231 7. Seismic inversion and characterization applied to geothermal energy All the inversion methods allow us to optimize the elastic properties. The stratigraphic inversion algorithm allows the reduction of random noise present in the initial seismic data, and therefore, does not include it into the inversion results. Seismic inversion parameters and QCs In practice, a reduced area is used to optimize the parameters, with the following criteria: • The number of iterations must be adjusted to obtain a maximum decrease in the objective function ending in a plateau (no more improvement by increasing iterations). • The residual seismic must be weak compared to the synthetic, proving that most of the information has been included in the elastic model. If possible, no (laterally) coherent signal must be present in the residuals. Especially for elastic inversion, additional controls are performed: • The residual seismic energy must be similar for all angle-stacks or honoring their relative quality. • The P-impedance and S-impedance, in seismic, must be equally updated and their frequency content comparable, or their difference explainable by a strong frequency difference between the angle-stacks, especially between the Near and the Far stacks. Extended to the entire seismic, a full-field inversion is performed and generally QC using the following criteria: • Control in sections (visuals) and control as maps (noise map, frequency map, energy map) to assess the enhancement or conformity of the synthetic data compared to the original seismic data. All information in the synthetic is contained in the elastic model. • Control of the convergence, both in terms of plateau and final values (%), expected to be compatible, approximately, with the signal to noise ratio observed in the original seismic. • Extraction at well locations (Figure 7.7), both participating or not (blind wells) in the calibration and modelling process, to assess the predictability of the inversion results. In practice, however, the inversion is often performed with all the wells in a final run, as the well data availability is often rare and valuable. The correspondence between the inverted properties and the properties computed at wells are often not a surprise for the inversion specialist, as it reflects the quality/ difficulties observed during the wavelet estimation. For the professional beginning the seismic characterization, this QC is a good starting point to assess how reliable the inversion results are before propagating valuable reservoir properties.
RkJQdWJsaXNoZXIy NjA3NzQ=