Systems and methods for using probabilities of lithologies in an inversion

1. A computer-implemented method for training a model that uses probabilities of lithologies as prior information in an inversion, the method being implemented in a computer system, the computer system including one or more physical computer processors and non-transient electronic storage that stores subsurface map data sets that correspond to different subsurface volumes of interest, wherein a given subsurface map data set corresponding to a given subsurface volume of interest specifies textural attribute values as a function of position within the given subsurface volume of interest, the method comprising:

obtaining, from the electronic storage, training data, the training data including (i) subsurface map data sets specifying textural attribute values as a function of position corresponding to a subsurface volume of interest, and (ii) known lithologies as a function of position within the subsurface volume of interest, such that the textural attribute values of the lithologies are known as a function of position within the subsurface volume of interest;

obtaining an initial seismic mapping machine learning model from the electronic storage;

generating a conditioned seismic mapping machine learning model by training, on the one or more physical computer processors, the initial seismic mapping machine learning model using the training data;

storing the conditioned seismic mapping machine learning model in the electronic storage;

obtaining a target subsurface map data set for the subsurface volume of interest;

applying the conditioned seismic mapping machine learning model to the target subsurface map data set to generate a classified lithology map data set, wherein the conditioned seismic mapping machine learning model receives as input the target subsurface map data set for the subsurface volume of interest and provides as output probabilities of lithologies for different locations within the subsurface volume of interest, the probabilities of lithologies for different locations within the subsurface volume of interest output by the conditioned seismic mapping machine learning model included within the classified lithology map data set;

applying the inversion to the classified lithology map data set to generate volumes of lithologies specifying parameters for individual lithologies as a function of position within the subsurface volume of interest, wherein the probabilities of lithologies for different locations within the subsurface volume of interest provided by the classified lithology map data set are used as the prior information in the inversion such that the output of the conditioned seismic mapping machine learning model is used as input to the inversion;

generating an image that represents the volumes of lithologies using visual effects to depict at least a portion of the parameters in the volumes of lithologies; and

displaying, on a graphical user interface, the image.

2. The computer-implemented method of claim 1 , wherein the classified lithology map data set estimates lithologies as a function of position within the subsurface volume of interest based on textural attribute values specified in the target subsurface map data set.

3. The computer-implemented method of claim 1 , wherein the training data is derived from well data and seismic data.

4. The computer-implemented method of claim 1 , wherein the initial seismic mapping machine learning model and the conditioned seismic mapping machine learning model comprise a random decision forest.

5. The computer-implemented method of claim 1 , wherein the inversion comprises a stochastic amplitude variation with angle of incidence (AVA) inversion.

6. The computer-implemented method of claim 1 , wherein the image indicates shale probabilities as a function of position within the subsurface volume of interest.

7. The computer-implemented method of claim 1 , wherein the image indicates sand probabilities as a function of position within the subsurface volume of interest.

8. A computer-implemented method for estimating lithologies as a function of position within a subsurface volume of interest, the method being implemented in a computer system, the computer system including one or more physical computer processors and non-transient electronic storage that stores subsurface map data sets that correspond to different subsurface volumes of interest, wherein a given subsurface map data set corresponding to a given subsurface volume of interest specifies textural attribute values of different lithologies as a function of position within the given subsurface volume of interest, the method comprising:

obtaining a target subsurface map data set for the subsurface volume of interest;

obtaining a conditioned seismic mapping machine learning model from the electronic storage, the conditioned seismic mapping machine learning model having been conditioned by training an initial mapping machine learning model using training data, wherein the training data includes (i) subsurface map data sets specifying textural attribute values as a function of position corresponding to the subsurface volume of interest, and (ii) known lithologies as a function of position within the subsurface volume of interest, such that the textural attribute values of the lithologies are known as a function of position within the subsurface volume of interest;

applying the conditioned seismic mapping machine learning model to the target subsurface map data set to generate a classified lithology map data set, wherein the conditioned seismic mapping machine learning model receives as input the target subsurface map data set for the subsurface volume of interest and provides as output probabilities of lithologies for different locations within the subsurface volume of interest, the probabilities of lithologies for different locations within the subsurface volume of interest output by the conditioned seismic mapping machine learning model included within the classified lithology map data set;

generating an image that represents the classified lithology map data set using visual effects to depict at least a portion of estimated lithologies within the subsurface volume of interest, wherein the estimated lithologies as a function of position within the subsurface volume of interest are determined based on application of an inversion to the classified lithology map data set, wherein the probabilities of lithologies for different locations within the subsurface volume of interest provided by the classified lithology map data set are used as prior information in the inversion such that the output of the conditioned seismic mapping machine learning model is used as input to the inversion; and

displaying, on a graphical user interface, the image.

