Estimating Ground-level NO₂ Concentrations Based On Geographically And Temporally Weighted Regression Model

China has been suffering from serious air pollution,which has imposed a bad influenced on human's health.NO₂ is an important air pollutant in atmosphere,and it can cause photo-chemical smog,acid deposition and atmospheric ozone variation.Environmental Protection Department of our country has built1497 stations to monitor the air pollution in different cities,and NO₂ has been one of the regular monitoring item.Ground-level NO₂ is closely related with human's health.Although ground-based monitoring stations can give relatively accurate measurements,it is not appropriate to use them in large scale research.Now,there has been scientific theory and mature algorithm for the retrieval of NO₂ concentrations from satellite data.Using satellite data to study the change of NO₂ in different scales has become an effective method.We used tropospheric NO₂ columns from OMI,meteorological data from WRF model and ground-based NO₂ measurements to analyze the relationship between tropospheric NO₂ columns,meteorological data and ground-based measurements.We used GTWR model and adaptive bandwidth to estimate ground-level concentrations,and compared GTWR result with the result of OLR,GWR,TWR,and computed population-weighted ground-level NO₂ concentration based on GTWR,and found: 1)GTWR had the best performance with the highest correlation coefficient and the lowest errors;2)TWR had a better performance than GWR,which represented that the relationship between NO₂ columns and ground-level NO₂ showed a bigger temporal variation;3)The performance of GTWR showed obvious seasonal difference and geographical difference.The estimation in winter had the best relationship with observed NO₂ concentrations while the estimation in summer had a worst one.The NO₂ columns in autumn had the best relationship with ground-level NO₂.3)Compared with the ordinary mean values of NO₂,population-weighted NO₂ concentrations could show the exposure of population to NO₂ pollution.The population-weighted NO₂ concentrations in Beijing-Tianjin-Hebei region,Shaanxi,Henan,Shandong was high and most provinces in eastern China has been influenced by NO₂ pollution in varying degrees.We also introduced other three method to estimate ground-level NO₂ concentrations: interpolation method,neural network and simulation from atmospheric chemical model WRF-Chem model.1)Interpolation method was simple and it could give the spatial distribution of ground-level NO₂ concentrations.But it was not easy to get accurate values or give details where monitoring stations were very sparse or not available.2)Neural network had a better performance than OLR model but it had a worse performance than GTWR.3)WRF-Chem could simulate the ground-level NO₂ and the vertical distribution of NO₂,based on which tropospheric NO₂ columns could be computed.The ground-level NO₂ concentration and NO₂ columns from WRF-Chem had a limited relation with ground-based measurements?R2=0.23,0.17 respectively?.