Conclusion

We propose a general workflow for moveout processing and inversion by means of time-warping. The method achieves event flattening and moveout parameter estimation in an automatic and efficient manner with minimal user inputs. We show that our method doesn't rely on multi-parameter scans but instead on data-driven slope measurements. Automatic event flattening can be achieved with the measured slopes and produces the information on moveout traveltimes for all corrected CMP events. Such information can then be used in a moveout inversion process for estimating effective parameters associated with any preferred choice of moveout approximation. Due to the cheap traveltime modeling with moveout approximation, we choose to solve this inverse problem using a global Monte Carlo optimization method. Using both synthetic and real-data examples, we demonstrate that the resulting moveout parameters estimated using our approach are represented by posterior probability distributions that contain valuable statistical information. Unlike the results from other estimation methods such as semblance-based picking and least-squares inversion, the posterior probability distributions are unbiased towards any particular solution but represent the entire space of possible solutions and their associated probability of occurrence. This information can be used in conjunction with results from other estimation methods for better assessment and quality control of estimated parameters associated with any selected reflectors. Our results also suggest that the solutions to the problem of moveout inversion from traveltime fitting for 2D CMP gathers are generally well-constrained with only a few parameters to estimate despite a trade-off between the second-order $W$ and quartic moveout coefficient $A$. On the other hand, for 3D CMP gathers associated with complex anisotropic subsurface, the solutions appear non-unique and that there are trade-offs among inverted parameters, especially the quartic coefficients ($A_i$). Only $W_i$ parameters that govern the NMO ellipse seem to be generally well-constrained by the traveltime data. Further limitations on the data such as noise and sparse measurements can hamper the inversion and lead to even inferior results as shown by more diffuse estimated posterior distributions at a high level of data uncertainty. Therefore, one has to be mindful to draw conclusions on representative model parameters to be used in subsequent stages of processing and imaging, especially for 3D anisotropic models.