Monitoring suspended particulate matter based on satellite-imagery and ground observation

Abstract

In recent years, due to the rapid urbanization all over the world, air pollution has become a serious problem in many countries. Especially in China, rapid industrialization associated with more fossil fuel consumption has caused serious suspended particulate matter pollution, and resulted in more frequent haze days. At present, many studies have been indicating that the size and the formation mechanism have a close relationship with public health. Therefore, a good understanding of the distribution of air suspended particles with different diameters is necessary. However, there are still some limitations for different size distributions of air suspended particles retrieval by remote sensing. Thus, this thesis firstly put forward a reliable method for monitoring dust distribution (diameter less than 1 mm) with the aid of ground-based plant leaf spectral data. A back propagation (BP) neutral network model was generated using spectral response functions and integrated remote sensing data to estimate dustfall weight in the city of Beijing. Compared with actual dustfall weight, validation of the results showed a satisfactory accuracy with a low RMSE of 3.6 g/m². Secondly, an algorithm was developed which incorporates haze monitoring and haze aerosol optical thickness (HAOT, particulate size between 0.001m to 10 m) retrieval based on MODIS data. From the comparison, this method can effectively make up for MODIS AOT products deficiency about missing data under haze weather condition. Then, the fine mode fraction (FMF) is a useful tool to separate the fine mode aerosol from the total aerosol. However, the spatial view of the FMF is still limited. Therefore, a lookup table-based spectral deconvolution algorithm (LUT-SDA) was proposed. This method was validated with ground-based data and had a high accuracy compared to the Aerosol Robotic Network (AERONET) FMF. Finally, assistant by LUT-SDA, a ground-level PM2.5 retrieval model was developed. This model had been applied to retrieval surface PM2.5 concentration over Beijing from December 2013 to June 2015 in cloud free day. The derived results ware compared with the monitoring values with R² = 0.64 and RMSE = 18.9 g/m³ (N = 921). This validation demonstrated that the developed model exhibits a good performance with a high accuracy.