Subsurface fluid-type likelihood using explainable machine learning

A method comprising:

• receiving pre-stack seismic data that includes a plurality of data elements that are seismically acquired for a plurality of geometric locations with respect to a subterranean reservoir;
• determining, using the pre-stack seismic data, a plurality of input features for each geometric location of the plurality of geometric locations, the plurality of input features representing at least one characteristic of the pre-stack seismic data;
• training a machine-learning model by using post-drill data about content and type of fluid in a historical subterranean reservoir and a computed plurality of historic input features that include (i) lithology and rock total compressibility of the historical subterranean reservoir, and (ii) entropy that is a measure of heat transfer of the historical subterranean reservoir;
• executing the trained machine-learning model on values of the plurality of input features to determine a likelihood of a type of fluid in the subterranean reservoir and to determine a list of features affecting the likelihood;
• and
• outputting the likelihood and the list of features.

A non-transitory computer-readable medium comprising instructions that are executable by a processing device for causing the processing device to perform operations comprising:

• receiving pre-stack seismic data that includes a plurality of data elements that are seismically acquired for a plurality of geometric locations with respect to a subterranean reservoir;
• determining, using the pre-stack seismic data, a plurality of input features for each geometric location of the plurality of geometric locations, the plurality of input features representing at least one characteristic of the pre-stack seismic data;
• training a machine-learning model by using post-drill data about content and type of fluid in a historical subterranean reservoir and a computed plurality of historic input features that include (i) lithology and rock total compressibility of the historical subterranean reservoir, and (ii) entropy that is a measure of heat transfer of the historical subterranean reservoir;
• executing the trained machine-learning model on values of the plurality of input features to determine a likelihood of a type of fluid in the subterranean reservoir and to determine a list of features affecting the likelihood;
• and
• outputting the likelihood and the list of features.