Authors: Miguel Corrales, Hussein Hoteit, and Matteo Ravasi

• Semi and fully data driven (easy integration of RPM).
• It relies on simple algebraic operations (SVD).
• Flexibility to select the number of optimal coefficients allows Seis2Rock to handle data with various degrees of noise (smaller number of optimal basis functions helps to manage noise levels).
• Formulates the problem as a series of post-stack inversion.
• When applied to field datasets, Seis2Rock relies on applying pre-processing steps to construct synthetic AVO gathers that closely mimic the field data.
• Finally, similar to any other data-driven method, Seis2Rock’s optimal basis functions may perform deficiently when applied on seismic datasets with geological settings differing from that of the training data (e.g, far away from well control).
•  

Figure 1. Descriptive summary of Training and Inference stages proposed in Seis2Rock.