We developed a method to generate two-week NO2 concentration maps with a high spatial resolution (10 m by 10 m) for the city of Zurich, Switzerland, based on statistical modelling. Our models utilize data from a dense passive diffusion sampler network consisting of 49 sites that measured 14-day mean NO2 concentrations in the year 2008. The regression analysis is based on Generalized Additive Models (GAMs) and a stepwise forward selection algorithm that leads to models relying on a small number of explanatory variables (2–3). The explanatory variables included in the regression analysis are spatially resolved information on traffic and heating systems related NOX and NO2 emissions, respectively, sky view factors, and topography (elevation). Deviance explained of the 26 models ranges from 0.66 to 0.79. 81% of the modelled and 77% of the predicted NO2 concentrations, respectively, deviate less than 25% from the observations. The modelling approach outlined in this paper augments the value of point measurements obtained from urban routine passive diffusion sampler networks by providing spatially resolved concentration fields. The derived maps allow a detailed assessment of NO2 levels in cities and can be used in applications such as public health protection. •We generated highly resolved two-week NO2 maps for the city of Zurich.•The used statistical models rely on NO2 passive sampler data and geoinformation.•Traffic activity is the most important predictor for our NO2 models in Zurich.•The statistical modelling augments the value of pointwise passive sampler data.•We analysed the NO2 concentration field w.r.t. annual limit value exceedances.