Aerosol modeling is central to both climate projections and air quality forecasts. In climate modeling, the total estimated aerosol radiative forcing uncertainty range (-2 to -0.6 W m-2 in IPCC AR6) is dominated by that of aerosol-cloud interactions (-1.7 to -0.3 W m-2). These interactions depend on the properties of the aerosol population as well as its evolution and transport over long distances. In particular, smoke from burning biomass undergoes drastic and poorly-constrained evolutionary changes throughout its lifetime. Smoke emissions are typically estimated from satellite observations that are vulnerable to sensor deficiencies or cloud cover, and, lacking a reference source of truth, emissions inventories have large discrepancies in emission amounts. Smoke composition, size distribution, and vertical placement are best quantified from limited and expensive in situ observations. These limitations, coupled with the complex and variable impact of smoke particles on cloud nucleation, create significant uncertainties in assessing smoke’s overall influence on radiative budgets and human health.In this dissertation, I address the task of reducing these uncertainties in the modeling of biomass-burning smoke. The southeastern Atlantic provides an excellent study region for smoke’s climatic impacts and evolution, as it has enormous smoke emissions from burning in southern Africa, coupled with a near-permanent offshore stratocumulus cloud deck. We evaluate WRF-CAM5, CESM, and E3SM with observations from three overlapping field campaigns in the southeastern Atlantic in 2017—ORACLES, CLARIFY, and LASIC. The wide spatial extent of the campaigns and the redundancy of several observations provide an excellent basis from which to understand smoke properties. I find that the models often lack important aerosol chemical processes in both the free troposphere and boundary layer, as well as inaccurate cloud properties and responses to aerosols. I implement sensitivity tests to two global earth system models, CESM and E3SM. As a result, I show that aerosol oxidation in the free troposphere is likely driving the loss of 25-50% of organic aerosol mass over timescales of ~4-12 days through this region, a process which E3SM and WRF-CAM5 initially lacks. I also demonstrate that dimethyl sulfide emissions and this oxidative loss are key factors representing the nearly tripling of sulfate by mass fraction in the boundary layer from the free troposphere. Modeled cloud droplet number concentration is shown to be highly sensitive to updraft biases. By doubling updraft strength in E3SM and CESM in line with aircraft measurements, I reduce the normalized mean bias against observed cloud droplet nucleation efficiency by ~25% on average.
I further improve smoke emissions for air quality forecasts by using Doppler weather radar to improve wildfire emissions during the record-breaking 2020 western United States wildfire season. Previous satellite-based emissions estimates during this period are plausibly underestimated by an order of magnitude or more during the largest and most intense fire periods. Correcting for emissions gaps in WRF-Chem reduces bias in AOD by 5-50% and decreases bias in surface PM2.5 from -25% to +3% over the entire western US. For both variables, improvements are larger when focusing on the strongest burning regions and time periods. This thesis outlines significant overall improvements in smoke chemical modeling that stand to improve both long-term climate projections and short-term air quality forecasting and early-warning systems.
