Evaluating the linearity of sulfate formation in response to changes in sulfur dioxide emissions within a regional air quality model

The Big Bend Regional Aerosol and Visibility Observational Study (BRAVO) was designed to investigate the sources of pollutants that contribute to haze at Big Bend National Park, located in southwestern Texas. One component of BRAVO was the use of the REMSAD (Regional Modeling System for Aerosols and Deposition) air quality model to estimate which source regions have the largest impact on Big Bend with regard to predicted sulfate concentrations. Sulfate particles (SO4) were the main component of visibility-impairing haze during the four month BRAVO field campaign, which was conducted during June - October 1999. Sulfate is formed when sulfur dioxide (SO2) is oxidized in the atmosphere through gas-phase or aqueous-phase reactions. The principal source of anthropogenic sulfur dioxide is coal-fired power plants. Four large regional emission source areas were evaluated for this study: the eastern US (14.0M tonnes SO2/yr), Texas (1.1M tonnes SO2/yr), Mexico (2.5M tonnes SO2/yr), and the western US (1.8M tonnes SO2/yr). In addition to these large regional sources, several smaller sub-regions were considered (e.g., northeastern Texas), as well as the Carbón I/II power plant, which is approximately 225 km southeast of Big Bend NP.

The contribution of each of these source regions was evaluated using a suite of REMSAD emissions sensitivity simulations. These sensitivity simulations were conducted by modifying sulfur dioxide and primary sulfate emissions from a source region, repeating the REMSAD simulation with these new emissions, and comparing the resulting sulfate prediction with the original base emissions simulation. The change in predicted sulfate arising from these emission sensitivity simulations defined a source region's contribution at Big Bend NP. Simulations where sulfur dioxide was removed from the source region of interest were termed “emissions-out” sensitivity simulations. A complementary set of “emissions-in” simulations was also evaluated, where sulfur emissions were retained for the source region of interest but removed elsewhere in the model domain. It should be noted that only sulfur dioxide and primary sulfate emissions were modified, and emissions of other species, e.g., volatile organic compounds, oxides of nitrogen, carbon monoxide, etc., were left at their normal levels.

It is important to investigate the equivalency of these two methods as either one could be regarded as a suitable approach for determining the impact of emissions from a source region on pollutant concentrations at a downwind receptor. Many processes within the model can be assumed to be linear and hence would yield the same results whether the emissions-out or the emissions-in inventory was used. These processes include, for example, dry deposition, as deposition fluxes scale in direct proportion to ambient concentrations. It is not obvious, however, that the rate of chemical transformation of sulfur dioxide to sulfate will respond in a linear fashion with regard to the emissions-out and emissions-in simulations, since the availability of photochemical oxidants (e.g., hydrogen peroxide and hydroxyl radical) are dependent upon second- and third-order chemical reactions.

This study examines the degree of linearity that exists in terms of simulated ulfate production as a function of sulfur dioxide emissions, and how non-linearities within the system are manifested. In general, the predicted attributions arising from the emissions-out and the emissions-in sensitivity simulations were very similar, with the emissions-in simulations resulting in slightly enhanced sulfate production, probably due to the increased availability of oxidants. During the four month BRAVO period, the overall relative enhancements associated with the emissions-in simulations were 2.8% for Mexican sources, 7.4% for Texas sources, 3.4% for eastern U.S. sources, 6.7% for western U.S. sources, and 4.2% for the boundary concentrations. This indicates that, in general, REMSAD behaves linearly with regard to gross emission modifications.