Clouds can typically be characterized in the VNIR/SWIR by higher reflectance than the underlying surface. Simple techniques which exploit this trait can be used to discriminate clouds from other features with considerable skill. The EO-1 polar-orbiting platform carries three primary sensors which make measurements of the earth's radiance in this wavelength region. Hyperion is an EO-1 hyperspectral sensor with over 200 contiguous narrow spectral bands from 0.4 to 2.5 µm and approximately 30 m spatial resolution with a swath width of 7.5 km.

Many applications using hyperspectral data require cloud-free conditions; a simple cloud cover detection algorithm can be a useful tool to these efforts. Also, the application of a cloud cover detection algorithm on-board the spacecraft, prior to downlink of collected scenes, can provide considerable savings in downlink resources and processing time.

This presentation begins with a simple cloud cover algorithm which utilizes VNIR (0.4 - 1.0 µm) channels only and expands to include more complete cloud characterization algorithms which also use both NIR and SWIR (1.0 - 2.5 µm) channels to discriminate cloud types and cloud/surface features. The algorithms are applied to a variety of Hyperion scenes which depict various cloud types, surface features and seasonal conditions. An attempt is made to contrast these algorithms in terms of computational complexity versus performance.