Emissions of ozone precursors from stationary sources: a critical review

This paper discusses and critiques methods used to estimate emissions of, and create both aggregate and detailed modeling inventories for, nitrogen oxides (NOx), volatile organic compounds (VOC) and carbon monoxide (CO), the main pollutants involved in ozone formation. Emissions of sulfur dioxide (SO2) and methods to project emissions into the future are also briefly discussed. Many improvements have been made in emissions inventories over the past decade. For example, the required use of continuous emission monitors (CEMS) has produced site-specific emissions estimates from almost all US electric utility power plants, which are the major stationary source of NOx. However, many data quality issues remain. For example, the overall quality of standardized emission factors is very poor. In addition, uncertainties have been introduced by use of simplistic assumptions on the existent level of emission control. Even the use of GEMS has not eliminated uncertainty in emissions from power plants, because methods to deal with missing data can introduce bias. Emissions data for Mexico are not comprehensive, making ozone modeling in US border regions difficult. Data for VOC speciation is outdated, and crude data is often used to disaggregate emissions to the fine level of spatial and temporal detail needed for atmospheric modeling. It is difficult to make general statements about the importance of each of these problems, because there are no reliable estimates of the overall uncertainty of emissions values, and because the impact of emission inventory errors is very site specific. The Emissions Inventory Improvement Program (EIIP) initiated by the US Environmental Protection Agency promises to enhance the quality of future inventories, mainly through communication of best practices among state agencies. Further inventory improvement efforts must be focused on problems that most strongly influence poor prediction of ozone concentrations. Targets for improvement could be based on comparison of photochemical modeling results to observed concentrations, coupled with other techniques that better explain source-receptor relationships. (C) 2000 Elsevier Science Ltd. All rights reserved.