228 Geophysics in Geothermal Exploration 7.2 Seismic inversion 7.2.1 About seismic conditioning Either to obtain full-stack or angle-stack data, for acoustic or elastic inversion respectively, gather reprocessing can be considered using up-to-date methodologies to enhance the final image, but must be “amplitude preserved”, meaning that no equalization or gain should have been applied to the data. As the inversion translates the amplitude into impedance changes, such processes can annihilate the vertical and lateral property changes normally observed through seismic inversion. This aspect is even more critical while considering several angle-stacks, as the amplitude preserved sequence ensures consistency between them. As the variation of the amplitude with the offset is meant to be translated into properties, any independent processing of each stack may ruin the desired estimation. In addition, while considering several stacks, seismic inversion is a computational process, and as such, all involved seismic data must be aligned. The NMO, often not perfect, must be completed by a mis-alignment correction (called trim statics on gathers, or Residual NMO in case of angle-stacks). The shift estimation in volumes through energy or correlation optimization is often preferred. The shifts are dynamic, and, therefore, are not constant vertically. The shift estimation can be post-processed, using smoothing or editions using an uncertainty analysis to prevent the generation of artifacts. 7.2.2 Wavelet extraction and optimization To link the modeled or optimized reflectivity with a seismic signal, the key operator, the wavelet, must be designed for each seismic dataset. A well-spread methodology in the industry is using a two-step procedure (Richard and Brac, 1988), with a statistical zero-phase wavelet extraction, followed by its optimization in phase and energy. For each seismic data involved in the process, the initial wavelet is extracted statistically by cross-correlation trace by trace, in a lateral zone and a given time window, if the signal correlates trace by trace while the noise, assumed to be random, does not correlate. To set up both the phase and energy for the wavelet, the wells are used and calibrated at the same time; the objective is to ensure the best match between synthetic traces, computed at well, and real traces, optimizing the wavelet parameters. To increase the robustness of such prediction, several traces can be considered, as displayed Figure 7.5. In this case, the best parameters can be displayed as histograms. After the process, the well-tie is often updated. In case of elastic inversion, the optimal location of the wells should then consider all the angle-stacks simultaneously.
RkJQdWJsaXNoZXIy NjA3NzQ=