A method includes acquiring seismic data of a region that utilizes multiple seismic energy sources and seismic energy receivers where the seismic data include blended seismic data for a number of emissions from a corresponding number of the multiple seismic energy sources; determining spatially distributed coherent noise properties for the region using the blended seismic data; via the spatially distributed coherent noise properties, modeling coherent noise as at least two coherent noise models for at least two of the emissions from a corresponding at least two of the multiple seismic energy sources; via the coherent noise models, attenuating coherent noise in a portion of the blended seismic data to generate coherent noise attenuated blended seismic data; deblending the coherent noise attenuated blended seismic data to generate deblended seismic data; and rendering an image of at least a portion of the region to a display using the deblended seismic data.

Systems and methods include a method for deblending signal and noise data. A shot domain for actual sources, a receiver domain for virtual sources, and a receiver domain for actual sources are generated from blended shot data. A dictionary of signal atoms is generated. Each signal atom includes a small patch of seismic signal data gathered during a small time window using multiple neighboring traces. A dictionary of noise atoms is generated. Each noise atom includes a small patch of seismic noise data gathered during a small time window using multiple neighboring traces. A combined signal-and-noise dictionary is generated that contains the signal atoms and the noise atoms. A sparse reconstruction of receiver domain data is created from the combined signal-and-noise dictionary. The sparse reconstruction is split into deblended data and blending noise data based on atom usage to create deblended shot domain gathers for actual sources.

Implementations described and claimed herein provide systems and methods for reservoir modeling. In one implementation, an input dataset comprising seismic data is received for a particular subsurface reservoir. Based on the input dataset and utilizing a deep learning computing technique, a plurality of trained reservoir models may be generated based on training data and/or validation information to model the particular subsurface reservoir. From the plurality of trained reservoir models, an optimized reservoir model may be selected based on a comparison of each of the plurality of reservoir models to a dataset of measured subsurface characteristics.

A method for seismic imaging includes receiving a multi-shot seismic data set that was collected using one or more streamers having recorders configured to detect seismic waves that propagate through a subterranean domain. The method also includes partitioning the multi-shot seismic data set into windows including a source dimension. The method also includes defining one or more first basis functions that describe the windows of the multi-shot seismic data set. The method also includes generating a model that describes a decomposition of the multi-shot seismic data set using the one or more first basis functions. The method also includes defining one or more second basis functions that describe a selected output data. The method also includes combining the one or more second basis functions with the model to produce a result for a source side wavefield and a receiver side wavefield.

Accordingly, there are disclosed herein geologic modeling methods and systems employing reference-based inversion of seismic image volumes. An illustrative method embodiment includes: (a) obtaining a measured seismic image volume; (b) determining a reference seismic image volume based on a reference model; (c) deriving a synthesized seismic image volume from a geologic model; (d) detecting at least one geologic model region where the synthesized seismic image volume and the measured seismic image volume are mismatched; (e) finding a reference model region where the reference seismic image volume best matches the measured seismic image volume; (f) replacing content of the at least one geologic model region with content of the reference model region to obtain an improved geologic model; and (g) outputting the improved geologic model.

A method includes, in a computer, generating a discretized model of the subsurface formation in space and time. The discretized model comprises at least one physical parameter of the formation and a relationship between the physical parameter and the physical property. For each spatial location and at each time in the discretized model, a time independent solution to the relationship is calculated. A context is defined of a selected number of grid cells surrounding each spatial location. Dimensionality reduction is performed on each context. Each dimensionality reduced context is input into the computer as a separate earth model to train a machine learning system to determine a relationship between the dimensionality reduced context and the physical property. The trained machine learning system is used to estimate the physical property at each spatial location and at each time.

A lithology prediction that uses a geological age model as an input to a machine learning model. The geological age model is capable of separating and recoding different seismic packages derived from the horizon interpretation. Once the machine learning model has been trained, a validation may be performed to determine the quality of the machine learning model. The quality may be improved by refining the training of the machine learning model. The lithology prediction generated by the machine learning model that utilizes the geological age model provides an improved lithology prediction that more accurately reflects the subterranean formation of an area of interest.

An example method is for producing a seismic wave velocity model of a subsurface area. The method includes receiving, by a processor of a computing system, from a seismic receiver, seismic data input comprising a recorded seismic wave field. The method includes receiving, by the processor, an initial 1D velocity model of the subsurface area. The method includes determining, by the processor, a Laplace-Fourier transform of the recorded seismic wave field. The method includes regenerating, by the processor, the current 1D velocity model to generate inverted data representing the subsurface area. The method may include performing, by the processor, an upscaling of a plurality of 1D velocity models to produce a 3D velocity model.

A computer-implemented method may include obtaining seismic data acquired in a time-domain for a subterranean region of interest. The method may further include obtaining a property model for the subterranean region of interest. The method may further include determining one or more time shifts using a segment dynamic image warping function based on the seismic data and the property model. The method may further include determining an adjoint source operator using the derived time shift and one-way wave equation. The method may further include updating the property model using a gradient solver in a data-domain reflection traveltime inversion. The method may further include outputting the updated property model for the subterranean region of interest. The method may further include generating a seismic image for the subterranean region of interest using the updated property model.

A method and apparatus for automated seismic interpretation (ASI), including: obtaining trained models comprising a geologic scenario from a model repository, wherein the trained models comprise executable code; obtaining test data comprising geophysical data for a subsurface region; and performing an inference on the test data with the trained models to generate a feature probability map representative of subsurface features. A method and apparatus for machine learning, including: an ASI model; a training dataset comprising seismic images and a plurality of data portions; a plurality of memory locations, each comprising a replication of the ASI model and a different data portion of the training dataset; a plurality of data augmentation modules, each identified with one of the plurality of memory locations; a training module configured to receive output from the plurality of data augmentation modules; and a model repository configured to receive updated models from the training module.