Spatially Varying Spectral Thresholds for Modis Cloud Detection

For many parameters derived from satellite imagery, the accurate detection of clouds is essential. The determination of cloud or no cloud for each pixel is known as a cloud mask. Cloud masks are used to filter out cloudy pixels for clear-sky retrievals such as land and sea surface temperature and total precipitable water, and also to detect clouds for cloud parameter retrievals such as cloud phase and cloud top pressure. The Moderate Resolution Imaging Spectroradiometer (MODIS) instruments on NASA’s Terra and Aqua polar-orbiting satellites provide multiple channels of high spatial resolution (250 m 1 km) data that are used to generate a cloud mask. The MODIS cloud mask is an Earth Observing System (EOS) standard product that is generated globally, both night and day. However, the MODIS cloud mask has performance limitations in certain situations, prompting an investigation into producing a MODIS cloud mask for regional applications. Research at the Global Hydrology and Climate Center (GHCC) located within the National Space Science and Technology Center (NSSTC) has produced a robust real-time GOES Imager and Sounder cloud mask (Guillory et al. 1998, Jedlovec and Laws 2001 and 2003, and Haines et al. 2004). The GHCC GOES cloud mask uses composite images to provide both spatially and temporally varying thresholds applied to several cloud tests. This paper discusses the adaptation of the GHCC GOES cloud mask to MODIS data. Following are descriptions of the EOS MODIS cloud mask and the GHCC GOES/MODIS cloud mask algorithm, and validation results generated from manual observations of the cloud mask and corresponding imagery.