During 5 different periods between summer 2009 and spring 2011, mobile measurements were carried out in the city of Aachen, Germany, in order to capture the spatial variability of particulate matter concentrations in urban and suburban environments. Results show a large spatial variability on a scale of tens of meters, mainly depending on traffic density and building structure. Spatial coefficients of variation exhibit larger spatial variability for PM10 than for PM2.5 and larger variability in traffic influenced inner city environments than in suburban areas. Based on the results of an extensive campaign, a regression model is developed for the prediction of PM10 and PM2.5 distributions over the city area. The three predictors for the regression model are an exponential PM concentration profile simulated on the basis of PM10 and PM2.5 traffic emissions, building density and green area density within radii of 50m and 100m. The model shows good agreement between measured and modeled PM levels during the campaign used for the model training with R2 values of 0.79 and 0.65, RMSE of 1.9μg/m3 and 1μg/m3 for PM10 and PM2.5, respectively. The model is further validated using data from the remaining measurement campaigns and modeling of PM levels at monitoring sites that were not used for the training of the regression model. For the total number of 59 monitoring sites, the regression model shows R2 values of 0.77 (PM10) and 0.61 (PM2.5) with RMSE of 2.3μg/m3 and 1.2μg/m3. The modeled concentrations are generally in better accordance with measured concentrations for PM10 than for PM2.5 concentrations. We attribute this to higher spatial homogeneity of PM2.5 levels compared to coarse particles. Inner city PM levels at traffic influenced sites are better reproduced by the model than suburban concentrations which exhibit the smallest spatial variability.
Read full abstract