The Campi Flegrei Blind Test: Evaluating the Imaging Capability of Local Earthquake Tomography in a Volcanic Area

Abstract

During the 1982–1984 bradyseismic crises in the Campi Flegrei area (Italy), the University of Wisconsin deployed a network of seismological stations to record local earthquakes. In order to analyse the potential of the recorded data in terms of tomographic imaging, a blind test was recently set up and carried out in the framework of a research project. A model representing a hypothetical 3D structure of the area containing the Campi Flegrei caldera was also set up, and a synthetic dataset of time arrivals was in turn computed. The synthetic dataset consists of several thousand P- and S-time arrivals, computed at about fourteen stations. The tomographic inversion was performed by four independent teams using different methods. The teams had no knowledge of either the input velocity model or the earthquake hypocenters used to create the synthetic dataset. The results obtained by the different groups were compared and analysed in light of the true model. This work provides a thorough analysis of the earthquake tomography potential of the dataset recording the seismic activity at Campi Flegrei in the 1982–1984 period. It shows that all the tested earthquake tomography methods provide reliable low-resolution images of the background velocity field of the Campi Flegrei area, but with some differences. However, none of them succeeds in detecting the hypothetical structure details (i.e. with a size smaller than about 1.5–2 km), such as a magmatic chamber 4 km deep and especially the smaller, isolated bodies, which represent possible magmatic chimneys and intrusions.

Anatomy of the Campi Flegrei caldera using Enhanced Seismic Tomography Models

“…In volcanic regions strong variations of Vp/Vs are expected both laterally and vertically, and the highest resolution is requested to reduce rough errors into the final models. To improve the initial 1D Vp/Vs model we first located the events using the initial 1D Vp model used in literature for the CF 9 , 23 , 24 and a constant Vp/Vs. Then, the hypocentres were separated in sets as function of the depth.…”

“…We calculated an average Vp/Vs value for each set of events using the Wadati diagrams 43 . We then obtained an optimized distribution of the Vp/Vs in depth using (a) the Wadati values obtained for each set of events, and (b) the information coming from the existing models e.g 9 , 18 , 23 , 24 . (see Fig.…”

“…However, the models proposed until now still do not find an agreement on the contribution of shallow magma bodies during the 1982–84 unrest 3 , 4 , 20 , 21 , 23 . Furthermore, existing seismic tomographies did not reach the sufficient resolution to corroborate the existence of small magmatic bodies 24 . Additionally, though the presence of a reinforced caprock was observed in the core samples 21 , its lateral extension is still poorly constrained.…”

Detailed lithospheric structure of an arc-continent collision beneath Taiwan revealed by joint inversion of seismological and gravity data

“…We use the linearized delayed traveltime tomography method described in Gautier et al (2006Gautier et al ( , 2008 and Priolo et al (2012) that originally inverts the velocity distribution and the hypocentre parameters simultaneously. Normalization or scaling of the derivative matrix is performed for better reconstruction of the different parameters.…”