Urban air quality is a major concern in the context of human health since cities are at the same time emission hot spots and home to a large fraction of the world's population. Airborne imaging spectrometers may be a valuable addition to traditional air pollution monitoring networks as they can be used to map the spatial distribution of air pollutants such as nitrogen dioxide (NO2) at high spatial resolution. Retrieving NO2 concentrations from such measurements requires information about the average path of the photons collected by the spectrometer, which depends on observation and solar geometry, surface reflectance, and atmospheric scattering by air molecules and aerosols. This is usually accounted for by an air mass factor (AMF) computed with a radiative transfer model (RTM). Since scattering processes are not homogeneous in the atmosphere, so-called box AMFs representing the AMFs for defined spatial grid boxes are calculated before being integrated to a total AMF. The actual 1D-layer AMFs (only account for vertical inhomogeneity of the atmosphere), traditionally used for NO2 retrievals, are not sufficient to resolve the high spatial variability of NO2 concentrations in cities. As a result, measured NO2 distributions of such measurements are much smoother than one would expect from looking at other mapping techniques in cities (e.g., high-resolution dispersion model simulations of NO2). Therefore, I study the impact of 3D and buildings in the radiative transfer calculation on the NO2 concentrations retrieved from ground-based and airborne spectrometers. In this study, AMFs are computed with the MYSTIC solver of the libRadtran RTM. MYSTIC uses a Monte Carlo technique to simulate photons journeys in the atmosphere and retrieve different radiative transfer quantities. The MYSTIC AMF module has been extended to a 3D radiative transfer code and is now able to account for complex ground features (i.e., buildings). With synthetic case studies, I demonstrate the importance of considering 3D features in the radiative transfer calculations. Considering the 3D path of photons in the retrievals affects the spatial structure of the sensitivity of ground-based and airborne instruments. Considering 3D radiative transfer features induces a horizontal smearing of the sensitivity, especially of features with high NO2 concentrations, as for example exhaust plumes or roads. Buildings reduce the instrument sensitivity to the near-surface NO2 and induce noticeable random noise that might explain part of the uncertainty observed in the retrieved NO2 maps. Moreover, I implemented the developed features in the Empa APEX NO2 retrieval algorithm and retrieved vertical column densities and high-resolution near-surface NO2 concentrations (NSC) from the APEX airborne imaging spectrometer. Finally I evaluate the quality of NO2 NSCs maps by comparing the results to in situ measurements. The new modules are essential for analyzing NO2 remote sensing data over cities as they will reduce systematic errors and spatially better allocate measurements. This pioneer research is intended to help the community identifying problems that might appear with the fast increase in horizontal resolution of satellites. The possible high-resolution maps obtained from trace gas remote sensing can support city authorities and different teams in the scientific community to access high-resolution ground NO2 concentration maps, to study urban air quality and to support urban planning.