As efforts to transition away from fossil fuels strengthen the hunt for new sources of low-carbon energy, scientists have developed a deep learning model to scan the Earth for surface expressions of subsurface reservoirs of naturally occurring free hydrogen.
Researchers used the algorithm to help narrow down the potential whereabouts of ovoids or semicircular depressions (SCDs) in the ground that form near areas associated with natural or "gold hydrogen" deposits. Though these circular patterns often appear in areas of low elevation, they can be hidden by agriculture or other vegetation. Recent discoveries of these circles in the U.S., Mali, Namibia, Brazil, France, and Russia have unveiled that they exist in greater numbers than previously thought.
To help uncover these nearly invisible semicircular depressions, two recent papers describe how lead authors Sam Herreid and Saurabh Kaushik, both postdoctoral scholars at the Byrd Polar and Climate Research Center at The Ohio State University, combined their model with global satellite imagery data to identify SCDs.
Their team compiled a list of known SCD locations to train their algorithm to search the globe. After using remote sensing data to analyze what these sites look like from above, they drew on geomorphic and spectral patterns to determine what sites around the world are most likely to be associated with SCDs related to geologic hydrogen.
Through their observations, the project found that AI demonstrates a unique ability to map out surface expressions of potential subsurface hydrogen reservoirs around the world, as well as establish a baseline for further investigation of hydrogen-associated sites. Their work was presented this week at poster sessions at the annual meeting of the American Geophysical Union.
To read more, click here.