Research On High-resolution Aerosol Remote Sensing Retrieval Over Urban Areas

Atmospheric aerosol plays an important role in the climate change and environmental pollution, especially for the current hot issues: global warming, air pollution and monitoring of haze. As a population gathering area, the climate change and air pollution of the urban area is concerned by people. Therefore, the aerosol retrieval in urban area is a hot research topic. However, due to the complex surface structure of urban areas, the traditional dark pixel algorithm and its improved algorithm method is not suitable to determine the surface reflectance for urban areas. And the spatial resolution of the aerosol products is generally low. For urban areas, the regional range is small, and the spatial variation of aerosol is intense, so the low spatial resolution of the data does not meet the demand for urban aerosol retrieval. In view of the problems existing in the current urban aerosol inversion, in this paper, the inversion method of high resolution aerosol in urban area were studied, the main research work is as follows:（1） The inversion algorithm based on segmentation was designed. According to the remote sensing image characteristics and the spectral characteristics of the features, firstly, the bands have little the influence of aerosol are choose, then the algorithm divide the pixels on the image into different sections though the apparent reflectance of each pixel at the bands. And then the segmented results are divided into the reference and non-reference two parts. For reference part, combined with the radiative transfer equation and lookup table, the aerosol optical thickness of the reference segments could be obtained by finding the clear pixels of each segment. For non-reference part, according to the results of the reference part and the correlation of features in geographical space, combined with atmospheric radiative transfer equation and lookup table, the aerosol optical thickness of the non-reference segments could be obtained by dividing the image into multiple grids.（2） Using Landsat 8 OLI high resolution remote sensing data, select Beijing, Xuzhou urban areas, the inversion experiments were carried out in the urban areas of Beijing, Xuzhou and Baotou by the algorithm based on segmentation and dark pixel algorithm. At the same time, the results of the inversion were verified and analyzed with the AERONET ground observation network data, and the two methods were compared and analyzed. Analysis shows that: for OLI data, in the urban areas, compared with the dark pixel algorithm, the algorithm based on segmentation has better applicability, and its accuracy is higher, among them, the absolute values of absolute error of the results of the new algorithm site are less than 0.133, while the values the dark pixel algorithm are more than 0.162.（3） The aerosol inversion experiments are carried out using the new inversion algorithm based on the medium resolution MODIS data. At the same time, combined with ground observation data, the results of the MOD04₃K product and the new method are verified. The verification shows that the deviation of MOD04₃K products in the urban areas is large, and the results of the new algorithm are more accurate compared to the aerosol products, but the deletion of the results in the urban areas was very serious. At the same time, the inversion results of the new method and aerosol products were analyzed, the results showed that the overall trends were consistent, and the difference between the two results was not large.Through the above research analysis, it shows that for high resolution inversion of aerosol to urban areas, the new algorithm both in the inverse range and accuracy are better than the dark pixel algorithm, and compared to the deep blue algorithm requiring very high for time resolution, the new algorithm is also applicable to low temporal resolution of the remote sensing image.