9. The computer-implemented method of claim 8 , wherein the classified lithology map data sets estimate lithologies as a function of position within the subsurface volume of interest based on textural attribute values specified in the target subsurface map data set.

10. The computer-implemented method of claim 8 , wherein the application of the inversion to the classified lithology map data set comprises using amplitude variation with angle of incidence (AVA) inversion on the classified lithology map data set.

11. The computer-implemented method of claim 8 , wherein the initial seismic mapping machine learning model and the conditioned seismic mapping machine learning model comprise a random decision forest.

12. The computer-implemented method of claim 8 , wherein the target subsurface map data set comprises textural attribute values derived from seismic data corresponding to the subsurface volume of interest.

13. The computer-implemented method of claim 8 , wherein the image indicates shale probabilities as a function of position within the subsurface volume of interest.

14. The computer-implemented method of claim 8 , wherein the image indicates sand probabilities as a function of position within the subsurface volume of interest.

15. A computer system for training a model that estimates lithologies as a function of position within a subsurface volume of interest, the system comprising:

one or more physical computer processors;

non-transient electronic storage that stores subsurface map data sets that correspond to different subsurface volumes of interest, wherein a given subsurface map data set corresponding to a given subsurface volume of interest specifies textural attribute values of different lithologies as a function of position within the given subsurface volume of interest;

memory; and

one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more physical computer processors, the one or more programs including instructions that when executed by the one or more physical computer processors cause the system to:

obtain training data from the electronic storage, the training data including (i) subsurface map data sets specifying textural attribute values as a function of position corresponding to a subsurface volume of interest, and (ii) known lithologies as a function of position within the subsurface volume of interest, such that the textural attribute values of the lithologies are known as a function of position within the subsurface volume of interest;

obtain an initial seismic mapping machine learning model from the electronic storage;

generate a conditioned seismic mapping machine learning model by training, on the one or more physical computer processors, the initial seismic mapping machine learning model using the training data;

store the conditioned seismic mapping machine learning model in the electronic storage;

obtain a target subsurface map data set for the subsurface volume of interest;

apply the conditioned seismic mapping machine learning model to the target subsurface map data set to generate a classified lithology map data set, wherein the conditioned seismic mapping machine learning model receives as input the target subsurface map data set for the subsurface volume of interest and provides as output probabilities of lithologies for different locations within the subsurface volume of interest, the probabilities of lithologies for different locations within the subsurface volume of interest output by the conditioned seismic mapping machine learning model included within the classified lithology map data set;

generate an image that represents the classified lithology map data set using visual effects to depict at least a portion of estimated lithologies within the subsurface volume of interest, wherein the estimated lithologies as a function of position within the subsurface volume of interest are determined based on application of an inversion to the classified lithology map data set, wherein the probabilities of lithologies for different locations within the subsurface volume of interest provided by the classified lithology map data set are used as prior information in the inversion such that the output of the conditioned seismic mapping machine learning model is used as input to the inversion; and

display, on a graphical user interface, the image.

16. The computer system of claim 15 , wherein the classified lithology map data set estimates a probability of a type of lithology as a function of position within the subsurface volume of interest to textural attribute values specified in the target subsurface map data set.

17. The computer system of claim 15 , wherein the application of the inversion to the classified lithology map data set comprises using amplitude variation with angle of incidence (AVA) inversion on the classified lithology map data set.

18. The computer system of claim 15 , wherein the initial seismic mapping machine learning model and the conditioned seismic mapping machine learning model comprise a random decision forest.

19. The computer system of claim 15 , wherein the target subsurface map data set comprises textural attribute values derived from seismic data corresponding to the subsurface volume of interest.