<p>Tropospheric ozone simulations have large uncertainties, but their biases, seasonality and trends can be improved with more accurate estimates of precursor gas emissions. We perform global top-down estimates of monthly NO<sub><i>x</i></sub> emissions using two OMI NO<sub>2</sub> retrievals (NASAv3 and DOMINOv2) from 2005 to 2016 through a hybrid 4D-Var/mass balance inversion. The 12-year averages of regional NO<sub><i>x</i></sub> budgets from the NASA posterior emissions are 37 % to 53 % smaller than the DOMINO posterior. Compared to surface NO<sub>2</sub> measurements, GEOS-Chem adjoint NO<sub>2</sub> simulations using the DOMINO posterior NO<sub><i>x</i></sub> emissions have smaller biases in China (by 15 %) and the US (by 22 %), but NO<sub>2</sub> trends have more consistent decreases (by 26 %) with the measurements (by 32 %) in the US from 2006 to 2016 when using the NASA posterior. The two posterior NO<sub><i>x</i></sub> emissions datasets have different strengths with respect to simulation of ozone concentrations. Simulations using NASA-based emissions provide better agreement for polluted conditions. Ozone from these shows the highest correlation (R<sup>2</sup> = 0.88) with annual MDA8 ozone trends and alleviates the double peak in the prior simulation of global ozone seasonality. The DOMINO-based emissions, on the other hand, are better for simulating ozone at remote sites, making them well-suited to generation of boundary conditions for regional models, and in capturing the interannual variability of daytime-average (R<sup>2</sup> = 0.72–0.81) and 24-hour average (R<sup>2</sup> = 0.88–0.96) surface ozone. We recommend using NO<sub><i>x</i></sub> emission datasets that have the best performance in the corresponding spatial domain and temporal focus to improve ozone simulations.</p>