For Pre-Collection scenes, the Automated Cloud Cover Algorithm (ACCA) is used to determine the cloud cover. The C Function of Mask (CFMask) is the algorithm used to populate fill, cloud, cloud confidence, cloud shadow, and snow/ice in the processing of Collection 1 Level-1 data products.
Both ACCA and CFMask algorithms perform numerous spectral tests to determine potentially cloud contaminated pixels, and both single bands (shortwave-infrared 2, thermal, red, green, blue) and spectral indices (Normalized Difference Vegetation Index (NDVI) and Normalized Difference Snow Index (NDSI)) are compared against static thresholds to determine potential clouds. CFMask builds upon these tests by using thermal data to contrast land and water temperature from cloud temperature to determine the confidence of each potential cloud pixel. Cloud confidence is assigned as "high", "medium", or "low" confidence.