摘要
It is difficult to estimate the surface-level particulate matter 2.5 (PM2.5) using satellite-based aerosol optical thickness (AOT) because the relevant correlation is influenced by many factors, such as retrieval AOT algorithms, meteorological parameters and heterogeneous reactions. In this study, we analyzed the influences of four meteorological parameters (temperature, air pressure, relative humidity and wind speed) and four gaseous pollutants (SO2, NO2, O-3 and CO) on ground PM2.5 estimation. We obtained AOT by Moderate Resolution Imaging Spectro-radiometer (MODIS) and Landsat 8 in city of Beijing, China from April to November in 2013. By correlation analysis, PM2.5 significantly correlated with AOT (R = 0.8), SO2 (R = 0.83) and relative humidity (R = 0.81), correlated somewhat with NO2 and O-3, while only weakly correlated with the air pressure. The accuracy is too low to build a linear model to predict ground PM2.5 with AOT alone. A more accurate linear model was obtained when incorporating gaseous pollutants and meteorological data (R-2 increased from 0.64 to 0.93, RMSE decreased from 31.2 to 14.1). If meteorological data were not available, using AOT and SO2 to estimate ground PM2.5 could also achieve good results (R-2 = 0.82, RMSE = 22.6). Comparing with linear models, using back propagation (BP) neutral network model to estimate ground PM2.5 was found to be more suitable as its RMSE was 6.6, being much smaller than that of the linear models.
学科与关键词
文献加载失败,请重试
重试