Paper submitted to EAGE Annual 2026
Summary
Full-waveform inversion (FWI) in land environments remains challenging due to several factors, including low signal-to-noise ratios at low frequencies, uncertainties in source-wavelet estimation, and the influence of surface topography. For shallow targets, limited near-offset coverage and strong near-surface heterogeneity often obscure primary reflections, reducing the effectiveness of conventional imaging algorithms (Reta Tang et al., 2023). Incorporating additional wavefield components beyond primary reflections can therefore play a critical role in improving illumination.
Although refracted energy is generally robust for velocity estimation, its depth of penetration is constrained by acquisition geometry—particularly maximum offset—as well as by the vertical velocity gradient. Beyond the depth limit sampled by refracted arrivals, FWI must rely predominantly on reflected energy. Effective background-velocity recovery from reflections, however, requires scale separation within the FWI gradient to achieve stable and efficient convergence.
In this study, we jointly invert for velocity and reflectivity using a multiparameter FWI (MP-FWI) formulation. The acoustic MP-FWI framework was introduced by Yang et al. (2021) and later extended to elastic media by Huang et al. (2025). Here, we adopt the acoustic formulation and incorporate the dynamic matching (DM) objective function (Huang et al., 2021) to enhance robustness, particularly at low frequencies. We first describe the methodology and then demonstrate its performance using a land seismic dataset from the Midland Basin.