Frontiers in Earth Science (Jan 2025)
Wavefield-reconstruction-based full waveform inversion on noisy data in seismic exploration
Abstract
Full waveform inversion (FWI) is commonly used in seismic exploration to calculate parameters of the medium, such as velocity, from the signal as it passes through the medium. To obtain an accurate result, FWI usually needs to have an initial model that is not too far from the true velocity model. However, using noisy low-frequency data to build the initial model can be a challenge for FWI in practice. To solve this problem, we propose a wavefield reconstruction method based on the first type of Rayleigh–Sommerfeld integral and apply the multiple reconstructed wavefield (MRW) to the gradient calculation. The MRW combines different reconstructed wavefields, and those wavefield components with similar properties are enhanced by superposition. The reflected waves, which are critical for updating the deep portions of the velocity model, are strengthened in the MRW to significantly reduce the negative effects of data noise when calculating FWI gradients. The MRW optimizes the gradient of the FWI, yielding high-quality results despite noise interference. Incorporating the MRW into the FWI effectively mitigates overfitting problems associated with noisy data and improves the robustness of the FWI.
Keywords
