资源环境科技发展态势分析平台

Background: Nitrogen dioxide (NO2) poses substantial public health risks in large cities globally. Concentrations of NO2 shows high spatial variation, yet intra-urban measurements of NO2 in Chinese cities are sparse. The size of Chinese cities and shortage of some datasets is challenging for high spatial resolution modelling. The aim here was to combine advantages of dispersion and land-use regression (LUR) modelling to simulate population exposure to NO2 at high spatial resolution for health burden calculations, in the example megacity of Guangzhou. Methods: Ambient concentrations of NO2 simulated by the ADMS-Urban dispersion model at 83 ?virtual? monitoring sites, selected to span both the range of NO2 concentration and weighting by population density, were used to develop a LUR model of 2017 annual-mean NO2 across Guangzhou at 25 m25 m spatial resolution. Results: The LUR model was validated against both the 83 virtual sites (adj R-2: 0.96, RMSE: 5.48 ?g m(?3); LOOCV R-2: 0.96, RMSE: 5.64 ?g m(?3)) and, independently, against available observations (n=11, R-2:: 0.63, RMSE: 18.0 ?g m(?3)). The modelled population-weighted long-term average concentration of NO2 across Guangzhou was 52.5 ?g m(?3), which contributes an estimated 7270 (6960?7620) attributable deaths. Reducing concentrations in exceedance of the China air quality standard/WHO air quality guideline of 40 ?g m(?3) would reduce NO2-attributable deaths by 1900 (1820?1980). Conclusions: We demonstrate a general hybrid modelling method that can be employed in other cities in China to model ambient NO2 concentration at high spatial resolution for health burden estimation and epidemiological study. By running the dispersion model with alternative mitigation policies, new LUR models can be constructed to quantify policy effectiveness on NO2 population health burden.