Study On PM_2.5 Inversion In China Mainland Based On Multi-source Remote Sensing Data And Its Relationship With Urbanization

Over the past decade,with the rapid development of China's economy and the acceleration of urbanization,frequent occurrence of smog has become a prominent environmental problem in China.It was only in recent years that China gradually realized the seriousness of PM_2.5 hazards.At the end of 2013,the national PM_2.5ground monitoring network was initially established.PM_2.5 monitoring is one of the important contents of environmental monitoring.Due to the scattered site and short monitoring sequence of conventional ground monitoring stations,the research on PM_2.5 health effects and environmental pollution control in China is relatively weak.The large-scale long-time remote sensing monitoring can effectively compensate for the shortage of ground station monitoring and provide rich data for PM_2.5 monitoring.Rapid urbanization is one of the important factors causing large-scale pollution of PM_2.5 in China.PM_2.5 inversion mainly uses satellite remote sensing aerosol products combined with meteorological data as data sources,but rarely takes into account the impact of urbanization information on PM_2.5 inversion.Nighttime Light Data?NTL?,as a good indicator of urbanization and human activities,can affect PM_2.5concentrations theoretically.Based on the above background,according to aerosol optical depth?AOD?and meteorological data,a PM_2.5 inversion model with night light data was constructed to evaluate the role of night light data in PM_2.5 prediction.Then the historical PM_2.5 concentration was inverted and its temporal and spatial variation characteristics were analyzed.Finally,the spatial distribution of urbanization factors was analyzed to provide a scientific basis for PM_2.5 pollution control and health research in mainland China.The specific work and main conclusions of this study are as follows:?1?In order to evaluate the role of nighttime light data in PM_2.5 prediction,the neural network model is based on the characteristics of nonlinear research,and the neural network-NTL model with nighttime light data is established based on the neural network model.The neural network-basic model of the two PM_2.5 inversion models,the inversion of PM_2.5 and ground station monitoring PM_2.5 verification shows that the neural network-NTL model?R²=0.750?is more than the neural network-basic model?R²=0.731?The inversion accuracy is higher and NTL can improve the performance of the inversion model.The nighttime light data makes the PM_2.5 inversion concentration distribution more seasonal due to the seasonal changes that reveal human activity.?2?Based on the assumption that the AOD-PM_2.5 relationship is relatively stable in the same month of each year,the 2016 model is used to estimate the PM_2.5concentration in the historical month of 2001-2016.Firstly,the PM_2.5 concentration in each month of 2015 was inverted and compared with the ground monitoring station PM_2.5.The results show that the accuracy is good,and the accuracy improvement of nighttime lighting data is also suitable in the historical concentration inversion.Further reversing the historical PM_2.5 concentration from 2001 to 2016,and analyzing the spatial and temporal distribution and variation characteristics of PM_2.5 in China from2001 to 2016.The results show that the distribution of PM_2.5 concentration in China is closely related to topography and human activities,and the overall distribution pattern of“five highs and five lows”is presented.The seasonal distribution characteristics show that the most serious pollution of PM_2.5 in China is winter,and the lightest is summer.The monthly distribution characteristics showed that the highest concentration was in December and January and the lowest in July and August.The annual mean concentration changes vary from region to region,but overall the concentration in autumn and winter is gradually increasing.Spatial autocorrelation analysis showed that the annual average concentration of PM_2.5 showed a strong positive spatial autocorrelation.?3?The q value of the urbanization factor geodetic detector of PM_2.5 distribution shows that population density is the largest factor affecting PM_2.5 concentration,followed by foreign direct investment,urban road area and per capita GDP.The explanatory power of the urbanization rate has gradually declined,and the explanatory power of the urban population and the area of garden green space in the municipal district has gradually increased.The results of geographically weighted regression show that the PM_2.5 in the north of the Yellow River and the northwestern China region are affected by urbanization factors,and PM_2.5 concentration is positively correlated with population density and urban road area in urbanization factors.The spatial distribution of population density is gradually increasing from south to north,and the impact of urban road area is gradually changing from west to east.Reinforcement;PM_2.5 concentration is negatively correlated with urbanization rate,proportion of tertiary industry,and garden green space in municipal districts.Among them,the negative correlation with urbanization rate is gradually increasing from southeast to northwest,and negative with the tertiary industry.The correlation gradually increases from northwest to southeast.Overall,the VIIRS NTL data has the potential to estimate PM_2.5 and can be a useful complementary source of data for estimating ground PM_2.5 distribution.The long-term monthly average PM_2.5 spatial distribution variation characteristics can provide a reference for the evaluation of historical energy-saving and emission reduction measures.Revealing the urbanization driving factors of PM_2.5 concentration distribution and analyzing its distribution status can provide a spatial reference for the government to implement cross-regional pollution prevention and control measures,and lay a foundation for improving the human settlement environment in China.