SENTINEL2 -- Sentinel-2 Products

Driver for Sentinel-2 Level-1B, Level-1C and Level-2A products. (GDAL >= 2.1). Starting with GDAL 2.1.3, Level-1C with "Safe Compact" encoding are also supported.

The SENTINEL2 driver will be used if the main metadata .xml file at the root of a SENTINEL2 data product is opened (whose name is typically S2A_OPER_MTD_SAFL1C_....xml). It can also accept directly .zip files downloaded from the Sentinels Scientific Data Hub

To be able to read the imagery, GDAL must be configured with at least one of the JPEG2000 capable drivers.

SENTINEL-2 data are acquired on 13 spectral bands in the visible and near-infrared (VNIR) and Short-wavelength infrared (SWIR) spectrum, as show in the below table:

Band nameResolution (m)Central wavelength (nm)Band width (nm)Purpose
B01 60 443 20 Aerosol detection
B02 10 490 65 Blue
B03 10 560 35 Green
B04 10 665 30 Red
B05 20 705 15 Vegetation classification
B06 20 740 15 Vegetation classification
B07 20 783 20 Vegetation classification
B08 10 842 115 Near infrared
B08A 20 865 20 Vegetation classification
B09 60 945 20 Water vapour
B10 60 1375 30 Cirrus
B11 20 1610 90 Snow / ice / cloud discrimination
B12 20 2190 180 Snow / ice / cloud discrimination

Level-1B

Level-1B products are composed of several "granules" of ~ 25 km across-track x ~ 23km along-track, in sensor geometry (i.e. non ortho-rectified). Each granule correspond to the imagery captured by one of the 12 detectors accros-track (for a total 290 km swath width). The imagery of each band is put in a separate JPEG2000 file.

Level-1B products are aimed at advanced users.

When opening the main metadata .xml file, the driver will typically expose N * 3 sub-datasets, where N is the number of granules composing the user product, and 3 corresponds to the number of spatial resolutions. There's one for the 4 10m bands, one for the 6 20m bands and one for the 3 60m bands. Caution: the number of such subdatasets can be typically of several hundreds or more.

It is also possible to open the metadata .xml of a given granule, in which case 3 subdatasets will be reported for each of the 3 spatial resolutions.

When opening a subdataset, the georeferencing is made of 5 ground control points for the 4 corner of the images and the center of image.

Level-1C

Level-1C products are organized in ortho-rectified tiles of 100 km x 100 km in UTM WGS84 projections. The imagery of each band is put in a separate JPEG2000 file.

When opening the main metadata .xml file, the driver will typically expose 4 sub-datasets:

All tiles of same resolution and projection are mosaiced together. If a product spans over several UTM zones, they will be exposed as separate subdatasets.

It is also possible to open the metadata .xml file of each tile (only for original L1C encoding, not supported on "Safe Compact" encoding), in which case the driver will typically expose the 4 above mentioned types of sub-datasets.

Level-2A

Similarly to Level-1C, Level-2A products are organized in ortho-rectified tiles of 100 km x 100 km in UTM WGS84 projections. The imagery of each band is put in a separate JPEG2000 file. The values are Bottom-Of-Atmosphere (BOA) reflectances. L2A specific bands are also computed:

When opening the main metadata .xml file, the driver will typically expose 4 sub-datasets:

All tiles of same resolution and projection are mosaiced together. If a product spans over several UTM zones, they will be exposed as separate subdatasets.

Metadata

Metadata of the main metadata .xml file is available in the general metadata domain. The whole XML file is also accessible through the xml:SENTINEL2 metadata domain.

Subdatasets are based on the VRT format, so the definition of this VRT can be obtained by querying the xml:VRT metadata domain.

Performance issues for L1C and L2A

Due to the way Sentinel-2 products are structured, in particular because of the number of JPEG2000 files involved, zoom-out operations can be very slow for products made of many tiles. For interactive display, it can be useful to generate overviews (can be a slow operation by itself). This can be done with the gdaladdo utility on the subdataset name. The overview file is created next to the main metadata .xml file, with the same name, but prefixed with _XX_EPSG_YYYYY.tif.ovr where XX=10m,20m,60m or PREVIEW and YYYYY is the EPSG code.

Trick: if the content of the zoom-out preview is not important for the use case, blank overviews can be created instantaneously by using the NONE resampling method ('-r none' as gdaladdo switch).

When converting a subdataset to another format like tiled GeoTIFF, if using the JP2OpenJPEG driver, the recommended minimum value for the GDAL_CACHEMAX configuration option is (subdataset_width * 2048 * 2 ) / 10000000 if generating a INTERLEAVE=BAND GeoTIFF, or that value multiplied by the number of bands for the default INTERLEAVE=PIXEL configuration. The current versions of the OpenJPEG libraries can also consume a lot of memory to decode a JPEG2000 tile (up to 600MB), so you might want to specify the GDAL_NUM_THREADS configuration option to a reasonable number of threads if you are short of memory (the default value is the total number of virtual CPUs).

Open options

The driver can be passed the following open options:

Note: above open options can also be specified as configuration options, by prefixing the open option name with SENTINEL2_ (e.g. SENTINEL2_ALPHA).

Examples

See Also:

Credits

This driver has been developed by Spatialys with funding from Centre National d'Etudes Spatiales (CNES)