| Atmospheric aerosol is a composite system of all kinds of solid and liquid particles which is suspended in the atmosphere,such as smoke,dust,Cloud drops,snow, and ice crystals, etc. Atmospheric aerosol is widely distributed, and has complex physical and chemical changes. Its impacts on the global climate and environmental quality are also very significant. Compared with ground-based remote sensing,satellite remote sensing has certain of superiority for its large spatial and temporal distribution. At present, there are many researches on the inversion of Aerosol Optical Depth (AOD) by satellite remote sensing. Among them, the comparatively mature and comprehensive researching method is the one based on MODIS satellite. For the algorithm which mainly focuses on how to accurately obtain the surface reflectivity to improve the accuracy of AOD inversion, has become the current problem which needs to be researched and resolved urgently by domestic and foreign scholars.The main contents and conclusions of this paper are as follows:(1)A new method to inversing AOD with MODIS data is proposed in this paper,which is based on the basic principle of AOD inversion and the pixel unmixing method to estimate the surface reflectivity. Firstly, screening the daily MODIS surface reflectance data by season to obtain the average surface reflectance value of spring, summer, autumn and winter separately. Then, selecting the four kinds of endmember (buildings, woodland, arable land and water) via visual interpretation.Next, determining the proportional relation in surface reflectance of each endmember between the Red or Blue band and the 2.1μm band. Finally, unmixing the surface reflectance products of different seasons by pixel to obtain the proportion of the 4 endmembers in each pixel. After that, combined with the apparent reflectance of MODIS LIB data of 2.1 μm band, the surface reflectance of red or blue band was determined eventually. After all these steps, the AOD inversion was carried out in the certain study area via using the 6S radiation transmission model to construct the lookup table and with surface reflectance data in red or blue band.(2)In this paper, a comparison was carried out between pixel unmixing method,the traditional dark pixel method and MODIS V5.2 algorithm. In general, the surface reflectance eliminated by pixel unmixing method is larger than the results of the other two methods, especially in the urban areas with concentrated buildings.However, in the dense forest area, the surface reflectance of the three methods is closer. Moreover, verifying the comparison results with ground observation data and the results show that the inversion results of AOD based on the pixel unmixing method have a good correlation with the results of the traditional dark pixel method and MODIS V5.2 algorithm. It means this algorithm is usable. The AOD inversion result based on the pixel unmixing method is smaller than those of the traditional dark pixel method and the MODIS V5.2 algorithm.(3)After verifying the results. with ground observation data,it shows that the inversion method based on pixel unmixing method has the highest correlation with ground observation data, and its R2 value is 0.8068, the R2 value of MODIS V5.2 algorithm is 0.8731 and the lowest correlation is received by the traditional dark pixel method,whose R2 value is 0.7037. What’s more,the proportion of the AOD inversion results of the three algorithms falling in the error range against the total inversion results are 78.31%, 60.52% and 35.52%. Therefore, compared with the MODIS V5.2 algorithm and the traditional dark pixel method, the pixel unmixing method has a better inversion result. |
PDF Full Text Request