Retrieval Of Aerosol Optical Depth And PM_2.5concentrations Estimation In The Guanzhong Region Based On MODIS Data

The Guanzhong region is the center of economic development in Shaanxi Province,which is sensitive to climate change and has a fragile ecological environment.With rapid economic development,hazy weather is frequent and air pollution is a prominent problem,with important impacts on people’s health and regional climate change.Aerosol Optical Depth（AOD）indicates the integration of extinction coefficient in the vertical direction and is often used to describe atmospheric turbidity.PM_2.5is fine particulate matter with aerodynamic diameter less than or equal to 2.5,which has become one of the top pollutants in some areas due to its light weight,high activity and other characteristics..The use of satellite data to quantitatively analyze the spatial and temporal distribution characteristics and impact factors of pollutants can compensate for the uneven distribution of ground monitoring stations and provide reference and scientific basis for regional air pollution prevention and control,monitoring and management.?In this paper,the Guanzhong region of Shaanxi Province was selected as the research area.We used the dark target method to retrieve aerosol optical depth from MOD/MYD02＿1km satellite remote sensing images from 2019 to 2021.The accuracy of the retrieval results was verified based on the MOD/MYD04＿3km AOD product,and its spatio-temporal variation characteristics were analyzed.The temporal and spatial correlations between AOD,meteorological data,elevation data,GDP,population density factor and ground-based monitoring PM_2.5concentrations were analyzed based on Pearson correlation coefficient,Jaccard index and Moran’s index.Based on the results of AOD retrieval,the optimized XGBoost（O-XGBoost）model with spatio-temporal characteristics is constructed to estimate PM_2.5concentrations in the Guanzhong region from 2019 to 2021 by introducing geographic location data,temporal data and other feature factors.The accuracy of the estimation results were compared with those of the multiple linear regression（MLR）model,random forest（RF）model,and XGBoost model,and the temporal and spatial distribution characteristics of PM_2.5concentrations were analyzed.The results of the study showed that:（1）The validation results of the AOD retrieved by the dark target method with the MOD04＿3km data products have good accuracy,with R²reaching 0.884.It indicates the applicability of the method in the Guanzhong region and the reliability of the inversion results,which can be used to analyze the spatial and temporal distribution characteristics of AOD in the region.（2）The aerosol optical depth in the Guanzhong region shows clear time-varying features.In terms of annual variation,the year-on-year AOD values decreased from2019 to 2021,the level of atmospheric pollution weakened,and the air quality gradually improved.In terms of seasonal variability,the highest AOD values were found in winter,followed by those in spring,autumn and summer.Coal burning exacerbates pollution in winter,dusty weather deteriorates air quality in spring,and meteorological conditions in autumn and summer are conducive to the diffusion of pollutants,making pollution lighter.（3）The spatial distribution of AOD in the Guanzhong region varies significantly.The Guanzhong Basin is an area with high pollution values,while the Qinling and Beishan mountain ranges to the north and south have lower AOD values.In terms of municipal administrative divisions,the AOD values from highest to lowest are as follows:Weinan City,Xianyang City,Xi’an City,Tongchuan City and Baoji City.The northern part of Xi’an,the southern part of Xianyang and the central part of Weinan are the most polluted,and the farther away from the city center,the less polluted.?（4）The PM_2.5concentration was found to be positively correlated with AOD,longitude,GDP and population density,and negatively correlated with day,latitude,BLH,RH,temperature and DEM.among which,PM_2.5concentrations were the most correlated with AOD.During model training,the AOD importance score was the highest and had the largest impact on the model results.?（5）Compared with MLR,RF,and XGBoost models,the O-XGBoost model used in this paper,which takes into account spatio-temporal factors and performs parameter optimization,has higher accuracy in estimation results,with R²reaching 0.891 and the RMSE and MAE of 13.531 and 8.924,respectively.It indicates that the influence of spatial and temporal factors in PM_2.5concentrations estimation cannot be neglected and the optimized XGBoost model is more suitable for PM_2.5concentrations estimation in the Guanzhong region.（6）PM_2.5and AOD show roughly the same spatial and temporal variation characteristics.In terms of temporal variation,PM_2.5concentrations in the Guanzhong region decreased year by year,with the worst pollution in winter,followed by spring,lighter in autumn,and the least polluted in summer.In terms of spatial distribution,the pollution is mainly concentrated in the Guanzhong basin,and the south and north of the Guanzhong region are mountain ranges,which are less polluted.