The composition of the atmosphere has changed substantially over Earth's history, with important implications for past climate. A number of case studies will be presented which employ coupled climate model simulations to assess the strength of these chemical feedbacks on the climate. The eruption of Mount Samalas in 1257 led to the largest stratospheric volcanic injection of aerosol precursor gases in the Common Era, however climate model simulations of the last millennium typically overestimate the resulting climatic cooling when compared with tree-ring proxy records. A novel configuration of the Met Office UM-UKCA climate model is presented which couples an atmosphere-ocean general circulation model to a rigorous treatment of the relevant atmospheric chemistry and microphysical aerosol processes. This permits the climate response to a particular stratospheric injection of reactive volatile gases to be quantified and for the first time to date applied to a historical volcanic eruption. This model configuration compares favourably to observational data for simulations of the 1991Mount Pinatubo eruption. Results from an ensemble of model simulations are presented, with different assumptions about the sulfur dioxide and halogen loadings based on a recent geochemical reconstruction. These show a muted climate response, in reasonable agreement with tree ring records. Emissions of halogenated compounds lead to an increase in the sulfur dioxide lifetime, widespread ozone depletion and a prolonged climatic cooling. Strong increases in incident ultraviolet radiation at Earth's surface also occur. Oxygen levels may have varied fromas little as 10% to as high as 35% in the Phanerozoic (541Ma - Present). An increase in atmospheric oxygen increases atmospheric mass which leads to a reduction in incident shortwave radiation at Earth's surface due to Rayleigh scattering. However, this is offset by an increase in the pressure broadening of greenhouse gas absorption lines. Dynamical feedbacks also lead to increased meridional heat transport, warming polar regions and cooling tropical regions. An increase in oxygen content using the HadCM3-BL and HadGEM3-AO climate models leads to a global mean surface air temperature increase for a pre-industrial Holocene base case, in agreement with idealised 1D and 2D modeling studies. Case studies from past climates are investigated using HadCM3-BL which show that in the warmest climates, increasing oxygen may lead to a temperature decrease, as the equilibrium climate sensitivity is lower. For the Maastrichtian (72.1 - 66.0Ma), increasing oxygen content leads to a better agreement with proxy reconstructions of surface temperature at that time irrespective of the carbon dioxide content. There is considerable uncertainty in the timing of the rise in atmospheric oxygen content from values around 1% in the Neoproterozoic (1000 Ma - 541 Ma) to the 10- 35% values inferred in the Phanerozoic with respect to two global glaciation episodes (717-635Ma). Results of simulations with HadCM3-BL which investigate the impact of oxygen content on the Neoproterozoic Snowball Earth glaciations are presented. These demonstrate that a smaller reduction in carbon dioxide content is required to initiate a Snowball Earth at low oxygen content. Geological evidence suggests the presence of a basaltic large igneous province before the Sturtian Snowball Earth episode. This could have caused episodes of paced explosive volcanism, injecting sulfate aerosol precursors into the stratosphere. Results of simulations to investigate the impact of different volcanic aerosol emission scenarios are presented. 500 Tg SO2 is investigated with a range of aerosol sizes. For aerosol size distributions consistent with the aerosol evolution in the aftermath of the Mount Pinatubo eruption, the Earth enters a Snowball Earth in between 30 and 80 years. Using a larger size of aerosols, consistent with a larger eruption, does not lead to a Snowball Earth. These simulations show that changes to the chemical composition of the atmosphere, whether reactive gases or bulk chemical composition may have played an important role in the past climate of Earth.

Trace gases and aerosols, or suspended liquid and solid material in the atmosphere, have significant climatological and societal impacts; consequently, accurate representation of their contribution to atmospheric composition is vital to predicting climate change and informing policy actions. Sensitivity analysis allows scientists and environmental decision makers alike to ascertain the role a specific component of the very complex system that is the atmosphere of the Earth. Anthropogenic and natural emissions of gases and aerosol are transported by winds and interact with sunlight, allowing significant transformation before these species reach the end of their atmospheric life on land or in water. The adjoint-based sensitivity method assesses the relative importance of each emissions source to selected results of interest, including aerosol and cloud droplet concentration. In this work, the adjoint of a comprehensive inorganic aerosol thermodynamic equilibrium model was produced to improve the representativeness of regional and global chemical transport modeling. Furthermore, a global chemical transport model adjoint equipped with the adjoint of a cloud droplet activation parameterization was used to explore the footprint of emissions contributing to current and potential future cloud droplet concentrations, which impact the radiative balance of the earth. In future work, these sensitivity relationships can be exploited in optimization frameworks for assimilation of observations of the system, such as satellite-based or in situ measurements of aerosol or precursor trace gas concentrations.