The environmental community has developed advanced numerical air quality models (AQMs) to understand the interactions among meteorology, emissions (both manmade and biogenic), and pollutant chemistry and dynamics. Emissions data from emissions models and regulatory inventories are one of the most important inputs for these air quality models. Scientists use air quality modeling for a number of purposes: for state and federal implementation plan development, for research on improved modeling methods, and most recently for air quality forecasting. In all of these cases, the trend has been to model larger regions, at a finer grid resolution, with more emissions sources, and for more purposes (e.g., ozone, particulates, toxics). These needs require a computationally efficient, user-friendly, and flexible emissions data processing system.

The MCNC Environmental Modeling Center (EMC) created the Sparse Matrix Operator Kernel Emissions (SMOKE) Modeling System to allow emissions data processing methods to integrate high-performance-computing (HPC) sparse-matrix algorithms. The SMOKE system is a significant addition to the available resources for decision-making about emissions controls for both urban and regional applications. It provides a mechanism for preparing specialized inputs for air quality modeling research, and it makes air quality forecasting possible. The SMOKE system continues to develop and improve at the University of North Carolina at Chapel Hill’s Institute for the Environment (IE).

The SMOKE prototype, available since 1996, was an effective tool for emissions processing in a number of regional air quality modeling applications. In 1998 and 1999, SMOKE was redesigned and improved with the support of the U.S. Environmental Protection Agency (EPA). The primary purposes of the first SMOKE redesign were support of (1) emissions processing with user-selected chemical mechanisms (described in more detail in Section 2.10, “Chemical speciation processing”) and (2) emissions processing for reactivity assessments (described in Section 2.13.3, “Creating the reactivity control matrix”). In 2002, SMOKE was enhanced to support driving the MOBILE6 model used to create on-road mobile emission factors and to support on-road and nonroad mobile toxics inventories, resulting in SMOKE v1.5. In 2003, SMOKE v2.0 was created to include all toxic inventories, including point and nonpoint (stationary sources reported at the county level) sources. SMOKE v2.1, released in 2004, included updated versions of the BEIS3 and MOBILE6.2 models, the ability to use humidity data when processing on-road mobile sources with MOBILE6, and support for polar stereographic output grids. In 2009, SMOKE v2.6 enhanced the processing of fire data, streamlined the processing for CAMx and UAM models, and added a new approach to processing CEM data. Version 2.7, SMOKE was enhanced to support processing the MOVES (MOter Vehicle Emission Simulator) emission rates output through the SMOKE system to model on-roadway and off-network mobile sources. As of version 4.0 SMOKE is enhanced to support hemispheric modeling by enabling the processing of global gridded emission inventories, such as EDGAR, for input to chemistry-tranport models.

SMOKE can process criteria gaseous pollutants such as carbon monoxide (CO), nitrogen oxides (NO_x), volatile organic compounds (VOC), ammonia (NH₃), sulfur dioxide (SO₂); particulate matter (PM) pollutants such as PM 2.5 microns or less (PM_2.5) and PM less than 10 microns (PM₁₀); as well as a large array of toxic pollutants, such as mercury, cadmium, benzene, and formaldehyde. In fact, SMOKE has no limitation regarding the number or types of pollutants it can process.

The purpose of SMOKE (or any emissions processor) is to convert the resolution of the emission inventory data to the resolution needed by an air quality model. Emission inventories are typically available with an annual-total emissions value for each emissions source, or perhaps with an average-day emissions value. The AQMs, however, typically require emissions data on an hourly basis, for each model grid cell (and perhaps model layer), and for each model species. (Refer to Appendix A. Glossary for definitions of these terms.) Consequently, emissions processing involves transforming an emission inventory through temporal allocation, chemical speciation, and spatial allocation, to achieve the input requirements of the AQM.