Nitrogen dioxide (NO₂) for the city of Plovdiv was estimated as an exposure surface using a land use regression model (LUR) developed for Plovdiv within the project. Briefly, a campaign was carried out to measure the concentrations of NO₂ in Plovdiv using passive diffusion tubes. The measurements were carried out at 40 points, and, given the resources available at that time, a monthly measurement was carried out in 26 points, and the remaining 14 were divided into two groups of seven, and for each group a measurement was carried out every other month. We conducted monthly measurements for a continuous six-month period in the second half of 2023. The tubes were placed on pillars of road signs, street lamps and others, at a height (~2.6 m), with distances from the roadway and intersections, in accordance with Directive 2008/50 of the European Union, Section A of Annex III [Directive 2008/50/EC]. These pillars were chosen so that they were close to the facades of the buildings, without the movement of air being affected by the aerodynamic shadow of the latter. To determine the spatial distribution of NO₂ throughout the territory of Plovdiv, a LUR was developed, based on a step-by-step manually supervised linear regression model using the method of least squares [Briggs 1997]. The model predicted the measured concentrations of NO₂ using information about the road infrastructure surrounding each measuring point and areas with different land uses in a coordinate system EPSG:7801 - BGS2005 / CCS2005. These environmental characteristics were derived for different Euclidean buffer radii around each point, from 50 to 1000 m. Sources of this information were OpenStreetMap (October 2023, https://www.openstreetmap.org) and the European Urban Atlas. The approach for building the model followed the ESCAPE manual [Brauer 2003; Cyrys 2012]. The LUR model explains the measured NO₂ values at 86% , and its tenfold cross-validation yields R² = 0.84; mean absolute error = 4.08 μg/m³; and RMSEA 5.02 μg/m³. The spatial autocorrelation test of the model residues showed that they were randomly distributed (Moran’s I = -0.033, z = -0.149, p = 0.882).

The NO₂ map showing the mean annual average spatial variation in NO₂ in Plovdiv in 2023 was created by generating a regular grid of points at a distance of 25 m from each other, covering the territory of Plovdiv, after which the values of the predictors from the final model for NO₂ were added to each point. The range of values of the predictor variables was limited to the observed one in the created model for NO₂, so as not to extrapolate NO₂ beyond the range for which the model has been validated. After that, the regression equation was applied to each point of the grid to calculate NO₂, and the calculated values were interpolated (rasterized) in a continuous surface within the boundaries of Plovdiv.

Downloads

• NO2_Plovdiv.tif