Cloud detection performance of spaceborne visible-to-infrared multispectral imagers.

We investigate the cloud detection efficiency of existing and future spaceborne visible-to-infrared imagers, focusing on several threshold tests for cloud detection over different types of ground surfaces, namely, the ocean, desert, vegetation, semibare land, and cryosphere. In this investigation, we used the CLoud and Aerosol Unbiased Decision Intellectual Algorithm (CLAUDIA), which was developed for unbiased cloud detection. It was revealed that imagers with fewer bands than the Moderate Resolution Imaging Spectroradiometer tend to have cloudy shifts. An imager without any infrared bands could yield cloudy shifts up to 17% over the ocean. To avoid false recognition of Sun glint as clouds, the 0.905 and 0.935  μm bands are needed in addition to the infrared bands. In reflectance ratio tests, the 0.87 and 1.6  μm bands can effectively distinguish clouds from desert. In the case of desert, thermal-infrared bands are ineffective when the desert surface temperature is low during winter. The 3.9 and 11  μm bands are critical for distinguishing between clear and cloudy pixels over snow-/ice-covered areas. The results and discussions of this research can guide CLAUDIA users in the optimization of thresholds. Here, we propose a virtual imager called the cloud detection imager, which has seven or eight bands for efficient cloud detection.