5.3
Design of Emission Data Processing for Including Emissions Uncertainties
Marc R. Houyoux, MCNC, Research Triangle Park, NC; and C. Frey, D. Loughlin, A. Holland, and G. Cano
The air quality modeling community widely acknowledges large uncertainties in emission inventories used for modeling, but these uncertainties are typically ignored for air quality modeling applications. Three reasons for ignoring the uncertainties are (1) a lack of data to quantify these uncertainties, (2) no available emissions data processor that can handle such data if it were available, and (3) lack of a methodology for addressing the previous two issues. This paper describes a methodology for integrating uncertainty into emissions data processing, summarizes the types of data required, and includes preliminary results from implementation of our method in an emissions data processor.
Our approach assumes uncertainties will be incorporated into air quality modeling analyses using Monte Carlo simulation. By integrating Monte Carlo into our approach, we are able use existing air quality models without modification to those models. In our approach, the emissions processor accepts data that describes the uncertainty in the emission factors and activity data on which the emission inventory is based. The approach also permits input of data describing uncertainty about other ancillary emission inputs, such as temporal adjustment factors, spatial allocation factors (gridding surrogates), and chemical speciation factors. The emission processor propagates these uncertainties through main stages of emissions data processing: inventory import, temporal allocation, spatial allocation, chemical speciation, and merging. The resulting model-ready emissions data is output as a collection of files instead of a single file, in which each file is used for one realization of the Monte Carlo simulation.
Our approach is being implemented into the Sparse Matrix Operator Kernel Emissions (SMOKE) modeling system. In this paper, we describe our general approach and expected benefits for air quality analyses. We also discuss details of the implementation, and present preliminary statistical results of the impact on emissions processing using our approach. These results are based on actual data that has been gathered to describe uncertainty in emission factors.
Session 5, concentration fluctuations, model uncertainty and evaluation
Wednesday, 22 May 2002, 9:00 AM-1:30 PM
Previous paper Next paper