Global Estimates of Biogenic Methane Production in Marine Sediments Using Machine Learning and Deterministic Modeling

Abstract We have developed a model of geospatially estimating carbon accumulation and methanogenesis in seabed sediments that uses more accurate sophisticated inputs to models than used previous estimates. Using this hybrid stochastic deterministic model, we estimate the maximum available for global seabed, subsequent microbial methane generated as function location depth (including gas hydrate stability zone). Global integration over present previously microbially reactive column yields total be ∼0.8–2.2 × 10 6 1.1–3.0 Pg C CH 4 , respectively. Our improvements accuracy include using machine learned estimates seafloor which modeling is most sensitive (e.g., organic carbon, heat flux, porosity). sophistication dependent (on 5 arc‐minute grid), new sediment compaction (allowing non‐linear geothermal gradients), variable age versus at each grid cell. A reservoir magnitude here consistent with recent IPCC suggestion long‐term sinks could explain imbalances reduction atmospheric CO 2 last 50 million years. technique provides foundation globally updateable learning parameters input geologic geochemical models, allowing observations update budgets methane, methane.