230 Geophysics in Geothermal Exploration for (3) while building a prior model (Figure 7.6). Indeed, the seismic information will be compared with the model during the inversion process, discarding too discrepant information, considered as noise. Figure 7.6 Conceptual amplitude spectrum of the prior model compared to the seismic spectrum. 7.2.4 Performing a seismic inversion Seismic inversion algorithm Different algorithms for seismic inversion are available and suit different objectives: • For the sparse-spike inversion, each trace is considered independently, introducing the entire trace. Computationally efficient, this technique is ineffective for reducing random noise. This inversion has no parameter but the number of iterations. • Model-based inversion is based on the objective function, or cost function, this inversion type has two parameters weighing the two terms: the seismic term, which controls the distance to the seismic, and the model term, which controls the distance to the prior model. For stratigraphic inversion algorithms, often preferred in projects for its noise reduction capability (Tonellot et al., 2001), the correlation length, added to the model term, controls the lateral continuity of impedance values along the correlation lines. The inversion is thus “multi-channel” when several traces are considered. It is “gridbased” when this comparison is consistent with a priori dip. In recent advances in seismic inversion projects, this dip is directly deduced from seismic independently from seismic interpretation.
RkJQdWJsaXNoZXIy NjA3NzQ=