Application of land use regression techniques for urban greening: An analysis of Tianjin, China

Abstract

The land cover types in a 4 km by 4 km space centered around an air quality monitoring station in the city of Tianjin were classified using the e-Cognition software. We identified 23 air quality monitoring sites and their surrounding land cover types. The suitability of using land cover types to predict traffic-related air pollution Nitrogen Dioxide (NO2) was tested through a machine learning land use regression (LUR) modeling technique. We found that vegetation was significantly but negatively associated with air pollution levels (correlation coefficient r = −0.5, p < 0.01) while highways and major roadways were positively and significantly associated with air pollution levels (r = 0.60, p < 0.01). The LUR model explained 84% of the total variance in measured NO2 concentrations. The associations of the land cover types with NO2 concentrations were then used in land cover retrofitting strategies to quantify the potential for reducing traffic-related air pollution. We identified that the improvements in air quality could reach 25% if specific urban greening strategies were fully implemented. The results of the study can help policy makers and environmental designers adopt effective air pollution reduction policies to improve air quality and protect public health in Tianjin and other urban regions in China.