GNGTS 2023 - Atti del 41° Convegno Nazionale

Session 1.2 GNGTS 2023 VSM supports the following types of data: InSAR (may include multiple files from different sensors and/or orbits), GNSS, levelling, EDM, tilt and strain. Data unit can be displacements (measured in meters) or mean velocities (measured in m/yr). In case of multiple datasets from different techniques, a joint inversion can be performed. In this case, weights for each dataset should be assigned by the user. The weights among data from different techniques are somehow arbitrary and combines with the number of datapoints and with the relative signal-to-noise ratio of each dataset. Weights can be chosen after several tests, checking the trend of improvement/worsening of the misfit of each dataset and the total one. The forward analytical models available in VSM are listed in Fig. 1. For all these models, the medium is an isotropic, homogeneous, and elastic half-space. The simplest source is a point-source spherical cavity with constant pressure applied on its boundary (Mogi, 1958), with update to finite volume from the McTigue (1987) formulation, useful when the depth is comparable with the source radius. The penny-shaped crack (Fialko et al., 2001) is suitable to represent sill-like intrusions and laccoliths, or hydrocarbon reservoirs. The spheroid is a bottle-like cavity with a 3D orientation in the half-space (Yang et al., 1988), usually employed to represent vertically or along-dip elongated batches of magma uprising in the crust. The Okada (1985) model is the typical representation of seismic faults, and it may also represent dikes and/or sill-like tabular reservoirs. The moment tensor source is a point-source composed of dipoles and double couples, that implicitly defines a large variety of sources (Davis, 1986). All the sources above may be used as forward models in VSM individually or combined to simulate their cumulative static effect. VSM performs non-linear inversions and implements two different sampling algorithms that can be chosen by the user. The first one is a global optimization algorithm based on the Neighbourhood Algorithm (NA, Sambridge, 1999), the second follows a probabilistic approach to parameters estimation based on the Bayes theorem (here called Bayesian Inference, BI). Both inversion algorithms have in common the generation of: i) a log file containing the details about the run; ii) synthetic; iii) generated ensemble of models; iv) optimal parameters; v) 1D and vi) 2D parameters marginal distributions; plots of vii) 1D and 2D distributions and viii) the parameters’ sampling. VSM requires a simple text file to run, reporting all the information needed in the correct order. The input consists basically in three parts as shown in the first block of Fig. 2, namely geodetic data input details, forward model setting, and inversion setting. There are three ways to launch VSM, with or without the input text file: - From a script with an input text file; - From the GUI (Fig. 3); - From a Jupyter Notebook. All the details about the VSM tool can be found in Trasatti (2022).