Nitrogen oxides (NOx = NO + NO2) play crucial roles in the formation of ozone, aerosol, and acid rain which are unfavorable to human health, climate, and ecosystem stabilities. NOx is emitted by both anthropogenic and natural sources, such as fossil fuel combustion, soil bacteria, lightning, etc. Accurate knowledge of NOx emissions is essential for relevant scientific research and air pollution control policies. The primary goal of this thesis is to evaluate current estimates of anthropogenic and natural NOx emissions over the United States and improve model’s prediction of surface ozone concentrations by using a 3-D Regional chEmistry and trAnsport Model (REAM) and various types of observations. The evaluation started from the DISCOVER-AQ 2011 campaign, where we analyzed the diurnal cycles of NO2 and O3 and found that the 2011 National Emission Inventory provided reasonable estimates of anthropogenic NOx emissions at the 36-km resolution but was unable to resolve NOx emission distributions at the 4-km scale over the Baltimore-Washington, D.C. region. We further extended the 36-km evaluation to the contiguous United States through comparisons of simulated and observed O3 peak values and peak time and suggested the underestimation of soil NOx emissions from the widely-used Yienger and Levy scheme and the overestimation of biogenic isoprene emissions from MEGAN. Then, we examined the anthropogenic NOx emission trend from 2003 – 2017 by using NO2 surface measurements and satellite NO2 tropospheric vertical column density (VCD) datasets and recommended the necessity of the selection of urban regions to infer anthropogenic NOx emission trend from satellite VCD products. By the above evaluations of surface NOx emissions, we finally explored the impacts of thunderstorms and lightning NOx on surface O3 concentrations.