Abstract
Oceanic dimethyl sulfide (DMS) is the largest natural source of atmospheric sulfur. DMS is biologically produced in seawater and emitted into the atmosphere, where its oxidation products contribute to aerosol formation with consequences for cloud albedo and the Earth's radiative budget and climate. Climate model projections of how DMS emissions change with global warming are largely uncertain, even contradictory. Here, we use machine-learning models trained with biome-resolved global observations to simulate seawater DMS concentrations (1850 to 2100) using physico-chemical and biological predictors from eight CMIP6 models. The scatter in current projections is largely reduced, and globally averaged seawater DMS concentrations are predicted to decrease in the coming decades. However, global DMS emissions will increase due to rising surface wind speeds and sea surface temperatures which contradicts the current AR6 assessment that the DMS flux will reduce in the future. Concurrence of increasing DMS emissions and declining anthropogenic sulfur dioxide emissions suggests an increase in the relative importance of DMS to sulfate aerosol formation and its climate cooling impact.