Answers are grounded in this paper and cite their sources.
Picking first breaks on seismic data has historically been a very demanding and time-consuming task. It may take several weeks or even months to pick first breaks for a single seismic survey. Trace counts for modern 3D seismic surveys can now reach into the billions. Manually picking first breaks on billions of traces is not feasible. Some automated methods for first break picking already exist, but typically do not perform well in the presence of noise and azimuthal anisotropy. The Permian dataset used in this study contains noisy traces and a ‘fill zone’ with strong anisotropy where most autopickers fail, requiring weeks of manual intervention. Using a combination of physics-based tomography and deep learning, we show that we can produce accurate first break picks in days rather than weeks, even in the presence of noise.
No takes yet. Share an insight, caveat, or question.
Consolvo et al. (2021) studied this question.