An Improved Retrieval Algorithm of Aerosol Optical Depth

The algorithm to retrieve the aerosol optical depth over land has been completely restructured to produce the collection 005 products based on the algorithm by Levy et al. But the accuracy of the MODIS aerosol optical depth (AOD) products still has very large differences for different seasons and geographic locations in China. In order to improve the accuracy of aerosol retrieval products, an easier and faster algorithm for retrieval of aerosol optical depth over land with MODIS 1B data is introduced. This algorithm deals with the surface reflectance relationships is the same way as MODIS V5.2 algorithm.In order to better represent aerosol properties in China, the size distribution and refractive index of aerosol have been improved. Considering the fine structure of the aerosol size distribution has a little effect on satellite remote sensing of aerosol optical depth, this algorithm use the Junge aerosol size distribution to approximate the aerosol size distribution in an actual atmosphere. The real and imaginary index of refractive is 1.5 and 0.005, respectively.The complex refractive index is assumed for all wavelengths (0.47, 0.55, 0.66 μm and 2.1 μm). In order to verify the accuracy and regional applicability of this algorithm, aerosol optical depth is derived with this algorithm using the MODIS 1B data at Taihu and Xianghe, and this retrieval result is compared with equivalent measurements from AERONET (AErosol RObotic NETwork) site (Level 2.0 data). The MODIS/AOD product and 1B data from September 2006 to June 2008 at Taihu (MODISI/AOD product and 1B data from May 2008 to July 2009 at Xianghe) has been matched with L2.0 AOD product from AERONET stations during the same period. Data from the AERONET are averaged within 30 min before and after the satellite's passing, and the MODIS data are averaged over a 10 km (15 km at Xianghe) area centered at the ground stations. The comparison results show that the standard deviation of the new algorithm inversion results and L2.0 AOD product at Taihu is 0.429. The standard deviation of the MODIS/AOD product and L2.0 AOD product at Taihu is 0.693. Accordingly, the standard deviations of two comparison experiment at Xianghe are 0.493 and 0.542, respectively. These results show that this algorithm retrieval results have good consistency with the sun photometer observational results at Taihu and Xianghe. The retrieval algorithm is more accurate than the current MODIS aerosol algorithm and its inversion results are reasonable. In addition, the settings for aerosol model and optical properties are simple and convenient in the new algorithm, which can effectively reduce the computational time for looking-up table and the iteration time for solving equations.