Emissions Modeling for Smraq: a Seasonal and Regional Example Using Smoke

The Seasonal Model for Regional Air Quality (SMRAQ) project is an effort to model ozone formation and transport from mid-May to mid-September in 1995 for the Eastern two-thirds of the continental United States. The modeling grid is 48 columns by 50 rows at a 54 kilometer resolution. The models used for the project are the Multiscale Air Quality SImulation Platform (MAQSIP) for ozone chemistry and transport, MM5 for meteorology, and the Sparse Matrix Operator Kernel Emissions (SMOKE) modeling system for emissions. The emissions inventory was derived from the Ozone Transport Assessment Group (OTAG) inventory for 1995. This paper describes the experiences and insight gained from emissions processing for the SMRAQ project. The huge emissions inventory required by the regional and seasonal features of SMRAQ unfavorably impacts emissions processing time, disk storage, and computer memory requirements. Additionally, the regional domain forces consideration of time zones, gridded meteorological data (such as temperature and wind), and other spatial data (such as land use and climate). These issues are described here in relation to the approach used for SMRAQ emissions processing. The improved computational efficiency provided by the SMOKE modeling system, which uses sparse matrix algebraic techniques to reduce both computational and storage requirements, permits efficient emissions processing for seasonal modeling of regional ozone formation and transport. Comparisons between SMOKE and EMS-95 are described that indicate SMOKE requires 32 times less CPU resources than EMS-95 for a five-day case based on the OTAG 1990 inventory. For the 120 day SMRAQ episode, SMOKE is estimated to be able to process all emissions on an IBM 590 workstation in 26.7 CPU hours.