GNGTS 2024 - Atti del 42° Convegno Nazionale

Session 3.3 GNGTS 2024 impedance I = ρ V ) and the transit time in a layer (via the ratio LV = L / V ). Since the Earth model must be consistent with both data sets, the velocity value must be the same for both inversion solutions. Another condition for the two inversion algorithms is the stability of the acoustic impedance I and the transit time LV against random noise, which we found in several tests with synthetic data. Therefore, each of these values is a well-constrained part of the two separate solutions, and we imposed that they are kept constant, while we perturb the values of velocity, density and thickness. Fig.1 – Earth model simulating a mud volcano. Although the model is 2D, the simulation and inversion are carried out in 1D only, so assuming just slow lateral variations. APPLICATION EXAMPLE To test the stability of this coupled inversion, we built an Earth model (Figure 1) that mimics a mud volcano. Its cone makes the water depth variable, while the water density and velocity are constant and known (1500 m/s and 1 gr/cc, respectively). Our target is the first layer below the seafloor, which consists of sediments with P velocity and density that vary laterally and reach a minimum in the center of the volcano. The basement is again homogeneous. We simulated by ray tracing only primaries and multiples from the seafloor and the sediment layer base, with an offset of 10 m between source and receiver. For the velocity inversion we need the two primaries of the latter ones, plus one or more multiples as peg-leg, intrabed or “simple” (Vesnaver and Baradello 2022a, b). The more, the better, because redundancy can reduce random noise due to picking errors and spurious events. For the amplitude inversion, we instead used only the primary and two reverberations between seafloor and sea surface to limit our solution space to the only two parameters we want to estimate, i.e., sediment velocity and density (Vesnaver and Baradello 2023). Including the amplitude of the other multiples is not so helpful: doing so, we would also have to calculate the velocity and density of the bedrock, leading to further unknowns and instabilities in our inversion.