Developing a land use regression model for NO2 concentrations in Guangzhou, China
Humstad, Kamilla Haugdahl
Health impacts of air pollution are widely recognised where exposure to NO2 increases the risk of respiratory diseases and mortality. High levels of ambient NO2 are common in megacities because of the rapid economic and industrial growth. Due to a large heterogeneity in observed pollutant concentrations, modelling concentrations is traditionally complex. Land use regression (LUR) techniques with the use of GIS have been developed where measurements of NO2 are utilised together with predictor variables to predict NO2 concentrations at unsampled locations. In this work a LUR model has been developed for Guangzhou, southern China. The resulting model was able to explain 96% of the variance in NO2 concentrations effectively demonstrating strong predictive power. The predictor variables included in the final model were green space in a 5000 m buffer, emissions from major roads in a 300 m buffer, green space in a 100 m buffer and the length of all roads in a 50 m buffer. Model validation presented low error in value prediction and confirmed model prediction ability. The LUR model was utilised to quantify the number of attributable deaths from exposure to NO2. A total of 7515 people was estimated to experience premature deaths of the total annual deaths in Guangzhou. The LUR model developed represents a highly relevant and valuable tool for assessing air quality issues in Guangzhou.