NARSTO Workshop 2003: Inverse Modeling To Estimate NH3 Emission SeasonalityAnd The Sensitivity To Uncertainty Representations

NARSTO Workshop on Innovative Methods
for Emission Inventory Development and Evaluation
University of Texas, Austin
October 14-17, 2003

Inverse Modeling To Estimate NH3 Emission SeasonalityAnd The Sensitivity To Uncertainty Representations

Alice B. Gilliland¹, Hae-Kyung Im², Michael L. Stein²
¹ NOAA Air Resources Laboratory, Atmospheric Sciences Modeling Division(on assignment to USEPA, National Exposure Research Laboratory)
² University of Chicago Department of Statistics

Inverse modeling has been used extensively on the global scale to produce top-down estimates of emissions for chemicals such as CO and CH4. Regional scale air quality studies could also benefit from inverse modeling as a tool to evaluate current emission inventories; however, underlying assumptions such as the linearity between emission and concentration changes can limit the applicability of inverse modeling. Ammonia (NH3) has been found to be a reasonable candidate because a strong linearity exists between NH3 emission adjustments and the response of modeled ammonium wet deposition. Further, the uncertainty in the emission estimates, especially on a monthly time scale, is quite large. While we anticipate that NH3 emissions from agricultural non-point sources have a strong seasonal pattern, the intra-annual variability of these primary NH3 sources is not yet understood well-enough to incorporate into current NH3 emission inventories. Along with the USEPA Community Multiscale Air Quality (CMAQ) model and NH4+ wet concentration data, an inverse modeling approach has been used to estimate monthly adjustments to the NH3 emission field over the Eastern United States. The first series of results, presented in Gilliland et al. [2003], offer the most comprehensive estimate of seasonal NH3 emission variability to date. These seasonal variations in NH3 emissions were shown to be essential for the prediction of nitrogen-containing compounds in that study. Further tests are now being conducted where a variety of uncertainty representations are considered in the inverse modeling calculations. These sensitivity tests will provide a more thorough range of emission adjustment estimates for each month and will test the rigor of the seasonal variability suggested by Gilliland et al. [2003].