Study On Atmospheric Path Radiation Retrieval And Application For Complicated Surface

The atmospheric path radiation is the value received by satellite sensors directly due to particle scattering in atmospheric transmission, such as atmospheric molecule, aerosol and ice crystals scattering. Because of atmospheric inhomogeneity and air pollution, the atmospheric path radiation value is inhomogenous for spatial distribution difference of atmospheric components and aerosol, including content, shape, character, particle radius and concentration. Furthermore, atmospheric path radiation values are generated by the interaction between different atmospheric components and aerosol in different position, so it can monitor atmospheric status and quality efficiently, which is very significant to research in depth.For remote sensing technology, atmospheric path radiation is the part of sky brightness radiation depending on the status of atmosphere and not including ground information, which can reflect atmospheric quality. For earth observation technology, atmospheric path radiation is the weak information which does not involve ground information compared with the strong background information (surface data), belonging to the atmospherics. Therefore, how to extract atmospheric path radiation exactly from remote sensing images under the condition of the strong background information (surface data) becomes a key issue to the accurate atmospheric correction and atmospheric environmental monitoring. Based on the validation model of the MODIS surface reflectance product using ASD spectrometer site measurements, the dissertation proposed a new optimization approach of computing inhomogenous atmospheric path radiation of MODIS images for complicated surface. Moreover, the matched applications of atmospheric path radiation have been studied in the end of dissertation. The main contributions and innovation-points of this dissertation are shown as following:(1) The dissertation proposed a new optimization model of MODIS reflectance based on ASD spectrometer site measurements. Based on characteristics of multi-date remote sensing images, the new optimization model combines the 6S model for MODIS and the field measurement model of ASD spectrometer effectively. The experiment results show the new model can optimize MOD09 product, mainly including restraining atmospheric impact and cloud to close the surface reflectance truth hand to hand. Additionally, the average error has decreased 55.42% after optimizing by the new model;(2) The dissertation proposed a new approach of recognizing water based on intelligence ant colony algorithm on MODIS remote sensing images. Based on intelligence ant colony algorithm, the new approach combining the spectrum-photometric and NDVI method, mainly figures out how to fix the break-flow of water-land boundary, recognize the large lake efficiently and repair the break-flow of the linear river. The experiment results show the new approach can recognize Taihu Lake clearly, also optimize the upper reaches of Huangpu and Qiantang River effectively. Compared with the data of National Fundamental Geographic Information System (NFGIS), difference ratio is down to 9.525%.(3) The dissertation proposed a new approach of computing atmospheric path radiation of water. Combining the dark object method and pixel-by-pixel imaging algorithm of satellite remote sensing images, the new approach can avoid the greater deviation of computing atmospheric path radiation for water pixel due to specular reflection of water, leading to improve the result-accuracy.(4) The dissertation proposed a new approach of recognizing cloud based on support vector machine (SVM) on MODIS remote sensing images. The new approach merges four general methods, namely cloud recognition using NDVI, the ratio of near-infrared reflectance, the characteristics of visible light reflectance and SVM method separately. The approach can detect the discrete thin-cloud pixels mainly using SVM under the condition that the large-scale thick-clouds have been recognized so that the result of recognizing cloud is ameliorated, so the accuracy of recognizing cloud is improved and the accumulative error of computing atmospheric path radiation in the next step is avoided obviously.(5) The dissertation proposed a new optimization approach of computing inhomogenous atmospheric path radiation. Based on pixel-by-pixel imaging algorithm of remote sensing images, the optimization approach of inhomogenous atmospheric path radiation has been constructed through analyzing solar radiation transmittance theory and studying on the approaches of computing inhomogenous atmospheric path radiation for different sorts. Finally, we completely code it using IDL, achieve how to generate the inhomogenous atmospheric path radiation images in computer using original radiation images and matched surface reflectance products. Compared with results by 6S model, the atmospheric path radiation value by the new approach has a little error, less than 10%, which meets the requirement of application. Moreover, the dissertation discusses how the error of surface reflectance impacts the accuracy of atmospheric path radiation value, and certifies atmospheric path radiation value can be as the data source for a retrieval approach of inhomogenous API.(6) The dissertation proposed a new retrieval approach of API using inhomogenous atmospheric path radiation based on a new estimation model of background value in the atmospheric path radiation image. Using API values in different sites in the same day, the new retrieval approach is built based on inhomogenous atmospheric path radiation images. The new approach can retrieve continuous inhomogenous local API values matched by atmospheric path radiation images. The experiment results show the accuracy of retrieved API value meets the requirement of atmospheric monitoring so that the new approach can be as a new application outlook for air pollution monitoring in the future.