GNGTS 2017 - 36° Convegno Nazionale

GNGTS 2017 S essione 3.1 555 Fig. 3 - a, b) Comparison between true, starting and predicted vertical velocity profiles corresponding to two horizontal coordinates, respectively. The black arrows in (a) indicate the reservoir layer. the reservoir interval (black arrows in Fig. 3a) we note that the inversion correctly estimates the significant decrease in Vp that marks the transition from the cap-rock to the reservoir layer, but fails to reproduce the corresponding increase in Vs . For the reservoir layer, the algorithm erroneously predicts a decrease in Vs and increase in Vp / Vs ratio instead of an increase in Vs and a corresponding strong decrease in Vp/Vs ratio. The good data reconstruction shown of Fig. 2 and the erroneous predictions of Fig. 3, indicate that the observed seismic data (marine data with single-component recording) does not contain enough information to constrain the Vs estimates, and for this reason, the final Vp / Vs model is not much modified from the initial to the reconstructed model. In other words, only the Vp information guide the inversion process. These outcomes were also confirmed by some additional FWI tests (not shown here) in which, for example, the initial P-wave models is kept fixed to the true model and only the Vp / Vs ratio is modified. Our preliminary results seem to indicate that EFWI of single-component marine seismic data, is not able to achieve reliable Vs and Vp / Vs estimations even if optimal starting models for both Vs and Vp are provided. Conclusions. In this work, we demonstrated the difficulty of shear waves velocity estimations for elastic full-waveform inversion (EFWI) of single-component seismic data, even if optimal starting models for the compressional and shear waves velocities are provided. To draw essential conclusions, we kept the inversion at a simple level. In particular, we limited our attention to the inversion of synthetic data computed on the Marmousi-2 model. We adopted a time-domain FWI algorithms that makes use of a multi-scale approach and a steepest-descent optimization procedure. Our study indicates that single-component (pressure) seismic data does not provide enough information to reliably constrain Vs and Vp / Vs estimations in a standard EFWI approach. Other EFWI approaches, such as target oriented or layer stripping methods, or the inclusion of multicomponent data in the observed data may help to improve the resolution and the reliability of the inversion results, which are certainly needed for reservoir characterization applications. Therefore, the challenge remains of how to use EFWI when only pressure measurements are acquired, as it commonly happens in marine data acquisition. The next step of our research will be to compare the results provided by EFWI to those yielded by a more conventional linear amplitude versus angle inversion of P-P wave reflection coefficients.