Methodology for enhancing properties of geophysical data with deep learning networks

A computer-implemented method of enhancing geophysical datasets, the method comprising:

• accessing a first geophysical dataset obtained via a first geophysical data gathering technique and a second geophysical dataset via a second geophysical data gathering technique, wherein the first geophysical data gathering technique is of a different type than the second geophysical data gathering technique, wherein the first geophysical dataset and the second geophysical dataset include at least one common attribute, wherein the first geophysical dataset is lacking at least one feature due to greater noise or due to a failure to obtain the at least one feature than in the second geophysical dataset;
• wherein the at least one feature comprises a higher frequency range;
• wherein the first geophysical dataset comprises towed streamer data, the towed streamer data including a lower frequency range of towed streamer data of greater reliability than the higher frequency range of towed streamer data of lower reliability or as being unavailable;
• wherein the second geophysical dataset comprises near-field hydrophone data,
• training a neural network using at least part or all of the second geophysical dataset in order for the trained neural network to reconstruct the at least one feature;
• and
• applying the trained neural network by inputting at least a part of the first geophysical dataset to the trained neural network in order for the trained neural network to reconstruct the at least one feature for the first geophysical dataset.