Study On The Methodology Of Using Nighttime Light Data To Retrieve Atmospheric Fine Particle’s Concentration

It is of great significance for the monitoring of haze by monitoring of aerosol with remote sensing. Using Aerosol Optical Thickness (AOT) to retrieve the surface atmospheric particulate concentration in day time is very common. And the results have high reliability. This study’s data source is NightTime Light (NTL) data from Defense Meteorological Satellite Program-the Operational Linescan System (DMSP-OLS). A model and method for the inversion of surface particulate matter concentration by nighttime light data is discussed, and inverse the concentration of atmospheric particulate matter at night based on this method.In this study, the Beijing city is selected as the experimental field. And the daily average PM2.5 concentration data of 23 "urban environmental assessment points" in the air pollution monitoring station of the Beijing city in the winter of 2013 and NTL data, moon phase data and meteorological data over the same period are collected. The quasi-synchronous LANDSTA-8 OLI image data of the Beijing area in September 1,2013 is also obtained. With theoretical analysising, the main factors affecting the daily NTL data were studied, and the various factors were treated in the different ways. A BP neural network model is used in the inversion. The input data of BP neural network is constructed, and the PM2.5 concentration of the data is retrieved by the NTL data. In order to avoid local optimization of BP network, the BP neural network is optimized by Particle Swarm Optimization (PSO) algorithm. In this paper, research work and conclusions are as follows:1) Based on correlation analysis of digital lunar phase, the sum of DN in every day NTL images and meteorological factors, the result shows that the lunar phase have a decisive influence on the sum of DN in every day NTL images.2) Using NTL images without the effect of lunary phase, four indexs of the NTL data were extracted. The four indexs all have correlation with the PM2.5 concentration after humidity correcting. And "non-saturated region brightness index" has the highest correlation coefficent.3) Based on the boundary of the Beijing city’s urban, four regions were created. The daily DMSP-OLS NTL data was divided into the four regions to obtain the inputs of BP neural network. Three models are established using the same data from the same period, to compare with the model of this paper.Main innovation points of this thesis are as follow:The remote sensing image of night light is used to retrieval atmospheric PM2.5 concentration. The correlation between night light data and PM2.5 concentration is discussed, and four night light indexs which are sensitive with PM2.5 concentration are extracted. By introducing a new data source to assess the status of air pollution, it provides a new way and an example for the PM2.5 concentration inversion and monitoring in the urban area.By eliminating the effect of lunar phase change, daily nighttime light data is comparable with respect to the influence of atmospheric variability.