Impact of spatial and temporal distribution of ammonia emissions from chemical fertilizer usage on predictions of air quality in the Midwest United States

Ammonia (NH3) is a precursor to ambient particulate matter (PM). PM is regulated to protect human health and visibility. Atmospheric deposition of NH3 and other reactive nitrogen compounds exacerbates surface water eutrophication, soil acidity and ecosystem damages. Accurate prediction of impacts of NH3 emissions on PM formation and reactive nitrogen deposition using chemical transport models (CTMs), require representative NH3 emission inventories that capture source specific spatial and temporal variability. Of particular interest are NH3 emissions from chemical fertilizer usage for agriculture that are highly uncertain due to dependence on local soil properties, weather conditions, farm-scale nitrogen management and other agricultural practices. The 2011 National Emission Inventory (NEI) has identified chemical fertilizer usage as the dominant source of anthropogenic NH3 emissions (> 40%) in the Midwest United States (Midwest). In this dissertation, a new approach to characterize NH3 emissions at high-spatial and high-temporal resolutions was developed and evaluated, as detailed in three objectives: First, a new hybrid emissions inventory was developed by combining a new spatial allocation method, called Improved Spatial Surrogate (ISS), combined with daily temporal allocation of emissions based on the biogeochemical Denitrification Decomposition (DNDC) model. In ISS, NH3 emissions from chemical fertilizer usage are gridded at 4 km x 4 km and 12 km x 12 km resolutions, using detailed crop maps and crop specific nitrogen loading data. Daily variations in NH3 emissions, characterized by DNDC capture influences of local soil and weather conditions and crop management practices. The ISS-DNDC approach captured larger spatial heterogeneity of NH3 emissions at sub-county resolution and identified seasonal emission peaks corresponding to chemical fertilizer usage in early summer and late fall, for the Midwest, which are not identifiable in the typical CTM inputs developed using the 2011 and earlier NEI modeling platforms. Second, performance and sensitivity of the DNDC model to input parameters were evaluated and DNDC prediction uncertainty was estimated for NH3 fluxes. This study is the first to evaluate DNDC predictions of NH3 fluxes from chemical fertilizer usage, over a growing season of an intensively fertilized crop, by comparison to measurements using the relaxed eddy accumulation (REA) method. For conditions in Central Illinois (located in the Midwest), DNDC replicated REA measured NH3 fluxes, with greater accuracy for a time period of 33 days after fertilizer usage (correlation coefficient (r) = 0.86 to 0.91), when NH3 fluxes were mostly from land to the atmosphere, in comparison to the entire growing season (r = 0.61 to 0.70), suggesting model deficiency in predicting depositional fluxes. Based on Monte Carlo simulations, predicted NH3 fluxes were identified to be sensitive to inputs of air temperature, precipitation, soil organic carbon, field capacity, soil pH, tilling date and nitrogen loading rate, timing and fertilizer application depth. By constraining variability of these inputs for observed conditions in Central Illinois, uncertainty in predicted NH3 fluxes was estimated between -87% to 61%, indicating need for denser measurements to cover the spatial heterogeneity of influential variables, and to support further model development, for regional upscaling of NH3 emissions. Third, the ISS-DNDC approach was evaluated for predictions of ambient NH3 and PM2.5 (PM with diameter < 2.5 µm), concentrations, for the Midwest, in May 2011, using the Comprehensive Air Quality Model with Extensions (CAMx). CAMx predictions obtained using ISS-DNDC emissions were compared to predictions obtained using baseline emissions from 2011 NEI modeling platform and evaluated for closure with ground-based observations. CAMx consistently underpredicted NH3 concentrations, potentially due to underestimates in NH3 emissions based on fertilizer sales data, or from differences in time scales of removal of NH3 through chemical reactions and deposition within CAMx and in the ambient environment. Implementation of ISS-DNDC emissions at 4 km x 4 km grid resolution, reduced bias in predictions of ambient NH3 concentrations by 33% compared to the baseline scenario. Implementation of ISS-DNDC emissions at 12 km x 12 km increased correlations with observations for PM2.5 by 24% compared to the baseline scenario. Resolving identified uncertainties in predictions of ambient NH3 and PM2.5 concentrations will further require: a) a combination of data and data assimilation methods to constrain total NH3 emissions based on ground-based, aerial and satellite observations and through implementation of biogeochemical models and b) closer investigation of CAMx behavior regarding NH3 consumption through chemical reaction and deposition; at different CTM grid resolutions