The atmospheric chemistry of cities depends strongly on the concentration of the hydroxyl radical (OH). OH removes organic molecules (τ = minutes - days), nitrogen oxides (NOx: NO + NO2; τ = 2 - 20 h), and sulfur dioxide (τ = 2 - 3 d) from the atmosphere and forms ozone and particulate matter in the process. The spatial and temporal variability of OH is not fully understood due in large part to its strong nonlinear dependence on the concentration of NOx. In this dissertation, I use measurements of the NO2 column from the Ozone Monitoring Instrument (OMI), a satellite-based UV/Visible spectrometer, and calculations from chemical transport models to investigate the relationship of OH and NOx in urban regions. I characterize the instrumental and model performance necessary to accurately infer OH concentration from space-based measurements of the NO2 column. I use the OMI observations and the WRF Chem model to investigate the spatial variability of the NO2 column over a variety of emission sources. I find, not surprisingly, that a result of the nonlinear dependence of OH on NOx is that the lifetime of NOx depends on the spatial distribution of NOx. Where the nonlinear NOx-OH feedbacks are most important, I find that a spatial resolution of 4 - 8 km is necessary for both model simulations and measurements to accurately capture variations of the NO2 column. Although this result is expected, quantitative analyses of the nonlinear coupling were not previously available and the calculations presented in this dissertation serve as a guide to interpretation of coarse-resolution measurements and models. I investigate the response of the NO2 column observed by OMI to variations of wind speed and day-of-week emission patterns. I find that these high spatial resolution measurements (13 × 24 km2 at nadir) contain independent information on both NOx emissions and removal.