A Study On Refined Exposure Assessment Of PM_2.5 Based On Land Use Regression Model

In the context of China’s rapid urbanization,high-density urban environments are resulting in increasingly complex scenarios of air pollution exposure assessment.However,since the traditional exposure assessment methods cannot fully adapt to the spatial-temporal dynamics of the current urban population and the spatial-temporal variation of air pollution.Therefore,this study makes efforts to improve exposure assessment in two aspects:adapting to the spatio-temporal dynamics of the population and revealing vertical exposure differences.On this basis,refined air pollutant exposure assessment models with spatio-temporal adaptation are constructed and evaluated.This will provide a more accurate scientific basis for air pollution health risk prevention in urban areas in China.The core urban area of Taiyuan,Shanxi Province,with many high-rise buildings,was selected as the study area.Surface concentration of PMi,PM2.5,PM 10,NO2 and BC pollutants were collected by vehicle-based mobile monitoring and vertical spatial data of these air pollutants were obtained by UAV vertical monitoring in multiple periods from 2019 to 2020.High-resolution ground-level air pollutant exposure assessment models were developed using a supervised forward multiple regression algorithm by combining regional traffic road network,land use,meteorological and ambient air monitoring data,etc.The vertical variation model of PM2.5 was developed using a spline smoothing function.On this basis,the three-dimensional land use regression model(3-D LUR)was proposed.The following are the major results:(1)In this study,a diurnal LUR modeling strategy and a refined dynamic exposure assessment method based on the diurnal LUR model were proposed.In the case of PM2.5,their improvements were evaluated separately in three exposure scenarios:short-term,seasonal and annual.The diurnal LUR modeling strategy showed the potential to improve the performance of the LUR models in the three exposure scenarios.The refined dynamic exposure assessment method can avoid exposure assessment errors within 15%,20%and 30%on average in short-term,seasonal and annual scenarios,respectively,and performs best in longer-term exposure scenarios.(2)Seasonal LUR models of PM1 and BC were developed based on the diurnal LUR modeling strategy.The tempo-spatial variations and differences in drivers of PM1 and BC within the Taiyuan urban area were revealed.It is found that the spatial and temporal variation of PM1 was mainly the variation of concentration,which was more influenced by meteorological factors.In contrast,BC was mainly the variation of spatial distribution,which was dominated by traffic factors.(3)A 3-D LUR model incorporating the nonlinear vertical variation of PM2.5 was proposed,explaining 53%to 62%of the horizontal variation and 91%of the vertical variation of the pollutant.This 3-D exposure assessment model can avoid a maximum of 5%concentration overestimation and 4%concentration underestimation.The results of the application of this model found 19.7 premature deaths per 10,000 adults attributable to PM2.5 exposure,and the difference with the traditional two-dimensional assessment was mainly in the vertical direction.