Daewon's concern for improving the fidelity between the driving NWP model and the "dependent" AQM simulation had been honed as part of the NAPAP program in the late 1980s and early 1990s, when he and the lead author worked closely together on the RADM model. When Daewon became the lead scientist for CMAQ development, our concern for such fidelity became more formally programmatic. Thus, development within the prototype MAQSIP model for PBL and cloud processes was architected from the outset to be "as coupled as possible" to the "best available" parameterizations within MM5 (McHenry et al.1996; McHenry and Binkowski,1996). This provided more confidence that the AQM could faithfully reproduce the critical PBL and cloud processes *just as they evolved* within the meteorological model, significantly reducing sources of error in AQM results.
Further, in order to meet the anticipated heavy regulatory requirements that CMAQ would face, Daewon was also concerned about the need for improved emissions inventories--along with the need for much faster emissions processing. As a result, Carlie Coats became the primary architect for the SMOKE emissions processing system (Coats, 1996) and Jeff Vukovich used it to develop inventories, initially for regulatory application (Vukovich, 1998). Because the SMOKE software was so much more efficient than previously available emissions processing systems, recently acquired raw inventories could be processed rather quickly into "model-ready" files, resulting in model-ready emissions inputs that could be considered adequate for "real-time operations."
Importantly, one of the core requirements of the overall Models-3 project was the development of a consistent software architecture that was flexible, extensible, and re-usable. This requirement for data-storage and data-exchange standards was encapsulated within the Models-3 I/O API (Coats et al., 1995). In turn,the I/O API was able to provide much more reliable and efficient data flows between all modeling systems involved in the Models-3 paradigm, including not only CMAQ but the emissions and meteorological model "drivers" as well (Coats et al.,1998).
In 1996, the lead author and colleagues submitted a grant proposal to EPA to implement a coupled met-emissions-aqm model in real-time, using the technology that was newly available. Unfortunately, that proposal was roundly rejected as both too complex and unneeded. Reviewers said that the Agency did not have any interest in real-time forecasting of air quality, and would not in the future. Still, both the lead author and Dr. Byun thought that a time would come when operational air quality model-based forecasting would have its day. Indeed, despite lacking a funded project from the EPA to under-gird the effort, Daewon encouraged me to persevere.
During 1997 and 1998, very modest funding along with significant "off-hours" time was acquired in order to support the effort. The very first runs were produced at Penn State University in partnership with Dr.Nelson Seaman, who donated output from his real-time MM5. Ironically, by the fall of 1998, work that had been proposed under the 1996 submitted Grant proposal was completed, a full year ahead of schedule and with almost no funding! Moreover, the initial results for ground level ozone showed both that the new technology was mature enough to support reliable operations and that the predictions were good enough to suggest that NAQP forecasting could become a useful or even routine tool (McHenry et al.,1999).
Once initial word "got out" that numerical air quality predictions were being made, various states and agencies already under requirement to issue public forecasts began hankering to get access to the results. The rest is history. This talk will focus on re-capitulating the development of the initial technology and its initial implementation as an adjunct to Models-3 development, along with the importance of Daewon's perspectives in helping operational NAQP forecasts to become a reality before their time.
Supplementary URL: