Paper submitted to EAGE Annual 2026

Summary

Full-Waveform Inversion (FWI) is a fully data-driven seismic inversion algorithm designed to produce subsurface velocity models with high-accuracy and high-resolution by iteratively matching the synthetic with observed data. The primary challenge in FWI is cycle skipping: when the phase or timing difference between synthetic and observed seismic data exceeds half of a wavelength at a giving frequency, the inversion tends to converge to a local minimum. In general, four factors control FWI quality: the source wavelet, the accuracy of the initial velocity model, the richness of low-frequency content in the seismic data and the robustness of the FWI algorithm.

The source wavelet is dependent on the gun signature during acquisition. A proper wavelet can be derived from the input data, though it can be part of the inversion. When a properly derived wavelet, synthetic data under acquisition geometry can be fully simulated with the matched phase and spectrum to that of the observed data. This factor ensures that no human-induced cycle skipping is introduced in the preprocessing.  

The accuracy of the initial velocity model and low-frequency content of the input data are inextricably linked. Low frequency data can tolerate less accurate initial models while a more accurate initial model can allow FWI started at higher frequency. Deriving of an accurate initial model is not always straightforward especially in the area with complex geologies. For example, in the Central US Gulf of America, complex salt geometries often obscure underlying reservoirs. Most of the existing velocity models were built from legacy streamer data using a conventional top-down workflow: sediment tomography (sometimes combined with diving wave FWI), followed by top-of-salt and base-of-salt interpretation with sediment flood and salt flood migration, and finally, tomography for subsalt area using surface offset gathers. This process is not only time-consuming, but it can often result in significant errors regarding salt geometry and subsalt velocity due to limited illumination and imaging angle.