Estimating Fine Particulate Matter (PM2.5)Concentrations Over Land In China By Using Remote Sensing Of Atmospheric Aerosols

Atmospheric aerosols are the solid and gas particles suspended in the air. As a key factor of the Earth’s budget and climate and environmental change, they play critical impacts on the regional and global atmospheric environment, climate and the ecosystem. The particulate matter with aerodynamic diameters of less than2.5μm, which is called fine particulate matter or PM2.5, is a hotspot in atmospheric environmental research due to its significant impacts on human health, environment and climate. With the rapid development of economy, high loadings of PM2.5in China induce severe and continuous air pollution episodes such as haze and smog events at an increasingly high frequency, making China be one of the most PM2.5polluted regions worldwide. Satellite observations such as aerosol optical depth (AOD) can be used for estimating ground-level PM2.5air-quality information. As a relatively new research field, it has extraordinary advantages in retrieving standardized and spatially continuous PM2.5mass concentrations in large area. But deficiency of accuracy and spatial and temporal applicability of the estimation models needs to be improved, especially for the research for China, which lacks of large-scale PM2.5ground-based monitoring data for a long time.This study is aiming at establishing a novel model both from site scale and regional scale to enhance the model accuracy for estimating PM2.5from satellite remote sensing observations. We first validated the aerosol optical depth retrievals over China from MODerate-resolution Imaging Spectroradiometer (MODIS) onboard Terra and Aqua, then compared spatial and temporal characteristics of ground-level PM2.5concentrations and AOD, and their correlations nationwide. After that, we developed a site-specific complex model and then a regional complex model by adopting PM2.5concentrations at a total of522monitoring sites throughout China during the recent two years along with satellite AOD and ancillary meteorological information. Also we proposed the zoning scheme of the model across China. The main research contents and results are as follows:(1) The validation of MODIS AOD over China reveals that the satellite retrievals are significantly correlated with AErosol RObotic NETwork (AERONET) AOD with the correlation coefficients above0.9. Overall the MODIS tends to slightly overestimate AOD, yet MODIS and AERONET AOD agree very well. It can be concluded that the quality of MODIS AOD are generally reliable for China. (2) Comparison between the distribution of MODIS AOD and PM2.5show the spatial patterns of annual mean PM2.5and AOD are similar. High AOD (>1.0) regions, e.g. the Northern China plain, the Pearl River Delta, the Yangtze River Delta and the Sichuan Basin, are also accumulated with much PM2.5with the annual mean values of above55μg/m3; low AOD (<0.2) regions also coincide with comparatively low PM2.5(<45μg/m3), e.g. the Qinghai-Tibet Plateau, the Yunnan-Guizhou Plateau, northeast of China and Fujian province. The seasonal variations show that PM2.5peaks in winter (86.84±67.38μg/m3) followed by autumn (63.36±49.55μg/m3), spring (60.62±46.34μg/m3) and summer (45.34±36.89μg/m3); while the mean AOD ranks as spring (0.74±0.52)> summer (0.60±0.52)> autumn (0.54±0.44)> winter (0.53±0.39). The map of correlations between PM2.5and AOD exhibits remarkable spatial and seasonal characteristics. Significant correlations of above0.45exist in most areas of the country in spring and winter. The correlations in summer and autumn both vary more in space. Overall they are higher in the east and lower in the west.(3) Estimating PM2.5by using the corrected relationship model between PM2.5and AOD over Beijing as a case study revealed that correlation of MODIS AOD and PM2.5show moderate to good agreement, with the R values being0.69,0.60and0.73for spring, summer and autumn, respectively. After using the vertical and RH correction, the correlation coefficients between MODIS AOD and ground PM2.5in summer improve by about11%(with PBLH correction) and4%(with both PBLH and XRH) correction), while the correlations in spring and autumn barely change.(4) The site-specific complex model was constructed by combining the scaling effects of PBLH and f(RH) on AOD and the impacts of meteorological variables on the PM2.5-AOD relationships. The model was established through two stages. First the site-specific optimization univariate model (UVM) was obtained based on comparing the original relationship model between PM2.5and AOD (OM) and the corrected regression model (TM); Then the UVM was combined with the multi-variate model (MVM) through selecting auxiliary meteorological variables the optimized subset selection. The results show that the model performance effectively improves comparing with the original model.(5) The zoning maps of the regional model were derived by applying multiple spatial clustering technologies. Numbers of sub-regions for each season are23,17and 16with diverse combinations of variables needed be included in the model. The regional complex model was validated. The results show that it has good fitting and prediction performance with the RMSE of predicted and actual PM2.5varying from24-43μg/m3for the selected regions in each season. The regional complex model was applied to generate the national.PM2.5estimation maps from MODIS AOD and auxiliary meteorological information. The seasonal maps of PM2.5derived from Terra and Aqua AOD are highly similar in spatial distributions, whose the means of PM2.5concentrations are the lowest in summer (43.5μg/m3and45.1μg/m3) followed by spring (51.6μg/m3and50.7μg/m3), autumn (55.5μg/m3and56.9μg/m3) and winter (328.9and344.4μg/m3). The North China and Sichuan Basin are found to be the regions that remain heavily polluted by the fine particulate matter throughout the whole year.In summary, this study proposed and established the complex model at both site and regional scale to estimate PM2.5concentrations by using remote sensing AOD data and auxiliary meteorological information. The model exhibits satisfactory performance and improved the spatial-temporal applicability, providing a useful method reference for the research of estimating PM2.5concentration based on remote sensing.