Satellite remote sensing of tropospheric nitrogen dioxide (NO₂) can provide valuable information for estimating surface nitrogen oxides (NO_x) emissions. Using an exponentially modified Gaussian (EMG) method and taking into account the effect of wind on observed NO₂ distributions, we estimate 3-year moving-average emissions of summertime NO_x from 35 US (United States) urban areas directly from NO₂ retrievals of the Ozone Monitoring Instrument (OMI) during 2005–2014. Following conclusions of previous studies that the EMG method provides robust and accurate emission estimates under strong-wind conditions, we derive top-down NO_x emissions from each urban area by applying the EMG method to OMI data with wind speeds greater than 3–5 m s⁻¹. Meanwhile, we find that OMI NO₂ observations under weak-wind conditions (i.e., < 3 m s⁻¹) are qualitatively better correlated to the surface NO_x source strength in comparison to all-wind OMI maps; therefore, we use them to calculate the satellite-observed NO₂ burdens of urban areas and compare with NO_x emission estimates. The EMG results show that OMI-derived NO_x emissions are highly correlated (R > 0.93) with weak-wind OMI NO₂ burdens as well as with bottom-up NO_x emission estimates over 35 urban areas, implying a linear response of the OMI observations to surface emissions under weak-wind conditions. The simultaneous EMG-obtained effective NO₂ lifetimes (~ 3.5 ± 1.3 h), however, are biased low in comparison to the summertime NO₂ chemical lifetimes. In general, isolated urban areas with NO_x emission intensities greater than ~ 2 Mg h⁻¹ produce statistically significant weak-wind signals in 3-year average OMI data. From 2005 to 2014, we estimate that total OMI-derived NO_x emissions over all selected US urban areas decreased by 49 %, consistent with reductions of 43, 47, 49, and 44 % in the total bottom-up NO_x emissions, the sum of weak-wind OMI NO₂ columns, the total weak-wind OMI NO₂ burdens, and the averaged NO₂ concentrations, respectively, reflecting the success of NO_x control programs for both mobile sources and power plants. The decrease rates of these NO_x-related quantities are found to be faster (i.e., −6.8 to −9.3 % yr⁻¹) before 2010 and slower (i.e., −3.4 to −4.9 % yr⁻¹) after 2010. For individual urban areas, we calculate the R values of pair-wise trends among the OMI-derived and bottom-up NO_x emissions, the weak-wind OMI NO₂ burdens, and ground-based NO₂ measurements, and high correlations are found for all urban areas (median R= 0.8), particularly large ones (R up to 0.97). The results of the current work indicate that using the EMG method and considering the wind effect, the OMI data allow for the estimation of NO_x emissions from urban areas and the direct constraint of emission trends with reasonable accuracy.