Dimethyl sulfide (DMS), arising from phytoplankton, is the largest natural source of sulfur in the atmosphere. The oxidation products of DMS, such as methanesulfonic acid and sulfuric acid, can contribute to cloud condensation nuclei. Studying the oxidation mechanism of DMS will help improve our modelling of the natural processes that contribute to the Earth's radiative balance. In the past decade, the discovery of a major new pathway in DMS oxidation (involving the formation of hydroperoxymethyl thioformate through autoxidation) has spawned renewed interest in the chemistry of DMS and resulted in new theoretical calculations, laboratory experiments, and field campaigns. However, these new studies still leave several unanswered questions and gaps in our understanding of DMS chemistry. This thesis utilises the recent literature to develop a comprehensive gas-phase DMS mechanism, which undergoes evaluation through five chamber experiments, and intercomparison with other near-explicit mechanisms in the literature. The new mechanism outperforms four mechanisms for 8 of the 14 DMS oxidation products studied. The developed mechanism is then extended to include aqueous and halogen chemistry, along with methanethiol chemistry, which has been identified as another substantial natural source of sulfur in the marine atmosphere. The extended mechanism is applied to tropical, temperate and polar field campaigns, and the uncertainties of the reactions are considered and propagated to provide uncertainties in the modelled concentrations of the three marine regimes. Additionally, the contributions of specific reactions to the uncertainty are determined to provide recommendations for reactions that should be studied further. Specifically, the photolysis of hydroperoxymethyl thioformate would reduce the uncertainty in OCS concentration, while reactions involving CH3SO2O2 would improve the modelling of gas-phase methanesulfonic acid and sulfuric acid in global models. In general, further experiments exploring reactions with OH could reduce the uncertainty of most DMS oxidation products by at least 10%. Finally, for the new mechanism to be useful to the global modelling community, the mechanism is reduced through an automated rates-based approach, and evaluated against the complete mechanism through randomly sampled box models of the marine boundary layer. This reduced mechanism performs well in reproducing the new reference mechanism (with a bias less than 11% for most oxidation products), and is recommended for use in global models. This work has produced a near-explicit, evaluated DMS reference mechanism that represents the current understanding of DMS oxidation. Uncertainty quantification has been conducted to illustrate the uncertainty in the mechanism and demonstrate the gaps in our current understanding that should be studied further. Finally, the reduction of this mechanism provides a simplified mechanism for use in global models, which is traceable to the complex reference mechanism.