Sentinel-2 Enhanced Cloud Mask Generation

Sentinel-2 Enhanced Cloud Mask Generation provides a more reliable identification of clouds in Sentinel-2 imagery than the cloud mask data included with the L1C and L2A products from ESA. The block processes a granule and generates an image with the same coverage indicating whether pixels are affected by clouds.

Technical information

The cloud identification process is based upon the Fmask algorithm, which identifies potential clouds, along with significant further modifications.

Feature	Info
Input parameters	The block is initiated with the identity of the granule to be processed.
Output format	The cloud mask is provided as an 8-bit single-channel raster in GeoTIFF format. The pixel size is 40 m and the map projection is the same as the source Sentinel-2 data. Pixel encoding:- nodata=0 / - cloud-free=40 / - cloud-affected=240
Training dataset	The improved cloud identification is based upon published methods.

Performance

The enhanced cloud identification generally provides better results than the ESA L1C cloud masks with much reduced occurrence of omission errors. The enhanced cloud identification also provides overall better results than the ESA L2A cloud masks with much reduced commission errors over bright targets, such as urban areas. Areas with mist or with clouds close to the ground can suffer from omission errors.

Use cases

Clouds pose a major challenge to the use of satellite imagery and can be detrimental to automated processing that rely on processes such as spectral analysis, change detection or object detection. The Sentinel-2 Enhanced Cloud Mask is spatially coincident with the source Sentinel-2 data and therefore can be used directly to identify cloud-free pixels in applications that process Sentinel-2 imagery.

More information

For more information about this processing block, please see the provider website.