore, partial-ring ruptures are modelled to emulate the collapse of Bárðarbunga caldera in Iceland. In this case, the correction over the seismic moment is attributable to the rupture area, maintaining the cumulative slip as constant. This applied correction improves the reconciliation of the seismic and geodetic moment for Bárðarbunga. For both cases, the inclusion of a curved source explains more accurately the observations and the conclusions are more realistic. Collaterally, I found evidence of network-dependent results, alongside intrinsic uncertainties in the location of these sources which needs to be taken into account for an improved source description.