A new waveform inversion workflow: Application to near-surface velocity estimation in Saudi Arabia

Synthetic data application

To evaluate the performance of the new workflow, a synthetic dataset was constructed where the true velocity model has two low velocity layers with a background velocity that smoothly increases with depth. The starting model used in the inversion is the background velocity without the two low velocity layers. The velocity is laterally invariant. The maximum offset of the survey is 8 km. The synthetic data are generated using constant density acoustic modeling engine. A typical shot gather (Figure 3) shows shingling as we move further away from source location. In the inversion, the source wavelet was known. We employed the workflow described above and ran the inversion using data with a peak frequency of 10 Hz. Inversion results are shown in Figure 4. The blue and red curves denote the true velocity model and the starting velocity model respectively, and the fuchsia curve is the inversion result. Notice that we were able to successfully recover both low velocity layers even when using relatively high frequencies. We also tried direct waveform inversion without wave-equation traveltime inversion using the same data. In this case, convergence to the right solution occurred only with peak frequencies of 7 Hz or lower.

synshot
Figure 3. Synthetic shot gather by acoustic modeling using true velocity model. [NR]

syninv
Figure 4. Inversion result with blue being true model, red being starting model and fuchsia being inverted model. [NR]

A new waveform inversion workflow: Application to near-surface velocity estimation in Saudi Arabia