Remote Sensing Estimation Method Of PM_2.5 Concentrations With High Spatial Resolution

Air pollution caused by fine particulate matter with a diameter smaller than 2.5μm(PM_2.5)can affect human’s health.Obtaining continuous spatial data of PM_2.5 concentrations can provide important data guarantee for environmental management,which is a popular topic in the current atmospheric environment research.Ground-based PM_2.5 monitoring stations are the main data source for PM_2.5 research in China,but their spatial distribution is uneven and can hardly reflect the spatial distribution of PM_2.5.In order to compensate for the uneven spatial distribution of ground-based PM_2.5 monitoring data,it is an effective way to simulate the spatial distribution of PM_2.5 concentrations by using the aerosol optical depth（AOD）product of satellite estimation as the main data source.However,due to the inversion algorithm,satellite platform,cloud and rain weather,the satellite aerosol products have low resolution and large data deficiencies.This reason limits the research of PM_2.5 concentrations estimation using satellite remote sensing AOD products.To solve above problems,this thesis aims to solve problem of missing AOD data and the effective estimation of PM_2.5 concentrations using satellite remote sensing technology with Beijing as the main research area.The main research contents are as follows.Aiming at the non-random missing problem of the currently acquired AOD data,this thesis proposes to apply the extreme gradient boosting algorithm to the missing value filling of 1km high-resolution AOD data to solve the unavoidable missing problem in the satellite data acquisition process.This method is conducive to finely reflecting the continuous spatial variation of atmospheric pollutants.The algorithm can effectively increase the coverage of AOD data,and ensure the high resolution and accuracy of the data while providing complete AOD data.Meanwhile,the accuracy of the obtained AOD data is verified by using ground-based AOD data to prove the accuracy of the algorithm in this paper.In order to reduce the effects of height and humidity on aerosol particle size and concentration,high humidity corrections are included to reduce the errors.Finally,the mixed effects model is used to demonstrate the applicability of the obtained AOD filled data in the PM_2.5 concentrations estimation study.To address the problems of sparse PM_2.5 monitoring stations and uneven distribution in the study area,this thesis proposes to use a long short-term memory network model to fit the AOD-PM_2.5 data relationship,which effectively improves the coverage and accuracy of PM_2.5concentrations estimation.It can comprehensively analyzes the spatial and temporal trends of PM_2.5 pollution in Beijing.Meanwhile,it is demonstrated the necessity of filling AOD data in PM_2.5 estimation study.Combining the filled AOD data and PM_2.5 ground observation data with meteorological data and land use data,a PM_2.5 concentration estimation model is constructed on an annual scale.Finally,the performance of the developed model is evaluated and compared with other commonly used PM_2.5 estimation models.The results show that the accuracy of the model is high.Through the study,high-resolution near-ground PM_2.5 concentrations are obtained on a large scale to compensate for the missing and insufficient data caused by the limited number of ground data monitoring stations.