The cloud cover percentage indicates the amount of area in the remote sensing image extent that is covered with clouds and therefore cannot provide information about the Earth surface conditions.The actual types of clouds included may depend on the product, but the CEOS definition includes cloud shadow. Next to that, from an optical remote sensing point of view, clouds can be roughly classified in: opaque/dense clouds, mainly composed of droplets that are highly reflective in the VIS region and generally located at low-medium altitudes and cirrus, consisting of a large number of thin non-spherical ice crystals that are normally translucent in the VIS region, relatively highly reflective in the SWIR spectrum, and located at high altitude.
The goal of cloud cover percentage is to provide a quality measure of usable information in a surface reflectance image. Earth observation product catalogs support it as a query parameter, to enable searching for products with a cloud cover percentage below a given threshold.
This simplifies for instance use cases that require only fully clear products (0% cloud cover), and may save download and processing resources by only handling images that have some valid pixels. For instance, by only using products with a cloud cover percentage smaller than 99.95%. The measure also gives an estimate of the number of valid observations in a given geographical area, allowing a quick assessment of whether minimal data requirements for a specific use case are met.
The measure is a percentage of actual observations in an image, so pixels where no data was recorded are not included. For derived products, cloud cover pixels are often also flagged separately from pixels where no data was recorded, but this may depend on the data provider. The definition specifically also includes cloud shadow pixels.
Reliable cloud cover percentages depend on good cloud and cloud shadow detection methods. Especially handling of translucent cirrus clouds is an open issue: a product that has a 100% cloud cover percentage due to cirrus clouds might still be usable for some cases, while for other cases they also render the product useless.
The used cloud detection algorithm will also affect the cloud cover percentage. A more strict algorithm will yield higher percentages compared to an algorithm that under detects clouds.
Due to these limitations, cloud cover percentages in product metadata have a fairly high error margin. The user should take this into account when determining optimal cloud cover percentage thresholds for the use case.