Land use Regression as Method to Model Air Pollution. Previous Results for Gothenburg/Sweden

In the past 20 years, considerable progress has been made to improve urban air quality in the EU. However, road traffic still contributesconsiderably to the deterioration of urban air quality to below standards, which requires a method to measure properly and model pollutionlevels resulting from road traffic. In order to visualize the geographical distribution of pollution concentration realistically, we applied the LandUse Regression (LUR) model to the urban area of Gothenburg.The NO2 concentration was already obtained by 25 samplers through the urban area during 7-20 May, 2001. Predictive variables such asaltitude, density, roads types, traffic and land use were estimated by geographic information system in buffers ranging 50 to 500 m-radii. Linearregression (α=5%) between NO2 and every predictive variable was calculated, and the most robust variables and without collinearity variableswere selected to the multivariate regression model. The final formula was applied using Kriging in a grid map to estimate NO2 levels.The average of measurements was 23.5 μg/m\ub3 (\ub1 6.8 μg/m\ub3) and 180 predictive variables were obtained. The final model explained 59.4% ofthe variance of NO2 concentration with presence of altitude and sum of traffic within 150 m around the sampler sites as predictor variables. Thecorrelation measured versus predicted levels of NO2 was r = 0.77 (