Estimation Of PM_2.5 Concentration In Beijing,Tianjin And Hebei Based On Fusion High Resolution AOD

In recent years,PM_2.5 has attracted wide attention due to haze events.However,the establishment of PM_2.5 monitoring network in China is relatively late and the initial time of each station is different,so there is a lack of long-term and continuous PM_2.5 concentration data.At present,aerosol optical thickness?AOD?is a commonly used method to predict PM_2.5concentration.As many cities in Beijing,Tianjin and Hebei have long been on the list of poor air environment,they have become the focus of environmental research.In this paper,the PM_2.5concentration in the Beijing-Tianjin-Hebei region was studied,and the changes of different time scales from December 2016 to November 2017 were analyzed.SARA algorithm is used to retrieve the AOD with resolution of 500 m,and it is fused with MODIS AOD products.After fusion,AOD was combined with NDVI,DEM,population and meteorological factors to retrieve high resolution PM_2.5 concentration using OLS,GWR and GTWR models.The main research contents and conclusions are as follows:?1?Meteorological factors and PM_2.5 concentration transport and diffusion have a very important impact.There are 26 meteorological stations in Beijing,Tianjin and Hebei,and 46additional stations are selected around them for inverse distance weight interpolation.PM_2.5and each meteorological factor have 24984 groups of data,which have passed the SPSS correlation test.PM_2.5 was negatively correlated with precipitation,wind speed and temperature,with correlation coefficients of-0.104,-0.228 and-0.312,and positively correlated with air pressure and humidity,with correlation coefficients of 0.196 and 0.273,respectively.The correlation between meteorological factors and PM_2.5 concentration in different intervals is analyzed.The overall correlation has not changed,but in some intervals the correlation is obvious,and in some intervals the correlation is not obvious.?2?Based on MOD02HKM,MOD03 and MOD09 data,a simple aerosol algorithm?SARA?was used to retrieve AOD with 500 m resolution in Beijing-Tianjin-Hebei.The results were compared with MOD04₃km and MOD04₁0km AOD data,and the AOD data of CE-318instrument on the roof of the College of Resources and Environmental Sciences of Hebei Normal University was used as a reference.The correlation coefficients R² of MOD04₁0km,MOD04₃km,SARA AOD and CE-318 AOD were 0.647,0.684 and 0.425,respectively.?3?MODIS,AOD fusion,due to the good quality of MODIS AOD data,but the lack of serious,SARA AOD can cover the entire study area.Based on MOD04₃km,MOD04₁0km and SARA AOD are resampled and corrected.The processed data is fused in the order of MO04₃km,MYD04₃km,MOD04₁0km,MYD04₁0km and SARA AOD to obtain full coverage AOD products with daily 3km resolution.The fusion product is verified with an accuracy between MODIS AOD and SARA AOD.?4?In order to understand the spatial distribution of PM_2.5 concentration in Beijing,Tianjin and Hebei,ordinary Kriging interpolation was used to analyze the seasonal,monthly and daily variation of PM_2.5 concentration at 80 PM_2.5 stations and 36 PM_2.5 stations around the study area.In terms of season,the Beijing-Tianjin-Hebei research area can be roughly divided into two areas south to north of Yanshan Mountain.PM_2.5 concentration shows the characteristics of low in the north,high in the south,simple in the north and complex in the south.The reason for this difference is the difference between the natural geographical environment and the human environment.In terms of monthly mean,the PM_2.5 concentration in Beijing-Tianjin-Hebei region in December 2016 was the highest in the whole study period,and the pollution situation was serious.During the study period,the PM_2.5 concentration in Beijing-Tianjin-Hebei region experienced a process of decreasing,decreasing again and rising slightly.In terms of daily mean,PM_2.5 concentration exceeded 75 ug/m³ for 92 days?limit standard?.?5?Ordinary least squares regression model regards PM_2.5 concentration as dependent variable,AOD,NDVI,DEM,population and meteorological data as explanatory variable.Regression results showed that population,NDVI,DEM and PM_2.5 concentration were negatively correlated,AOD and PM_2.5 concentration were positively correlated,and the maximum coefficient was the most important parameter in the model.In this paper,the statistic significant unsteady regression model of dependent variable and explanatory variable is suitable for GWR and GTWR analysis.?6?The concentration of PM_2.5 was inverted by GWR and GTWR models,and the optimal bandwidth was calculated by AICc method.By comparing the results of OLS,GWR and GTWR,taking R²,AICc,sum of squares of residuals and Sigma as the parameters,it shows that GTWR model has the best fitting effect.