Cloud Mask

A cloud mask is a quality layer that flags pixels contaminated by clouds, thin cirrus, or sometimes haze. Masks can be conservative (flagging many pixels) or permissive (missing thin clouds).

Masks are crucial because clouds change reflectance in ways that break indices, classifications, and change detection. The most common failure mode is not obvious thick clouds, but thin cloud and edge effects that slightly shift values.

For monitoring, it often helps to treat cloud-masked pixels as missing data rather than filling them with zeros or naive interpolation. Downstream models should be designed to handle gaps explicitly.