Discriminating cirrus cloud from MODIS imagery over oceans utilizing gradient technique

The presence of thin cirrus clouds often contaminates remote sensing retrieval products. In this paper, a simple algorithm to discriminate cirrus-contaminated pixels using MODIS imagery over oceans is proposed. The apparent reflectance values of both the 1.375 µm and 1.240 µm channels were used because of their difference sensitivity characteristics. The algorithm is based on the gradient of the line connecting the 1.375 µm and 1.240 µm channels of a log-log graph of the apparent reflectance values against their MODIS wavelengths. The graph of the gradient is dependent on the atmospheric constituents in the study area. The gradient of the graph is lower in the dust and aerosol-dominated area than in the cirrus cloud area. A substantial increase in the 1.375 µm channel led to a low value of the gradient of the cirrus-contaminated pixels. The threshold for discriminating cirrus-contaminated pixels from MODIS imagery was obtained by using histogram analysis. This algorithm was tested on a number of MODIS datasets acquired over the Korean peninsula, Mediterranean Sea and Atlantic Ocean. In conclusion, cirrus cloud-contaminated pixels were successfully detected.