Modeling Secondary Organic Aerosol Formation From Emissions of Combustion Sources

Atmospheric aerosols exert a large influence on the Earth’s climate and cause adverse public health effects, reduced visibility and material degradation. Secondary organic aerosol (SOA), defined as the aerosol mass arising from the oxidation products of gas-phase organic species, accounts for a significant fraction of the submicron atmospheric aerosol mass. Yet, there are large uncertainties surrounding the sources, atmospheric evolution and properties of SOA. This thesis combines laboratory experiments, extensive data analysis and global modeling to investigate the contribution of semi-volatile and intermediate volatility organic compounds (SVOC and IVOC) from combustion sources to SOA formation. The goals are to quantify the contribution of these emissions to ambient PM and to evaluate and improve models to simulate its formation.

To create a database for model development and evaluation, a series of smog chamber experiments were conducted on evaporated fuel, which served as surrogates for real-world combustion emissions. Diesel formed the most SOA followed by conventional jet fuel / jet fuel derived from natural gas, gasoline and jet fuel derived from coal. The variability in SOA formation from actual combustion emissions can be partially explained by the composition of the fuel.

Several models were developed and tested along with existing models using SOA data from smog chamber experiments conducted using evaporated fuel (this work, gasoline, fischertropschs, jet fuel, diesels) and published data on dilute combustion emissions (aircraft, on- and off-road gasoline, on- and off-road diesel, wood burning, biomass burning). For all of the SOA data, existing models under-predicted SOA formation if SVOC/IVOC were not included.

For the evaporated fuel experiments, when SVOC/IVOC were included predictions using the existing SOA model were brought to within a factor of two of measurements with minor adjustments to model parameterizations. Further, a volatility-only model suggested that differences in the volatility of the precursors were able to explain most of the variability observed in the SOA formation.

For aircraft exhaust, the previous methods to simulate SOA formation from SVOC and IVOC performed poorly. A more physically-realistic modeling framework was developed, which was then used to show that SOA formation from aircraft exhaust was (a) higher for petroleumbased than synthetically derived jet fuel and (b) higher at lower engine loads and vice versa.

All of the SOA data from combustion emissions experiments were used to determine source-specific parameterizations to model SOA formation from SVOC, IVOC and other unspeciated emissions. The new parameterizations were used to investigate their influence on the OA budget in the United States. Combustion sources were estimated to emit about 2.61 Tg yr-1 of SVOC, IVOC and other unspeciated emissions (sixth of the total anthropogenic organic emissions), which are predicted to double SOA production from combustion sources in the United States.

The contribution of SVOC and IVOC emissions to global SOA formation was assessed using a global climate model. Simulations were performed using a modified version of GISS GCM II’. The modified model predicted that SVOC and IVOC contributed to half of the OA mass in the atmosphere. Their inclusion improved OA model-measurement comparisons for absolute concentrations, POA-SOA split and volatility (gas-particle partitioning) globally suggesting that atmospheric models need to incorporate SOA formation from SVOC and IVOC if they are to reasonably predict the abundance and properties of aerosols.

This thesis demonstrates that SVOC/IVOC and possibly other unspeciated organics emitted by combustion sources are very important precursors of SOA and potentially large contributors to the atmospheric aerosol mass. Models used for research and policy applications need to represent them to improve model-predictions of aerosols on climate and health outcomes. The improved modeling frameworks developed in this dissertation are suitable for implementation into chemical transport models.