Journal of Geo-information Science ›› 2020, Vol. 22 ›› Issue (10): 2038-2050.doi: 10.12082/dqxxkx.2020.190394

Previous Articles     Next Articles

Retrieval of Aerosol Optical Depth over Taiwan Island Using Visible Channels of Sentinel-3A OLCI

WANG Feng1,2,3(), WANG Xiaoqin1,2,3,*(), DING Yu1,2,3   

  1. 1. The Academy of Digital China (Fujian), Fuzhou 350003, China
    2. National & Local Joint Engineering Research of satellite-spatial Information Technology, Fuzhou University, Fuzhou 350108, China
    3. Key Laboratory of Spatial Data Mining & Information Sharing of MOE, Fuzhou 350108, China
  • Received:2019-07-24 Revised:2020-09-16 Online:2020-10-25 Published:2020-12-25
  • Contact: WANG Xiaoqin;
  • Supported by:
    National Key Research and Development Program of China(2017YFB0504203);Central Guide Local Science and Technology Development Projects(2017L3012)


OLCI (Ocean Land Colour Instrument) sensor, as a successor of MERIS (Medium Resolution Imaging Spectrometer) sensor, has better temporal resolution, spacial resolution and image width, so it has potential advantages in aerosol retrieval, but there are few studies on aerosol monitoring using OLCI data. In order to expand the application field and scope of Sentinel-3AOLCI image data, a high-precision OLCI cloud detection algorithm based on the multi-channel advantage and observation geometry characteristics of OLCI was first proposed, which can effectively and accurately detect thin and thick clouds. Secondly, dynamic lookup tables based on observed geometric features was constructed. Compared with the traditional lookup table, the efficiency of AOD retrieval using dynamic lookup table is improved significantly. The surface reflectance of MODIS and OLCI red and blue channels were also transformed in an equivalent manner through spectral convolution and obtain the fixed relation of surface reflectance of OLCI red and blue channels. Finally, OLCI Aerosol Optical Depth (AOD) retrieval at 550 nm over The Taiwan Island was carried out, and the spatial distribution of AOD we estimated is highly similar to MODIS (Moderate-resolution Imaging Spectroradiometer) AOD. The validation results of OLCI AOD and AERONET (AErosol Robotic Network) level 2.0 AOD at 550 nm show that the accuracy performance of different seasons and stations is different, and then compared with the same period MOD04_3K AOD at 550 nm products, the correlation between the retrieved OLCI AOD and the measured AOD at AERONET stations is more significant (R2=0.8199), the Root Mean Squared Error (RMSE) decreases from 0.175 to 0.113, the Relative Mean Error (RME) decreases from 33.6% to 26.7%, and 67.5% of OLCI AODS fall within Expected Error (EE), which is significantly greater than that of MOD04_3K AOD falling within the prediction error range (55.7%). Error analysis showed that when the actual AOD value is low, the misjudgment of the relationship between surface reflectance of red and blue channels will lead to relatively obvious relative error of AOD retrieval, which means that there are some potential difficulties in retrieving high-precision AOD in areas with good air quality, but for areas with severe pollution, OLCI AOD has significant precision advantage.

Key words: OLCI, Aerosol Optical Depth (AOD), cloud detection, dynamic look up table, Taiwan Island