Estimation Of PM_2.5 Concentration And Analysis Of Spatial And Temporal Distribution Based On AOD Data

In recent decades,air pollution had become more and more serious with the rapid development of cities,and the study of PM2.5 concentration had attracted widespread attention.However,surface air quality monitoring sites were mainly concentrated in large and medium-sized cities with limited numbers and short historical observation time.It was difficult to provide wide coverage and long-term continuous PM2.5concentration observation data.The aerosol optical depth（AOD）retrieved by remote sensing was used to estimate PM2.5 concentration,which made up for the spatial faults and information gaps at the ground air quality monitoring sites,and provided a basis for decision-making by relevant departments in the future for PM2.5 concentration prevention and control.In addition,because fine particles（PM2.5）could stay in the atmosphere for a long time and transmit long-distance,it was easy to adhere to various harmful and toxic substances.It was harm to health.Therefore,the precise temporal and spatial distribution of PM2.5 concentration was very important for epidemiological research.Henan Province was used as an experimental area.This study selected the 2015-2019 MOD04_L2（dark blue algorithm DB）and MOD04₃K（dark target algorithm DT）AOD data of Aqua satellites.According to the characteristics of DT algorithm for dark surface areas and DB algorithm for bright surface areas,DB AOD was integrated into DT AOD data based on different land use types.The DB/DT AOD data set was generated.In order to determine the suitable model for fitting the PM2.5 concentration in Henan Province,this paper combined meteorological factors to construct a random forest,BP neural network,and geographically weighted regression model of DB/DT AOD and PM2.5 concentration.The performance of the model was evaluated by 10-fold cross validation.The accuracy of the fitting model of DB/DT AOD data,DT AOD data and PM2.5 concentration was compared and analyzed.The model with the best performance evaluation effect was used to estimate the PM2.5 concentration.According to estimated and measured PM2.5 concentration,the temporal and spatial distribution characteristics of PM2.5 concentration in Henan province from 2015 to 2019 was analyzed.The main conclusions were as follows:1.In the five-year period from 2015 to 2019,the R2 of the random forest model was the largest.Except for 2016,the RMSE of the random forest model was the smallest in the rest of the year.The performance of the model was better than the BP neural network and the geographically weighted regression model.2.The accuracy of the DB/DT AOD,DT AOD data and the PM2.5 concentration random forest model was compared.According to the model evaluation results 2015-2019,the R2 of DB/DT AOD data was larger than DT AOD data,but the RMSE increased.It could be concluded that DB/DT AOD data to build a random forest model of PM2.5 concentration in Henan Province had better fitting effect than DT AOD data,but the RMSE increased.3.The spatial distribution characteristics of the estimated PM2.5 concentration in the five-year period were roughly similar to the actual measured value.The high-value areas of PM2.5 concentration were mainly distributed in the northern cities of Henan Province.The low-value areas were mainly distributed in the southern cities of Henan Province.The spatial distribution characteristics were high in the north and low in the south.The Moran autocorrelation tests from 2015 to 2019 were all positive.The air pollution in Henan Province had a strong spatial correlation during the five years,and the intensity of spatial agglomeration was high.4.From the annual scale,the PM2.5 concentration from 2015 to 2019 had been declining year by year,and the magnitude of the change had weakened,showing a stable trend.From the seasonal scale,the characteristics of seasonal changes were obvious.The PM2.5 concentration was high in winter and low in summer,followed by spring and autumn in the five year.Regardless of the annual and seasonal scales,the fluctuation range of PM2.5 concentration in cities in Henan Province was decreasing year by year.From monthly perspective,the lowest monthly average PM2.5concentration in 2015,2016,and 2019 was August,and the highest was January.The lowest monthly average PM2.5 concentration in 2017 and 2018 was July and the highest is January.The PM2.5 concentration was high in months of winter,and low in months of summer,which was the same as the seasonal characteristics.