GNGTS 2019 - Atti del 38° Convegno Nazionale

242 GNGTS 2019 S essione 1.4 The geophysical inputs explained in Sect. 2 are used to define the prior distribution in Eq. 2, providing geometrical and density information. Note that all the voxels above the TUC surface have fixed label and density because topography and bathymetry were taken from the GEBCO08 1’ grid, while the sedimentary layers were taken from the CRUST 1.0 model. The a-priori geometrical information entered as the admissible depth ranges of the TMC, TLC and Moho surfaces, computed by using the available geophysical data and their uncertainties and used to set up the penalty functions s 2 i ( L i ) and q 2 ( L i , L j ) in Eq. 3. In the areas lacking in local seismic information, an additional input was given by the global crustal model at 1° × 1° spatial resolution reported in Huang et al. (2013). As for the a-priori probability distribution of the model density, the mean and standard deviation of Upper Crust (UC), Middle Crust (MC) and Lower Crust (LC) are defined reproducing the statistics of the CRUST 1.0 global values, integrated with local values inferred from DSS seismic velocity data, while the uppermost mantle layer is designed with a mean density according to the PREM model. To introduce a density spatial regularization in the final model, constraints on the maximum lateral and vertical variations between two adjacent voxels are introduced, under the condition of increasing or decreasing density with depth. Estimated crustal model. The final estimation of the model is retrieved by optimizing the target function of Eq. 3 for different sets of input parameters. A set of 10 3 combination of the deterministic parameters present in Eq. 3, i.e. the ones weighting the prior terms and controlling the strength of the geometry and density constraints, are used to compute different solutions. Then, to find the best one, a direct comparison of the target function values is not useful, because of the missing normalization. Therefore, four indexes are defined and computed per each estimated solution to evaluate the quality of gravity fitting, the level of density smoothing (considering both lateral and vertical variation) and the level of discontinuity surfaces smoothing. The best solution is chosen by a two-step procedure: Fig. 2 - Depth maps of the Top of the Upper Crust (TUC), Top of the Middle Crust (TMC), Top of the Lower Crust (TLC) and Moho Discontinuity (MD) for the 6°×4° area centred at the JUNO detector location. Negative values mean surfaces above the zero-level.