Picking the first breaks from seismic data is often a challenging problem and still requires significant human effort. We have developed an iterative process that applies a traditional seismic automated picking method to obtain preliminary first breaks and then uses a machine learning (ML) method to identify, remove, and fix poor picks based on a multitrace analysis. The ML method involves constructing a convolutional neural network architecture to help identify poor picks across multiple traces and eliminate them. We then further refill the picks on empty traces with the help of the trained model. To allow training samples applicable to various regions and different data sets, we apply moveout correction with preliminary picks and address the picks in the flattened input. We collect 11,239,800 labeled seismic traces. During the training process, the model’s classification accuracy on the training and validation data sets reaches 98.2% and 97.3%, respectively. We also evaluate the precision and recall rate, both of which exceed 94%. For prediction, the results of 2D and 3D data sets that differ from the training data sets are used to demonstrate the feasibility of our method.
A new method of data-domain full traveltime inversion (FTI) is proposed to estimate the near-surface velocity model using early arrivals in seismic shot gathers. Data-domain FTI is capable of generating a background velocity model from which the predicted early arrivals can kinematically match the observed ones. Such a match is measured and quantified in terms of the crosscorrelation function between the computed and observed traces. Our method aims to find an optimal estimated velocity model that minimizes the crosscorrelation computed from the selected early arrivals. The early arrivals are isolated via a sequence of operations, including the [Formula: see text]-[Formula: see text] scan, autopicking, multidomain quality control, and guide interpolation. Because windows, rather than exact arrival times, are constructed, the difficulties encountered while picking precise arrivals are reduced. In addition, the gradient of data-domain FTI is derived based on an amplitude-constrained optimization problem, which makes the gradient essentially different from that derived with the Born approximation in which no constraint is used. The constraint requires the inversion to honor traveltime information only, and it thus ignores any amplitude changes caused by velocity variations. This method is validated using 3D synthetic as well as field data sets. The results show that data-domain FTI, combined with the early arrival selection workflow, is able to generate reasonable background velocities that kinematically match the predicted early arrivals with the observed ones, and the associated depth-domain images are clearly improved.
Residual statics estimation is an important step in seismic data processing. We develop a methodology for determining residual statics with common offset gathers (COG). The methodology is divided into two steps: 1) inverting the total residual statics of each trace in COG; 2) decomposing the total residual statics into the shot and the receiver residual statics. In the first step, the total residual statics can be inverted from equations derived from the refraction traveltimes in the common offset domain constrained by L1 norm regularization. In the second part, the total residual statics of each trace is decomposed into shot and receiver residual statics with zeroth-order Tikhonov regularization. The synthetic examples show that the method can help determine large magnitude residual statics and mitigate the effect of picking errors. The method is applied to real data and the results show improved quality of CMP stacks.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.