A Global Marine Methanethiol Climatology Estimated Using Machine Learning

Methanethiol (MeSH), a reduced sulfur compound, has received far less attention than dimethyl sulfide (DMS) despite its potential importance for atmospheric sulfur cycling and climate-relevant aerosol processes. Compared with DMS, gas-phase oxidation of MeSH yields more SO₂ and has a shorter atmospheric lifetime, suggesting a disproportionate influence on new particle formation, aerosol growth, and cloud condensation nuclei (CCN) abundance in the marine atmosphere. Current global estimates of marine MeSH emissions have relied on scaling DMS concentration climatologies using empirical MeSH:DMS ratios, implicitly assuming co-variability between the two compounds.

Here, we present global monthly marine MeSH emissions derived using a machine-learning framework constrained by 27 years of satellite observations, ocean reanalysis products, and shipboard measurements. Key satellite predictors include chlorophyll-a (Chl-a), phytoplankton functional types (PFTs), phytoplankton size classes (PSCs), and photosynthetically active radiation (PAR). Our approach directly predicts MeSH concentrations from environmental drivers, independent of DMS distributions. The regression models were trained and validated using MeSH sea-surface concentration measurements from multiple oceanographic field campaigns.

We estimate a global annual marine MeSH emission of 5.06 Tg S yr^-1. Regional emissions were analyzed by dividing the global ocean into nine Longhurst biomes. The largest contributions originate from the Southern Westerlies (29.05%), Pacific Trades (15.22%), and Coastal Ocean regions (14.03%). Both seawater MeSH concentrations and emissions exhibit pronounced seasonal variability, with peak global emissions occurring in October and a minimum in June. These results provide a satellite-based global climatology of marine MeSH emissions and establish a basis for assessing its impacts on atmospheric chemistry and global climate.