AAAR 32nd Annual Conference
September 30 - October 4, 2013
Oregon Convention Center
Portland, Oregon, USA
Abstract View
Spatial and Temporal Assessment of a Hybrid Source Apportionment Model Using Nonlinear Optimization
CESUNICA IVEY, Heather Holmes, Yongtao Hu, James Mulholland, Armistead Russell, Georgia Institute of Technology
Abstract Number: 526 Working Group: Source Apportionment
Abstract A hybrid source apportionment model has been developed that takes into account both observed concentrations and measurement uncertainties to estimate source impacts that better reflect the observations. The model employs an effective variance approach to represent model and measurement uncertainties. Sequential quadratic programming (SQP) is used to estimate source impact adjustment factors by minimizing the error between measured and modeled concentrations. Adjustment factors are then applied to improve the original source impact estimates.
This method is employed to generate daily adjusted source impacts for continental United States for the month of January 2004. First, Chemical Speciation Network (CSN) observation data that are available at every third or sixth day are temporally interpolated to produce daily values. Daily observations and modeled daily source impacts are then used for optimization, generating spatially-sparse, daily datasets of adjustment factors. Finally, a kriging technique is employed to spatially interpolate the adjustment factors to a uniform grid covering CONUS that has a 36-km resolution in horizontal. Resulting datasets include temporally and spatially dense source impacts for 33 source categories including diesel and gasoline vehicles, wood burning, natural gas, coal combustion, etc. Optimization and interpolation methods are evaluated by data withholding and comparison to daily concentrations from the Southeastern Aerosol Research and Characterization (SEARCH) Network and EPA funded supersites, including sites in Atlanta, GA; Birmingham, AL; and St. Louis, MO. These data are independent of those used for model development.
The impact of this work extends to epidemiological studies, where daily source apportionment estimates are utilized in time-series epidemiologic analyses to investigate pollutant emissions and adverse health outcomes. The increased spatial and temporal resolution of air quality metrics improves exposure estimates used in health studies.