Dual Origins and Accelerated Accumulation of Natural Hydrogen in the Songliao Basin: Insights from Explainable Machine Learning

Natural hydrogen has garnered increasing attention as a clean energy source in the global energy transition. However, accurately interpreting the formation mechanisms of natural hydrogen remains challenging because of its diverse origins and limitations of traditional analytical methods. This study investigates the origin and accumulation patterns of hydrogen in the Songliao Basin, northeast China, by integrating geochemical measurements with explainable machine learning models trained on an extensive database compiled from 1,034 published natural hydrogen samples worldwide. Results reveal that hydrogen in the Songliao Basin principally originates from a dual system comprising mantle degassing and crustal processes, including water-rock reactions and water radiolysis occurring on the basin basement. The strikingly depleted hydrogen isotopes (–806.9‰ to –623.7‰) further suggest significant post-genetic fractionation during rapid advective migration of hydrogen. These advective conditions result from tectonic inversion reactivating deep faults, which significantly enhances permeability and creates high-flux conduits for deep-sourced hydrogen to ascend, accumulate and mix with thermogenic and biogenic methane in shallower reservoirs. Our calculations illustrate that tectonic inversion-related fault reactivation decreases hydrogen transport time along deep faults by an order of magnitude (from ~105 to ~104 years). This acceleration allows rapid hydrogen accumulation within limited windows of suitable trapping and sealing conditions. These findings highlight the critical relationship between structural evolution and hydrogen accumulation patterns, demonstrating tectonic inversion as an essential control on natural hydrogen systems and providing a predictive framework for exploration in sedimentary basins.