GRASS logo


i.sentinel.mask - Creates clouds and shadows masks for Sentinel-2 images.


imagery, satellite, Sentinel, cloud detection, shadow, reflectance


i.sentinel.mask --help
i.sentinel.mask [-rtsc] [input_file=name] [blue=name] [green=name] [red=name] [nir=name] [nir8a=name] [swir11=name] [swir12=name] [cloud_mask=name] [cloud_raster=name] [shadow_mask=name] [shadow_raster=name] cloud_threshold=integer shadow_threshold=integer [mtd_file=name] [metadata=name] [scale_fac=integer] [--overwrite] [--help] [--verbose] [--quiet] [--ui]


Set computational region to maximum image extent
Do not delete temporary files
Rescale input bands
Compute only the cloud mask
Allow output files to overwrite existing files
Print usage summary
Verbose module output
Quiet module output
Force launching GUI dialog


name of the .txt file with listed input bands
input bands
input bands
input bands
input bands
input bands
input bands
input bands
name of output vector cloud mask
Name of output raster cloud mask
name of output vector shadow mask
name of output raster shadow mask
cloud_threshold=integer [required]
threshold for cleaning small areas from cloud mask (in square meters)
Default: 50000
shadow_threshold=integer [required]
threshold for cleaning small areas from shadow mask (in square meters)
Default: 10000
name of the image metadata file (MTD_TL.xml)
Name of Sentinel metadata json dump
rescale factor
Default: 10000

Table of contents


i.sentinel.mask allows to automatically identify clouds and their shadows in Sentinel-2 images.

The following figures show the difference between the standard cloud mask as provided in Sentinel-2 SAFE products and the cloud detection results obtained with i.sentinel.mask (shadow detection not activated in this example; see below for an example with cloud and shadow masks):

Fig: Standard cloud mask as provided in Sentinel-2 SAFE products
(example: Indonesia)

Fig: Cloud detection with i.sentinel.mask
(example: Indonesia)

The implemented procedure consists essentially of an algorithm based on values thresholds, comparisons and calculations between bands which leads to two different rough maps of clouds and shadows. These require further improvements and elaborations (e.g. transformation from raster to vector, cleaning geometries, removing small areas, checking topology, etc.) carried out in the different steps of the procedure.

Fig: Module General WorkFlow

Fig: Cloud detection procedure

Fig: Shadow detection procedure

The algorithm has been developed starting from rules found in literature (Parmes et. al 2017) and conveniently refined.

Regarding the detection of shadows, the algorithm has been developed to identify only the shadows of clouds on the ground. Obviously, some misclassifications can occur. Often shadows and water have in fact, similar reflectance values which can lead to erroneous classification of water bodies as shadows. Therefore, in order to increase the accuracy of the final shadow mask, a control check is implemented. Clouds and shadows are spatially intersected in order to remove misclassified areas. This means that all those shadow geometries which do not intersect a cloud geometry are removed.

Fig: "Cleaning" procedure of the shadow mask

The algorithm works on reflectance values (Bottom of Atmosphere Reflectance) therefore, the atmospheric correction has to be applied to all input bands (see i.sentinel.preproc or i.atcorr)

All necessary input bands (blue, green, red, nir, nir8a, swir11, swir12) must be imported in GRASS and specified one by one or using an input text file. The text file has to be written following the syntax below: variable=your_map

Tha variables names (blue, green, red, nir, nir8a, swir11, swir12) have to be written precisely like in the example above (e.g. not Blue, nor BLUE but blue), no spaces, empty lines or special characters are permitted.

The final outputs are two different vector maps, one for clouds and one for shadows.

The metadata file (MTD_TL.xml or S2A_OPER_MTD_L1C_TL_MPS__*.xml) is required only if both masks (cloud and shadow) are computed. The module retrieves from this file the sun azimuth and zenith necessary for the shadow mask cleaning phase (see the scheme above)

If flag -s is given all selected bands are rescaled using the specified scale factor [scale_fac=integer]. By default the scale factor is set to 10000, the QUANTIFICATION_VALUE from the metadata of Sentinel-2 images.

The module takes the current region settings into accout. To ignore the current region and set it from the whole image, the flag -r has to be given.

The module allows to compute only the cloud mask or both cloud and shadow masks. If flag -c is given, only the cloud procedure will be performed. The computation of cloud mask is mandatory for shadow mask creation. In fact cloud map is used during the cleaning phase of the shadow mask in order to remove misclassifications.

If the input_file is given, the mtd_file or metadata can also be specified in the this file.


The example illustrates how to run i.sentinel.mask for a Sentinel-2A image (S2A_MSIL1C_20180713T155901_N0206_R097_T17SPV_20180713T211059.SAFE) in the North Carolina location.
Obviously, the image has been imported and atmospheric correction has been performed before running i.sentinel.mask .

i.sentinel.mask -r input_file=path/input_cloud_mask.txt cloud_mask=cloud shadow_mask=shadow cloud_threshold=25000 shadow_threshold=5000 mtd_file=path/MTD_TL.xml

The input text file:


Use -r to set the computational region to the maximum image extend.

Figure1 (left): Sentinel-2A Band 02 - Figure2 (right): Sentinel-2A Band 02 with computed cloud and shadow masks


i.sentinel.mask works for Sentinel-2 images whose names follow both the New Compact Naming Convention (e.g. S2A_MSIL1C_20170527T102031_N0205_R065_T32TMQ_20170527T102301.SAFE) and the Old format Naming Convention (e.g. S2A_OPER_PRD_MSIL1C_PDMC_20160930T155112_R079_V20160930T095022_20160930T095944.SAFE). Therefore, both the MTD_TL.xml and S2A_OPER_MTD_L1C_TL_MPS__*.xml file can be provided as input for the computation of shadow mask. Both files can be found in the GRANULE folder of the downloaded *.SAFE product.
For further information about the naming convention see ESA Sentinel User Guide.


Parmes, E.; Rauste, Y.; Molinier, M.; Andersson, K.; Seitsonen, L. 2017: Automatic Cloud and Shadow Detection in Optical Satellite Imagery Without Using Thermal Bands - Application to Suomi NPP VIIRS Images over Fennoscandia. Remote Sens., 9, 806. (DOI)



Overview of i.sentinel toolset, i.sentinel.import, i.sentinel.preproc, r.import, r.external


Roberta Fagandini, GSoC 2018 student
Moritz Lennert
Roberto Marzocchi


Available at: i.sentinel.mask source code (history)

Main index | Imagery index | Topics index | Keywords index | Graphical index | Full index

© 2003-2020 GRASS Development Team, GRASS GIS 7.8.4dev Reference Manual