Spatial-Temporal Estimation Of PM_2.5 Concentrations In Beijing-Tianjin-Hebei Region Based On Two-Stage Model

In recent years,China's rapid economic development and environment have shown a serious uncoordinated development trend.According to the daily air detection data released by China,the overall air quality in China is relatively poor,especially in the Beijing-Tianjin-Hebei region,where the population density is large and the main industry is second industry,which has become the region with the most severe haze pollution in China and the world.This environmental issue has aroused widespread concern from the society and the public,and the regional governance of haze pollution is also imminent.Fine particulate matter is one of the main causes of haze and the primary pollutant in winter in the Beijing-Tianjin-Hebei region.Therefore,obtaining accurate and comprehensive temporal and spatial distribution characteristics of PM_2.5 can provide a scientific basis for further understanding of the causes and changes of air pollution in the Beijing-Tianjin-Hebei region,and it can also provide a certain scientific direction for the temporal-spatial changes of regional atmospheric diffusion conditions and the treatment of air pollution.Relevant research shows that the use of satellite remote sensing to monitor air pollution can obtain environmental monitoring data with a wider coverage area than ground monitoring stations,and this advantage makes up for data acquisition channels for sparsely distributed ground monitoring stations or non-site distribution areas such as mountainous areas.The aerosol optical thickness obtained by satellite remote sensing is one of the important parameters of aerosol,and it can reflect the status of atmospheric pollution.A statistical model is established by the aerosol optical thickness and the ground-level PM_2.5.5 concentration.The time and space changes of the ground-level PM_2.5 can be obtained by this model.The lack of PM_2.5 concentration data provides a scientific reference for understanding the characteristics of fine particle concentration distribution in the Beijing-Tianjin-Hebei region.The research period of this thesis is from 2013 to 2016 and the study area is the Beijing-Tianjin-Hebei region.This thesis utilizes hourly PM_2.5 concentration data and AOD data of the Aqua satellite and the Terra satellite which are acquired by the National Air Quality Real-time Monitoring Platform and MODIS remote sensor.The resolution of the MODIS AOD is 10km.The two-stage nesting of the linear mixed effect model reflecting the temporal change of the relationship between PM_2.5-AOD and the geographically weighted regression model reflecting the spatial change of the relationship between the two was constructed using the daily average data of AOD and PM_2.5 concentration after fusion processing model.Compared with a single data source,the fused MODIS AOD data can improve the AOD coverage of the Beijing-Tianjin-Hebei region,but the AOD data obtained by remote sensing still does not meet the continuous distribution in time and space,and cannot fully reflect the overall PM_2.5 concentration of the region.For grids where there lacks AOD data in time and space,the generalized additive mixture model is used to further interpolate the PM_2.5 concentration.The over-fitting phenomenon and fitting accuracy of the above models are evaluated by the ten-fold cross-validation method and the determination coefficients,root mean square prediction error,relative prediction error and other statistical indicators.Finally,according to the model,the PM_2.5concentration in the Beijing-Tianjin-Hebei region is predicted,and the spatial and temporal distribution characteristics of PM_2.5 concentration in the four years from2013 to 2016 are obtained.The research results show that:?1?The R² and slope of the10-fold cross-validation fitting using the LME only for the relationship between PM_2.5-AOD are low.The mean value of R² is 0.82 and the mean value of slope is 0.84by using LME+GWR,indicating that the model fitting results are good and the model over-fitting is relatively light.The two-stage model can be used to more accurately predict the ground-level PM_2.5 concentration in the study area.?2?After adding the ground inversion information processing,the PM_2.5 concentration grid can basically cover the entire study area and the time series is more complete,and the average CV R² of the generalized additive mixed model used for PM_2.5 concentration information supplement is 0.66.The phenomenon of over-fitting is light and the model results are reliable.?3?Based on the model inversion and supplementation,the spatial and temporal variation information of PM_2.5.5 concentration in Beijing-Tianjin-Hebei region from 2013 to 2016 was obtained.From the perspective of time distribution,the PM_2.5pollution in the region during the study period was the most serious in winter and lighter in summer;from the spatial distribution It can be seen that the PM_2.5concentration is higher in the central and southern plains,and the pollution in the northern mountainous areas is lighter.From 2013 to 2016,the average PM_2.5concentration was 81.65?g/m³,72.13?g/m³,61.42?g/m³ and 59.50?g/m³,showing a decreasing trend year by year,indicating that the air pollution control in Beijing-Tianjin-Hebei region has achieved remarkable results.This study innovatively uses a two-stage model and a generalized additive mixed model to simulate the ground-level PM_2.5 concentration in Beijing-Tianjin-Hebei region.On the basis of previous research,the simulation accuracy of the spatial and temporal changes of fine particle concentration in the study area is further improved.