A four-dimensional data assimilation method for air quality modeling

In this thesis the development and testing of a Four-Dimensional Data Assimilation (FDDA) technique for comprehensive air quality models (AQMs) is addressed. The method is used as a top-down tool to validate the emissions inventories used in model applications, based on ambient measurements. The FDDA approach consists of an iterative optimization procedure where an AQM is coupled to an inverse model, and adjustment of the emissions minimizes the difference between ambient measurements and model-derived concentrations. Formal direct sensitivity analysis of three-dimensional AQMs is used to derive the response fields that are used by the inverse modeling technique to suggest adjustments in emission strengths.; The method was tested using a modeling domain centered about Atlanta's metropolitan area and with pseudo-observations of gas-phase species. Perturbed emissions were adjusted about their known original value in the test scenarios. Next, proposed changes to NO_x, volatile organic compounds (VOC) and CO emissions from different sources in Atlanta were estimated using actual observations. The speciation profiles of VOC emissions were also analyzed, and the effect of the adjustments applied to VOC emissions on possible changes in ambient VOC reactivity was computed. The change in reactivities has direct implications on VOC control strategy selection. Finally, the FDDA technique was used to derive emissions adjustments for both ozone and particulate matter (PM) precursors simultaneously using a modeling domain encompassing the eastern United States. Emission adjustments for three different episodes were estimated. Sources investigated included categories of NO_x, VOCs, CO, SO ₂, NH₃ and fine organic and elemental carbon aerosol emissions. Ambient measurements used include gas-phase inorganic and organic species, and speciated fine PM.; In general, the results obtained here are consistent with what are viewed as the current deficiencies in emissions inventories as derived by other techniques, such as tunnel studies. Results indicate that, for the base case inventories used here, emissions of SO₂ and CO appear to be estimated reasonable well, while NO_x, VOC, NH₃ and organic and elemental fine PM emissions require more significant revision. Convergence and stability properties of the method are also discussed, as well as implications on the quality of the ambient data used to drive the computations.