Constraining anthropogenic emissions and impacts of ethanol and other oxygenated VOCs: a combined modeling and field observational approach

Ethanol is an abundant volatile organic compound with important atmospheric chemical implications, both through its direct oxidation (as a major contributor to urban OH reactivity) and as a precursor of acetaldehyde and in turn peroxyacetyl nitrate (PAN), by which it contributes to long-range transport of NO_x and increased tropospheric ozone production. It is emitted both by plants and from anthropogenic activity (including industry, solvent use, fuel, and agriculture), but is chronically underestimated in atmospheric chemistry models. Here, we seek to mitigate this underestimate in GEOS-Chem, a global chemical transport model, by incorporating novel sources of ethanol and constraining their emissions using field observations collected across the United States in various recent field campaigns. We correlate measured ethanol concentrations to those of tracers with known emission profiles (e.g. nonanal from cooking, D5 siloxane from personal care products, etc.) using multivariate regression analysis to apportion ethanol to each tracer's individual source. We show that urban summertime ethanol has different dominant sources in the major US cities sampled across field campaigns -- e.g., agriculture in Chicago, volatile chemical products (VCPs) in Los Angeles, cooking in Las Vegas and Salt Lake City, and traffic in New York. These ethanol sources are poorly represented in the current version of GEOS-Chem, in which biogenic emissions dominate the global ethanol budget and VCP and cooking sources are omitted entirely. Based on our source apportionment from field data, we add ethanol emissions from agriculture, traffic, VCPs, and cooking to GEOS-Chem, along with additional species from each source (including updated mechanisms for newly added cooking and VCP tracers), and perform new simulations to compare with the field datasets. Incorporating these emissions into GEOS-Chem eliminates the model ethanol bias in US cities and improves model performance in simulating formaldehyde and PAN, though it results in overestimations of acetaldehyde. Finally, we assess the importance of ethanol to global budgets of PAN, ozone, and OH, and we extend our hybrid field/modeling analysis to other oxygenated volatile organic compounds typically underestimated by models, including methanol, acrolein, and ethylene glycol, and make recommendations for inclusion of their emissions and chemistry in GEOS-Chem.