Modelling the vertical gradient of nitrogen dioxide in an urban area

Land use regression models environmental predictors to estimate ground-floor air pollution concentration surfaces of a study area. While many cities are expanding vertically, such models typically ignore the vertical dimension. We took integrated measurements of NO2 at up to three different floors on the facades of 25 buildings in the mid-sized European city of Basel, Switzerland. We quantified the decrease in NO2 concentration with increasing height at each facade over two 14-day periods in different seasons. Using predictors of traffic load, population density and street configuration, we built conventional land use regression (LUR) models which predicted ground floor concentrations. We further evaluated which predictors best explained the vertical decay rate. Ultimately, we combined ground floor and decay models to explain the measured concentrations at all heights. We found a clear decrease in mean nitrogen dioxide concentrations between measurements at ground level and those at higher floors for both seasons. The median concentration decrease was 8.1% at 10 m above street level in winter and 10.4% in summer. The decrease with height was sharper at buildings where high concentrations were measured on the ground and in canyon-like street configurations. While the conventional ground floor model was able to explain ground floor concentrations with a model R2 of 0.84 (RMSE 4.1 μg/m3), it predicted measured concentrations at all heights with an R2 of 0.79 (RMSE 4.5 μg/m3), systematically overpredicting concentrations at higher floors. The LUR model considering vertical decay was able to predict ground floor and higher floor concentrations with a model R2 of 0.84 (RMSE 3.8 μg/m3) and without systematic bias. Height above the ground is a relevant determinant of outdoor residential exposure, even in medium-sized European cities without much high-rise. It is likely that conventional LUR models overestimate exposure for residences at higher floors near major roads. This overestimation can be minimized by considering decay with height. [Display omitted] •NO2 concentrations are typically 8 to 10% lower 10 m above the street than at street level.•The higher the NO2 concentration at ground level, the sharper the vertical decay.•Canyon-like street configurations trap NO2 at ground level, causing sharper vertical decay.•Land use regression (LUR) models can consider decay by including height as a predictor.•When predicting NO2 at higher floors, 3D LUR models outperfor