Manual first arrival picking is a long-term (weeks or months), repetitive and monotonous task, which might add a subjective touch to processing. Moreover, a picking procedure performed by different specialists may differ significantly. Besides, breaking areas into several segments for separate picking procedures provides edge effects on their borders, which might lead to error ocurrence. All such picking errors may manifest themselves as linear or segmental stripes on velocity maps.
Automated or semi-automated first arrival picking may essentially mitigate the above-mentioned errors and provide areally homogeneous corrections.
This method uses multi-step calculations of the first arrival point on a registered seismic trace. First arrival time correction is carried out on each step. The final step ensures overall correction of first arrival time. It is based on selection of similar values of apparent velocities on traces close to the calculated one, using Kohonen neural networks. First arrival picking method has proved to be highly efficient on different data.
Specialists of seismic data processing department introduced their work to a wide audience at Geomodel 2021 conference in September 2021. Conference abstracts.