Secondary organic aerosols (SOA) are large contributors to fine particle mass loading and number concentration and interact with clouds and radiation. Several processes affect the formation, chemical transformation, and removal of SOA in the atmosphere. For computational efficiency, global models use simplified SOA treatments, which often do not capture the dynamics of SOA formation. Here we test more complex SOA treatments within the global Energy Exascale Earth System Model (E3SM) to investigate how simulated SOA spatial distributions respond to some of the important but uncertain processes affecting SOA formation, removal, and lifetime. We evaluate model predictions with a suite of surface, aircraft, and satellite observations that span the globe and the full troposphere. Simulations indicate that both a strong production (achieved here by multigenerational aging of SOA precursors that includes moderate functionalization) and a strong sink of SOA (especially in the middle upper troposphere, achieved here by adding particle‐phase photolysis) are needed to reproduce the vertical distribution of organic aerosol (OA) measured during several aircraft field campaigns; without this sink, the simulated middle upper tropospheric OA is too large. Our results show that variations in SOA chemistry formulations change SOA wet removal lifetime by a factor of 3 due to changes in horizontal and vertical distributions of SOA. In all the SOA chemistry formulations tested here, an efficient chemical sink, that is, particle‐phase photolysis, was needed to reproduce the aircraft measurements of OA at high altitudes. Globally, SOA removal rates by photolysis are equal to the wet removal sink, and photolysis decreases SOA lifetimes from 10 to ~3 days. A recent review of multiple field studies found no increase in net OA formation over and downwind biomass burning regions, so we also tested an alternative, empirical SOA treatment that increases primary organic aerosol (POA) emissions near source region and converts POA to SOA with an aging time scale of 1 day. Although this empirical treatment performs surprisingly well in simulating OA loadings near the surface, it overestimates OA loadings in the middle and upper troposphere compared to aircraft measurements, likely due to strong convective transport to high altitudes where wet removal is weak. The default improved model formulation (multigenerational aging with moderate fragmentation and photolysis) performs much better than the empirical treatment in these regions. Differences in SOA treatments greatly affect the SOA direct radiative effect, which ranges from −0.65 (moderate fragmentation and photolysis) to −2 W m−2 (moderate fragmentation without photolysis). Notably, most SOA formulations predict similar global indirect forcing of SOA calculated as the difference in cloud forcing between present‐day and preindustrial simulations.