PM_2.5 Concentration Estimation Based On GTWR-XGBoost Model And Its Relationship With Urbanizatio

With the rapid recovery of China’s economy and the acceleration of China’s industrialization and urbanization,the rapid development has brought serious air pollution problems to most parts of the country,especially many cities in China with PM_2.5 as the primary pollutant.It is of great significance to study the estimation method of PM_2.5concentration with high time resolution and high precision for timely and accurate air quality prediction and prevention and mitigation of air pollution.In this study,Sichuan Province was taken as the research area,and Himawari-8 AOD hourly products,ERA5 reanalysis meteorological data,DEM and ground monitoring site data were selected as estimation variables.The GTWR-XGBoost combined model was constructed to estimate the hourly concentration of PM_2.5 in Sichuan Province and analyze its spatial and temporal changes.At the same time,the monthly average gridded PM_2.5 concentration data of Sichuan Province inverted were inverted by the model,the influence of different urbanization index factors on the change of PM_2.5 concentration was analyzed by geographical detector.The results are summarized as follows:（1）Through correlation analysis and collinear analysis of the PM_2.5 hourly concentration estimation dataset,it was concluded that PM_2.5 concentration was positively correlated with AOD,PS and RH,and negatively correlated with BLH,CVH,T2M,D2M,U₁₀,V₁₀,RAIN and DEM.And there are multiple variables with obvious collinear relationships,and then the variable factors without collinear relationships are obtained through multiple linear regression analysis.The variable factors retained after the analysis were substituted into the GTWR-XGBoost model and the cross-validation accuracy of the other eight control models,respectively,and the GTWR-XGBoost model had the strongest ability to explain the variation of PM_2.5 concentration and the smallest error(the fitting results R²,MAE and RMSE were0.96,2.69μg·m^-3 and 5.38μg·m^-3,respectively;The results R²,MAE and RMSE were 0.89,6.14μg·m^-3 and 8.92μg·m^-3,respectively),which were the best models for hourly estimation of PM_2.5 concentration.（2）From 2015 to 2020,the annual average concentration of PM_2.5 in Sichuan Province showed a downward trend year by year,and the decrease was the most obvious from 2019 to2020,and the annual average PM_2.5 concentration dropped to 35.40μg·m^-3,reaching the boundary of the secondary limit of China’s ambient air quality standard(35.00μg·m^-3).Although the seasonal variation varied in different years,the seasonal average PM_2.5concentration showed an inverted"U"shape on the whole.There were obvious differences in PM_2.5 concentrations in different seasons,and the concentration values ranged from47.18～70.29μg·m^-3,34.54～52.89μg·m^-3,32.93～50.66μg·m^-3 and 25.83～35.16μg·m^-3,respectively,with significant seasonal variations:winter>spring>autumn>Summertime.However,the degree of decline of PM_2.5 concentration in different regions of Sichuan Province varied,with the decline trend being particularly significant in the central and eastern regions with high PM_2.5 concentration,the continuous and slow decline in the west,and the PM_2.5concentration in only Aba,Ganzi,Liangshan and Panzhihua sporadic areas showing an upward trend.The overall distribution pattern of PM_2.5 concentration in Sichuan Province is still characterized by high in the east and low in the west,high in the north and low in the south,as well as the agglomeration pattern of high value aggregation and low value aggregation.（3）From 2015 to 2020,the global Moran’s I index value of PM_2.5 concentration in Sichuan Province was distributed between 0.4508 and 0.5437,and the high/low clustering index value of PM_2.5 concentration was greater than 0.05,which concluded that the spatial distribution of PM_2.5 in Sichuan Province was concentrated in high concentration areas,and the urban agglomeration in central and eastern Sichuan Province centered on Chengdu has formed stable and continuous PM_2.5 continuous pollution areas.The relationship between urbanization index factors and PM_2.5 concentration was analyzed by using geographic detectors,and the ability of urbanization index factors to explain the variation of PM_2.5concentration was explored,and single factors such as urban built-up area,built-up area green area,urban population at the end of the year,urbanization rate,population density and total natural gas supply played a key role on PM_2.5 concentration value,and the interaction between secondary industry proportion,per capita GDP and urbanization rate factor played an important role in PM_2.5 concentration.The research results provide a scientific reference for putting forward targeted countermeasures and suggestions for PM_2.5 pollution control.