Bayesian estimation of 2-D P-velocity models from active seismic arrival time data: imaging of the shallow structure of Mt. Vesuvius (Southern Italy)

A multi-2-D non linear traveltime tomography of the shallow (3–4 km deep) structure of MtVesuvius volcano was performed. Data have been collected during two recent active seismicexperiments using a total of 17 on-land shots and about 140 three-component digital seismographs.A newly developed technique for imaging the volcano velocity structure has beenapplied, based on an adaptive model space investigation where the number of grid nodes isprogressively increased (multi-scale approach). The optimal model parametrization is chosenaccording to the minimum of the Akaike Information Criteria (AIC) parameter. This correspondsto finding the best compromise between the data misfit and simplicity of the model. Themodel parameter estimate is performed through the computation of an a posteriori probabilitydensity function (pdf), defined following the Bayesian approach. The maximum likelihoodmodel is searched by an optimization technique which combines the genetic and simplex algorithms.The evaluation of the a posteriori pdf is based on traveltime computations using raytracing techniques. Constraints on the model parameters are inserted in the form of prior pdfand error maps are inferred from cross-sections of the posterior probability around the foundbest fit solution. The retrieved images of Mt Vesuvius volcano show variable P-velocities inthe range 1700–5800ms−1. A fairly detailed image of the top of the Mesozoic carbonate rocksforming the basement of the volcanic area is obtained. A 9 km long, 1 km deep depressionwas detected at the N side of the volcano. The presence of a shallow high velocity body is evidencedunderneath the Mt Somma caldera and it can be interpreted as a sub- or palaeovolcanicstructure.