GNGTS 2018 - 37° Convegno Nazionale

246 GNGTS 2018 S essione 1.3 Fig. 1 - a) RICEN Experiment layout, b) An example of seismic geophone, c) The RICEN working team. the crater floor; b) two 1D orthogonal seismic arrays deployed along NNE–SSW and WNW– ESE directions and crossing the 400 m crater surface. The goal of this work is to present a high resolution multi-parametric image of the shallow Solfatara crater analyzing P-wave velocity and attenuation tomographic models. In particular, we present 1) a 3D velocity model [4]; 2) two bi-dimensional velocity sections multi-2D interpreted [5]; 3) a 3D attenuation model. We compare the obtained 3D images with an electrical resistivity section [3] and temperature and CO 2 flux measurements. Data and Method. The RICEN experiment consisted in three successive geophysical surveys carried out at the Solfatara volcano respectively in September 2013, May and November 2014, each one lasting one week. The pilot phase consisted of the joint acquisition, from both sparsely distributed three components geophones inside the crater, and seismic stations placed on a regular grid of 115 × 90 m 2 area sampled by a regular grid of 240 vertical sensors (named 2D array) in front of the Fangaia (Fig. 1, a). In the second phase (RICEN2) besides the two joint acquisitions mentioned previously, hosted a 2D profile, with NNE–SSW direction, which was performed on May 21, 2014. The same configuration was used for the last phase (RICEN3) with the difference that the 2D profile, performed on November 11, was oriented orthogonally to the one performed during the RICEN first act (Fig. 1, a). Active seismic data were obtained using a Vibroseis Truck soil energizator, which operated in the frequency range 5–125 Hz. Seismic waveforms were recorded by 4.5 Hz vertical component geophones (Fig. 1, b). In this study, for the 3D models, we analyse the data collected during the first experiment by the 2D array. The P-wave first arrivals have been first detected through a Neural Network implemented into the ProMAX SeisSpace software trained on a limited, manually picked dataset of source- receiver couples. Thereafter, the picking dataset has been manually validated on the basis of a visual inspection of seismic signals. Beside the 3D array, for the 2D seismic tomography we considered the data from the two orthogonal seismic profiles. The first array was about 430 m long and oriented NNW-SSE while