A method for training a backpropagation-enabled segmentation process is used for identifying an occurrence of a sub-surface feature. A multi-dimensional seismic data set with an input dimension of at least two is inputted into a backpropagation-enabled process. A prediction of the occurrence of the subsurface feature has a prediction dimension of at least 1 and is at least 1 dimension less than the input dimension.

A method for training a backpropagation-enabled regression process is used for predicting values of an attribute of subsurface data. A multi-dimensional seismic data set with an input dimension of at least two is inputted into a backpropagation-enabled process. A predicted value of the attribute has a prediction dimension of at least 1 and is at least 1 dimension less than the input dimension.

Systems and methods are disclosed for generating subsurface data as a function of position and time. Exemplary implementations may include obtaining a first initial subsurface model and a first set of subsurface parameters, obtaining training subsurface property data and a first training subsurface dataset, generating a first conditioned subsurface model, and storing the first conditioned subsurface model.

Systems and methods are disclosed for generating subsurface property data as a function of position and time. Exemplary implementations may include obtaining a first initial subsurface property model and a first set of subsurface property parameters, obtaining training input data and a first training subsurface property dataset, generating a first conditioned subsurface property model, and storing the first conditioned subsurface property model.

Methods and apparatus are described for performing a 4D monitor marine seismic survey that repeats a previous survey. A number of sources may be used during the 4D monitor survey that differs from a number of sources that were used during the previous survey. Shot points from the previous survey are repeated by the 4D monitor survey, and additional shot points may be produced during the 4D monitor survey that were not produced during the previous survey. Embodiments enable efficiency and data quality improvements to be captured during 4D survey processes, while preserving repeatability.

A system and method is provided for restoring a 3D tomographic model of the Earth's subsurface geology from the present-day to a past restoration time. Whereas at the present time all faults represent active discontinuities, at a past restoration time some faults have not yet formed. Accordingly, the restored model divides the fault network into τ-active faults (discontinuous surfaces for faults that intersect the layer deposited at the past restoration time) and τ-inactive faults (continuous surfaces for faults that do not intersect the layer deposited at the past restoration time). A new 3D restoration transformation is also provided that uses linear geological constraints to process the restoration model in less time and generate more accurate geological images.

An illustrative fracture mapping system includes: a data acquisition unit collecting measurements deformation measurements during a fluid injection phase of a hydraulic fracturing operation; and a processing system implementing a formation mapping method. The formation mapping method includes: obtaining an initial fracture map having a location and geometry for one or more fractures activated during the fluid injection phase, the fracture map corresponding to a given time during the hydraulic fracturing operation; deriving from the initial fracture map a time series of fracture maps for times preceding the given time; and storing the time series on a nonvolatile information storage medium. The deriving may be performed in an iterative fashion to obtain each fracture map in the time series from a subsequent fracture map.

Disclosed is a method for detecting a fluid, including at least one step of: measuring by at least one sensor of a wave propagating in an environment of the wave, in order to obtain at least one measured signal, the wave being particularly a mechanical wave; splitting the measured signal over a plurality of split time intervals with a predefined duration in order to obtain samples of the measured signal; computing the temporal coherence of the samples; and determination of the presence of the fluid using the computed temporal coherence.

Disclosed are methods, systems, and computer-readable medium to perform operations including: generating, using a source wavelet and a current velocity model, modeled seismic data of the subterranean formation; applying a pre-condition to a seismic data residual calculated using the modeled seismic data and acquired seismic data from the subterranean formation; generating a velocity update using the source wavelet and the pre-conditioned seismic data residual; updating, using the velocity update, the current velocity model to generate an updated velocity model; determining that the current velocity model satisfies a predetermined condition; and responsively determining that the updated velocity model is the velocity model of the subterranean formation.

Common transmission points can be used to monitor a seismic reservoir. First and second common transmission points (CTPs) are received. For each of the first CTP gather and the second CTP gather, the traces before the CTP can be aggregated, and the traces crossing the CTP can be aggregates. The aggregated before CTP traces from the first CTP gather can be compared with the aggregated before CTP traces from the second CTP gather to determine a first time difference. The aggregated cross CTP traces from the first CTP gather can be compared with the aggregated cross CTP traces from the second CTP gather to determine a second time difference. A third time difference can be determined based at least partially on the first time difference and the second time difference.